Coupled Phenomena in Environmental Geotechnics

COUPLED PHENOMENA IN ENVIRONMENTAL GEOTECHNICS
PROCEEDINGS OF THE INTERNATIONAL SYMPOSIUM, ISSMGE TC 215, TORINO,

ITALY, 13 JULY 2013
Coupled Phenomena in
Environmental Geotechnics
From theoretical and experimental research
to practical applications
Editors
Mario Manassero, Andrea Dominijanni, Sebastiano Foti &
Guido Musso
Dipartimento di Ingegneria Strutturale, Edile e Geotecnica,
Politecnico di Torino, Torino, Italy
Cover photo credits:

On top: microscopic representation of the clay-water system by Mario Manassero
Below: FESEM (Field Emission Scanning Electron Microscopy) micrographs of clay particles by
Markus Tuller & Rossella Viola
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business
2013 Taylor & Francis Group, London, UK
Typeset by MPS Limited, Chennai, India
Printed and bound in Great Britain by CPI Group (UK) Ltd, Croydon, CR0 4YY.
All rights reserved. No part of this publication or the information contained herein may be
reproduced, stored in a retrieval system, or transmitted in any form or by any means,
electronic, mechanical, by photocopying, recording or otherwise, without written prior
permission from the publishers.
Although all care is taken to ensure integrity and the quality of this publication and the
information herein, no responsibility is assumed by the publishers nor the author for any
damage to the property or persons as a result of operation or use of this publication
and/or the information contained herein.
Published by:
CRC Press/Balkema
P.O. Box 11320, 2301 EH, Leiden, The Netherlands
e-mail: Pub.NL@taylorandfrancis.com
www.crcpress.com www.taylorandfrancis.com
ISBN: 978 1 138 00060 5 (Hardback + CD Rom)

ISBN: 978 1 315 88490 5 (eBook)
Coupled Phenomena in Environmental Geotechnics Manassero et al (Eds)

2013 Taylor & Francis Group, London, ISBN 978 1 138 00060 5
Table of contents
Organisers
Preface
Committees
XI
XIII
XV
Special papers
Keynote lectures
Recent advances in understanding and improving the performance of lining and capping systems
for landfill and mining applications
R.K. Rowe
Coupled modelling of barriers for radioactive waste disposal

A. Gens
21
Application of solid potassium permanganate to oxidize VOC vapours: Batch and column experiments
M.G. Mahmoodlu, S.M. Hassanizadeh, N. Hartog & A. Raoof
35
General reports
Membrane behavior in engineered bentonite-based containment barriers: State of the art
C. Shackelford
45
Liquid and gas flow in municipal solid waste

W. Powrie, R.P. Beaven, D.S. Holmes & A.P. Hudson
61
Experimental evidence of anisotropy in municipal solid waste

D. Zekkos
69
Impact of subgrade water content on cation exchange and hydraulic conductivity of

geosynthetic clay liners in composite barriers
C.H. Benson
Advances in the analysis of thermo-active foundations
L. Laloui, T. Mimouni & F. Dupray
79
85
Bio-geo-chemical processes for improvement of soil engineering properties with focus on microbially
induced calcite precipitation
J.D. DeJong, B.C. Martinez, T.R. Ginn & D.C. Nelson
Multiphase aspects of soil contamination by immiscible petroleum hydrocarbons
P. Delage
Electrokinetic remediation of soils at complex contaminated sites: Technology status, challenges,
and opportunities
K.R. Reddy
Cement bentonite cutoff walls for polluted sites
K. Soga, K. Joshi & J.C. Evans
103
119
131
149
Specific lectures
Osmotic phenomena in bentonites
A. Dominijanni, M. Manassero & S. Puma
169
Effect of mechanical settlement on biogas transfer in MSW landfills

L. Oxarango, J.P. Gourc, G. Stoltz & A.J. Tinet
181
Mass loss and volume change: From sand-salt analogues to MSW

J.R. McDougall, I.R. Fleming, R. Thiel, P. Dewaele, D. Parker & D. Kelly
189
Modeling the coupled chemo-hydro-mechanical behavior of compacted active clays

G. Musso, G. Della Vecchia & E. Romero
199
Soil effective thermal conductivity from energy pile thermal tests

A. Bouazza, B. Wang & R.M. Singh
211
Modification of clay consolidation properties using microbial gas production

A.M. Puzrin
221
Injection of zero-valent iron micro- and nano-particles for groundwater remediation: Laboratory
tests and transport modelling
T. Tosco, F. Gastone & R. Sethi
223
Modelling gas and water flow in environmental problems

S. Olivella
231
Migration of aggressive solutions through cement-bentonite slurry walls

E. Fratalocchi, V. Brianzoni, M. Di Sante, F. Mazzieri & E. Pasqualini
243
General papers
Landfill bottom and sides lining systems
Comparison of natural, polymer treated, and sodium activated Ca-bentonites
G. Di Emidio, R.D.F. Verastegui & A. Bezuijen
Dynamic shear behavior of geosynthetic-soil interface and numerical implementation considering
chemical effect
C.W. Kwak, I.J. Park, J.B. Park & J.K. Kim
Influence of membrane behavior on solute diffusion through GCLs
M. Malusis, J. Kang & C. Shackelford
Hydraulic behaviour of activated calcium bentonite mixed with polyionic charged polymers
for landfill and earthwork applications
A. Razakamanantsoa, I. Djeran-Maigre & G. Barast
255
261
267
275
Interaction between clogging of a leachate drainage layer and leakage through a composite liner
R.K. Rowe, Y. Yu & M.S. Hosney
283
Some geotextiles properties useful for HDPE geomembrane puncture protection

G. Stoltz, D. Croissant & N. Touze-Foltz
291
Landfill waste characterization

Geotechnical characterization of waste dredged sediments for Algerian dams
M.A. Bourabah, B. Serbah, N. Abou-Bekr & S. Taibi
299
Dynamic characterization of municipal solid waste by SDMT

F. Castelli & M. Maugeri
307
Correlation between compressibility and degradation for a PTMB waste

M. Conte, J.P. Gourc, P. Carrubba & L. Oxarango
313
A laboratory landfill simulator for physical, geotechnical, chemical and microbial characterization
of solid waste biodegradation processes
X. Fei, D. Zekkos & L. Raskin
Leachate recirculation using horizontal trenches considering the effect of MSW settlement
S.J. Feng & X. Zhang
VI
321
329
2011 East Japan earthquake and tsunami Geoenvironmental challenges

T. Katsumi, T. Inui, A. Takai, K. Endo, H. Sakanakura, T. Yasutaka, Y. Otsuka, H. Suzuki, H. Sakamoto,
M. Okawara & H. Imanishi
335
Composition and degradation effects on the compaction characteristics of municipal solid wastes
H.F. Pulat & Y. Yukselen-Aksoy
341
Shear strength of municipal waste materials from two landfills in Serbia
c
D. Rakic, L. Caki
& S. Cori
347
Stability and settlement analysis of landfills and natural slopes

HydroMechanical coupled model for unsaturated phenomena in a rainfall-induced landslide
W. Arairo, I. Djran-Maigre, F. Prunier & A. Millard
357
A model for anaerobic degradation of municipal solid waste

Y.M. Chen, R.Y. Guo, Y.C. Li & L.T. Zhan
365
Water retention and shrinkage properties of a cemented dredged material

A.C.F. Chiu, X.P. Hu, C.W.W. Ng & G.H. Lei
371
J -integral as a useful fracture parameter for analysis of desiccation cracking in clayey soils
S. Costa, J. Kodikara & J. Xue
377
Soil mass stability analysis using Euler-Lagrange coupling

M.S. S erbulea, A. Andronic, D.M. Manoli & A. Priceputu
383
Piggyback liner detection by means of spatially constrained inversion of SWM data

S. Stocco, C. Castiglia & D. Mainero
387
Three-dimensional rotational-translational mechanism for the stability analysis of landfill

H. Wang & M. Huang
393
Linings for radioactive waste and capping systems for landfills and polluted sites
Environmental restoration with deep roots herbaceous species
N. Era & T. Verrascina
401
Micropore vs. macropore flow: Implications for landfill final cover design
M.V. Khire & D.S. Saravanathiiban
407
Coupled thermo-hydraulic pulse tests on two reference Belgian clay formations

A. Lima, E. Romero, A. Gens, J. Vaunat & X.L. Li
413
Improved correction factor for indirect determination of permeability function

Priono, H. Rahardjo, E.C. Leong, J.Y. Wang & K. Chatterjea
419
Capillary barrier system for landfill capping

H. Rahardjo, A. Satyanaga, F.R. Harnas, J.Y. Wang & E.C. Leong
425
Geosynthetic clay liner gas permeability relationship with moisture content and suction under
pre-conditioning stresses
M.A. Rouf, R.M. Singh, A. Bouazza & R.K. Rowe
435
Modeling of strain localization around the radioactive waste disposal galleries

F. Salehnia, R. Charlier & S. Levasseur
443
Constitutive model for unsaturated soils based on the effective stress

H. Shin & S.R. Lee
451
Crack control of landfill liner and cap materials using nano-alumina powder
M.R. Taha & O.M. Taha
459
Underground energy issues

Coupled phenomena induced by freezing in a granular material
F. Casini, A. Gens, S. Olivella & G.M.B. Viggiani
VII
467
Numerical study of the heating-cooling effects on the geotechnical behaviour of energy piles
A. Di Donna, F. Dupray & L. Laloui
475
Effect of groundwater flow on the THM behavior of an energy pile

F. Dupray, M. Baehler & L. Laloui
483
Thermal, hydraulic and mechanical performances of enhanced grouting materials

for borehole heat exchanger
S. Erol & B. Franois
491
Technical and economic feasibility study of different thermal energy supply systems scenarios
for multi-residential buildings
R. Katzenbach, F. Clauss, D. Casini & H. Frchtenicht
501
Long term performance of heat exchanger piles

C.G. Olgun, S.L. Abdelaziz & J.R. Martin
511
A device for studying simultaneous heat and moisture movement through soils
B.H. Rao
519
Soil thermal behavior of a horizontal ground source heat system

H.R. Thomas, B.D.P. Hepburn & M. Sedighi
525
Natural and anthropogenic bio-chemical processes within soils and rocks

Mechanical and hydraulic effects of deep roots planting on slope stability
M. Cecconi, V. Pane, P. Napoli & C. Zarotti
533
Chemo-mechanical weathering of calcarenites: Experiments & theory

M.O. Ciantia, C. di Prisco & R. Castellanza
541
Chemo-mechanical behaviour of a low activity clay

A. Ferrari, P. Witteveen & L. Laloui
549
Exploiting physico-chemical modification of soils in closed face tunnelling

S.A. Jefferis & A.S. Merritt
557
Combining biomineralization with chemically enhanced drainage in soils

P. Minder & A.M. Puzrin
565
Characterization of polluted sites and related aquifers

Characterization of LNAPL distribution in whole domains subject to precipitation by the simplified
image analysis method
G. Flores, T. Katsumi, T. Inui & H. Ramli
573
Interaction of metal adsorption and solubilization in a lateritic clay

S.G. Gabas & M.E.G. Boscov
579
Factors affecting heavy metal leaching from excavated rocks with natural contamination
T. Inui, T. Katsumi, A. Takai & M. Kamon
587
Novel techniques for simulating and monitoring impact of contaminants on geoenvironment

K.R. Iyer, P. Pathak & D.N. Singh
593
Studying the effects of contamination on the geotechnical properties of clayey soil

M.O. Karkush, A.T. Zaboon & H.M. Hussien
599
Batch equilibrium test methodology for estimating the retardation factor (Rd ) using a
hexachlorocyclohexane (HCH) solution
L.S. Macedo, F.P. Mano & G. Mondelli
609
The resistivity piezocone for high resolution geo-environmental site investigation:

A Brazilian example
M.T. Riyis & H.L. Giacheti
617
A case study of contaminant transport and fate at a landfill site in China

M. Sedighi, H.R. Thomas, H.J. Xie, Y.M. Chen & C. Guan
VIII
625
Effect of geometrical sand drains on consolidation of soft contaminated marine clays using Oedometer
M.V. Shah & A.V. Shroff
631
Geostatistical analysis of groundwater nitrates distribution in the Plaine dAlsace

R.-L. Spacagna, C. de Fouquet & G. Russo
637
Degradation extraction and inerting systems for the reclamation of polluted sites
The use of lime in the stabilization and solidification of lateritic soil contaminated with
tannery effluent
A.O. Eberemu
647
Preliminary results on the stabilization of dredged sediments from the Port of Taranto
A. Federico, A. Murianni, E. Miccoli, C. Vitone, M. Nobile & G. Intern
655
Sustainable remediation of contaminated sites

C.N. Mulligan, S. Dumais & R. Noel-de-Tilly
663
Prediction of time to reduce tsunami sediment salinity by rainfall after the

Great East Japan Earthquake
K. Tada, H. Komine & S. Murakami
Green and sustainable remediation of contaminated Indian Ridge Marsh site in Chicago, USA
E.N. Yargicoglu & K.R. Reddy
671
675
Active and passive barriers for polluted sites

Critical aspects related to Fe0 and Fe0 /pumice PRB design
S. Bilardi, P.S. Calabr & N. Moraci
685
Flow numerical modeling for efficiency assessment of vertical barriers in landfills

E. Koda, P. Osinski & T. Koanka
693
Design, installation and performance of containment system at a petroleum-release site

A. Sanzeni, M. Morando, E. Crescini & A. Guerini
699
Hydraulic barrier performance of soil bentonite mixture cutoff wall

A. Takai, T. Inui, T. Katsumi, M. Kamon & S. Araki
707
Author index
715
IX

Organisers
SYMPOSIUM ORGANISED BY
UNDER THE AUSPICES OF
INTERNATIONAL SOCIETY
OF SOIL MECHANICS
AND GEOTECHNICAL ENGINEERING
INTERNATIONAL GEOSYNTHETICS SOCIETY
ENDORSED BY
POLITECNICO DI TORINO
XI

Preface
Environmental Geotechnics deals with a wide variety of applications, such as the characterization of polluted
sites and landfill waste, the design of containment systems for subsoil pollutant control, radioactive waste
disposal, geo-energy exploitation and bacteria-driven soil modification, among others. In order to obtain reliable
and effective predictions of the actual behavior and performance of these very complex systems, theoretical
and experimental research and advanced design procedures need to take into account the coupled hydro-biochemo-mechanical phenomena that occur at very different scales. Future progress in the scientific state of the
art and substantial advancements in standard practices will therefore be closely related to the development
of shared knowledge among different disciplines. The extension and refinement of theoretical modelling and
the experimentation capabilities stimulated by geo-environmental applications more in general provide the
framework for substantial advancements in the soil and rock mechanics fields.
The International Society for Soil Mechanics and Geotechnical Engineering (ISSMGE) has contributed to
these developments through the activities of an ad hoc Committee (TC 215 Environmental Geotechnics
formerly TC 5). The committee was established under the ISSMGE presidency of Prof. M. Jamiolkowski (1994
1997) and has been very active ever since. Several very lively conferences, symposia and workshops have been
organised culminating in the 6th Edition of the International Conference on Environmental Geotechnics (TC
215 ICEG) which was held in New Delhi, India (2010).
Following this tradition, the international symposium, organised by ISSMGE TC 215 in July 2013 in Torino
(Italy) has focused on Coupled Phenomena in Environmental Geotechnics from theoretical and experimental
research to practical applications. The conference has been considered an opportunity to discuss and share
knowledge, skills and front-edge research activities in the field. By including contributions not only from the
geotechnical community, but also from related and complementary disciplines, the conference has gathered new
experimental evidence, contributions to theoretical developments and innovative applications.
The present volume collects the special lectures and the papers that have been presented at the symposium,
which cover a wide range of fundamental and applied research on geo-environmental engineering topics. Four
sessions of the symposium have dealt with landfills: waste characterization, stability problems, lining and capping
systems. Three sessions have been devoted to polluted sites and their interaction with aquifers both in terms of
characterization and remediation strategies. The remaining two sessions have focused on the emerging topics of
energy issues and bio-chemical processes.
Mario Manassero
Andrea Dominijanni
Sebastiano Foti
Guido Musso
XIII

Committees
Organising Committee
Stefano Aversa
Daniele Cazzuffi
Claudio Soccodato
Claudio Scavia
Mariachiara Zanetti
Mario Manassero
Andrea Dominijanni
Sebastiano Foti
Guido Musso
Italy, President of Italian Geotechnical Society

Italy, President of IGS Italian Chapter
Italy, Secretary of Italian Geotechnical Society
Italy, Director of DISEG
Italy, Director of DIATI
Italy, Symposium Chairman
Italy
Italy
Italy
Scientific Committee
Mario Manassero
Abdelmalek Bouazza
Andrea Dominijanni
Sebastiano Foti
Antonio Gens
Michele Jamiolkowski
Stephan Jefferis
Takeshi Katsumi
Italy, TC Chair
Australia, TC Vice Chair
Italy, TC Secretary
Italy
Spain
Italy
UK
Japan
Rolf Katzenbach
Edward Kavazanjian
Renato Lancellotta
Guido Musso
Erio Pasqualini
R. Kerry Rowe
Charles Shackelford
Kenichi Soga
Germany
USA
Italy
Italy
Italy
Canada
USA
UK
Woon-Hyung Kim
Eugeniusz Koda
Hideo Komine
Aidar B. Konusbaev
Maria Lurdes Lopes
Mario Manassero
Desiree Marin
Catherine Mulligan
Vadim G. Ofrikhter
Marina Pantazidou
Fernando Pardo Santayana
Krishna R. Reddy
Sergio Reyes
Antonio Roque
R. Kerry Rowe
Charles Shackelford
Luis Sopena
Bertrand Soyez
Attila Szabo
Marc Van Den Broeck
Peter Van Impe
Albert Yeung
Dimitrios Zekkos
Korea
Poland
Japan
Kazakhstan
Portugal
Italy
Ecuador
Canada
Russia
Greece
Spain
USA
Argentina
Portugal
Canada
USA
Spain
France
Hungary
Belgium
Belgium
Hong Kong
USA
ISSMGE TC 215 Members

Dali Naidu Arnepalli
Kazem Badv
Craig H. Benson
John Black
Maria Eugenia Gimenez Boscov
Abdelmalek Bouazza
Yunmin Chen
Hang-Seok Choi
Roger Clark
Dimitris Coumoulos
John Cowland
Manoj Datta
Gemmina Di Emidio
Andrea Dominijanni
Evelina Fratalocchi
Antonio Gens
Jean-Pierre Gourc
Georg Heerten
Liming Hu
Stephan Jefferis
Cristina Jommi
Takeshi Katsumi
Edward Kavazanjian
India
Iran
USA
USA
Brazil
Australia
China
Korea
UK
Greece
Hong Kong
India
Belgium
Italy
Italy
Spain
France
Germany
China
UK
Italy
Japan
USA
XV
Special papers
Keynote lectures

Recent advances in understanding and improving the performance of lining

and capping systems for landfill and mining applications
R. Kerry Rowe
GeoEngineering Centre at Queens-RMC, Department of Civil Engineering, Queens University, Kingston, Canada
ABSTRACT: The available evidence suggests that both geosynthetic clay liners (GCLs) and composite liners
with a geomembrane (GMB) over a clay liner have performed extremely well at controlling leakage in field
applications for a couple of decades. However there have also been some problems reported and recent research
has allowed us to have a much better understanding of the key design and construction factors affecting good
and poor performance. This paper examines some of these issues including factors affecting GCL performance
such as the water retention curve of GCLs, subgrade grain size and initial water content, GCL water content
and normal stress on the GCL, the effect of daily thermal cycles on hydration, GCL panel shrinkage, and cation
exchange. Factors affecting composite liner performance examined include the potential for desiccation of the
clay liner under a sustained thermal gradient, GMB/GCL interface transmissivity, wrinkles in the GMB when
the ballast layer is placed over the composite liner, and the potential interaction between wrinkles and GCL
panel overlaps. Recent insights regarding leakage through composite liners are discussed. Although a number
of potential issues with liner performance are discussed, it is concluded that all can be addressed by appropriate
design, material selection, construction, and operations.
INTRODUCTION
2007; Rowe 2012a). However there have also been

challenges when the liners have been exposed to conditions not anticipated in the design and where, until
very recently, there has been a paucity of research. An
example of this is landfills that, due to their operations
(e.g., as bioreactors) or the nature of the waste being
disposed (e.g., combustion ash, aluminum production
wastes etc.), are resulting in temperatures (Table 1)
on or near the liner that raise concerns regarding their
long-term and, in some cases, even short-term viability
(Jafari et al. 2013). High liner temperatures can also be
found in mining applications involving, for example,
brine ponds and some heap leach pads (Table 1). There
is a need for research to examine the effect of these
temperatures and the exposure to different chemicals
on short- and long-term liner performance.
Liners in landfill and mining covers/caps have
exhibited both good and poor performance over the
years. If intact, GMBs are excellent barriers to the
advective migration of fluids (liquids and gases) but
once they have a hole(s) or opening (e.g., due to inferior welding), they can readily transmit relatively large
volumes of fluid through the hole(s) and it is for this
reason that they are often used in conjunction with a
clay liner to form a composite system (e.g., Rowe et al.
2004; Rowe 2005, 2012a).
When used alone, but also in some cases when used
as part of a composite liner, GCLs used in covers have
had mixed success (sometimes performing very well
and sometimes not) in both landfill and mining applications (e.g., Melchior 2002; Adu-Wusu et al. 2002;
Liners have been an essential component of modern municipal solid waste (MSW) and hazardous
waste landfills for 2030 years and more recently
are becoming an essential component of many mining undertakings where applications include ponds
for storage of liquids generated by mining, and liners for tailings dams and heap leach pads. Likewise,
covers have been an important part of landfill design
for decades. In mining, for many years there has been
a recognized need for engineered covers for minimizing impacts due to acid generating waste rock however
there is growing need for covers for other mine waste
(e.g., arsenic bearing gold mine tailings).
High density polyethylene (HDPE) geomembranes
(GMBs) have been used in landfill liners and covers, and for liners in ponds and heap leach pads for
mining applications, but there has also been growing
use of linear low density polyethylene (LLDPE) GMB
liners in heap leach applications. Geosynthetic clay
liners (GCLs) and compacted clay liners (CCLs) have
been used alone in covers and in some bottom liners,
however GCLs used in landfill applications are most
commonly used with a GMB to form a composite liner.
Given the long history of the use of liners in landfill
applications and the research and monitoring that has
been conducted with respect to that application, it is
known that well designed and constructed composite
bottom liners have performed very well in landfills
(Bonaparte et al. 2002; Rowe 2005; Mitchell et al.
to provide a continuous layer of GCL. As manufactured, the most commonly used GCLs comprise
a lower carrier geotextile, a layer of bentonite,
and an upper cover geotextile. The GCL is held
together by needle-punching the upper cover geotextile fibres through the bentonite and into the carrier
geotextile. Many GCLs meet this broad description.
For example, one single Canadian manufacturer has
over 50 different GCLs that they manufacture fitting
this description, with the different products having
different characteristics suitable to different design/site
conditions. Understanding why GCLs perform
remarkably well in many situations and not well in
some others requires an understanding of the many factors that can affect the performance of a GCL which
include (but are not limited to): (a) the type of bentonite, (b) whether or not there is a polymer in the
bentonite, (c) the mass per unit area, MA , of bentonite,
(d) the type and MA of the geotextiles used, (e) the
amount of needle-punching, (f) whether or not the
needle-punched fibres are thermally fused to the carrier geotextile, (g) the presence or absence of a geofilm
bonded to the GCL, the nature of the geofilm, and how
the geofilm is bonded to the carrier geotextile, (h) the
characteristics of the GCL panel overlap, (i) whether
or not the GCL is part of a composite liner, (j) the
presence of wrinkles in the GMB, (k) the initial water
content and particle size distribution of the soil above
and/or below the GCL, (l) geochemical interactions
between the bentonite and the pore water in the soil
adjacent to the GCL, (m) possible interaction of the
bentonite with the fluid to be retained, (n) the amount
of cover soil over the GCL, (o) the level of exposure
to thermal cycles, (p) sustained thermal gradients, and
(q) the stress on the GCL. The following subsections
will explore some aspects of the performance of GCLs
and will include consideration of the role of some of
the factors noted above. Ongoing research will shed
light on other factors not discussed here.
Table 1. Temperature on (or near) liners for different environments (after Rowe, R.K. (2012a). Short and long-term
leakage through composite liners, Canadian Geotechnical
Journal, 49(2): 141169.)
Environment
Temp. ( C)
Ref.a
Normal MSW landfills

(limited moisture addition)
Wet landfills (e.g. bioreactor
landfills) where there is a
significant amount of moisture
Unusual MSW landfills
3040
1,2,3,4,5
4060
5,6,7
6080b
5060c
46
5090b
6570c
85d
5
5
8
5
5
9,10
>143e
70
7093
9,10
11
12
Ash monofills
MSW with aluminum production

waste and leachate recirculation
Nickel heap leach pad
Ponds for highly saline fluid
a
Reference: 1. Brune et al. (1991); 2. Rowe (2005); 3. Koerner
& Koerner (2006); 4. Needham & Knox (2008); 5. Authors
files; 6. Yoshida & Rowe (2003); 7. Koerner et al. (2008); 8.
Klein et al. (2001); 9. Calder and Stark (2010); 10. Stark et al.
(2012); 11. Abdelaal et al. (2011); 12. Lichtwardt & Comer
(1997).
b
No monitors on liner so liner temperature is unknown,
temperature given is in waste about 3 m above liner.
c
Leachate temperature.
d
Temperature in leachate collection pipes.
e
Temperature in waste.
Renken et al. 2005; Meer & Benson 2007; Benson

et al. 2010; Scalia & Benson 2011; Hosney & Rowe
2013) begging the question as to what are the factors
affecting good versus poor performance?
This paper seeks to examine some recent advances
in the understanding of factors affecting the performance of liners in both bottom liner and capping
systems for landfill and mining applications and
the implications of this new understanding regarding
strategies that might be considered for improving liner
performance in these applications. Since the theme of
this paper is recent advances, emphasis will be placed
on research published in the last few years and research
about to be published, although other publications
will be referenced when required to place the current
advances in context. Also, because of space limitations, attention will be focused on advances in understanding related to leakage through GCLs both when
used alone and when used in composite liners. The
many recent advances with respect to understanding
GMB performance, diffusion through liners, and liner
stability relevant to the performance of both bottom
liners and capping systems are not addressed here.
2
2.1 Water retention curve of GCLs

The performance of a GCL as a barrier to fluids (either
liquid or gas) is intimately linked to the uptake of
moisture by the bentonite in the GCL and its resulting degree of saturation. It is often implicitly assumed
by designers that the GCL will be hydrated when it is
needed to act as a fluid barrier. However whether or not
this assumption is realized will depend on many factors. One key factor is the water retention curve (WRC)
of the GCL itself (e.g., Daniel et al. 1993; Barroso
et al. 2006; Southen & Rowe 2007; Acikel et al. 2011;
Beddoe et al. 2010, 2011). A GCLs WRC describes the
relationship between the water content of the GCL and
the suction in the bentonite. The off-the-roll GCL has
a very low water content and very high suction. When
the GCL is placed on a subgrade with lower suction,
the GCL will uptake (suck) water from that soil. This
will cause the GCL water content to increase and its
suction to decrease. One can expect that the uptake of
water would continue until the suction in the GCL had
GEOSYNTHETIC CLAY LINERS
GCLs are manufactured and delivered to site on rolls

that are laid down with adjacent panels overlapped
Table 2. GCL Gravimetric water content of four different

GCL products at saturation at 2 kPa (adapted from Beddoe
et al. 2011).
decreased sufficiently to come into equilibrium with

the suction in the soil in direct contact with the GCL
(i.e., the adjacent soil). Thus intuitively one may expect
that the final equilibrium water content of the GCL will
depend on the WRC of the GCL (discussed in this section) and the WRC of the adjacent soil together with
the initial water content of this soil which will control
its suction (Section 2.2).
Under isothermal conditions one can expect the
GCL to start at a very high suction (corresponding
to its off-the-roll water content) and to move up what
is called the wetting curve as it takes up moisture until
it comes into equilibrium with the adjacent soil. As the
GCL takes up moisture, the bentonite swells and is constrained by the needle-punching between the cover and
carrier geotextiles and the overburden stress. When
the overburden stress is low, most of the resistance
is provided by the needle-punched fibres. If the needle punched fibres are not well anchored they can be
pulled out of the carrier geotextile and there is a permanent change in the GCL. The extent of this change
will depend on how well the fibres are anchored (e.g.,
thermally fused or not). As the GCL pulls out fibres,
the reduction in resistance allows the void ratio of the
GCL to increase and hence more water is required to
saturate the GCL than if the fibres had not pulled out
(or if there were a higher overburden stress).
If a GCL has experienced an increase in water content by moving up the wetting curve and is then heated
(as discussed in later sections), the heat can drive the
moisture out of the GCL and the suction increases as
the water content decreases along what is called the
drying curve. In general, the drying curve is different to the wetting curve and the greater the pullout of
fibres during wetting the greater one might expect the
difference between the wetting and drying curves to
be. Thus to fully understand the hydration of a GCL,
one needs to know the WRC in wetting. To understand
how the GCL will respond to thermal cycles, one also
needs to understand the WRC in drying.
Despite its importance to the uptake of moisture,
the WRC of GCLs has received very little study,
probably because of the inherent difficulty of experimentally obtaining the WRC of the bentonite when
encased in geotextiles and the wide range of suctions that need to be investigated (Beddoe et al. 2010).
Beddoe et al. (2011) obtained the wetting and drying WRCs of four GCL products under 2 kPa normal
stress. They showed that this wetting curve varied quite
substantially depending on the characteristic of the
GCL as manufactured. At a given suction, the gravimetric water content was lowest for a GCL (denoted
as GCL2) which was thermally treated to fuse the
needle-punched fibres to the scrim reinforced nonwoven (i.e., a combination of a woven and nonwoven)
carrier geotextile. This GCL reached a higher degree
of saturation at a given moisture content than the other
GCLs tested. The wetting curves for GCL1 (which was
thermally treated to fuse the needle-punched fibres
to the woven carrier geotextile) and GCL4 (with a
simple needle-punched nonwoven carrier and only
Water content at saturation

Product
Mean (%)
Std. deviation (%)
GCL1
GCL2
GCL3
GCL4
166
130
205
194
10
5
16
8
mechanical bonding) crossed at a suction of 10 kPa.

GCL3 (with a woven carrier and only mechanical
bonding) had the highest water content at any given
suction.
There is always variability in manufactured products like GMBs, GCLs, and geotextiles. Recognizing
this, GCLs are produced to meet minimum average roll
values (MARV). The effect of the GCL characteristics,
as manufactured, and the related variability is most evident for a GCL that is fully hydrated at low stress (e.g.,
below a GMB with little or no cover over the GMB).
When saturated, there was a significant difference in
the average water content and variation in saturated
water content for the four most commonly used GCLs
in North America (Table 2). At saturation, GCL2 had
by far the lowest water content and the lowest variability (standard deviation) because the needle-punched
fibres were most effectively and consistently bonded
to the carrier geotextile thereby minimising fibre pullout as the bentonite hydrates and providing the best
confinement of the bentonite. For GCL1, the thermal
treatment resulted in the next lowest saturated water
content but the bonding to the woven carrier was not
as consistent as that to the scrim-reinforced nonwoven
carrier of GCL2. GCL4 had a substantially higher saturated water content than GCL1 or GCL2. GCL3 had
the highest mean water content and by far the highest
variability indicating that the mechanical bonding of
the needle-punched fibres to a woven GCL was least
effective and most inconsistent in terms of constraining the bentonite as it tried to swell at low stress.
Since it is the degree of saturation, rather than the
actual water content, that most affects GCL performance and since both the hydraulic conductivity and
diffusion coefficient are lower with a lower bulk void
ratio (Petrov & Rowe 1997; Lake & Rowe 2000a,b),
it follows that the lower the saturated water content
the better the performance as a hydraulic and diffusive
barrier (other things being equal).
Swelling of the GCL at low stress results in some
pull-out of fibres that are not well anchored. This is
manifested as a difference in the drying and wetting
WRCs and the degree of hysteresis if the GCL is subjected to drying after hydration. The hysteresis was by
far the least for GCL2 for which there was very little difference between the wetting and drying curves.
Thus in field applications where the GCL may be
Table 3. Hydration of GCLs with time resting on silty sand

at different initial water contents, wfdn (adapted from Rayhani
et al. 2011).
exposed to wetting and drying cycles, such as when

a GCL is below a GMB that is left exposed, one might
infer that GCL2 will be less susceptible to the moisture
changes that result in shrinkage of the GCL and hence
predict less shrinkage for GCL2 than the other three
GCLs in a real field situation.
Southen & Rowe (2007) demonstrated that the
WRC for a GCL was dependent on the applied stress
based on tests at 0.5 kPa and 100 kPa. This is important
in modelling moisture loss and possible desiccation of
GCLs when subject to a sustained thermal gradient
discussed later. At present there is very limited data
on GCL WRCs at stresses greater than 2 kPa. Thus,
more research is required to obtain WRCs for different
GCLs at different stress levels.
2.2
Isothermal moisture uptake by a GCL
When a GCL is placed on soil containing water, the

suctions in the bentonite will induce moisture migration from the soil to the GCL until the suction in the
GCL is in equilibrium with the suction in the adjacent
soil. The rate at which this process takes place and the
time to equilibrium will be dependent on the hydraulic
conductivity of the soil adjacent to the GCL. This
hydraulic conductivity will depend on the water content of the soil and can be expected to reduce as water
is transferred from the soil to the GCL. The suction at
which equilibrium is reached will depend on the WRCs
of both the GCL and soil from which it is hydrating.
Despite its importance, this has received remarkably
little attention. Daniel et al. (1993) reported that an initially air dry GCL reached water content, w, of 88%
after 4045 days resting on sand at 3% gravimetric
water content. Eberle & von Maubeuge (1998) showed
that an initially air dry GCL resting on a sand with
a water content, wfdn , of 810%, reached w = 100%
in less than 24 hours and w = 140% after 60 days.
However, the effect of GCL type, subsoil grain-size
distribution and water content has only recently been
published (Rayhani et al. 2011; Anderson et al. 2012;
Chevrier et al. 2012).
Rayhani et al. (2011) examined the moisture uptake
of three GCLs (those denoted as GCL1, GCL2 and
GCL4 in the Beddoe et al. 2011 study and described
above) in a composite liner under isothermal conditions at 22 C. Tests were conducted for the GCLs
resting on a silty sand subgrade (SM: 35% non-plastic
fines and standard Proctor optimum water content,
wopt , of 11.4%; Table 3) and a poorly graded sand
subgrade (SP: 5% non-plastic fines, wopt = 10.3%;
Table 4). Because of the presence of an overlying
GMB, all water up-take was from the subgrade below.
When the GCLs were placed on the silty sand
subgrade at a water content corresponding to field
capacity (wfdn = 21%), GCL2 was essentially hydrated
in 5 weeks (Sr = 97%, Table 3) and GCL1 and GCL4
had Sr of about 90%. All GCLs had fully hydrated
after 30 weeks, however there was a significant difference in the final water content, wf , between GCL2
(wf = 116%) and GCL1 (wf = 141%) and GCL4
Week
10
GCL
Subgrade
wfdn (%)
Degree of saturation, Sr
(%)
GCL1
GCL1
GCL1
GCL1
GCL2
GCL2
GCL2
GCL2
GCL4
GCL4
GCL4
GCL4
5
10
16
21
5
10
16
21
5
10
16
21
25
52
60
91
32
63
66
97
32
50
57
89
25
57
68
100
33
69
70
100
41
61
68
94
20
24
62
72
100
34
74
75
100
50
66
73
99
30
30
wf
(%)
24
62
73
100
34
74
77
100
55
68
76
99
34
86
102
141
40
85
88
116
83
102
114
149
Table 4. Hydration of GCLs with time resting on poorly

graded sand at different initial water contents (adapted from
Rayhani et al. 2011).
Week
10
GCL
Subgrade
wfdn (%)
Degree of saturation
(%)
GCL2
GCL2
2
10
26
68
27
75
20
27
76
30
30
wf
(%)
27
78
31
90
(wf = 149%). When the GCLs were placed on silty

sand at wfdn = 10% (i.e., just below wopt ), GCL2
hydrated to Sr = 74% (wf = 85%) in 20 weeks, GCL1
took 30 weeks to reach Sr =62% (wf = 86%), and
GCL4 took 30 weeks to reach Sr = 68% (wf = 102%).
Thus when placed on silty sand at wfdn just below
wopt , the GCLs would only be partially hydrated from
the subgrade at low stress. If the stress on the GCL
increased (e.g., due to placing sufficient cover soil or
material to be contained by the liner), one would expect
an increase in Sr (Siemens et al. 2012).
For the same soil initial water content wfdn = 10%,
the GCL placed on the sand (SP; Table 4) hydrated
faster and to a slightly higher final water content than
when on the silty sand (SM; Table 3). When GCL2
was placed on soil close to its residual water content (5% for the silty sand and 2% for the sand) it
only hydrated to Sr = 34% (wf = 40%) and Sr = 27%
(wf = 31%) respectively, highlighting the need for the
GCL to be placed on soil with sufficient moisture to
allow reasonable hydration of the GCL.
Rayhani et al. (2011) reported that the hydration
was not significantly affected by whether the cover or
carrier GCL rested on the subgrade, and inferred that
the WRC of the GCLs were not affected by whether
the carrier geotextile was placed up or down.
the GCL has a high degree of saturation prior to contact

with leachate. The work described above has indicated
that whether or not this will happen is highly dependent
on the initial moisture content and grain size characteristics of the subgrade below the GCL. The work cited
above also shows that the degree of saturation when
this criterion is met can be highly dependent on the
type of GCL. For example at w = 98102%, GCL1,
GCL3 and GCL4 had Sr of 6773%, 52%, and 68%
respectively (see Tables 3 and 5). When GCL2 was at
w = 116% it had Sr = 100% but at a similar w = 114%
GCL 4 only had Sr = 76% (see Table 3). Thus, while
w = 100% would appear to be a reasonable requirement for the scrim-reinforced and thermally treated
GCL2 it appears quite inadequate for the other GCLs
(especially GCL3).
Table 5. Hydration of GCLs with time resting on clayey

sand at different initial water contents (adapted from
Anderson et al. 2012).
Week
GCL
Subgrade
wfdn (%)
Sr
(%)
GCL1
GCL1
GCL1
GCL1
GCL2
GCL2
GCL2
GCL2
GCL3
GCL3
GCL3
GCL3
5
10
15
20
5
10
15
20
5
10
15
20
17
26
28
29
18
26
27
29
13
21
22
24
22
22
w
(%)
16
59
67
90
18
67
75
90
12
48
52
92
24
85
98
130
22
79
88
106
23
90
99
174
2.3 Modelling isothermal moisture uptake

Siemens et al. (2012) used the WRCs from Beddoe
et al. (2011) together with the isothermal hydration
data from Rayhani et al. (2011) to calibrate a numerical model for GCL hydration. They then performed
sensitivity analyses to examine the effect of different subgrades (with different WRC and wfdn ). These
results showed that for a given GCL WRC, the final
equilibrium water content, and hence Sr , was governed
by the initial suction in the subgrade (which depended
on the subgrade WRC and wfdn ). Based on this observation, they proposed a method for assessing the likely
equilibrium water content of the GCL (and hence Sr ).
In essence, knowing the WRC of the subgrade (which
can be inferred from its grain size distribution) and
wfdn , one can infer the suction in the subgrade and
then, using the wetting WRC of the GCL (e.g., from
Beddoe et al. 2011), one can deduce the value of wf
for that suction and hence calculate the likely value
of Sr (knowing the likely saturated water content for
that GCL). Conversely, knowing the desired value of
Sr , one can calculate the required equilibrium value of
wf for the GCL and hence, from the WRC, the corresponding suction. Using this suction and the WRC
for the subgrade one can then deduce the value of wfdn
required to achieve the desired hydration of the GCL.
To apply the approach above, one needs the wetting
WRC of the GCL. This has only been developed for
a limited number of GCLs and mostly at low stress
(2 kPa); there is a need for wetting WRC for a range
of hydrating stresses. This information void leaves
open the question as to what effect does the stress on
the GCL at the time of hydration have on the WRC and
hence the value of Sr that can be achieved for a given
subgrade? To provide some insight, Siemens et al.
(2013) used the WRC data from Beddoe et al. (2011)
and that at 0.5 kPa and 100 kPa from Southen & Rowe
(2007), the consolidation characteristics of GCLs from
Lake & Rowe (2000a), and the GCL hydraulic conductivity data at various confining stresses from Rowe &
Hosney (2013) in conjunction with the numerical procedure developed by Siemens et al. (2012) to provide
the basis for a numerical investigation of the possible
Anderson et al. (2012) examined the hydration of

GCL1, GCL2 and GCL3 resting on a clayey sand
(Table 5). The clayey sand had 21% fines (12%
clay size) and wopt = 11.3%. When the subsoil had a
water content of 20%, all GCLs hydrated to a Sr of
about 90%. When the subsoil water content was at a
little below optimum (wfdn = 10%), GCL2 hydrated
to a Sr = 67% (w = 79%) compared to Sr = 59%
(wf = 85%) for GCL1 and Sr = 48% (wf = 90%) for
GCL3. At a subsoil water content of 5% there was very
little hydration and after 22 weeks Sr was only 18, 16
and 12% for GCL2, GCL1, and GCL3 respectively.
Anderson et al. (2012) compared their results with
those obtained by Rayhani et al. (2011) under otherwise similar conditions and concluded that the subsoil
affected both the rate and degree of hydration. For
example, the time it took to reach a value of Sr = 60%
ranged from 2 weeks over the sand (SP), to 5 weeks
over the silty sand (SM), to 8 weeks over the clayey
sand (SC).
Also for otherwise comparable conditions, the final
equilibrium value of Sr was (slightly) higher for the
GCL over the sand (SP) than over the silty sand (SM)
which was higher than for the GCL over the clayey
sand (SC). These differences are due to the different
WRCs of the three subgrades.
Chevrier et al. (2012) obtained similar findings with
respect to the effect of subgrade grain size and initial water content but at a stress of 4 kPa which was
slightly greater than the 2 kPa used in the Rayhani et al.
(2011) and Anderson et al. (2012) studies. Chevrier
et al. (2012) also demonstrated that the rate of hydration was temperature dependent, with faster hydration
at 45 C than at 20 C than at 5 C.
The French Committee for Geosynthetics
(Fascicule, 1998) requires that the water content of
the bentonite in GCLs used in liners should be at least
100%, presumably with the objective of ensuring that
Table 6. Hydration of GCLs resting on silty sand when

subjected to daily thermal cycles of 2060 C; values given
are immediately following the heating period (adapted from
Rowe et al. 2011b).
effects of higher confining stress on the hydration

behaviour of GCLs. Their results indicated that hydration of a GCL at a higher confining stress would result
in a higher Sr (other things being equal), with the
magnitude of this difference being dependent on wfdn .
These findings provide encouragement for laboratory
studies of the effect of stress on the WRC of GCLs.
Since adequate hydration of a GCL is key to good
performance as a barrier, their performance in covers
and liners can be improved by paying attention to the
water content of the subsoil upon which they are placed
and the selection of an appropriate GCL, especially
under more challenging conditions.
2.4
Effect of moisture uptake and stress on gas

permeability
An example of the linkage between the GCL WRC

and moisture uptake and the potential engineering performance of GCLs as a gas barrier has been recently
reported by Rouf et al. (2013). They examined the gas
permeability of a GCL somewhat similar to GCL2 discussed earlier (the main difference being the use of
powdered Trugel bentonite in this case and granular
Wyoming bentonite in GCL2). Experiments were performed for a range of water contents and suctions for
stresses of 2 kPa and 20 kPa.
At low water content (w < 50% and suctions
>1.6 MPa) the gas permeability was very high and not
much affected by stress. However as the water content
increased, there was not only a substantial (orders of
magnitude) decrease in gas permeability but also a significant effect of the difference in stress between 2 kPa
and 20 kPa. This was likely due to the change in the
WRC with applied stress which resulted in a higher
degree of saturation and lower void ratio at a given
water content at 20 kPa than at 2 kPa.
2.5
Week
GCL
Subgrade
wfdn (%)
Sr
(%)
GCL1
GCL1
GCL1
GCL1
GCL2
GCL2
GCL2
GCL2
GCL4
GCL4
GCL4
5
10
16
21
5
10
16
21
5
16
21
15
23
24
52
16
15
20
66
13
10
45
7
w
(%)
16
25
22
81
14
14
24
100
9
11
85
23
35
30
113
16
16
27
117
14
17
127
the values obtained under isothermal conditions, especially for subsoil with wfdn of 10% and 16% (Table 3).
However for wfdn = 21% (just below field capacity),
even with daily thermal cycles Sr was 100% for GCL2,
81% for GCL1, and 85% for GCL4 (Table 6). Thus the
effect of the daily cycles was very sensitive to wfdn .
A key consideration with respect to possible shrinkage of the GCLs is the change in water content during
a daily cycle (i.e., between the end of the cool period
and the end of the hot period). After a few weeks
of daily cycles, for wfdn = 16% GCL1 experienced a
13% change in water content during a cycle compared to 10% for GCL4 and 2% for GCL2. These
differences are related to the difference in the WRC
discussed earlier. Thus, Rowe et al. (2011b) predicted
that, for the conditions examined and to the extent that
shrinkage is dependent on a change in moisture over
a daily cycle, the susceptibility to shrinkage would be
GCL1 GCL4 > GCL2.
Anderson et al. (2012) examined the performance of
GCL2 and GCL3 over the same clayey sand subgrade
as examined for isothermal conditions and discussed
in the previous section (Table 5). For wfdn of 5%, 10%
and 15%, after 6 weeks of daily thermal cycles the
GCLs at the end of the heating cycle had Sr of only 9
14% (Table 7). Comparing results for GCL2 in Tables
6 and 7, it appears that the clayey sand suppressed
the hydration of the GCLs even more than the silty
sand at the end of the heating cycle. This effect is even
more evident comparing results for subsoil with wfdn
of 2021% where, for the silty sand, GCL2 was fully
hydrated (Sr = 100%) but for the clayey sand it only
reached Sr = 56%.
For a GCL in an exposed composite liner, the cases
of isothermal hydration at 20 C and daily thermal
cycles from 20 to 60 C represent two extremes in
one sense but neither represents a worst case for GCL
shrinkage and neither is likely to be realised given that
GCL moisture uptake with daily thermal cycles
Isothermal conditions, as examined above, represent

ideal conditions for hydration. If a GCL is used as part
of a composite liner that is left exposed to the sun,
the GCL will be subject to thermal cycles. Rowe et al.
(2011b) examined the same GCLs and silty sand subsoil as Rayhani et al. (2011) (discussed in a previous
section) but subjected the GCL to daily thermal cycles.
Based on observations at the Queens University environmental liner test site (QUELTS) located north of
Kingston, Ontario, Canada, in the summer the GMB
can heat to up to 60 C on a sunny day and cool to 20 C
in the evening. For the laboratory study this was idealized as a daily cycle where a GMB above the GCL
resting on the silty sand was heated to 60 C for 12
hours and then allowed to cool to 20 C over the next
12 hours.
For GCLs over silty sand with wfdn of 5%, 10%
and 16%, after 7 weeks of daily thermal cycles the
GCLs at the end of the heating cycle had Sr of only 9
24% (Table 6). These values are very low compared to
Table 7. Hydration of GCLs resting on clayey sand when

subjected to daily thermal cycles of 2256 C; values given
are immediately following the heating period (adapted from
Anderson et al. 2012).
Week
GCL
Subgrade
wfdn (%)
Sr
(%)
GCL2
GCL2
GCL2
GCL2
GCL3
GCL3
GCL3
GCL3
5
10
15
20
5
10
15
20
11
11
12
21
9
9
9
25
11
13
14
56
9
10
10
48
Table 8. Summary of observed GCL panel separation

(gap + original design overlap of 150 mm). Data from
Koerner & Koerner (2005) and Thiel et al. (2006).
GCL1
Slope
Separation
(mm)
Exposure
(months)
w
(%)
W/W
N/W
N/W
N/N
N/N
N/N
N/N
22
18
4
34
18
4
24
450
350
450
1350
450
600
300
60
15
2
36
5
2
2
13
16
16
66
16
20
18
92
Cover geotextile/carrier geotextile; W = woven geotextile;

N = nonwoven geotextile.
(2010) discussed a number of such cases as briefly

summarized below.
GCLs (similar to GCL3 and GCL4) that had experienced significant shrinkage in several cases reported
in Table 8 were installed on a 3H:1V (18 ) slope with a
silty clay subgrade having wfdn = 5% (CETCO 2006).
Water was sprayed on the GCL after placement to raise
the GCL water content to 4045% and it was then
covered with a GMB. Over a 10 month period, during which GMB temperatures were reported to reach
over 60 C, the GCL water content was reported to
have gone from an initial high of about 45% to a low
of 11%. The GCL panels only experienced shrinkage
strains of 0.6%0.9% and overlap movements of only
25 mm to 37 mm. In this case the low shrinkage was
probably due to a number of factors including the low
initial water content of the subsoil (which limited the
rate and amount of hydration in cool periods) and the
fact that the GMB did not appear to have been welded
so there was potential for moisture loss at the GMB
panel edges to the atmosphere that could be expected
to have reduced the moisture cycles.
Gassner (2009) installed a GCL very similar to
GCL1 on a 55 m long 3H:1V slope and covered it
with a GMB and white geotextile. After 18 months
of exposure in Melbourne, Australia, the inspection of
1 m of the overlap one-third of the way down the slope
indicated shrinkage of only 50 mm to 80 mm and the
300 mm overlap used in this case appeared adequate.
Thiel & Rowe (2010) reported that four different
GCL test plots were constructed on a 40 m long 3H:1V
slope at one of the sites where there had been significant panel separation on relatively flat areas (Table 8).
After it was covered with a GMB, the overlaps below
the GMB were periodically inspected. Over a period
of 7 months there was reported to be no notable panel
shrinkage, but a high value of Sr , flowing water, and
internal bentonite erosion resulting from the flowing
water were reported.
It is apparent that the shrinkage is a result of wet-dry
cycling. However the fact that in some case there was
shrinkage that caused a large gap between the GCL
panels and in other cases there was no significant
the weather and cloud cover are likely to change from

day to day. A worse case, for example, is if there is a
period of cloudy weather where the GCL can hydrate
almost isothermally followed by a hot sunny day when
the moisture can evaporate or be driven back into
the soil, followed by another period of cooler/cloudier
weather; thereby allowing larger changes in the water
content of the GCL when heated.
2.6 GCL shrinkage

GCL manufacturers recommend that the GCL be
placed under dry weather conditions and be covered
with a ballast layer (typically 0.3 m or more of cover
soil) shortly after placement. When covered by a GMB
to form a composite liner this recommendation is
often more honoured in the breach that the observance.
There can be a variety of reasons for this but a common one is that the cover soil on the side slopes cannot
be adequately placed quickly for the length of slope
that has been installed and so the liner system is left
exposed for months and sometimes many years. Covering the composite liner in a timely manner avoids a
number of potential problems not the least of which
is the potential for shrinkage of the GCL below the
exposed GMB.
The potential for shrinkage was reported by Thiel &
Richardson (2005) for cases where a GMB was opened
to discover that the GCL had shrunk to the point that
the GCL panels that had originally been overlapped
by 150 mm had shrunk sufficiently to leave a significant gap between the panels. Koerner & Koerner
(2005) and Thiel et al. (2006) reported other cases
(Table 8) which corresponded to shrinkage of the overlap by 3001350 mm and shrinkage strains ranging
from 7% to 31%. These cases demonstrated that there
could be significant shrinkage in periods of as little
as two months on both relatively flat (24 ) slopes as
well as side slopes of 18 to 34 . However there are
also cases where GCLs have been installed and no
significant shrinkage was observed. Thiel and Rowe
shrinkage of a needle-punched GCL with a thermally

treated scrim-reinforced nonwoven carrier geotextile
and granular bentonite (GCL2) was less than that for a
needle-punched GCL with a simple nonwoven carrier
and granular bentonite (GCL4). It is significant that,
in their experiments, the shrinkage strain required to
cause the loss of a 150300 mm panel overlap could be
developed in as little as about five wet-dry cycles and
this helps explain why panel overlap has been lost in as
little as two months in some field situations (Table 8).
Thus the conditions that would allow this number of
significant hydration cycles become important.
The uptake of moisture and the potential of a change
in water content are highly dependent on the initial
water content and the WRC of the soil below the
GCL, and the WRC of the GCL. The thermal cycles
experienced by the GCL also will depend on weather
conditions as discussed earlier.
Shrinkage may be suppressed by bonding between
GCL panels.This could occur due to gluing of panels
together when supplemental bentonite between GCL
panels gets wet and then dries (Brachman et al. 2010),
although this is not a reliable method of suppressing
shrinkage (e.g., it will be lost when the supplemental
bentonite at a location is re-hydrated). A potentially
more reliable means of suppressing shrinkage is to
heat tack the overlaps. Limited evidence suggests that,
when well heat tacked, there is sufficient bond to resist
at least some wet-dry cycles (Thiel and Thiel 2009;
Rowe et al. 2010a; Joshi et al. 2011), although it is
not yet clear to what extent the bond would be sufficient to resist overlap movement under the conditions
that resulted in the loss of panel shrinkage reported in
Table 8.
The laboratory studies reported above have provided insight regarding the relative performance of
different GCLs for idealized conditions; however, to
date, there has been no published study examining
the field performance of different GCLs side by side
under similar field conditions. Brachman et al. (2007)
described the construction of the Queens University environmental liner test site (QUELTS) north of
Kingston, Ontario in 2006 to examine the relative performance of four commonly used North American
GCLs (GCLs 14 discussed earlier) in six different exposed composite liner configurations (some
smooth, some textured GMBs) under as identical conditions as possible (Figure 1). This experiment was
terminated in 2011 after approximately 5 years. The
GMBs and GCLs placed in 2006 were removed in May
2012 and seven new test sections were constructed and
have been monitored since that time (Figure 2). Papers
describing the findings from QUELTS are in preparation; suffice it to say that some GCLs experienced
significant shrinkage while one experienced very little
shrinkage in almost 5 years of exposure.
Based on the available evidence to date, it is
expected that GCL shrinkage is not a concern if
the GCL is placed in accordance with good construction practice and the composite liner is quickly
covered with a ballast layer (e.g., cover soil or leachate
shrinkage indicates that this is a complex issue. The

potential for GCL shrinkage has been the subject of
a number of laboratory studies (Thiel et al. 2006;
Bostwick et al. 2010; Rowe et al. 2010a, 2011a;
Brachman et al. 2010; Joshi et al. 2011) as briefly
summarized below.
Experiments have been conducted on GCLs placed
in a pan and subjected to artificial wet-dry cycling by
adding a prescribed amount of water and then drying the GCL in an oven at 60 C (Thiel et al. 2006;
Bostwick et al. 2010; Rowe et al. 2011a). These experiments showed that all eight GCLs tested have the
potential to shrink significantly, although the method
of manufacture did have an effect with some types of
GCL shrinking more than others. However when the
GCL is used as part of an exposed composite liner in
the field, the moisture cycles experienced by the GCL
will depend on the WRCs of the GCL and the subgrade (which will depend on the type of GCL and grain
size distribution of the subgrade), the initial water content of the subgrade, and the length and nature of the
thermal cycles as discussed above.
Bostwick et al. (2010) reported that the dry mass per
unit area of a GCL did not significantly affect shrinkage provided that the bentonite was evenly distributed.
However, much higher shrinkage was reported for
GCLs with an uneven distribution of bentonite (generally samples with a low average mass per unit area).
GCLs with a lower mass per unit area appear to have
an increased probability of variability in the distribution of bentonite and, GCLs with little to no bentonite
in some areas experience higher shrinkage under otherwise similar test conditions. Thus the level of quality
control needed to ensure sufficient bentonite at every
location on a GCL roll appears to increase as the
average mass of bentonite per unit area is decreased.
Bostwick et al. (2010) also noted that the maximum
shrinkage of 14.4% observed for GCL4 in their study
was well below the maximum of 23% observed by
Thiel et al. (2006) for nominally the same product
using the same test method, suggesting that there
can be a difference in shrinkage of different rolls of
nominally the same product under otherwise similar
conditions which, presumably, is related to difference in the variability of the product being produced
at different times but having the same designation.
Bostwick et al. (2010) and Rowe et al. (2011a) found
that some products exhibited considerable variability
while other products had quite small variability.
Rowe et al. (2011a) examined eight different GCLs
subjected to laboratory wetting and drying cycles.
They reported that the off-the-roll water content could
affect the initial shrinkage but did not significantly
change the final shrinkage (i.e., after many cycles).
They also reported that when GCLs with granular bentonite were wetted to a water content 60% in each
cycle, the actual water content did not have much effect
on the final equilibrium shrinkage but it did change
the rate of shrinkage. A higher water content in the
hydration cycle resulted in the maximum shrinkage
being reached in fewer cycles. They found that the
10
0.751 m of soil above the GCL. High k (108

106 m/s) was reported for samples with water content
less than 85%, while the k ranged between 1010
109 m/s for samples with a water content greater than
100% (with the lowest values being obtained from well
hydrated samples exhumed after 11 years). Thus the
water content of the sample appears to be an important
factor.
Scalia & Benson (2011) exhumed GCLs from six
composite liners with 0.31.1 m of cover soil (thickness depended on site) in landfill final covers after
4.7 to 6.7 years. The SI dropped to 711 mL/2g (i.e.,
typical of calcium bentonite) at four sites whereas it
was 1222 mL/2 g at the two other sites. There was an
almost 5 order of magnitude variation in k over the six
sites (9.3 1012 k 2.1 107 m/s).
Rowe & Abdelatty (2012a) examined cation
exchange, and its effect on k, for a GCL resting (a)
directly on Ca-rich soil (pore water: 17001800 mg/L
Ca2+ ) (b) on a 0.3 m foundation layer (pore water:
200300 mg/L Ca2+ ) over the Ca-rich soil, and (c)
on the foundation layer alone. All soils were silty sand
at an initial water content of 13.9%.
For the GCL on the foundation layer alone, the GCL
w increased to a final equilibrium value of 113% after
500 days. There was no change in the k of the GCL.
For the GCL directly on the Ca-rich soil, w increased
to a maximum of 86% and then decreased to 67% due
to cation exchange under isothermal conditions, showing that simple chemical interaction with the subgrade
can result in a decrease in GCL water content (other
things being equal). A similar trend was observed for
the case where the GCL was on the foundation layer
over the Ca-rich soil but the maximum w was 96% and
it reduced to 80% at equilibrium.There was no desiccation cracking of the GCLs. The value of k (at 15 kPa)
increased from the initial value of 3 1011 m/s to
2 1010 m/s when resting directly on the Ca-rich
soil and 7 1011 m/s (after 3 years) when on the
foundation layer over the Ca-rich soil.
Hosney & Rowe (2013) examined the field performance of three needle-punched GCL products covered
with up to 1 m of cover soil (gravelly sand) as part
of a cover over an abandoned gold mine in Nova
Scotia, Canada. One GCL (GCL1 discussed earlier)
had sodium bentonite sandwiched between a woven
and a nonwoven geotextile. The other two GCLs had
polymer enhanced sodium bentonite and a nonwoven
cover geotextile but different carriers: one a scrim
reinforced nonwoven geotextile, the other a woven
geotextile laminated with a polypropylene film. GCL
samples were exhumed after 1 and 2 years. Almost all
the moisture uptake occurred in the first year and there
was very little change at year 2. All exhumed GCLs
had w > 80%.
After 2 years, GCL1 with untreated sodium bentonite experienced almost complete cation exchange
(SI dropped from 26 to 810 mL/2g) but still maintained k 5 1011 m/s at locations where there was
0.7 m of cover soil above the GCL. The SI of GCL
with polymer enhanced bentonite decreased from 24
Figure 1. Queens University environmental liner test site

(QUELTS) north of Kingston Ontario, Canada during construction in 2006.
Figure 2. QUELTS II following re-construction in 2012.
collection system). If the GCL must be left exposed,

its performance in covers and liners can be improved
by (a) placing panels with 300 mm of overlap, (b) heat
tacking the overlaps, and (c) using a GCL that has
shown low shrinkage in the field. These three strategies (especially if used together) will minimize the risk
of panel separation. However, at this time, there is no
assurance that they will prevent panel separation under
worst case conditions; the composite liner should still
be covered as quickly as possible.
2.7 Cation exchange

Cation exchange is a well-recognised mechanism
whereby the sodium ions in the smectite, which is
the component of sodium bentonite responsible for
its low hydraulic conductivity, are exchanged with
other cations in the permeating fluid, the pore water
of the soil adjacent to the GCL or, especially for
activated sodium bentonite, in calcium bearing minerals in the bentonite itself. There have been instances
where cation exchange (probably coupled with wetdry cycles) has resulted in substantial increases in
hydraulic conductivity, k, from the low (typically
k 5 1011 m/s) values for GCLs as they come off
the roll. For example, Benson et al. (2007) reported
that after 15 months in a landfill final cover, a GCL
similar to GCL4 discussed above and with 760 mm
of soil above the GCL experienced an increase in
k to between 1.8 107 m/s and 6.9 107 m/s and
a decrease in swell index (SI ) from 24 mL/2g to
715 mL/2g. These changes were attributed to the
exchange of Ca2+ and Mg2+ in the adjacent soils for
Na+ in the bentonite together with dehydration.
Meer & Benson (2007) reported findings for GCLs
exhumed from four different landfill covers, with
11
to 14 mL/2g and k 3 1011 m/s when there was

0.7 m of cover soil above the GCL. The SI of the
GCL with polymer enhanced bentonite and a carrier
geotextile coated with a geofilm decreased from 25
to 1315 mL/2g but because the k is controlled by the
thin geofilm, the k value did not change and remained
around 5 1012 m/s (virgin k value). After two years
of exposure to the natural climatic cycles in this humid
climate, all GCLs were performing adequately despite
cation exchange when the cover soil 0.7 m. When
the cover above the GCL was reduced to 0.5 m, GCL1
experienced a similar change in SI but k increased to
1 1010 m/s.
The work described above highlights the complex
interaction between variables such as the availability
of exchangeable cations (especially Ca2+ and Mg2+ ),
wfdn of the subgrade, w of the exhumed GCL, the
amount of cover soil above the GCL, and the type of
GCL, on the k of GCLs used in covers. While there
are certainly cases where there has been a very substantial increase in k of the GCL in covers, there are
also cases where it has performed very well, suggesting that performance is related to both local conditions
(soil and climate) coupled with the details of the design
(including the type of GCL used).
In addition to ensuring the subgrade below the GCL
has adequate moisture to allow reasonable hydration
of the GCL, as discussed earlier, the performance
of GCLs in low stress applications (especially those
where they could be subjected to wet-dry cycling)
can be improved by (a) ensuring there is sufficient
cover soil above the GCL (the amount needed will
depend on local conditions/climate), and (b) selecting
the appropriate GCL for the application.
the cracking and the size of the desiccation cracks,

and (c) the stress on the GCL (higher stress increases
the ability of the GCL to self-heal; other things being
equal). The ability to rehydrate to a low k may also be
reduced by the chemical composition of the permeant
(Petrov & Rowe 1997) even if there was little initial
cation exchange.
The size of the desiccation cracks can be expected
to affect how well it can self-heal. The larger the cracks
the more difficult it will be for the bentonite to selfheal, especially after some cation exchange and at low
stress. It is well recognised that desiccation of a GCL is
caused by wetting followed by drying; however it is not
well recognised that the size of the desiccation cracks
is a function of a number of factors. For example, Take
et al. (2012a) showed that the size of cracks developed
in a GCLs bentonite core depended on both the type of
GCL and the rate of drying. They examined the desiccation of GCL2 and GCL4 under isothermal drying at
20 C and 60 C. As discussed earlier, GCL2 had more
effective anchorage of its needle-punched fibres than
GCL4 (due to the combination of needle-punching to
the scrim reinforced carrier geotextile and the thermal
bonding of the fibres in GCL2). As a result, GCL2
had much thinner cracks when desiccated than GCL4
(which relied on mechanical anchorage to a nonwoven carrier geotextile for fibre anchorage). Also, for
GCL4, the rate of drying was important to crack size,
with slow drying giving significantly larger cracks
than rapid oven drying at 60 C. The effect of drying
rate was much less for GCL2.
A laboratory study of the potential for desiccation
of a GCL in a composite liner due to sustained thermal gradient was reported by Southen & Rowe (2004,
2005). This work was extended to consider double
composite liners by Azad et al. (2011). These studies showed experimentally that, under some circumstances, desiccation would occur whereas in others
there was no desiccation. Factors affecting whether
or not there was desiccation included the: (a) initial
subsoil water content (higher wfdn decreased the potential for desiccation); (b) type of GCL; (c) bentonite
MA (higher MA appeared to reduce potential desiccation); (d) carrier geotextile thickness (higher MA
appeared to reduce potential desiccation); (e) temperature gradient (lower gradient decreased the potential
for desiccation).
The numerical model DESICCATE (Zhou & Rowe
2003) has been successfully used to model the experimental data discussed above (Southen & Rowe 2011;
Azad et al. 2012) and used to examine the effects of
factors such as liner temperature, overburden stress,
subsoil initial water content, grain size of the subgrade,
and depth to the water table (Hoor & Rowe 2013).
Based on the studies to date, and more research
is required, it appears that the potential for desiccation can be reduced, and hence the performance of
liner systems improved, by ensuring an adequate water
content of subsoil and by one or more of the following: (a) limiting the temperature gradient (e.g., by
avoiding leachate recirculation or disposing of waste
GCL DESICCATION IN COMPOSITE LINERS
A GCL may desiccate as a result of one or more wetdry cycle. This may occur because the GCL is in an
exposed composite liner (i.e., the mechanisms giving
rise to shrinkage discussed earlier), the GCL is in a
cover liner without adequate cover soil to protect it
from significant wet-dry cycles due to climatic cycles,
or because it is in a composite bottom liner that initially
hydrates (as discussed in Sections 2.22.5) and is then
dried by the thermal gradient generated by hot waste
(e.g., municipal solid waste where there is leachate
recirculation or disposal of combustion ash). When
it desiccates, the GCL k value will be high but, provided that there is not too much cation exchange, it can
quickly reduce again to low values (Southen & Rowe
2005) because of the ability of the sodium bentonite to
swell and self-heal on re-wetting (i.e., when it comes
into contact with the fluid that is to be contained).
However as indicated by some of the cases cited in the
previous section, when desiccation is combined with
cation exchange the self-healing capacity is reduced
or lost, with the magnitude of the effect depending on
(a) the amount of cation exchange, (b) the extent of
12
work of Mendes et al. (2010) and Rowe & Abdelatty

(2012b, 2013).
For otherwise similar conditions (50 kPa, smooth
GMB, and permeation with water), Mendes et al.
(2010) showed that two calcium bentonite GCLs
had almost identical values of despite a two
order of magnitude difference in k. Specifically, for
the GCL with k = 5.8 108 m/s (MA = 5730 g/m2 )
was 3.0 1011 m2 /s while for the GCL with
k = 6.9 1010 m/s (MA = 10590 g/m2 ) was 2.8
1011 m2 /s. These values of were only slightly
larger than two values obtained for two different
sodium bentonite GCLs (viz: = 2.4 1011 m2 /s
for a GCL with k = 1.6 1011 m/s (MA = 7400 g/m2 )
and = 2.6 1011 m2 /s for a GCL with k = 3.2
1011 m/s (MA = 5410 g/m2 ) respectively). Comparing the extreme cases, a 360,000% increase in k of the
GCL only corresponded to an increase in by 25%.
Thus k and do not appear to be correlated suggesting
that is controlled by factors other than k.
Rowe & Abdelatty (2012b, 2013) arrived at a similar general finding by a very different route. They
performed four experiments to examine contaminant
transport through a hole in a GMB overlying a GCL
(at 100 kPa). These tests were of much larger diameter than previous GMB/GCL transmissivity tests.
Rowe & Abdelatty (2012b) initially permeated the
simulated barrier system with distilled water and from
the measured flows at steady state Rowe & Abdelatty
(2013) deduced 2.3 1011 m2 /s. The permeant
was changed to a NaCl solution and, at steady state,
had decreased to 1.1 1011 m2 /s. Thus, despite the
fact that the k of the GCL increased up to a factor
of 8 from about 5 1011 m/s when permeated with
water to 24 1010 m/s when permeated with a 0.14
M NaCl solution, the value of decreased by a factor of 2; a totally opposite trend. This is attributed
to an improvement in the interface characteristics of
the GMB/GCL interface as the GCL consolidated in
response to the chemical interaction that increased k.
This also demonstrates that for the same GCL, the values of k and are uncoupled when the permeant is
changed. This is good news for GCLs since Rowe &
Abdelatty (2013) showed that for the conditions they
examined an increase in k due to clay-leachate interaction did not significantly increase leakage through
the composite liner and the GCL continued to fulfill
its primary function of minimizing leakage though a
hole in the GMB.
The available evidence suggests that even with good
construction, the GMB/CCL interface transmissivity is several orders of magnitude higher than for a
GMB/GCL (Rowe 2005, 2012a).
such as combustion ash that would increase liner

temperatures Hoor & Rowe 2013), or (b) removing
heat (Rowe et al. 2010b), or (c) providing insulation
(Hoor & Rowe 2012).
COMPOSITE LINERS OVER A DRAINAGE

LAYER
GCLs may be used as part of a primary composite

liner in a double liner system. When a GCL is to
be placed over a relatively uniform gravel or geonet/
geocomposite drainage layer (e.g., a leak detection
layer), there are two issues that need careful consideration in design. First, there is the issue of hydration.
Consideration should be given to how the GCL will
hydrate when placed on a material that typically has
very low water content as placed. For example, if the
GCL is hydrated from the leachate it may increase the
hydraulic conductivity of the GCL (Petrov & Rowe
1997). Second, there is the potential for internal erosion of the bentonite in the GCL when it rests on a
gravel or geocomposite/geonet layer. Rowe & Orsini
(2003) showed that different types of GCL had different susceptibility to internal erosion with GCLs having
a scrim reinforced carrier performing better than those
with either a woven or nonwoven carrier.
One design solution to these issues is to place the
GCL on a well graded foundation layer that can act
as a filter to prevent internal erosion and also provide
a source of moisture for hydration of the GCL. An
added benefit is that the layer provides at least partial
insulation that could reduce the temperature on the secondary liner (Rowe & Hoor 2009) and hence reduce
the many negative impacts of higher temperature (e.g.,
higher leakage, higher diffusion, reduced GMB service life, potential desiccation of the secondary clay
liner e.g., Rowe 2005). This will increase cost but
improve long-term performance.
GMB/GCL INTERFACE TRANSMISSIVITY
Rowe (2012a) reviewed the published data relating

to GMB/GCL interface transmissivity, . Considering
permeation with water, published values fall in the
range 6 1012 m2 /s 2 1010 m2 /s with an
average of about 4 1011 m2 /s for reinforced GCLs
over a range of stresses from 770 kPa. For GCLs
containing sodium bentonite, the average was about
2 1011 m2 /s (at 50 kPa). A study of the effect of
applied stress in the range 25200 kPa (Barroso et al.
2010) did not indicate a strong relationship between
stress and for the cases considered. It also did not
appear to matter greatly whether the GMB is smooth
or textured in terms of for the conditions examined
by Barroso et al. (2008).
Perhaps the most interesting recent finding is the
lack of correlation between and k arising from the
6 WRINKLES
Although design drawings show composite liners with
GMBs directly and continuously on the underlying
clay liner, in reality when it comes time to cover the
13
GMB with the ballast or leachate collection layer,

there will usually be wrinkles in the GMB (Figure 3). Although the presence of wrinkles in GMBs
has long been recognised and Rowe (1998) developed
an equation for calculating leakage through GMBs
with wrinkles, it is only recently that the geometry and
connectedness of wrinkles has been well documented
(Take et al. 2007; Chappel et al. 2012a,b; Rowe et al.
2012a,b). Wrinkles arise from the thermal expansion
of the GMB when heated (e.g., by the sun). They may
align with geometrically regular features (e.g., creases
arising from manufacture of blown film GMBs, seams
which run parallel to the roll, or GCL panel overlaps),
or they may occur at random locations (all of these are
evident in Figure 3). The height and width of wrinkles
as well as the percent area of a liner below wrinkles and
length of connected wrinkles have been documented
for a range of sites, times of day, and times of year in
the references cited above.
The studies cited above were conducted in eastern
Canada. For the conditions examined, it appears that if
the GMB were to be covered by the leachate collection
layer before 8:00 am or after 4:00 pm there would be
about 6 to 7 connected wrinkles per hectare each with
connected length L 200 m. The length of connected
wrinkle would increase with more solar radiation with
the maximum connected wrinkle length often being
observed (on a sunny day) at around 1:00 pm. For an
unrestrained area of GMB of about 0.2 ha, there could
be a connected wrinkle with a length of up to 2000 m
at around 1:00 pm. Assuming a similar situation for
each part of the GMB covered there could be five such
connected wrinkles per hectare and a relatively high
probability that at least one hole in the GMB will align
with a the long connected wrinkle per hectare. As the
restrained area that is covered at a given time increases
so too does the length of the longest connected wrinkle.
For example, Chappel et al. (2012b) reported the case
of a GMB area of 0.61 ha where the longest connected
wrinkle was 5330 m.
The studies cited above have indicated that, at least
in eastern Canada, wrinkles in HDPE GMBs induced
by thermal expansion may reach heights of 0.2 m, and
occasionally more, but are typically about 0.06 m in
height and wrinkle widths may reach up to 0.5 m but
are usually in the 0.20.3 m range.
Figure 3. Wrinkles in GMB at QUELTS. In this photo GMB

rolls run east-west (left to right) and GCL panels run north
south (top to bottom) (modified from Rowe et al. 2012b and
Rowe 2012a).
loss of overlap because fluid (leachate or gas) could

easily migrate through a hole anywhere in the wrinkle
and then laterally below the wrinkle to a point where
the wrinkle intersects the place where overlap is lost
and again composite liner action is lost. This further
highlights the need to avoid loss of panel overlap.
Where significant wrinkles (i.e., with height
>30 mm at the time they are covered) coincide with
panel overlaps, the overlap below the wrinkle will be
unloaded after the waste or cover soil is placed. Thus
even if there is no loss of panel overlap, a significant reduction in panel overlap could represent a weak
point for leakage as noted by Brachman et al. (2011).
Current work at Queens University is examining the
potential for leakage through panel overlaps using
unique equipment (geosynthetic landfill liner simulators with internal diameters up to 1 m). This equipment
allows consideration of different overlaps, presence
and absence of supplemental bentonite, heat tacking
etc. on the hydraulic performance of overlaps both
parallel and perpendicular to the wrinkle for applied
stresses up to 250 kPa.
8
7
OVERLAPS
LEAKAGE THROUGH COMPOSITE LINERS
It has been well recognised for many years (e.g.,

Giroud & Bonaparte 1989; Rowe 1998) that leakage
through a composite liner will depend on the head on
the liner, the number and size of holes in the GMB,
and the k of the clay component of the liner. Less well
recognised has been the importance of the interface
transmissivity, , between the GMB and the GCL and
the role of wrinkles, although there have been some
significant advances in understanding with respect
to these two parameters over the last few years as
discussed in previous sections. Rowe (2005, 2012a)
examined the use of the Rowe (1998) equation for calculating leakage through the primary liner (allowing
The potential for shrinkage of GCL panels causing a

loss of overlap was discussed earlier. Any such gap
would be problematic if it were beneath a hole in the
GMB since the composite liner action would be lost at
that location. However even more problematic would
be the situation where there was a hole in a wrinkle that
either ran directly above the overlap (as may happen in
some cases; Figure 3) or if the wrinkle were to cross
the point where there was a loss of overlap (which
can be easily envisaged based on Figure 3). In this
case the hole no longer needs to align directly with the
14
above the GMB is sufficient to ensure composite

liner action, the leakage can be reduced by using
a composite liner involving a GMB/GCL/CCL. In
this case the GCL provides good interface properties thereby minimizing leachate flow at the
GMB/GCL interface and the CCL controls the leakage where leachate can migrate between the GMB
and GCL. Rowe (2012a) showed that the performance of the GMB/GCL/CCL system was substantially better than either a GMB/GCL or GMB/CCL
system.
for wrinkles) in double lined landfills and compared

calculated and observed values. In 2005 there was a
paucity of data on connected wrinkle length and the
equation was used to infer what winkle length was
required to explain the observed leakages. In 2012,
the data on wrinkles discussed in an earlier section
was available and Rowe (2012a) showed that the Rowe
(1998) equation explains the observed leakage in landfills for heads and connected wrinkle lengths typical
of those observed in construction provided the liner is
not covered in the middle of a sunny day.
Rowe & Hosney (2010) and Rowe (2012a) have
examined the effect of some of the key factors discussed above on the calculated leakage through composite landfill bottom liners. Some of their conclusions
are summarized below assuming good liner construction and liner temperature less than 40 C. The justification for these comments is not repeated here and the
interested reader is referred to the original papers for
that information.
CONSTRUCTION ISSUES
Composite liner performance is highly related to construction issues (Rowe 2012a, b). With respect to the
issues discussed in this paper, the performance of
composite liners can be improved by considering the
following.
An increase in liner temperature (Table 1) can

greatly increase the calculated leakage, with the
leakage at 60 C and 80 C being about double and
triple that at 20 C respectively, other things being
equal.
Well-constructed composite liners substantially
reduce leakage compared to a GMB, GCL or CCL
alone.
Leakage through a composite liner with a
GMB/GCL is substantially less than through a
GMB/CCL.
Advection begins to dominate contaminant transport when leakage, Q, exceeds about 200 lphd and
careful consideration must be given to the potential
impact on groundwater for these situations.
Assuming typical landfill design parameters (see
Rowe 2012a), the calculated leakage through
GMB/GCL composite was Q 14 lphd for L 200 m
and Q 50 lphd for L 700 m. For a GMB/CCL
(0.6 m thick), the corresponding leakage was
Q 83 lphd for L 200 m and Q 580 lphd for
L 700 m.
Provided that the connected wrinkle length is kept
relatively low (L 125 m), then a GMB/GCL composite liner can be expected to give Q < 50 lphd
(Darcy flux <0.002 m/a) for leachate heads less
than 3 m and hence remains modest. These leakages are considered small and usually of no practical
consequence for MSW landfills. At these leakages,
diffusion is likely to be the dominant contaminant
transport mechanism for contaminants (e.g., volatile
organic compounds) that can readily diffuse through
a GMB.
Provided that the connected wrinkle length is kept
relatively low (L 125 m), then a GMB/CCL composite liner can be expected to give Q < 200 lphd
for leachate head 1.5 m.
Leakage through composite liners used in lagoons
may be substantially larger than through landfill liners. This is because of the much lower stress on the
liner and, typically, much higher heads. If the ballast
15
Ensure a suitably prepared CCL surface for placing the GMB (without stones, clods etc.) since this
will influence the interface transmissivity and hence
potential leakage.
Ensure no linear features/irregularities in surface of
CCL (e.g. tyre tracks).
Keep the CCL surface sufficiently moist to avoid
desiccation cracks before the GMB is placed.
Ensure that the subgrade below the GCL is firm and
unyielding (>90% standard Proctor density) with
no abrupt changes in elevation (e.g., ruts). It should
be proof rolled with a smooth drum roller immediately prior to placement of the panels. There should
be no foreign matter or stones loose on the surface
or penetrating out of the subgrade >10 mm.
The subgrade upon which the GCL is to be placed
should be at an appropriate water content to allow
hydration of the GCL.
Cover the composite liner with the protection layer
and drainage or ballast layer quickly to avoid (a)
desiccation cracking of a CCL which, as shown by
Rowe (2012a) can substantially increase leakage, or
(b) significant GCL shrinkage and opening of a gap
between GCL panels. Of these the most critical is
the need to cover a GMB over a CCL since severe
desiccation can occur in a matter of hours on a hot
sunny day when the GMB can heat to over 60 C and
perhaps as much as 70 C.
Minimize wrinkles by ensuring that the GMB is
placed early in the morning or late in the day
when the number of wrinkles is smallest (Chappel et al. 2012a,b; Rowe et al. 2012a,b; Take et al.
2012b).
To ensure the desired performance, good inspection is

required during construction to check that the design
conditions are actually achieved and factors such as
those highlighted above (plus other issues not discussed here e.g. see Rowe 2012b) are addressed.
The old adage: You get what you inspect, not what
you expect should always be kept in mind.
10
CONCLUSIONS
panels that can occur when subjected to thermal cycles

(e.g., when a composite liner is left exposed to heating
by the sun). There are strategies that can be adopted
to minimize the risk of panel separation in cases
where the liner must be left exposed as discussed
herein. Additional insights regarding issues arising
from leaving composite liners exposed, and mitigation
measures if they must be left exposed, will be provided
when the findings from the studies at QUELTS are
published.
GCLs used in covers are particularly prone to problems if not properly designed and constructed because
of the combination of low stress and potential for
wet-dry cycles. The risk of problems can be substantially reduced by (a) ensuring that the subgrade
below the GCL has an adequate water content to
allow reasonable hydration of the GCL, and (b) ensuring there is sufficient cover soil above the GCL, and
(c) selecting the appropriate GCL for the application
(some types of GCLs will perform much better than
others).
Presently available evidence suggest that, under
some circumstance, GCLs (or CCLs) used in a composite liner where the temperature may exceed 35 C
could be desiccated by the thermal gradient. Although
more research is required, it appears that the potential for desiccation can be reduced, and hence the
performance of liner systems improved, by ensuring an adequate water content of subsoil, and one
or more of the following: (a) limiting the temperature gradient (e.g., by avoiding leachate recirculation
or disposing of waste such as combustion ash that
would increase liner temperatures), or (b) by removing
heat, or (c) by providing insulation to control the liner
temperature.
When GCLs are used in primary liners as part of
double liner systems, special consideration should be
given to how the GCL will hydrate and the potential for
internal erosion of the bentonite in the GCL if the GCL
rests over a uniform gravel or geonet/geocomposite
drainage layer. One design solution to these issues is
to place a well graded foundation layer between the
GCL and the leak detection system. This layer can act
as a filter to prevent internal erosion and a source of
moisture for hydration of the GCL. In addition, this
layer will provide at least some insulation to reduce
the temperature on the secondary liner.
For a well-constructed composite liner, the interface transmissivity, between the GMB and clay liner
is at least as important as the hydraulic conductivity, k
of the liner. Recent research suggests that k and are
not correlated and hence that is controlled by factors
other than k of the GCL (or CCL). One study showed
that, for the conditions examined, permeation of a composite liner with a salt solution caused k to increase by
almost an order of magnitude but decreased by a factor of two and there was negligible change in leakage
through the hole in the GMB.
Well-constructed composite liners substantially
reduce leakage compared to a GMB, GCL or CCL
alone, however the leakage through composite liners
Recent advances in the understanding of factors affecting the performance of liners in both bottom liner and
capping systems for landfill and mining applications
have been explored with respect to GCL and composite liner performance. Some of the implications for
improving liner performance are highlighted below.
The available evidence suggests that both GCLs
and composite liners have performed extremely well
at controlling leakage in field applications for a couple of decades. However there have also been some
problems and recent research has allowed us to have a
much better understanding of the key design and construction factors affecting performance. There are also
operational issues that can affect the performance of
the system. Particular attention must be given to factors that could cause the liner temperature to exceed
40 C since they can have many negative effects on liner
performance (e.g., higher leakage, higher diffusion,
reduced GMB service life, and potential desiccation
of the clay liner). Special design will be required for
liners to be used in an environment where the liner
temperature will exceed 3540 C.
Factors that can affect the performance of a GCL
include: (a) the type of bentonite, (b) the mass per
unit area of bentonite, (c) the type and mass per unit
area of the geotextiles used, (d) the amount of needlepunching, (e) whether or not the needle-punched fibres
are thermally fused to the carrier geotextile, (f) the
presence or absence of a geofilm bonded to the GCL,
the nature of the geofilm, and how the geofilm is
bonded to the carrier geotextile, (g) the characteristics
of the GCL panel overlap, (h) whether or not the GCL
is part of a composite liner, (i) the presence of wrinkles
in the GMB, (j) the initial water content and particle
size distribution of the soil above and/or below the
GCL, (k) geochemical interactions between the bentonite and the pore water in the soil adjacent to the
GCL, (l) interaction of the bentonite with the fluid to
be retained, (m) the amount of cover soil over the GCL,
(n) the level of exposure to thermal cycles, (o) thermal
gradients, and (p) and the stress on the GCL.
A key parameter affecting GCL performance as a
fluid barrier (to liquid or gas) is the degree of saturation. A GCL should have a high degree of saturation
before it is required to restrict the migration of either
gas or liquids (especially liquids that could chemically
interact with the bentonite). The degree of saturation
of the GCL will depend on the type of GCL (since
they have different water retention curves) and the
grain size distribution and initial water content of the
subgrade (or cover soil if no GMB is present). As discussed, techniques have been developed that allow the
estimation of a target water content required to achieve
a desired degree of saturation of the GCL for a given
subgrade.
GCLs should be covered with at least 0.3 m of ballast (e.g., leachate collection system or cover soil)
shortly after GCL placement to minimize the risk of
a number of problems including shrinkage of GCL
16
ACKNOWLEDGEMENTS
is highly dependent on the length of connected wrinkles that intersect a hole in the GMB. Wrinkles present
at the time the GMB is covered with the ballast layer
increase the area of the underlying clay liner in contact
with leachate and hence leakage. Even more problematic would be the situation where there was a hole in
a wrinkle intersecting a location where there has been
loss of GCL panel overlap (e.g., due to panel shrinkage). In this case the hole in the GMB does not need
to align directly with the loss of overlap because fluid
(leachate or gas) could easily migrate through a hole
anywhere in the wrinkle and then laterally below the
wrinkle to a point where the wrinkle intersects the
place where overlap is lost. In this case, composite
liner action is lost. This further highlights the need to
avoid loss of panel overlap.
When a composite liner involves a CCL, the performance is highly dependent on the interface transmissivity between the GMB and CCL and desiccation
of the surface of the CCL either before or after the
GMB is placed. The available evidence suggests that
even with good construction the GMB/CCL interface
transmissivity is orders of magnitude higher than for
a GMB/GCL. As a consequence, leakage through a
composite liner with a GMB/GCL is substantially less
than through a GMB/CCL.
This paper has discussed a number of potential
problems with liner performance. However all can
be avoided by appropriate design, material selection,
construction and operations. There are many examples where liners have exhibited excellent long-term
performance.
11
The research reported in this paper is supported

by an NSERC Strategic Grant. The support of the
Killam Trust in the form of a Killam Fellowship
to Dr. Rowe is greatly appreciated. The author is
very grateful to: his colleagues in the GeoEngineering Centre at Queens-RMC, especially Drs. Richard
Brachman, Andy Take and Greg Siemens; industrial partners, Terrafix Geosynthetics Inc., Terrafix
Environmental Technologies Inc., TAG Environmental Inc, Solmax International, Ontario Ministry of
Environment, AECOM, AMEC Earth and Environmental, Golder Associates Ltd., Canadian Nuclear
Safety Commission, CTT Group, Knight Pisold and
Thiel Engineering for their advice and support with
various aspects of this research that forms the basis
for much of the information presented. The author is
also very grateful for the assistance of M. Hosney in
the preparation of the paper and to L. Ashe, M. Hosney
and A. Rentz for their review of the manuscript. The
views expressed herein are those of the author and not
necessarily those of the people who have assisted with
the research or review of the manuscript.
REFERENCES
Abdelaal, F.B., Rowe, R.K., Smith, M., & Thiel, R. 2011. OIT
depletion in HDPE geomembranes used in contact with
solutions having very high and low pH. In Proceedings of
the 14th Pan-American conference of Soil Mechanics and
Geotechnical Engineering, Toronto, October, paper #483,
CD-ROM, 7p.
Acikel, A. S., Singh, R. M., Bouazza, A., Gates, W. P., &
Rowe, R.K. 2011. Water retention behaviour of unsaturated geosynthetic clay liners. In Proceedings of the 12th
International Conference of International Association
for Computer Methods and Advances in Geomechanics
(IACMAG), Melbourne, May, 2011, 626630.
Adu-Wusu, C., Renken, K., Yanful, E.K., & Lanteigne, L.
2002. Engineered covers on acid-generating waste rock
at Whistle mine, Ontario. In Proceedings of the 55th
Canadian Geotechnical Conference, Niagara Falls,
Ontario, 131138.
Anderson, R., Rayhani, M.T., & Rowe, R.K. 2012. Laboratory
investigation of GCL hydration from clayey sand subsoil.
Geotextiles and Geomembranes 31: 3138.
Azad, F., Rowe, R.K., El-Zein, A., & Airey, D. 2011. Laboratory Investigation of Thermally Induced Desiccation of
GCLs in Double Composite Liner Systems. Geotextiles
and Geomembranes 29 (6): 534543.
Azad, F., El-Zein, A., Rowe, R.K., & Airey, D. 2012. Modelling of thermally induced desiccation of geosynthetic
clay liners in double composite liner systems. Geotextiles
and Geomembranes 34: 2828.
Barroso, M., Touze-Foltz, N., & Saidi, F.K. 2006. Validation
of the use of filter paper suction measurements for the
determination of GCL water retention curves. In Proceedings of the 8th International Conference on Geosynthetics,
Yokohama, pp. 171174.
Barroso, M., Touze-Foltz, N., & von Maubeuge, K. 2008.
Influence of the Textured Structure of Geomembranes on
the Flow Rate Through Geomembrane-GCL Composite
Liners. In proceedings of the EuroGeo4, CD-ROM, 8p.
NOTATION
CCL
GCL
GMB
HDPE
k
L
lphd
MA
MARV
MSW
Q
QUELTS
SI
Sr
w
wf
wfdn
wopt
WRC
Compacted clay liner

Geosynthetic clay liner
Geomembrane
High density polyethylene
Hydraulic conductivity (m/s)
Length of connected wrinkle (m)
Litres per hectare per day
Mass per unit area of GCL (g/m2 )
Minimum average roll value
Municipal solid waste
Leakage through a composite liner
(lphd)
Queens University Environmental
Liner Test Site
Swell index
Gravimetric water content of GCL
Equilibrium gravimetric water content
of GCL
Gravimetric water content of subgrade
Standard Proctor optimum water content
of subgrade
Water retention curve
GMB/GCL interface transmissivity
(m2 /s)
17
Barroso, M.C.P., Lopes, M.G.D.A., & Bergamini, G. 2010.

Effect of the waste pressure on fluid migration through
geomembrane defects. In Proceedings of the 9th International Conference on Geosynthetics, Guaruja, Brazil,
959962.
Beddoe, R.A., Take, W.A., & Rowe, R.K. 2010. Development
of suction measurement techniques to quantify the water
retention behaviour of GCLs. Geosynthetics International
17(5): 301312.
Beddoe, R.A., Take, W.A., & Rowe, R.K. 2011. Water
retention behaviour of geosynthetic clay liners. Journal of Geotechnical and Geoenvironmental Engineering
137(11): 10281038.
Benson, C., Thorstad, P.A., Jo, H.Y., & Rock, S.A. 2007.
Hydraulic performance of geosynthetic clay liners in a
landfill final cover. Journal of Geotechnical and Geoenvironmental Engineering 133(7): 814827.
Benson, C., Kucukkirca, E., & Scalia, J. 2010. Properties of
geosynthetics exhumed from a final cover at a solid waste
landfill. Geotextiles and Geomembranes 28(6): 536546.
Bonaparte, R., Daniel, D., & Koerner, R.M. 2002.Assessment
and recommendations for improving the performance of
waste containment systems. EPA Report, EPA/600/R02/099.
Bostwick, L.E., Rowe, R.K., Take, W.A., & Brachman, R.W.I.
2010. Anisotropy and directional shrinkage of geosynthetic clay liners. Geosynthetics International 17(3):
114.
Brachman, R.W.I., Rowe, R.K., Take, W.A., Arnepalli, N.,
Chappel, M.J., Bostwick, L.E., & Beddoe, R. 2007.
Queens composite geosynthetic liner experimental site.
In Proceedings of the 60th Canadian Geotechnical
Conference, Ottawa, 21352142.
Brachman, R.W.I, Gudina, S., Rowe, R.K., & Take, W.A.
2010. Adhesion from supplemental bentonite placed at
GCL overlaps. In Proceedings of the 63rd Canadian
Geotechnical Conference, Calgary, 13591364.
Brachman, R.W.I., Joshi, P., Rowe, R. K., & Gudina, S.
2011. Physical response of geomembrane wrinkles near
GCL overlaps. In Proceedings of the Geo-Frontiers 2011,
Dallas, 11521161.
Brune, M., Ramke, H.G., Collins, H., & Hanert, H.H. 1991.
Incrustations process in drainage systems of sanitary landfills. In Proceedings of the 3rd International Landfill
Symposium, Santa Margherita di Pula, Cagliari, Italy, 14
18 October 1991. Edited by T. H. Christensen, R. Cossu, &
R. Stegmann. CISA, Environmental Sanitary Engineering
Centre, Cagliari, Italy, 9991035.
Calder, G.V., & Stark, T.D. 2010. Aluminum reactions and
problems in municipal solid waste landfills. Journal of
Hazardous, Toxic, and Radioactive Waste Management
15(1), 18.
CETCO 2006. TR-339 GCL Panel shrinkage lab and field
trials
Chappel, M.J., Brachman, R.W.I., Take, W.A., & Rowe,
R.K. 2012a. Large-scale quantification of wrinkles in a
smooth, black, HDPE geomembrane. Journal of Geotechnical and Geoenvironmental Engineering 138 (6):
671679.
Chappel, M.J., Rowe, R.K., Brachman, R.W.I., & Take, W.A.
2012b. A comparison of geomembrane wrinkles for nine
field cases. Geosynthetics International 19(6): 453469.
Chevrier, B., Cazaux, D., Didier, G., Gamet, M., &
Guyonnet, D. 2012. Influence of subgrade, temperature
and confining pressure on GCL hydration. Geotextiles and
Geomem-branes 33: 16.
Daniel, D.E., Shan, H.-Y. & Anderson, J.D. 1993. Effects
of partial wetting on the performance of the bentonite
component of a geosynthetic clay liner. In Proceedings of

the Geosynthetics 93, IFAI, St. Paul, MN, 3: 14821496.
Eberle, M.A., & von Maubeuge, K. 1998. Measuring the
in-situ moisture content of geosynthetic clay liners
(GCLs) using time domain reflectometry (TDR). In Proceedings of the 6th International Conference on Geosynthetics, Atlanta. 1: 205210.
Fascicule n 12, 1998. Recommandations gnrales pour
la ralisation dtanchit par Gosynthtiques Bentonitiques. French Geosynthetic Committee, Paris, France.
Gassner, F. 2009. Field observation of GCL shrinkage at a site
in Melbourne Australia. Geotextiles and Geomembranes
27 (5): 406408.
Giroud, J.P., & Bonaparte, R. 1989. Leakage through liners constructed with geomembranes Part II: Composite
liners. Geotextiles and Geomembranes 8: 71111.
Hoor, A., & Rowe, R.K. 2012. Application of tire chips
to reduce the temperature of secondary geomembranes
in municipal solid waste landfills. Waste Management
32:901911.
Hoor, A., & Rowe, R.K. 2013. Potential for desiccation of
geosynthetic clay liners used in barrier systems. Journal
of Geotechnical and Geoenvironmental Engineering (in
press).
Hosney, M.S., & Rowe, R.K. 2013. Changes in GCL properties after two years in a cover over arsenic-rich tailings.
Canadian Geotechnical Journal (in press).
Jafari, N.H, Stark, T., & Rowe, R.K. 2013. Service life of a
landfill liner system subjected to elevated temperatures
Journal of Hazardous, Toxic, and Radioactive Waste (in
press).
Joshi, P., Brachman, R.W.I., & Rowe, R.K. 2011. Heat-tacked
overlap strength of four GCLs. In Proceedings of the
14th Pan-American conference of Soil Mechanics and
Geotechnical Engineering, Toronto, October, paper #752,
CD-ROM, 7p.
Klein, R., Baumann, T., Kahapka, E., & Niessner, R. 2001.
Temperature development in a modern municipal solid
waste incineration (MSWI) bottom ash landfill with
regard to sustainable waste management. Journal of
Hazardous Materials 83: 265280.
Koerner, R.M., & Koerner, G.R., 2005. GRI White Paper #5
on In-Situ Separation of GCL Panels Beneath Exposed
Geomembranes. Geosynthetic Research Institute.
Koerner, G.R., & Koerner, R.M., 2006. Long term temperature monitoring of geomembranes at dry and wet landfills.
Geotextiles and Geomembranes 24(1): 7277.
Koerner, R.M., Koerner, G.R., Eith, A.W., & Ballod, C.P.
2008. Geomembrane temperature monitoring at dry and
wet landfills. In Proceedings of the Global Waste Management Symposium, Copper Mountain, Colo., 710
September 2008. National Solid Wastes Management
Association, Washington, D.C (CD-ROM).
Lake, C.B., & Rowe, R.K. 2000a. Swelling characteristics of
needlepunched, thermally treated GCLs. Geotextiles and
Geomembranes 18(2): 77102.
Lake, C.B., & Rowe, R.K. 2000b. Diffusion of sodium and
chloride through geosynthetic clay liners. Geotextiles and
Geomembranes 18(2): 102132.
Lichtwardt, M.A., & Comer, A.I. 1997. Geosynthetics in
salinity-gradient ponds. Geotechnical Fabrics Report
15(3): 2428.
Meer, S., & Benson, C. 2007. Hydraulic conductivity of
geosynthetic clay liners exhumed from landfill final
cover. Journal of Geotechnical and Geoenvironmental
Engineering 133(5): 550563.
Melchior, S. 2002. Field studies and excavations of geosynthetic clay liner in landfill covers. Clay Geosynthetic
18
Barriers, Zanzinger, Koerner and Gartung (Eds.),

A.A. Balkema Publishers, Lisse, pp. 321330.
Mendes, M.J.A., Touz-Foltz, N., Palmeira, E.M., & Pierson,
P. 2010. Influence of structural and material properties
of GCLs on interface flow in composite liners due to
geomembrane defects. Geosynthetics International 17(1):
3447.
Mitchell, J.K., Alvarez-Cohen, L., Atekwana, E., Burns, S.E.,
Gilbert, R.B., Kavazanjian, E., ORiordan, W.H., Rowe,
R.K., Shackelford, C.D., Sharma, H.D., & Yesiller, N.
2007. Assessment of the performance of engineered waste
containment barriers. Report of the Committee to Assess
the Performance of Engineered Barriers to National
Research Council of the U.S. National Academies, ISBN13: 978-0-309-10809-6 (121p.).
Needham, A., & Knox, K. 2008. Long-term basal temperatures at Beddington farmlands landfill and temperature
influences on HDPE liner service life. In Proceedings of
the EuroGeo4, Paper #325,CD-ROM 7p.
Petrov, R.J., & Rowe, R.K. 1997. Geosynthetic clay liner
compatibility by hydraulic conductivity testing: Factors
impacting performance Canadian Geotechnical Journal
34(6):863885.
Rayhani, M.T., Rowe, R.K., Brachman, R.W.I., Take, W.A., &
Siemens, G. 2011. Factors affecting GCL hydration under
isothermal conditions. Geotextiles and Geomembranes
29(6):525533.
Renken, K., Yanful, E.K., & Mchaina, D.M. 2005. Field
performance evaluation of soil-based cover systems to
mitigate ARD for the closure of a potentially acidgenerating tailings storage facility. British Columbia
Reclamation Symposium. Abbotsford, BC. Sept. 1922,
2005.
Rouf, M.A., Singh, R.M., Bouazza, A., & Rowe, R.K. 2013.
Geosynthetic clay liner gas permeability relationship
with moisture content and suction under pre-conditioning
stresses. In Proceedings of the International Symposium
on Coupled Phenomena in Environmental Geotechnics,
Torino, Italy, June.
Rowe, R.K. 1998. Geosynthetics and the minimization of
contaminant migration through barrier systems beneath
solid waste. In Proceedings of the 6th International
Conference on Geosynthetics, Atlanta, Ga., 2529 March
1998.Industrial Fabrics Association International, St.
Paul, Minn.1:27103.
Rowe, R.K. 2005. Long-term performance of contaminant
barrier systems. 45th Rankine Lecture, Geotechnique
55(9): 631678.
Rowe, R.K 2012a. Short and long-term leakage through
composite liners. The 7th Arthur Casagrande Lecture,
Canadian Geotechnical Journal 49(2): 141169.
Rowe, R.K. 2012b. Design and construction of barrier systems to minimize environmental impacts due to municipal
solid waste leachate and gas. In Proceedings of the 3rd
Indian Geotechnical Society: Ferroco Terzaghi Oration,
Indian Geotechnical Journal 42(4): 223256.
Rowe, R.K., & Abdelatty, K. 2012a. Effect of a calcium-rich
soil on the performance of an overlying GCL. Journal of
Geotechnical and Geoenvironmental Engineering 138(4):
423431.
Rowe, R.K., & Abdelatty, K. 2012b. Modeling contaminant transport through composite liner with a hole in the
geomembrane. Canadian Geotechnical Journal 49 (7):
773781.
Rowe, R.K., & Abdelatty, K. 2013. Leakage and contaminant
transport through a single hole in the geomembrane component of a composite liner. Journal of Geotechnical and
Geoenvironmental Engineering (in press)
Rowe, R.K., & Hoor, A. 2009. Predicted temperatures and

service lives of secondary geomembrane landfill liners.
Geosynthetics International, 16(2):7182.
Rowe, R.K., & Hosney, M.S. 2010. A systems engineering
approach to minimizing leachate leakage from landfills. Theme lecture, In Proceedings of the 9th International Conference on Geosynthetics, Guaruja, Brazil,
501510.
Rowe, R.K., & Hosney, M.S. 2013. Laboratory investigation
of GCL performance for covering arsenic contaminated
mine waste. Geotextiles and Geomembranes (in review).
Rowe, R.K., & Orsini, C. 2003. Effect of GCL and subgrade type on internal erosion in GCLs. Geotextiles and
Geomembrane 21(1):124.
Rowe, R.K., Quigley, R.M., Brachman, R.W.I., & Booker,
J.R. 2004. Barrier Systems for Waste Disposal Facilities,
Taylor & Francis Books Ltd (E & FN Spon) London, 587p.
Rowe, R.K., Bostwick, L.E., & Thiel, R. 2010a. Shrinkage
characteristics of heat-tacked GCL seams. Geotextiles and
Geomembranes 28(4):352359.
Rowe, R.K., Hoor, A., & Pollard, A. 2010b. Examination
of a method for reducing the temperature of MSW landfill liners. Journal of Environmental Engineering 136(8):
794803.
Rowe, R.K., Bostwick, L.E., & Take, W.A. 2011a. Effect of
GCL properties on shrinkage when subjected to wet-dry
cycles. Journal of Geotechnical and Geoenvironmental
Engineering 137(11): 10191027.
Rowe, R.K., Rayhani, M.T., Take, W.A., Siemens, G., &
Brachman, R.W.I. 2011b. GCL Hydration under Simulated Daily Thermal Cycles. Geosynthetics International
18(4): 196205.
Rowe, R.K., Yang, P., Chappel, M.J., Brachman, R.W.I., &
Take, W.A. 2012a. Wrinkling of a geomembrane on a compacted clay liner on a slope. Invited paper for Special Issue
on Geosynthetics, Geotechnical Engineering, Journal of
the South East Asian Geotechnical Society 43(3): 1118.
Rowe, R.K., Chappel, M.J., Brachman, R.W.I., & Take, W.A.
2012b. Field monitoring of geomembrane wrinkles at a
composite liner test site. Canadian Geotechnical Journal
49(10): 11961211.
Scalia, J., & Benson, C. 2011. Hydraulic conductivity of
geosynthetic clay liners exhumed from landfill final
covers with composite barriers. Journal of Geotechnical
and Geoenvironmental Engineering 137(1): 113.
Siemens, G.A., Take, W.A., Rowe, R.K., & Brachman,
R.W.I. 2012. Numerical investigation of transient hydration of unsaturated geosynthetic clay liners. Geosynthetics
International 19(3): 232251.
Siemens, G.A., Take, W.A., Rowe, R.K., & Brachman, R.W.I.
2013. Effect of confining stress on the transient hydration
of unsaturated GCLs. In Proceedings of the 18th International Conference on Soil Mechanics and Geotechnical
Engineering, Paris, September 2013.
Southen, J.M., & Rowe, R. K. 2004. Investigation of the
behavior of geosynthetic clay liners subjected to thermal gradients in basal liner applications. Journal of
ASTM International 1(2) ID JAI11470, Online; Available:
www.astm.org. and also published by ASTM in hardcopy
in Advances in Geosynthetic Clay Liner Technology,
Mackey, R.E. and von Maubeuge, K. (Eds), pp. 121133.
Southen, J.M., & Rowe, R.K. 2005. Laboratory investigation of GCL desiccation in a composite liner subjected to
thermal gradients. Journal of Geotechnical and Geoenvironmental Engineering 131(7): 925935.
Southen, J.M., & Rowe, R.K. 2007. Evaluation of the water
retention curve for geosynthetic clay liners. Geotextiles
and Geomembranes 25(1): 29.
19
Southen, J.M., and Rowe, R.K. 2011. Numerical modelling

of thermally induced desiccation of geosynthetic clay
liners observed in laboratory experiments. Geosynthetics
International 18 (5): 289303.
Stark, T.D., Martin, J.W., Gerbasi, G.T., Thalhamer, T., &
Gortner, R.E., 2012. Aluminum waste reaction indicators in a subtitle D landfill. Journal of Geotechnical and
Geoenvironmental Engineering 138(3): 252261.
Take, W.A., Chappel, M.J., Brachman, R.W.I., & Rowe, R.K.
2007. Quantifying geomembrane wrinkles using aerial
photography and digital image processing. Geosynthetics
International 14(4): 219227.
Take, W.A., Rowe, R.K., Munro, M., Kerr, A., & Schreiner,
J. 2009. Development of X-ray imaging techniques to
investigate the internal shrinkage mechanism of GCLs.
In Proceedings of the Geosynthetics 2009, Salt Lake City,
U.S.A., pp. 1928.
Take, W.A., Potvin, J., Siemens, G.A., & Kerr, A. 2012a.
Comparison of desiccation cracking patterns between airand oven-dried GCL specimens. EuroGeo 2012 3: 7176.
Take, W.A, Watson, E., Brachman, R.W.I., & Rowe, R.K.
2012b. Thermal expansion and contraction of geomembrane liners subjected to solar exposure and backfilling.
Journal of Geotechnical and Geoenvironmental Engineering 138(11): 13871397.
Thiel, R., & Richardson, G. 2005. Concern for GCL shrinkage
when installed on slopes. In Proceedings of the GRI-18
at GeoFrontiers, Austin, Texas, GII Publications, Folsom,

Pa., paper 2.31.
Thiel, R., & Rowe, R.K. 2010. Technical developments
related to the problem of GCL panel shrinkage when
placed below an exposed geomembrane. In Proceedings
of the 3rd International Symposium on GCLs, Wrzburg,
Germany, September 2010, 93102.
Thiel, R., & Thiel, C. 2009. GCL Shrinkage A Possible
Solution. GFR 27(1): 1021.
Thiel, R., Giroud, J.P., Erickson, R., Criley, K., & Bryk, J.
2006. Laboratory measurements of GCL shrinkage under
cyclic changes in temperature and hydration conditions.
In Proceedings of the 8th International Conference on
Geosynthetics, Yokohama, Japan, 1822 September 2006.
Millpress Science Publishers, Rotterdam, the Netherland
1: 2144.
Zhou, Y., & Rowe, R.K. 2003. Development of a technique
for modelling clay liner desiccation. International Journal
for Numerical and Analytical Methods in Geomechanics
27 (6):473493.
Yoshida, H., & Rowe, R.K. 2003. Consideration of landfill
liner temperature. In Proceedings of the 8th International
Landfill Symposium, Santa. Margherita di Pula, Cagliari,
Italy, 15 October 2001. CISA, Environmental Sanitary
Engineering Centre, Cagliari, Italy. (CD Rom, 9p).
20

Coupled modelling of barriers for radioactive waste disposal

A. Gens
Universitat Politcnica de Catalunya (Barcelona Tech), Barcelona, Spain
ABSTRACT: The paper first presents an overview of the issues associated with radioactive waste disposal and
follows with a brief description of deep geological disposal for high level radioactive waste. The performance
assessment of these underground repositories requires a good understanding of the basic phenomena that affect
the succession of barriers separating the waste from the biosphere. A key safety element in the design of these
facilities is an engineered barrier, made up of highly expansive clay, that surrounds the canister containing the
waste. This barrier is subjected to a complex set of coupled thermo-hydro-mechanical (THM) phenomena that
must be correctly modelled and understood. The paper presents a coupled THM formulation that incorporates
the most relevant processes occurring in the barrier and their mutual interactions. The formulation is then applied
to the modelling of an in situ test that simulates repository conditions at full scale. The tests involves heating and
hydration of the engineered barrier. It is shown that the modelling results represent satisfactorily the observations
from the test, including those obtained dafter a partial dismantling stage. This exercise demonstrates that the
basic phenomena appear to be well understood and that the formulation and associated computer code are useful
tools for the coupled analysis of this type of THM processes.
INTRODUCTION AND BACKGROUND
they contain most of the radioactivity (over 95% of

the total), the waste quantities are relatively low, of
the order of 10,000 tonnes per year (Mc Combie
et al. 2000). Generally, HLW waste is left to cool
off before disposal for a number of years.
Radioactivity is a natural phenomenon that arises

mainly from the disintegration of the unstable nuclei
of some elements but there are some other natural
sources like cosmic rays and their interaction with the
atmosphere. However, the issue of nuclear waste is
associated basically with human use of radioactivity
for industrial, medical and military purposes as well
as for energy production. These activities produce a
wide variety of wastes that there are often classified
in the following classes:
In this paper, attention will be focused on HLW

where the performance of coupled modelling plays a
key role in many aspects. They are practically exclusively the product of energy generation in nuclear
power plants. They are mainly constituted by spent fuel
rods or solidified high-level waste from reprocessing.
According to the recent IAEA (2012) data, there are
at present 436 nuclear power plants in operation with a
net installed capacity of about 371 GW (gigawatts). As
shown in Figure 1, total nuclear power production has
remained approximately steady (2518 terawatt-hour
TWh in 2011) but the share of electricity production
has dropped to 11% from a maximum of 17% in the
early 1990s as the world electricity production has
increased significantly since then.
Although 67 nuclear power reactors are officially
listed as under construction (mainly in Eastern Europe
and Asia), this figure is misleading because a number
of projects are often cancelled or left uncompleted.
Recent events in Japan have significantly affected the
use and development of nuclear power and the subject remains hugely controversial. However, no matter
what the future holds for nuclear energy production,
the issue of the safe management of HLW must be
addressed.
In this paper, the basic principles of deep geological repositories and its performance assessment are
Low Level Waste (LLW). Low activity waste is normally associated with radionuclides of short halflife and comprises the bulk of waste. All nuclear
activitie generate amounts of LLW in significant
quantities. Also, wastes resulting from mining and
ore processing normally belong to this category.
Worldwide they make up 90% of the waste volume
but contain only 1% of the total radioactivity of all
radioactive wastes.
Medium or Intermediate Level Waste (MLW
ILW). This is an intermediate class of waste that
contains non-negligible amounts of radioactivity
and normally requires shielding. They normally
result from enrichment and fuel fabrication, reactor
operation, reprocessing, and nuclear plant decommissioning. Heat emissions are low.
High Level Waste (HLW). High activity wastes that,
in addition to many short half-life radionuclides,
also contain large amounts of long-lived radionuclides. They are also strong heat emitters. Although
21
Figure 2. Conceptual scheme of a deep geological repository for high-level nuclear waste.
Figure 1. Nuclear energy production and share of electricity

production in the world, 19902011 (Schneider & Froggatt
2012).
method of low level waste (LLW) disposal, some countries are also considering deep geological disposal of
all waste types, including LLW, because they consider
the additional cost involved is compensated by the
perceived enhanced safety of deep geological disposal.
All disposal designs for HLW resort to the multibarrier concept to achieve the required degree of waste
isolation. If one considers the potential path of a
radionuclide from inside the canister to the biosphere,
it is clear that it will need to cross several barriers, the
canister itself, the backfill (engineered barrier) and the
host rock (geological barrier). Each one of those elements will provide a degree of safety to the overall
disposal system. Originally it was thought that each
barrier should be designed in such a way to provide
sufficient isolation on its own, so that a simultaneous
failure of all barriers would be required for significant
radioactive releases to occur. In fact, this is too restrictive and, in cases involving long-lived wastes, possibly
impossible to achieve. It is more realistic to consider
all the barriers acting together in a unified disposal
system.
A typical scheme for an underground mined repository involves the sinking of deep shafts down to a
depth of several hundred meters (Figure 2), the depth
is, of course, controlled by local geological conditions
(Chapman & Mc Kinley 1987). The shafts provide
access to a network of horizontal drifts that constitute the main repository area. Part of those drifts will
be access tunnels and part will be devoted to nuclear
waste disposal. A concrete plug often separates the disposal area from the access tunnel. The space between
canisters and the host rock is generally (but not always)
filled by a suitable material to constitute an engineered barrier. The material most usually considered is
compacted swelling clay, normally some kind of bentonite on its own or mixed with other materials like
sand although cement-based materials and crushed
salt, for repositories located in salt rock, are also being
considered for some specific applications.
The bentonite barrier fulfils several important functions. In the first instance, a very low hydraulic
conductivity restricts water penetration and retards
significantly solute transport due to its low diffusion
coefficient and to additional sorption effects. It should
also provide a favourable chemical environment and be
able to self-heal if subjected to physical perturbation
summarily addressed first follows by a brief review of

the basic thermo-hydro-mechanical (THM) behaviour
of clay-based engineered barriers, a frequent component of the multi-barrier system of these facilities.
A coupled formulation designed to encompass the
THM phenomena involved and their interactions is
then described. Finally, the performance of the formulation and associated computer code are applied
to the analysis of a large scale test performed in an
underground laboratory simulating actual repositories
conditions. Although chemical processes interacting
with THM phenomena also play a part in the evaluation
of repository safety, they are not considered herein.
2
DEEP GEOLOGICAL REPOSITORIES
Any disposal system should in principle guarantee that

the waste is removed from the human environment, the
waste is isolated and contained over long periods of
time (depending on waste type) and only small release
rates will occur once the complete isolation period is
over. A variety of procedures have been put forward
to solve the problem of radioactive waste disposal:
a) space disposal, b) ice sheet disposal, c) ocean bed
disposal, d) disposal beneath the seabed, e) nuclear
transmutation, and f) geological disposal.
Currently, only options e) and f) are the subject
of significant research activity. Transmutation of the
most harmful long-lived radionuclides does offer the
possibility to reduce the requirements applied to a
long-term disposal facility. This approach requires
carrying out chemical separation of very radioactive materials, going well beyond present reprocessing
activities. Whatever the result of this research, there
will always remain significant quantities of HLW
waste to be disposed of in some other fashion. Therefore, the construction of deep geological repositories
remains the favoured option for practically all countries. The aim is to free future generation from the
burden created by our present activities.
In fact, many countries have opted for the deep
geological disposal not only of HLW but of non-heat
emitting long-lived medium level waste (MLW) as
well. Although shallow burial is the most common
22
Figure 4. Example of performance assessment division into

near field and far field.
Figure 3. Mean annual dose computed for the reference scenario of the Spanish reference concept for high level nuclear
waste (Astudillo 2001).
conservative hypotheses and by the consideration of a

sufficiently wide range of possible scenarios.
The contribution of geomechanical numerical analyses is pervasive in the study of the likely performance
of a number of the various barriers. To carry out the
assessment in an effective way, it is convenient to
identify a number of subsystems that are analyzed
separately. An example of a first-level subsystem classification is depicted in Figure 4. It can be observed
that the output of a subsystem constitutes the input
of the next one. A conceptual model is built for each
subsystem that includes the most relevant processes,
the main parameters and the interaction between phenomena. The subsystem is quantitatively analysed by
means of appropriate numerical models. Finally the
results of each subsystem are integrated in the description of the overall behaviour of the entire system. This
division between different subsystems must be made
considering what are the phenomena and time scales
relevant to each particular component. Often there are
large differences between the processes that operate in
the various subsystems. A useful conceptual distinction refers to the division between near field and far
field. In a rough way, the near field may be defined as
the part of the disposal system that is directly affected
by the presence of the waste. It usually includes the
canister, the buffer or barrier and the adjacent rock.The
far field extends from the boundary of the near field
(not a precise location) to the region near the surface
that may interact with the biosphere. In this context,
the potential contribution of geotechnical numerical
analysis is especially strong in the analyses affecting
the near field, especially in the early-stage transient
period.
From the geomechanical point of view attention is
concentrated on the barrier and rock, canisters and
waste matrices are the concern of materials science.
A number of phenomena require specific numerical
analysis; e.g. the coupled thermo-hydro-mechanical
and chemical (THM and THMC) behaviour of the
engineered barrier (Guimaraes et al. 2007, Snchez
et al. 2012), the identification of the magnitude
and role of the Excavation Damaged Zone, EDZ,
(Vaunat & Gens 2004), the thermohydraulic behaviour
of the host rock (Gens et al. 2007) or the migration of
gas through the barrier on to the rock (Olivella and
Gens 2000, Olivella and Alonso 2008). Due to the
such as cracking and fissuring events. The adjacent

rock interacts with the barrier and it also plays a significant role in the safety performance of the repository by
providing the next barrier of isolation and retardation.
PERFORMANCE ASSESSMENT
To ensure the adequacy of a design of an underground

repository, it is necessary to undertake a thorough
performance assessment exercise that allows the evaluation of safety in a comprehensive way. To this end,
all the processes and phenomena that may affect the
performance of the repository in a significant way
must be considered in a systematic manner. This safety
evaluation is a complex process due to the variety of
materials and components of a repository and to the
large number of interacting processes that potentially
play a role.
The output of the performance assessment exercise may be expressed in different ways. A typical
one is in terms of the computed individual dose that
may be expected as a function of time. An example
is given in Figure 3 for the reference scenario of the
Spanish reference concept for high level nuclear waste
(Astudillo 2001). It can be noted that not only the total
dose is given but also the individual contributions of
each radionuclide. Two immediate observations can be
made: i) computations are extended to extremely long
times, 106 years in this case, and ii) the maximum
doses obtained are very low compared with the natural radiation and are also well below the frequently
specified limit value of 104 Sv/year. Naturally, the
computed doses increase when considering the effect
of external factors. Figure 3 is also useful to indicate
that the central phenomenon to be examined is the
transport of the radionuclides to the biosphere and
that the analyses of the large variety of phenomena
involved should be performed with this final aim in
mind. Naturally, complexity and uncertainty in many
areas prevent exact predictions. This difficulty, however, may be overcome (at least partially) by the use of
23
importance of the problem, performance of large-scale

tests in underground laboratories is also a characteristic feature of this field (Gens 2003). They are
carried out to advance the understanding of the phenomena and to evaluate the validity of the models
used. Again, the role of numerical analyses is a very
prominent feature of this work (e.g. Gens et al. 1998;
Gens et al. 2009; Thomas et al. 2009). In this paper,
attention will be focused on the analysis of the THM
behaviour of clay-based barriers and its interaction
with the near field host rock. The barrier behaviour
will be examined and numerically analysed with reference to a large-scale experiment carried out in the
Grimsel Underground Laboratory (Switzerland).
4
Figure 5. Thermo-hydraulic processes occurring in the near

field.
ENGINEERED BARRIER BEHAVIOUR
Drying of the bentonite will cause shrinking of the

material whereas hydration will produce swelling that
may be quite strong in bentonite barriers. Because
the barrier is largely confined between canister and
rock, the main result of hydration is the development
of swelling pressures, in a process quite akin to a
swelling pressure test. The magnitude of the stresses
developed is critically dependent on the emplacement
density of the bentonite and may reach values of
several MPa.
The crucial feature of the THM behaviour described
is that all those phenomena are strongly coupled, interacting with each other in a complex manner. As an
example, consider the phenomenon of vapour transport. Evaporation and condensation depend on the
value of suction (hydraulic variable) and temperature
(thermal variable). Transport itself is a mixture of
advection and diffusion that is influenced by temperature (thermal), degree of saturation (hydraulic) and
porosity (mechanical). In fact vapour transport cannot
be considered on its own, but as a branch of a cycle
closed by the movement in opposite direction of liquid
water. The flow of liquid water also depends on temperature (via water viscosity), degree of saturation and
porosity. Another example of interaction is the main
heat transfer mechanism, heat conduction. This is basically controlled by thermal conductivity that, in turn,
depends on degree of saturation (hydraulic effect) and
porosity (mechanical effect). Other coupling examples could be similarly discussed. The complexity of
THM behaviour increases further when the interaction
of the bentonite barrier with the host rock is taken into
account (Gens et al. 2002).
Therefore, in order to encompass the processes
outlined above, the THM formulation for saturated
and unsaturated porous media should incorporate the
following phenomena:
The near field is an area of complex phenomena and

interactions. The swelling clay making up the barrier is compacted so, initially, it is in an unsaturated
state. After placing the canisters, the main actions
that affect the bentonite barrier (at least in the short
term) are the heating arising from the canisters and
the hydration from the surrounding rock. At the inner
boundary, the barrier receives a very strong heat flux
from the canister. The dominant heat transfer mechanism is conduction that occurs through the three
phases of the material. A temperature gradient will
therefore develop in the near field and heat dissipation will be basically controlled by the thermal
conductivity of the barrier and host rock. Maximum
temperatures envisaged in repository design can be
quite high. Some designs limit the maximum temperature to 100 C but other concepts allow significantly
higher temperatures.
In the inner zone of the barrier, the heat supplied
by the heater results in a temperature increase and in
strong water evaporation that induces drying of the
bentonite. Degree of saturation and water pressure
will reduce significantly in this region. Vapour arising from bentonite drying will diffuse outwards until
finding a cooler region where vapour will condense,
causing a local increase in water saturation. Vapour
diffusion is a significant mechanism of water transfer mechanism and, to a much lesser extent, of heat
transport. Due to low water pressures existing initially
in the unsaturated material that constitutes the backfill, hydration will take place with water moving from
the host rock to the barrier. The distribution of water
potential is also affected by the phenomena of bentonite drying and vapour transport described above.
Hydration will eventually lead to saturation of the barrier, but saturation times can often be very long due
to the low permeability of the bentonite and/or host
rock. Figure 5 shows a scheme of the main thermohydraulic processes occurring in the bentonite barrier
and rock in the near field. In the Figure, the rock has
been assumed saturated but, in fact, it can desaturate
in some circumstances.
In addition to the thermo-hydraulic behaviour, there
are important mechanical phenomena also occurring.
i) Heat transport: Heat conduction, heat advection by

liquid water, water vapour and gaseous air.
ii) Water flow: Liquid advection and water vapour
diffusion.
iii) Air flow: Gas advection and dissolved air diffusion
iv) Mechanical behaviour: behaviour of porous
materials dependent on stresses, suction and
temperature.
24
In the next section a thermo-hydro-mechanical

(THM) formulation incorporating all those phenomena in a coupled manner is outlined. More details are
given in Gens & Olivella (2000).
to the solid phase and is denoted by jwg . It corresponds

to the total flux minus the advective part caused by
solid motion. When solid deformation is negligible,
then j= j.
5.2
5.1
COUPLED THM FORMULATION
Balance equations
Mass balance of solid present in the medium is

written as:
General
Several THM formulations have been described in

the literature, e.g. Olivella et al. (1994), Gawin et al.
(1995), Thomas & He (1995, 1997), Khalili & Loret
(2001) and Gatmiri &Arson (2008).The equations that
govern this problem can be classified into three main
groups: balance equations, constitutive equations and
equilibrium relationships (Gens 2010). Equations for
mass balance are established following the compositional approach. That is, mass balance is established
for water, air and mineral species instead of using solid,
liquid and gas phases. Equation for balance of energy
is established for the medium as a whole.
The following convention is established for writing
the equations defining the formulation: superscripts
refer to species (w for water and a for air) and subscripts refer to phases (s for solid, l for liquid and g
for gas). The notation used is as follows:
where s is the mass of solid per unit volume of solid

and js is the flux of solid. From this equation, an
expression for porosity variation is obtained as:
Equation (2) expresses the variation of porosity

caused by volumetric deformation and solid density
variation. To obtain it, the material derivative with
respect to the solid
: porosity
i: non-advective mass flux
: stress tensor
: mass content per unit volume of phase, i.e., =
E: specific internal energy
: density
q: advective flux
b: body forces vector
Sl : degree of saturation of liquid phase
j: total mass flux
u: solid displacement vector
: mass fraction
Sg : degree of saturation of gas phase
jE : energy fluxes due to mass motion
has been used.

Water is present in liquid and gas phases. The total
mass balance of water is expressed as:
where f w is an external supply of water. An internal

production term is not included because the total water
mass balance is considered. The use of the material
derivative leads to:
Volumetric mass of a species in a phase (e.g. water

in gas phase gw ) is the product of the mass fraction of
that species, gw , and the bulk density of the phase, g ,
i.e. gw = gw g .
The total mass flux of a species in a phase (e.g. flux
of air present in gas phase jgw) is, in general, the sum
of three terms:
The final objective is to find the unknowns from the

governing equations. Therefore, the dependent variables will have to be related to the state variables
in some way. For example, degree of saturation will
be computed using a retention curve that should be
expressed in terms of temperature, liquid pressure and
gas pressure.
Porosity appears in this equation of water mass
balance not only as a coefficient, but also in a term
involving its variation caused by a variety of processes.
It is also implicit in variables that depend on porosity
(e.g. intrinsic permeability). The way of expressing the
derivative term as a function of the state variables is
via the solid mass balance equation. This allows taking
the nonadvective flux: iwg , i.e. diffusive/dispersive,

flux
the advective flux caused by fluid motion: gw qg ,
where qg is the Darcys flux,
the advective flux caused by solid motion:
Sg gw u/dt where du/dt is the vector of solid velocities, Sg is the volumetric fraction of pores occupied
by the gas phase and is porosity.
The sum of the nonadvective and fluid motion advective fluxes is separated from the total flux in order to
simplify the algebraic equations. This flux is relative
25
formulation are liquid pressure, gas pressure, displacements and temperatures. The governing equations are
finally written in terms of the unknowns when the
constitutive equations are substituted in the balance
equations.
Another series of relationships that relates dependent variables with unknowns are the equilibrium
restrictions. They are obtained assuming chemical
equilibrium for dissolution of a particular species in
a phase. This assumption is appropriate because those
processes are fast compared to the transport phenomena that take place in porous media and, therefore, they
are not rate controlling. The concentration of vapour
in the gas phase and the amount of air dissolved in the
liquid phase are governed by expressions of this type.
The constitutive equations and equilibrium restrictions adopted are given in detail in Gens & Olivella
(2000) and Gens et al. (2009). Only some aspects are
highlighted here:
correctly into account the influence of porosity variation in the balance equation for water. It should be
noted that in equation (5) the material derivatives can
be approximated as eulerian if the assumption of small
strain rate is adopted while the volumetric change is
not neglected. This is consistent with the classical way
of deriving the coupled flow-deformation equations.
Following the same approach, it is straightforward
to obtain the mass balance of air taking into account
that air is the main component of the gas phase and
that it may also be present in the liquid phase as
dissolved air.
The momentum balance reduces to the equilibrium of

stresses if inertial terms are neglected:
Balance of momentum for fluid phases and dissolved species are reduced to constitutive laws:
Darcys law and Ficks law.
Fouriers law. Thermal conductivity depends on
porosity and state of hydration of the material
through a geometric mean expression.
Darcys law. Hydraulic conductivity depends on
porosity, on degree of saturation and, via water
viscosity, on temperature.
Ficks law is adopted for vapour diffusion. A tortuosity coefficient, , is included to take into account
the effect of a porous medium on the diffusion
coefficient.
The mechanical constitutive law explicitly considers the effects of strains, suction and temperature
changes.
where is the stress tensor and b is the vector of body

forces.
The equation for internal energy balance for the
porous medium is established taking into account the
internal energy in each phase (Es , El , Eg ):
where ic is energy flux due to conduction through

the porous medium, the other fluxes (jEs , jEl , jEg ) are
advective fluxes of energy caused by mass motions
and f Q is an internal/external energy supply. The use
of the material derivative leads to an equation formally
similar to the mass balance of water.
Only one equation is therefore required to express
the balance of internal energy in the porous medium
as a whole. The fluxes in the divergence term include
conduction of heat and advection of heat caused by the
motion of each species in the medium.A non-advective
mass flux causes an advective heat flux because a
species inside a phase moves and transports energy.
In contrast to the case of the movement of a contaminant in a groundwater system, the diffusive term
for heat transport (conduction of heat) is much larger
than the term concerning hydromechanical dispersion
(non-advective flux caused by the velocity of fluids).
For this reason, this term is usually neglected.
5.4 Numerical implementation and computer code

The formulation outlined above has been discretized
in space (finite elements) and time (finite differences)
in order that it can be used for numerical analysis. The
basic formulation and numerical discretization constitute the bases of a computer code, CODE_BRIGHT
that has been used to perform the analysis reported
below (Olivella et al. 1996).
DESCRIPTION OF THE LARGE-SCALE

IN SITU TEST
The large-scale in situ test (named FEBEX) simulates,

at full scale, the Spanish repository concept for HLW
that envisages placing the waste-containing canisters
in horizontal drifts surrounded by an engineered barrier made up of compacted bentonite. The test has been
performed in the Grimsel Test Site located in the Swiss
Alps. To install the experiment, a 2.28 m diameter circular tunnel was excavated in the GTS underground
laboratory using a TBM machine. The tunnel is 70.4 m
long and the final 17.4 m section was selected for the
5.3 Constitutive equations

A fundamental part of the formulation is constituted
by the set of constitutive laws and equilibrium restrictions. The constitutive equations establish the link
between the state variables (or unknowns) and the
dependent variables. The global unknowns in this
26
Figure 6. Heater placed in the axis of the excavated tunnel and surrounded by an engineered barrier made up of
compacted bentonite blocks.
Figure 8. Location of boreholes for installing instrumentation in the rock.
pore water pressures, total pressures and displacements. Of special importance were the observations
of hydraulic variables: relative humidity and pore
pressures. Relative humidity was determined using
PCT-55 Wescor psychrometers and capacitive transducers manufactured by Vaisala. In some places, they
were supplemented by vibrating wire transducers to
observe positive pore water pressures. Capacitive sensors proved to be much more robust and reliable
than psychrometers. It should be noted that there is
a direct relationship between relative humidity and
total suction given by Kelvins equation (shown later).
Therefore, measuring relative humidity (and temperature) is equivalent to measuring total suction. In
addition, a number of boreholes were drilled into the
rock prior to the emplacement of the heaters and bentonite blocks in order to install monitoring devices
(Fig. 8). Temperature, pore pressures, total stresses and
displacements were observed in the rock.
After the installation of the test, four months were
allowed to elapse before switching the heaters on
to achieve equilibrium conditions in the monitoring
system. During this time there was some hydration
of the barrier from the rock but, obviously, no thermal loading. During the heating stage, the test was
temperature-controlled with a prescribed maximum
temperature of 100 C at the contact between liner and
bentonite. To achieve this, the sequence of operations
was as follows: i) the heaters were switched at a constant power of 1200 W each during 20 days. The aim
was to identify the thermal response of the system and
to adjust the power control algorithm, ii) afterwards,
the power of each heater was increased to 2000 W
and kept constant until a temperature of 100 C at the
contact between liner and bentonite was reached (this
occurred after 33 days), and iii) subsequently, the system operated under temperature control, i.e. power
was adjusted to keep a constant 100 C maximum
temperature on the liner/barrier contact
The heating stage under this temperature controlled
conditions lasted for exactly five years.Afterwards, the
Figure 7. Layout of the large-scale in situ test. Instrumented

sections are indicated.
performance of the test. In the test area, two 4,300 W

heaters were placed in the axis of the horizontal drift.
The heaters were 4.54 m long and 0.90 m in diameter and were intended to simulate the release of heat
by HLW. The space between the rock surface and the
heaters was backfilled using blocks of compacted bentonite (Fig. 6). Finally, the test area was sealed by a
2.7 m long concrete plug. Figure 7 shows a schematic
layout of the test.
The host rock is good quality Central Aare granite
crossed, in the test zone, by a lamprophyre dyke. The
clay is a bentonite (wL = 98106, wP = 5056) with
a high smectite content, in the range of 88%96%,
with small and variable quantities of accessory minerals such as quartz, calcite and feldspars. The cation
exchange capacity is 100102 meq/100 g (42% Ca,
33% Mg, 23% Na, 2% K). The bentonite blocks were
compacted to a dry density of 1.7 g/cm3 at an average
water content of 14.4%. Because there were small gaps
between blocks and between the clay barrier and the
rock, the overall dry density of the emplaced barrier
was 1.6 g/cm3 .
The test is heavily instrumented, a total of 632 sensors were installed in the engineered barrier and in
the rock. The instrumented bentonite sections are indicated in Figure 7. The following parameters were measured: temperatures, relative humidity (total suction),
27
Figure 9. a) Bentonite retention curve. b) Relationship

between thermal conductivity and degree of saturation.
heaters were switched off and, after a cooling period,

the experiment was partially dismantled. The dismantling phase is considered in more detail in the final
part of the paper. Additional information on the test is
given in Huertas et al. (2006).
Figure 10. Variation of temperatures with time in the engineered barrier: observations and computed results. a) section
D1, b) section I.
MODELLING RESULTS
Here only some typical results of the modelling compared with test observations over the five years of the
heating stage are presented. The analyses require a
significant number of parameters that, for the most
part, were determined independently from laboratory
and in situ tests. For illustration purposes the retention
curve of the bentonite and the relationship between
thermal conductivity and degree of saturation together
with laboratory experimental results are depicted in
Figure 9. More information on the analyses carried
out and the material parameters adopted are given in
and in Snchez & Gens (2006) and Gens et al. (2009).
7.1 Bentonite barrier

Thermal. Figure 10 shows the evolution of the temperature in two sections of the test. In each section,
temperatures at different radial distances from the axis
of the tunnel are plotted. More than one set of observations for a particular radial distance and section are
shown because temperature sensors were sometimes
placed along several radial lines. The repeatability
of results indicates the degree of approximation to
thermal axisymmetric conditions and, indirectly, the
reliability of the observations.
28
It can be noted that, after the initial stage, temperatures generally rise very slowly throughout the
barrier. The analysis reproduces quite well the observations, although this is helped by the fact that it is
a temperature-controlled test. Because practically all
energy is transported by conduction, the good agreement at different radial distances indicates that the
value of thermal conductivity is well captured, even
when the barrier is subjected to changes in water
content and degree of saturation.
Hydraulic. Of more interest and more challenging
are the observations and predictions of the hydraulic
changes inside the barrier. Figure 11 shows the variation with time of relative humidity, measured with
capacitive sensors, in two sections, a cool one away
from the heater (C) and a hot one between the two
heaters (H). Relative humidity is plotted because it is
the actual parameter measured by the capacitive transducers. It is recalled that total suction, , is related to
relative humidity, RH, via the psychrometric (Kelvins)
equation:
where Mw is the molecular weight of water.

It can be observed that in the cool section C there is
a monotonic increase of relative humidity corresponding to the progress of hydration caused by water drawn
from the host rock. The inner region, however, is far
from being saturated after five years. The hydraulic
behaviour is more complex in the hot section close
to the heaters. In the zone near the rock and in the
middle of the barrier, there is a monotonic increase of
relative humidity reflecting again the process of hydration induced by the rock water. However, in the region
closer to the tunnel axis, it can be observed that there
is first an increase of relative humidity immediately
followed by drying and, finally, a gentle increase of
relative humidity occurs.
Assisted by the examination of the numerical analysis results, it is possible to identify the processes
underlying this behaviour. The first relative humidity
increase is due to a vapour front (driven by heating)
passing through the observation point. Afterwards,
temperature increase causes evaporation and, therefore, drying of the material. After approximately one
year, general hydration takes over causing the final
gradual increase of relative humidity. It can be noted
that vapour movement plays an important role in the
evolution of the hydraulic state of the barrier in the
hot zone; indeed, a significant part of the early hydration of the middle and outer barrier zones is due to the
condensation of vapour coming from the inner region.
Naturally, those hydraulic changes cause variations in
thermal conductivity that, in turn, affects heat conduction. It is interesting to note that the analysis is able
to reproduce very satisfactorily the thermo-hydraulic
observations; even from a quantitative point of view,
predictions agree quite well with observations.
Figure 11. Variation of relative humidity with time in

the engineered barrier: observations and computed results.
a) section C (cool zone), b) section H (between heaters).
Mechanical. Because of the very limited movements that the confined bentonite barrier can undergo,
the main mechanical parameter considered is total
stress. It would be expected that stresses in the barrier will increase very significantly during hydration
because of the high swelling potential of a heavily compacted bentonite. As Figure 12 demonstrates, such an
increase is indeed observed reaching quite high values of total stresses at the end of the test, of the order
of 24 MPa. Unfortunately, it is not easy to measure
reliably total stresses as the scatter of observations
clearly indicates. It is difficult, therefore, to check in
detail the agreement with the results from the analysis.
29
Figure 13. Variation of temperatures with time in the host

rock: observations and computed results. Borehole SF21.
Figure 12. Variation of total radial stress with time in

the engineered barrier: observations and computed results.
Section E1.
It is however possible that the slower development of

the measured swelling stresses as compared to the predictions from the analysis may be due, at least in part,
to the initial liner/barrier and rock/barrier gaps, not
accounted for in the modelling. In any case, it can be
noted that the predicted final total stress in the barrier
is similar to that measured in quite a number of total
stress cells.
7.2
Host rock (granite)
The host rock, granite, is subjected to more limited heat

loading and, therefore, the observed effects are bound
to be more muted. However, to complete the picture
of near field behaviour, it is necessary to examine the
behaviour of the rock mass during the test.
Thermal. Figure 13 shows, as an example, the evolution of the temperature measured in borehole SF21.
Four temperature sensors at different distances from
the tunnel axis were installed in the borehole. A rapid
increase of temperature is again observed at the start
of the test followed by a constant but gentle rise.
Observed temperatures in the rock reached a maximum value of about 40 C. The values and evolution
of the temperatures are well predicted in practically all
observation points, indicating again that the thermal
conductivity estimate was quite accurate.
Hydraulic. The initial pore pressure field in the
granite was determined in a comprehensive hydrogeological investigation performed before the test that
provided the initial hydraulic conditions for the analysis (Ortuo et al. 2005). Those pore pressures were
subsequently affected by the excavation of the tunnel
for the installation of the experiment. By modelling
this excavation, the new pore pressure distribution was
Figure 14. Variation of pore pressures with time in the host

rock: observations and computed results. Borehole SF24.
largely reproduced. During the thermal test, it was

observed that the pore pressures in the granitic rock
were little affected by heating. What Figure 14 shows
is basically the recovery of pore pressures after excavation. The same process is in fact predicted by the
numerical analysis.This is quite different from the very
strong hydrothermal response of argillaceous rocks in
similar circumstances (Gens et al. 2007).
Mechanical. Both displacements and total stresses
have been measured in the rock. The measured radial
displacements are shown in Figure 15; they are quite
30
Figure 17. Layout of the experiment after dismantling.
swelling pressure in the bentonite contributes to the

rock stress increase, the major factor in rock stress
development is the rise of temperature. Stresses computed in the numerical analysis appear to provide a
good estimate of the magnitude of the observed stress
increase although it is difficult to judge the accuracy
of the predicted evolution.
Figure 15. Variation of radial displacements with time in the

host rock: observations and computed results. Borehole SI1.
OBSERVATIONS DURING THE

DISMANTLING OF THE EXPERIMENT
Exactly after five years of heating, Heater 1 (the one

closest to the concrete plug) was switched off. After
allowing 33 days for cooling, the bentonite was carefully removed up to the end section of Heater 1, leaving
the rest of the barrier in place. At the same time,
Heater 1 was removed from the experiment. Those
operations lasted for an additional 75-days period.
Subsequently, a new shotcrete plug was constructed
and the test on the remaining part of the experiment
continued (Fig. 17). A detailed description of the procedures during dismantling is presented in Brcena
et al. (2003).
This partial test dismantling allowed the direct and
detailed observation of the state of the barrier after
five years of heating and hydration. In addition, large
amounts of quantitative data could be obtained; this
information provides an opportunity to check the performance and reliability of the numerical model in a
completely predictive mode. At selected sections, a
large number of specimens were cored out from the
barrier and dry density and water content were determined immediately in a field laboratory in order to
minimize disturbance and humidity loss. This comprehensive set of values of dry density and water content
helped to provide a complete picture of the state of the
bentonite barrier. Thanks to the high density of determinations, contours of dry density and water content
can be plotted in the sampled sections. An example is
presented in Figure 18 where it can be observed that the
process of hydration and associated density changes is
basically axisymmetric.
Figure 16. Variation of total stresses with time in the host

rock: observations and computed results. Borehole SG2.
small (of the order of 0.1 mm) and they develop in

response to the increase of temperature. The predicted
radial displacements are also small but larger than the
observed ones. It is likely, therefore, that the thermal
expansion of the rock was overestimated in laboratory
tests performed to determine thermal dilation.
Measured normal stresses in borehole SG2 are plotted in Figure 16. Again, the difficulty of measuring
stresses is reflected in the rather erratic nature of
the observations. It is clear, though, that the magnitude of stress increase reduce as the distance to the
tunnel becomes larger. Although the development of
31
Figure 18. Contours of (a) water content and (b) dry density
measured during dismantling. Hot section near the middle of
Heater 1.
Figure 19 shows the all measured values of water

content and dry density as a function of distance to
the tunnel axis for a section near the middle of Heater
1; it is therefore representative of a hot region of the
test. It can be observed that, as expected, water content
increases as the rock is approached, consistent with the
process of natural hydration from the rock. It is also
interesting to observe that the part of the barrier close
to the heater is still below its initial value of water
content, a consequence of the strong drying that has
occurred in this area. In spite that a certain amount of
water has already reached the inner region, it is still
insufficient, after five years, to compensate for the
initial drying. It is also noticeable that the barrier as a
whole is very far from full saturation at the end of the
heating stage.
The distribution of dry density also shows the
expected patterns. Close to the rock the clay has
expanded, exhibiting values of dry density well below
its initial value. In contrast, in the zone near the heater
the dry density has increased. Because of the confined
nature of the test, the variation of dry density in the
Figure 19. Dismantling observations and computed results:

a) dry density, b) water content. Hot section near the middle
of Heater 1.
inner part is compensated by the reduction of dry density in the outer part. It should be noted that the change
of dry density (i.e. porosity) is the combined effect
of expansion due to temperature increase (thermal
effect), suction changes (hydraulic effect) and stress
increase due to the development of swelling pressure
(mechanical effect).
Water content and dry density values are also plotted for a cool section (Fig. 20) where the temperature
increase is very limited. There are some differences
in the patterns of observations. Water content again
increases as the rock is approached but, now, there is
also a net (but small) gain of water content near the
32
predictions, the numerical analysis has simulated the

cooling stage and the excavation performed prior to
sampling operations. Dismantling operations lasted
for two months; the time required to reach each particular sampling section has also been incorporated in
the analysis.
CONCLUDING REMARKS
Nuclear energy production generates high level

radioactive waste that must be isolated for very long
periods. At present, the most widely adopted method
for disposal is the construction of deep underground
repositories in suitable geological media. The performance and safety assessment of these facilities require
a good understanding and adequate modelling facilities for all components of the system. Often, one
of these components is an engineered barrier composed of compacted expansive clay that fills the space
between the canisters containing the waste and the
host rock. Understanding the complex THM phenomena occurring in the barrier in the early life for the
repository require adequate modelling. The paper has
presented a formulation that encompasses, in a coupled manner, all the relevant THM processes and their
interactions. It has been shown that the formulation
and computer code has enabled quite a successful
modelling a large scale test that simulates the THM
behaviour of the clay barrier under full scale conditions. Consequently, it can be stated that formulation
and computer code constitute a useful numerical tool
for tackling this type of THM problems.
ACKNOWLEDGEMENTS
The authors gratefully acknowledge the financial and
technical support of ENRESA and the Commission of
the European Union through several research projects.
The contribution of the Ministry of Science and Innovation of Spain through grant BIA2011-27217 is also
gratefully acknowledged. Much of this work has been
performed in cooperation with many members from
the UPC research group and elsewhere as noted in the
references. The author is very grateful to them all.
Figure 20. Dismantling observations and computed results:

a) dry density, b) water content. Cool section.
centre of the tunnel. There has been no drying in this

region but the amount of hydration is very small, even
after five years, because of the larger thickness of the
barrier in this section. The dry density also shows a
significant reduction close to the rock but a somewhat
smaller increase in the inner part where the change
appears to be more uniform. Again, the net volume
change of the section is quite small.
It is interesting to observe that the results of the
numerical analysis show a very good agreement with
the measurements. The patterns of the hot and cool
sections are reproduced very well and the quantitative agreement between observations and predictions
is quite close. It should be noted that to obtain those
REFERENCES
Astudillo, J. 2001. El almacenamiento geolgico profundo
de los residuos radiactivos de alta actividad. Madrid:
ENRESA.
Barcena, I., Fuentes-Castillana, J. & Garcia-Sieriz, J 2003.
Dismantling of heater N 1 at the FEBEX in-situ
test. Descriptions of operations, Technical Publication
09/2003, Madrid: Enresa.
Chapman, N.A. & Mc Kinley, I.G. 1987. The geological
disposal of nuclear waste. John Wiley: Chichester.
Gawin, D., Baggio, P. & Schrefler, B.A. 1995. Coupled heat,
water and gas flow in deformable porous media. Int.
Journal for numerical methods in fluids 20: 969987.
33
Gatmiri, B & Arson, C. 2008. -STOCK, a powerful tool

of thermohydromechanical behaviour and damage modelling of unsaturated porous media. Computers and
Geotechnics 35: 890915.
Gens, A. 2003. The role of Geotechnical Engineering for
nuclear energy utilisation. Proc. 13th. Europ. Conf. on
Soil Mechanics and Geotech. Eng., Prague, 3: 2567.
Gens, A. 2010. Soil-environment interactions in geotechnical engineering. 47th Rankine Lecture. Gotechnique 60:
374.
Gens, A., Garca-Molina, A. J., Olivella, S., Alonso, E. E. &
Huertas, F. 1998. Analysis of a full scale in situ test simulating repository conditions. Int. J. Numer. Anal. Methods
Geomech. 22: 515548.
Gens, A., Guimares, L. do N., Garcia-Molina, A. & Alonso,
E.E. 2002. Factors controlling rock-clay buffer interaction
on a radioactive waste repository. Engineering Geology
64: 297308.
Gens,A. & Olivella, S. 2000. Non isothermal multiphase flow
in deformable porous media. Coupled formulation and
application to nuclear waste disposal. In Smith & Carter
(eds.), Developments in Theoretical Soil Mechanics: 619
640, Rotterdam: Balkema.
Gens A., Snchez, M., Guimares, L. do N., Alonso, E.E.,
Lloret, A., Olivella, S., Villar, M.V., Huertas, F. 2009.
A full-scale in situ heating test for high-level nuclear
waste disposal: observations, analysis and interpretation.
Gotechnique 59: 377399.
Gens, A., Vaunat, J., Garitte, B. & Wileveau, Y. 2007. In situ
behaviour of a stiff layered clay subject to thermal loading: observations and interpretation. Gotechnique 57:
207228.
Guimaraes, L do N., Gens, A. & Olivella, S. 2007. Coupled thermo-hydro-mechanical and chemical analysis of
expansive clay subjected to heating and hydration. Transport in porous media 66: 341372.
Huertas, F., Faria, P., Farias. J., Garca-Siriz, J.L.,
Villar, M.V., Fernndez, A.M., Martn, P.L., Elorza, F.J.,
Gens, A., Snchez, M., Lloret, A., Samper, J. & Martnez,
M.A. 2006. Full-scale Engineered Barrier Experiment.
Updated Final Report Technical Publication 05-0/2006,
Madrid: Enresa.
IAEA 2012. Nuclear power reactors in the world. Reference
data series No 2. Vienna: International Atomic Energy
Agency.
Khalili, N. & Loret, B. 2001. An elasto-plastic model for
nonisothermal analysis of flow and deformation in unsaturated porous media: formulation. International Journal
of Solids and Structures 38: 83058330.
Mc Combie, C., Pentz, D.L., Kurzeme, M. & Miller, I.

2000. Deep geological repositories: a safe and secure
solution to disposal of nuclear wastes, GeoEng2000, An
International Conference on Geotechnical and Geological Engineering: 1: 686691. Melbourne: Technomic
Publishing.
Olivella, S. & Alonso, E.E. 2008. Gas flow through clay
barriers. Gotechnique 58: 157176.
Olivella, S., Carrera, J., Gens, A. & Alonso, E.E. 1994. Nonisothermal Multiphase Flow of Brine and Gas through
Saline media. Transport in Porous Media 15: 271293.
Olivella, S. & Gens, A. 2000. Vapour transport in low permeability unsaturated soils with capillary effects. Transport
in Porous Media 40: 219241.
Olivella, S., Gens, A., Carrera, J. & Alonso, E.E. 1996.
Numerical Formulation for a Simulator (CODE_BRIGHT)
for the Coupled Analysis of Saline Media. Engineering
Computations 13: 87112.
Ortuo, F., Carretero, G., Martnez-Landa, L. & Carrera,
J. 2005. Hydraulic characterisation of the FEBEX granite: test performance and field interpretation. In Alonso
E.E. & Ledesma, A. (eds.), Advances in Understanding
Engineered Clay Barriers, 133141, London: Taylor and
Francis.
Snchez, M. & Gens, A. 2006. FEBEX Project. Final
report on thermo-hydro-mechanical modelling, Technical
Publication 05-2/2006, Madrid: Enresa.
Snchez, M., Gens, A. & Olivella, S. 2012. THM analysis
of a large-scale heating test incorporating fabric changes.
Int. J. Numer. Anal. Methods Geomech. 36: 391421.
Schneider M. & Frogatt A. 2012. The world nuclear industry
status report 2012. Paris, London: MSC.
Thomas, H.R., Cleall, P.J., Dixon, D. & Mitchell, H.P 2009.
The coupled thermal-hydraulic-mechanical behaviour of
a large-scale in situ heating experiment. Gotechnique 59:
401413.
Thomas, H.R. & He,Y. 1995. Analysis of coupled heat, moisture and air transfer in a deformable unsaturated soil.
Gotechnique 45: 677689.
Thomas, H.R. & He, Y. 1997. A coupled heat-moisture
transfer theory for deformable unsaturated soil and its
algorithmic implementation. Int. Journal Numer. Methods
in Eng. 40: 34213441.
Vaunat, J., & Gens, A. 2004. Aspects of modelling geotechnical problems in hard soils and soft argillaceous rocks.
9th Int. Symp. Numerical Models in Geomech., Ottawa,
3743.
34

Application of solid potassium permanganate to oxidize VOC vapours:

Batch and column experiments
M.G. Mahmoodlu
Utrecht University, Department of Earth Sciences, The Netherlands
S.M. Hassanizadeh
Soil and Groundwater Systems, Deltares, Utrecht
N. Hartog
Soil and Groundwater Systems, Deltares, Utrecht
KWR Watercycle Research Institute, Nieuwegein, The Netherlands
A. Raoof
ABSTRACT: Volatile organic compounds (VOCs) may cause major contamination problems in groundwater
and soil. Their presence in air can create a hazard to public health. However, limited remedial options exist in
controlling the vapour transport of these compounds in the unsaturated zone. In this study, batch and column
experiments were carried out to investigate the oxidation of TCE, ethanol, and toluene vapour in air by solid
potassium permanganate under different condition. The kinetic of chemical oxidation has been investigated
widely for dissolved VOCs in the saturated zone. But, so far there have been few studies on the use of in-situ
chemical oxidation (ISCO) of vapour phase contaminants. Results of batch experiments revealed that solid
potassium permanganate is able to transform the vapour of these compounds into harmless oxidation products.
The oxidation rates for TCE and ethanol were higher than for toluene. The oxidation process was modeled
using a kinetic model, linear in the gas concentration of VOC [ML3 ] and relative surface area of potassium
permanganate grains (surface area of potassium permanganate divided by gas volume) [L1 ]. The second-order
reaction rate constants (cm s1 ) for TCE, ethanol, and toluene were found to be equal to 2.0 106 , 1.7 107 ,
and 7.0 108 , respectively. Column experiments were carried out to study the impact of water saturations on
the removal efficiency and longevity of potassium permanganate. Results of column experiments showed that
the efficiency and longevity of potassium permanganate increased with water saturations. Results also showed
that the efficiency of potassium permanganate depends on the solubility of target compound in water.
INTRODUCTION
vapours can migrate upwards to the surface through

diffusion and produce elevated concentrations within
indoor air spaces (Berscheid et al., 2010). Exposure
to some VOCs might affect central nervous system
and internal organs, and might cause symptoms such
as headache, respiratory tract irritation, dizziness and
nausea, known as the Sick Building Syndrome (SBS)
(Yu, and Lee, 2007).
We have chosen TCE, ethanol, and toluene, as
model VOCs (target compounds) for chlorinated solvents, biofuel, and mineral oil, respectively, for the
reasons explained below.
TCE is one of the most common man-made chemicals found in soil (Albergaria et al., 2012). It has
been widely used as a dry cleaning solvent, degreasing agent, and chemical extraction agent. Since TCE
is carcinogenic, its movement from contaminated
Volatile organic compounds (VOCs) are defined as

organic compounds with boiling points (at 1 amt)
below 260 C (De Nevers, 2000). VOCs have high
vapour pressures under normal conditions, so they can
easily vaporize into the atmosphere or form vapour
plumes in the soil (Kim et al., 2007). VOCs are present
in some household products and automobile liquids
(Berscheid et al., 2010).
Releases of VOCs to the environment have occurred
through surface spills, leaking underground storage
tanks, and inadequate disposal practices (Berscheid
et al., 2010). Small quantities of VOCs may contaminate large volumes of water. When released as
free product, VOCs may migrate downward to significant depths through the soil. In addition, VOC
35
groundwater and soil into the indoor air of overlying

buildings is of serious concern (EPA, 2011).
Ethanol is being increasingly used in (renewable) fuel alternatives and as replacement for methyl
tertiary-butyl ether (MTBE), which, despite helping to accomplish Clean Air Act goals, has caused
widespread water contamination problems (Johnson
et al., 2000; Capiro et al., 2007). Ethanol can reduce
the biodegradation rate of light non-aqueous phase
liquid (LNAPL) such as benzene, toluene, ethylbenzene, and xylene isomers (BTEX) in groundwater and soil (Mackay et al., 2007; Freitas et al.,
2010).
Toluene is found frequently in indoor environments.
Toluene is mainly used as an additive to improve the
octane number of gasoline (Yu and Lee, 2007).
One of most common treatment techniques for
unsaturated zone polluted with VOCs is soil vapour
extraction (SVE). This is a long-term operation and
does not convert a contaminant to less toxic compounds. A promising alternative is in-situ oxidation
of VOC that can lead to favourable results in less time.
Oxidation of VOCs may convert hazardous contaminants to harmless compounds. The oxidizing agents
most commonly used for the treatment of hazardous
contaminants are potassium permanganate, ultraviolet radiation, ozone, chlorine dioxide, hydrogen peroxide, sodium persulfate, and Fentons reagent (H2 O2
oxidation in the presence of ferrous iron, Fe2+ ).
Among these oxidants, potassium permanganate
has been receiving increased attention for the treatment of liquids, slurry soils, and sludges polluted with
VOCs (Kao et al., 2008). Early laboratory studies have
indicated that dissolved potassium permanganate can
remediate a variety of organic compounds, chlorinated
alkanes (Waldemer, and Tratnyek, 2006), chlorinated
ethylenes (Huang et al., 1999; Hood et al., 2000;
Yan and Schwartz, 2000; Waldemer and Tratnyek,
2006; Kao et al., 2008; Urynowicz, 2008), oxygenates
(Jaky et al., 2000; Damm et al., 2002; Waldemer,
and Tratnyek, 2006), BTEX (Gardner,1996; Rudakov
and Lobachev, 2000; Waldemer, and Tratnyek, 2006),
substituted phenols (Jin et al., 2003; Waldemer, and
Tratnyek, 2006) and PAHs (Forsey, 2004), in aqueous phase. However, the potential of solid potassium
permanganate to oxidize VOC vapours in unsaturated
zone is currently unknown.
In this study, we demonstrate the ability of solid
potassium permanganate to oxidize VOC vapours.
Currently, the literature lacks data on the reaction
between solid oxidants such as permanganate and
vapour phase of contaminants. Therefore, we planned
a series of batch and column experiments with four
objectives: (1) to evaluate the ability of solid potassium
permanganate to oxidize vapour phase contaminants,
(2) to determine kinetic parameters for TCE, ethanol,
and toluene oxidation by solid potassium permanganate, (3) to evaluate the ability of solid potassium
permanganate as a horizontal permeable reactive layer
to oxidize the vapour of three VOCs under saturated
conditions, and (4) to investigate the impact of water
Figure 1. A schematic view of the column.
saturation on the removal efficiency and longevity of

potassium permanganate.
2
MATERIALS AND METHOD
2.1 Materials
The contaminants used in this study were TCE,
ethanol, and toluene (from Sigma-Aldrich, Merck
and ACROS companies, respectively). Solid potassium permanganate of 99% purity was obtained
from Sigma-Aldrich. The sand used in this study
was extracted from river bed in Papendrecht (Filcom
Company, Netherlands) and sieved 0.51 mm.
The deionized (DI) water was used to provide
the different saturations for the combination of
solid potassium permanganate and sand in column
experiments.
Since TCE reaction experiment produces proton
(Equation 1), we used NaHCO3 and CaCO3 (Merck)
as basic salts to control pH during the experiment.
A 10-mL transparent glass vial (Headspace-vial,
Grace), which was capped with a magnetic cap
(Tin plate magnetic crimp cap, Chromacol) and hard
septum (DK GRY, Agilent) was used as batch.
A cylindrical column glass of 5 cm length and 2 cm
internal diameter and a steel stainless lid was employed
to perform experiments involving the vapour intrusion and a horizontal reactive barrier (Figure 1). The
column was divided into two parts by a glass filter.
Column was equipped with three valves for injecting
pure phase of VOC, adding air, and taking samples.
36
Table 1. Initial
experiment.
2.2 Experimental procedure

2.2.1 Batch experiment
We used 4.8 106 , 3.5 106 , 2.3 106 moles of
TCE, ethanol, and toluene vapours, respectively.
These chemical were calculated in 1.5 mL of gas
samples (under normal conditions). Oxidation reactions for TCE, ethanol, and toluene vapours, respectively, by solid potassium permanganate are as follows:
VOC
experimental
[C]o
Exp. (mol)
conditions
for
MKMnO4 Vgas A
(g)
(cm3 ) (cm2 )
batch
So
(cm1 )
1
2
3
6 105 2.703
3 105 2.703
6 105 1.351
10.5
10.5
10.5
2703 257.43
2703 257.43
1351.5 128.71
Ethanol 1
2
3
6 105 2.703
3 105 2.703
6 105 1.351
10.5
10.5
10.5
2703 257.43
2703 257.43
1351.5 128.71
Toluene 1
2
3
6 105 2.703
3 105 2.703
6 105 1.351
10.5
10.5
10.5
2703 257.43
2703 257.43
1351.5 128.71
TCE
[C]o : Initial concentration of compound, Vgas : Volume of

the gas in batch, So : Initial relative surface area
The required amount of potassium permanganate for

oxidizing above-mentioned amount of TCE, ethanol,
and toluene were estimated to be 1.25, 2.22, and
4.44 mg, respectively. These were calculated based on
reactions in an aqueous environment, assuming a full
dissolution of crystals. In dissolved form, potassium
permanganate may be fully available for oxidation.
But, in the solid form only the surface of potassium
permanganate grains is in contact with the gas phase.
Accordingly, more potassium permanganate is needed
to avoid limitation in the degradation rate. Hence,
excess amount of potassium permanganate (namely,
2.703 g) was used for each batch.
Potassium permanganate grains were put inside
12-mL transparent glass vials, which were capped
with a hard septum to prevent any leakage. VOC
vapour was injected using a gas tight syringe (2.5 mL
Hamilton, SGE) and 16 mm disposable needles
(0.5 mm, Terumo).
In order to get kinetic parameters, three batch experiments at three different initial amounts of vapour and
potassium permanganate were performed for all compounds (Table 1). All experiments were carried out in
duplicate. For each experiment, we prepared several
identical batches and each batch was allocated to a
given sampling time.
A control experiment was also performed in duplicate for each compound to ensure that the loss of
target compound due to leakage was negligible over
the course of the experiments. To prepare a control
batch, a 12-ml transparent glass vial was capped with
a hard septum. Then VOC vapour was injected into
the vial. All experiments were carried out in a vertical
rotary shaker, at room temperature, 20 2 C, and air
humidity of 37 2%, which is also the initial humidity
inside the vials.
combination of required materials, the mixture was

packed and shaked for 20 min. Finally, the mixture
was place on the glass filter and the column was
immediately capped by a steel stainless lid.
To provide the vapour of target compounds, 2.5 ml
pure phase of target compounds was injected into the
bottom part of column. Control columns with 30 g sand
and same saturation conditions were carried out for all
compounds.
To prevent the photodecomposition of potassium
permanganate, all columns were wrapped by aluminum foil. The columns were kept at room temperature (22 1 C).
2.3
Sampling and measurements
2.3.1 Batch experiment

Reaction and control batches were periodically sampled using a gas tight syringe until no detectable
concentration was found in the reaction vial. To eliminate the effect of pressure drop due to sampling, each
vial was used only once.
The concentrations of target compounds, TCE,
ethanol, and toluene were measured by a gas chromatograph (GC). Gas samples of 2 mL were taken using
the headspace syringe of the GC from each vial. Then,
samples were injected into the GC. The GC (Agilent
Technologies 6850) equipped with a capillary column
(0.25 mm 60 m), a flame ionization detector, and a
purge and trap.
Specific surface area of potassium permanganate
was measured using 10-point Brunauer-Emmett-Teller
(BET) method by a Nova 3000 from Quantachrome.
Performance of this machine was controlled using reference 173 from Community Bureau of Reference.
Samples were degassed at 120 C overnight before
measurements. The relative surface area was calculated as the surface area per volume of gas (Table 1).
To calculate the amount of the potassium permanganate consumption, at the end of each experiment,
potassium permanganate grains were dissolved in
DI water and its concentration was measured using
a UV-Spectrophotometer (UV-1800, Shimadzu) at a
wavelength of 525 nm.
2.2.2 Column experiment

Different combinations of mass of solid potassium permanganate grains (20 g), dry sand (10 g), and various
water saturations (0, 20, 40, and 60%) were placed on
the glass filter.
To provide the desired saturation, both solid potassium permanganate and sand were put into a small
plastic container. This mixture was moisted with the
required amount of DI water. To prepare a well mixed
37
2.3.2 Column experiment

At various times 1.5 ml sample from headspace of
reaction and control columns were taken using a 2.5 ml
gas tight syringe (SGE Analytical Science, Australia).
At the same time, in order to eliminate the effect of
pressure drop due to sampling, the same volume of
air (1.5 ml) was injected in upper part of the column.
Then, the gas sample was injected to a 10-ml transparent glass vial which was capped with a magnetic
cap and hard septum (Magnetic Bitemall; Red lacquered, 8mm center hole; Pharma-Fix-Septa, Silicone
blue/PTFE grey; Grace Alltech). Sampling vials were
immediately put into the tray of a gas chromatograph
(GC). Then, the concentrations of target compounds
were measured by the GC. In the GC, the headspace
method was used to determine the concentration of
target compounds.
Total organic carbon content (TOC) of the dried
and sieved (grain diameter <250 m) soil sample were
measured in an elemental analyzer (Fisons Instruments
NA 1500 NCS). We also measured dissolved organic
carbon using Geolab GL-WV-014 method (Shimadzu
TOC-5050A).
3
RESULTS AND DISCUSSION
3.1 Target compound oxidation in batch system

Figure 2 depicts the normalized concentration (C/C0 )
of the target compounds as a function of time, where
C denotes the observed concentration of the target
compound for a given time and C0 is the initial
concentration of the target compound. Oxidation of
target compounds shows an exponential trend, as indicated by the fitted formula in the graph (Figure 2).
These results also show that solid potassium permanganate was able to rapidly oxidize the vapour phase
of TCE and ethanol. Toluene was also degraded but
less rapidly. No VOC intermediates or by-products
were found in vapour samples. During the experiment, potassium permanganate crystals turned into
dark color, which is the color of a coating layer of
produced manganese dioxide (MnO2 ). Moreover, we
compared results of oxidation ofVOCs in vapour phase
with the oxidation of VOCs in aqueous phase that
we have performed and also reported in the literature (Waldemer, and Tratnyek, 2006). We found that
the oxidation process in aqueous phase takes place in
shorter time. Otherwise the oxidation in both phases
is similar. So, for example, the oxidation rate for TCE
was much higher than for the other two compounds.
However it takes place in short time.
3.2
Figure 2. Degradation date for TCE, ethanol, and toluene

vs. time using from batch Experiment 1 (Exp. x-y: x denotes
number of experiment and y is the experiment repetition).
potassium permanganate, such as the surface area and

mass of potassium permanganate. Such an equation
may be written as:
where k denotes the reaction rate constant, C is the

vapour concentration of compound [ML3 ], t is time
[T], and S is the relative surface area of solid potassium
permanganate [L1 ], which is defined as:
Kinetic analysis of batch data
Since the degradation of three reactants showed an

exponential trend (Figure 2), we assumed that the
kinetics followed a first-order reaction rate. We also
assumed that only surface of solid potassium permanganate reacts with compounds. So, to a proper calculation of the reaction rate coefficient, we should have
an equation that involves the physical properties of
where A is the surface area of the potassium permanganate [L2 ] and V is the volume of the gas
phase [L3 ].
First, we assumed the relative surface area not to
alter significantly during the course of the reaction.
38
Table 2. Reaction rate constants for oxidation of TCE,

ethanol, and toluene in vapour phase at 20 C.
Compound
K (cm s1 )
R2
TCE
Ethanol
Toluene
2.0 106
1.7 107
7.0 108
9.6 101
9.8 101
9.9 101
R2 = The square of the correlation coefficient.
To validate this assumption, we determined the amount

of potassium permanganate that was consumed, at the
end of the experiment. This was done by dissolving
potassium permanganate grains which were used in
the experiments in DI water and then using a spectrophotometer to determine its mass. We compared
this to the initial mass and the result showed that there
was no significant consumption of solid potassium
permanganate during our experiments.
With S set equal to S0 Equation 4 can be solved to
obtain:
where S0 denotes the initial relative surface area of

potassium permanganate [L1 ]. According to Equation 6, plot of 1/S0 LnC/C0 versus time should yields
a straight line. Then, the reaction rate constant k can
be obtained from the slope of this line. We used all
data from Experiments 1, 2, and 3 to estimate k for
each compound. Figure 3 shows that the plots of 1/S0
LnC/C0 versus time for all compounds. The square of
linear correlation coefficients (R2 ) obtained for TCE,
ethanol, and toluene were 9.6101 , 9.9101 , and
9.9 101 , respectively. Results suggest that the oxidation of the target compounds can be modelled by
Equation 6 as long as sufficient relative surface area
of potassium permanganate is available. These results
also show the accuracy of reaction rate constants which
are given in Table 2.
We compared results of our experiments with the
oxidation of VOCs in aqueous phase reported in the
literature (Waldemer, and Tratnyek, 2006). We found
that the oxidation process in both phases follows a
first-order model. This comparison also revealed that
the oxidation rates for VOCs in vapour phase are much
smaller than for aqueous phase. However, in both
phases, the reaction rate for TCE is higher than for
ethanol and toluene. As mentioned above, as a result
of oxidation, MnO2 is produced and coats the grains.
This may affect the efficiency of oxidation process. To
analyze the effect of surface coating, we accounted for
the reduction of the relative surface area during the
experiment. This was done by supplementing Equation 4 with an equation relating the relative surface
area to the MnO2 concentration. This resulted in the
following equation:
Figure 3. Plot of 1/S0 LnC/C0 (denoted by y) vs. time

(denoted by x) for evaluating k following Equation 6. Data
are from batch experiments.
where is the number of moles of MnO2 produced

per mole of target compound (based on stoichiometry reaction) and is the coating factor (number
of moles of produced MnO2 per area of potassium
permanganate grains) [ML2 ].
3.3 Impact of water saturation on the removal
efficiency and longevity of potassium
permanganate
Figures 46 depict the normalized concentration
(C/C0 ) of the target compounds as a function of time
for different saturation values. Oxidation of a target
compounds reveals that the concentration of target
compound gradually increases with time. The concentration of VOC vapours reaches to the concentration
of control experiment due to the coating effect of produced MnO2 during the experiment. The results for
different water saturations showed that the efficiency
and longevity of potassium permanganate increased
with saturation, so that, at the highest saturation
(Sw = 60%), we observed a high removal efficiency
for all target compounds. Results also showed that the
efficiency of potassium permanganate depends on the
solubility of the target compound in water. Dissolution
39
Figure 4. Oxidation of TCE vapour under different water

saturations in column experiments.
Figure 5. Oxidation of ethanol vapour under different water
Since ethanol is fully miscible in water, it is

expected that more ethanol vapour (mass of ethanol)
can be removed during the vapour migration.
of VOC vapours into water around and between grains

can increase the residual time of VOC in the reactive
layer (combination of potassium permanganate, sand,
and water). Since potassium permanganate is soluble
in water (6.38 g/100 mL; 20 C), it also gives a chance
to the dissolved potassium permanganate to remove
VOC from water. As shown in Figures 46, the removal
efficiency of potassium permanganate for ethanol is
much higher than for TCE and toluene.
3.4 Intermediate products in column experiments

We found intermediates products during TCE and
ethanol oxidation in column. At high saturation conditions (Sw = 0.6) for TCE experiment, an inorganic
40
In our case, proton can be produced through Equation

1. Accumulation of produced proton in close system
may create acidic conditions. Then, produced proton
can react with potassium permanganate and produce
manganese heptoxide and water (Equation 9).
To control pH during the oxidation of TCE vapour

(Sw = 0.6), we used sodium bicarbonate (NaHCO3 )
and calcium carbonate (CaCO3 ) as basic salts. Both
salts could control pH in TCE oxidation and prevent
Equation 9 to take place. However, results showed
that sodium bicarbonate can increase the removal efficiency potassium permanganate more that calcium
carbonate.
We also measured two unknown peaks (compounds) in the GC results for ethanol experiment
(Sw = 0.6). To determine these compounds, three
gas samples were analyzed by Chromatography-Mass
Spectrometry (GC-MS).
Results showed that first peak belongs to ethyl
acetate and second one to 1,1-diethoxyethane. We
also found that ethanol can react with potassium
permanganate and produces ethanoic acid based on
the following two reactions:
or
Above Equations show that reaction can occur in both

neutral and acidic conditions. Equation 10 may produce water but Equation 11 gives H+ and decreases
pH. Ethanoic acid, by-product of Equation 10 and 11,
again might react with ethanol. If reaction take places
based on the equation 10, ethyl acetate (C4 H8 O2 ) can
be produced otherwise 1,1-diethoxyethane (C6 H14 O2 )
will be formed.
We found that they observed only in high saturation conditions (Sw = 0.6). During experiment, first
these peaks became larger and then disappeared. It
seems that the concentration of these two peaks positively correlated to concentration of ethanol and water
saturations.
Figure 6. Oxidation of toluene vapour under different water

compound, with the formula manganese heptoxide

(Mn2 O7 ) was observed after one hour. Manganese
heptoxide is produced as dark-green oil by the addition
of concentrated sulfuric acid to potassium permanganate (Equation 8). The reaction initially produces
permanganic acid, HMnO4 (structurally, HOMnO3 ),
which is dehydrated by sulfuric acid to form its
anhydride, Mn2 O7 .
CONCLUSION
In this study, we investigated the oxidation of both

chlorinated and non-chlorinated hydrocarbon vapours
by solid potassium permanganate under room temperature and humidity conditions. Results showed that
potassium permanganate is able to oxidize the vapour
41
Forsey, S.P., 2004. In Situ Chemical Oxidation of Creosote/

Coal Tar Residuals: Experimental and Numerical Investigation. Ph.D. Thesis, University of Waterloo.
Freitas, J.G., Fletcher, B., Aravena, R., Baker, J.F., 2010.
Methane production and isotopic fingering in ethanol fuel
contaminated sites. Groundwater 48, 844857.
Gardner, K.A., 1996. Permanganate Oxidations of Aromatic
Hydrocarbons in Aqueous and Organic Solution. PhD
Thesis, University of Washington.
Hood, E.D., Thomson, N.R., Grossi, D., Farquhar, G.J.,
2000. Experimental determination of the kinetic rate
law for the oxidation of perchloroethylene by potassium
permanganate. Chemosphere 40, 13831388.
Huang, K.C., Hoag, G.E., Chheda, P., Woody, B.A., Dobbs,
G.M., 1999. Kinetic Study of Oxidation of Trichloroethylene by Potassium Permanganate. Environ. Eng. Sci. 16,
265274.
Jaky, M., Szammer, J., Simon-Trompler, E., 2000. Kinetics
and mechanism of the oxidation of ketones with permanganate ions, J. Che. Soc., Perkin Trans. 2, 15971602.
Jin, Z., Gui-bai, L., Jun, M., 2003. Effects of chlorine content and position of chlorinated phenols on their oxidation
kinetics by potassium permanganate, Environ. Sci. 15,
342345.
Johnson, R., Pankow, D., Bender, C.P., Zogorski, J., 2000.
Target compounds-to what extent will past releases contaminate community water supply wells?. Environ. Sci.
Technol. 34, 210A217A.
Kao, C.M., Huang, K.D., Wang, J.Y., Chen, T.Y., Chien,
H.Y., 2008. Application of potassium permanganate as
an oxidant for in-situ oxidation of trichloroethylenecontaminated groundwater: A laboratory and kinetics
study. Hazard. Mater. 153, 919927.
Kim, H.H., Ogata, A., Futamura, S., 2007. Complete Oxidation of Volatile Organic Compounds (VOCs) Using
Plasma-Driven Catalysis and Oxygen Plasma. Int. J.
Plasma Environ. Sci. Technol. 1, 4651.
Mackay, D., De Sieyes, N., Einarson, M., Feris, K., Pappas,
A., Wood, I., Jacobson, L., Justice, L., Noske, M., Wilson,
J., Adair, C., Scow, K., 2007. Impact of ethanol on the natural attenuation of MTBE in a normally sulfate-reducing
aquifer. Environ. Sci. Technol. 41, 20152021.
Mahmoodlu, M.G., H., Niels, Hassanizadeh, S. Majid, and
Raoof, A., 2013. Oxidation of volatile organic vapours in
air by solid potassium permanganate. Chemosphere (In
Press).
Pant, P., Pant, S., 2010. A review: Advances in microbial
remediation of trichloroethylene (TCE). Environ. Sci. 22,
116126.
Rudakov, E.S., Lobachev, V.L., 2000. The first step of oxidation of alkylbenzenes by permanganates in acidic aqueous
solutions, Russ. Chem. Bull. 49, 761777.
Urynowicz, M.A., 2008. In-situ chemical oxidation with
permanganate: assessing the competitive interactions
between target and nontarget compounds. Soil & Sediment Contam. 17, 5362.
Waldemer, R.H., Tratnyek, P.G., 2006. Kinetics of contaminant degradation by permanganate. Environ. Sci. Technol.
40, 10551061.
Yan,Y.E., Schwartz, F.W., 2000. Kinetics and mechanisms for
TCE oxidation by permanganate. Environ. Sci. Technol.
34, 25352541.
Yu, K.P., Lee, G.W.M., 2007. Decomposition of gas-phase
toluene by the combination of ozone and photocatalytic
oxidation process (TiO2 /UV, TiO2 /UV/O3 , and UV/O3 ).
Appl. Catal. B: Environ., 75, 2938.
of TCE, ethanol, and toluene. This study also shows

that TCE and ethanol in vapour phase can be rapidly
oxidized by solid potassium permanganate. Toluene,
however, is degraded much slower.
A linear kinetic oxidation model, based on the concentration ofVOC in gas and a constant relative surface
area, effectively predicted the rate of TCE, ethanol, and
toluene degradation. Results revealed that the reaction
rate constants (cm s1 ) for TCE, ethanol, and toluene
are 2.0 106 , 1.7 107 , and 7.0 108 , respectively. Results also showed that the amount of used
potassium permanganate for all three compounds was
only a small fraction of initial amount (around 4%).
Results of column experiments showed that the
efficiency and longevity of potassium permanganate
increased with water saturations and solubility of
target compound.
These findings will be helpful in designing a horizontal permeable reactive barrier with solid potassium
permanganate in unsaturated zone for vapour intrusion. The performance of such methodology may be
affected by temperature of the soil matrix. Also, one
has to consider the health effects of by-product gases
such as ethane (Pant and Pant, 2010) and methane
(Freitas et al., 2010) that can be produced under anaerobic biodegradation of TCE and ethanol, respectively.
ACKNOWLEDGMENTS
The authors would like to thank Jan Kubiak
(Wageningen University), Michiel Kienhuis (Utrecht
University) and Pieter Van Rijwijk (NIOO-CEME) for
his technical assistance, Dr. Kotai Laszlo (Hungarian
Academy of Sciences) for providing critical comments
throughout the course of this research and Tom Bosma
(Utrecht University) for his thoughtful review of this
manuscript.
REFERENCES
Albergaria, J.T., Alvim-Ferraz, M.dc.M., Delerue-Matos, C.,
2012. Remediation of sandy soils contaminated with
hydrocarbons and halogenated hydrocarbons by soil
vapour extraction. Environ. Manage. 104, 195201.
Berscheid, M., Burger, K., Hutchison, N., Muniz-Ghazi, H.,
Renzi, B., Ruttan, P., Sterling, S., 2010. Proven Technologies and Remedies Guidance: Remediation of Chlorinated Volatile Organic Compounds in Vadose Zone Soil.
California Department of Toxic Substances Control.
154 pp.
Capiro, N.L., Stafford, B.P., Rixey, W.G., Bedient, P.B.,
Alvarez, P. J.J., 2007. Fuel-grade ethanol transport and
impacts to groundwater in a pilot-scale aquifer tank. Water
Res. 41, 656664.
Damm, J.H., Hardacre, C., Kalin, R.M., Walsh, K.P., 2002.
Kinetics of the oxidation of methyl tert-butyl ether (MTBE)
by potassium permanganate. Water Res. 36, 36383646.
De Nevers, N., 2000. Air Pollution Control Engineering,
2nd Edition, McGraw-Hill International, Boston. EPA,
2011. Toxicological Review of Trichloroethylene. CAS
No. 79-01-6.
42
General reports

Membrane behavior in engineered bentonite-based containment barriers:

State of the art
C. Shackelford
Colorado State University, Fort Collins, Colorado, USA
ABSTRACT: Membrane behavior in clays is a coupled phenomenon that results in solute restriction and
chemico-osmosis. Thus, the containment function of engineered clay-based barriers used for geoenvironmental
applications (e.g., landfills, waste impoundments, contaminated groundwater control) can be enhanced if such
barriers exhibit membrane behavior. In particular, membrane behavior has been shown to be potentially significant in bentonite, such that engineered barriers comprised wholly or partly of bentonite (e.g., geosynthetic clay
liners [GCLs], compacted clay liners [CCLs], and soil-bentonite [SB] cutoff wall backfills) may offer improved
containment performance relative to the expected performance based on conventional contaminant transport
analyses that neglect membrane behavior. Thus, the purpose of this paper is to provide a state-of-the-art review
of the available information pertaining to the existence and magnitude of semipermeable membrane behavior
in engineered, bentonite-based containment barriers. After illustrating conceptually the reduction in contaminant concentration emanating from engineered containment barriers that exhibit membrane behavior, a brief
description of the expressions governing liquid and chemical mass fluxes through clay membranes is provided.
This brief description is followed by a detailed description of the experimental considerations required for measuring membrane behavior in the laboratory. Of particular note is the distinction between the use of open and
closed systems for performing such experiments, with emphasis on the relative advantages of closed systems,
followed by a detailed description of the closed-system approach for conducting membrane testing. Example
results from different studies involving closed-system membrane testing of a GCL, two SB backfills, and a
bentonite-amended CCL then are presented to illustrate the types of data typically obtained from such testing.
The effects of four independent variables then are illustrated, including salt concentration, bentonite content, salt
cation valence, and effective confining stress. Although the data are limited and exceptions do exist, membrane
behavior in bentonite-based barriers is shown to increase with decreasing concentration of simple monovalent
salt solutions (e.g., KCl) and with increasing bentonite content and/or effective confining stress. However, the
use of salt solutions containing multivalent cations (e.g., Ca2+ ) has been shown to adversely affect the existence
and magnitude of membrane behavior. Finally, recommendations are given for further research and relevant
conclusions are drawn.
INTRODUCTION
2012, 2013, Shackelford et al. 2001, 2003, Shackelford

& Malusis 2002, Van Impe 2002, Dominijanni &
Manassero 2003, 2005a,b, 2008, 2012a,b, Manassero
& Dominijanni 2003, Mazzieri et al. 2003, 2005, 2010,
Shackelford & Lee 2003, Van Impe et al. 2003, 2005,
Yeo 2003, Henning 2004, Lu et al. 2004. Domimijanni
2005, Garavito 2005, Yeo et al. 2005, Dominijanni
et al. 2006, 2013, Henning et al. 2006, Evans et al.
2008, Kang 2008, Kang & Shackelford 2009, 2010,
2011, Whitworth & Ghazifard 2009, Di Emidio 2010,
Shackelford 2011, 2012, Bohnhoff 2012). The results
indicate that membrane behavior can exist in clay
containment barriers, but this behavior is likely to
be significant only in clay barriers that contain high
swelling smectite minerals (e.g., montmorillonite),
such as sodium bentonite (Shackelford et al. 2003).
Accordingly, the objective of this paper is to provide a state-of-the-art review with respect to the
existence and magnitude of membrane behavior in
The primary objective of clay barriers used for geoenvironmental containment applications, such as the
containment of solid wastes and liquids (e.g., landfills,
liquid impoundments, etc.) and the containment or
control of contaminated groundwater, is to minimize
the migration of contaminants into the surrounding
environment. With respect to this objective, the existence of semipermeable membrane behavior, whereby
solutes are restricted from migration through the clay
barrier, represents a potentially significant beneficial
aspect that presently is not considered in practical
applications. In this regard, the potential existence
and magnitude of such membrane behavior in clay
barriers has been the subject of significant research
over the past 15 years (Keijzer et al. 1999, Keijzer
2000, Malusis 2001, Malusis & Shackelford 2001,
2002a,b,c, 2004a,b, Malusis et al. 2001a,b, 2003,
45
efficiency coefficient for , and the term membrane

efficiency to refer to the value of expressed in
percent.
Although negative values of (or ) have been
reported in some cases due to atypical circumstances
resulting from processes such as diffusion-osmosis
(e.g., Olsen et al. 1990), the vast majority of (or )
values typically range from zero for clays exhibiting
no membrane behavior to unity (100%) for perfect
or ideal membranes that prohibit passage of all
solutes. Membranes with efficiencies less than 100
% (i.e., < 1) are referred to as imperfect or nonideal. Imperfect membranes also have been referred
to as semipermeable, selectively permeable,
partially permeable, or differentially permeable
(Shackelford 2011). These other terms result because
perfect membranes are still permeable to the solvent
water molecules (H2 O), whereas imperfect membranes also allow some solutes to migrate through the
larger pores of the clay. Clays that exhibit membrane
behavior generally are non-ideal, because the variation in pore sizes inherent in most clays results in only
some of the pores within the clay being restrictive.
the various types of bentonite-based barriers commonly used for geoenvironmental containment applications. The focus of the review is on barriers that are
comprised wholly or partially of sodium bentonites,
and more specifically on those bentonite-based barriers that are engineered for containment applications as opposed to natural smectitic clay formations
(aquitards, aquicludes) that may behave as semipermeable membranes (e.g., Greenberg et al. 1973). These
engineered barriers generally include geosynthetic
clay liners (GCLs), soil-bentonite (SB) backfills for
vertical cutoff walls, and bentonite amended natural
clays that are used as compacted clay liners (CCLs)
(Shackelford 2012).
2
2.1
DEFINITIONS & MOTIVATION

Definitions & terminology
Membrane behavior in clays represents the ability of

clays to exclude dissolved (aqueous miscible) chemical species, or solutes, from entering the pores of the
clays, thereby restricting the migration of the solutes
through the clays. In the case of charged inorganic
solutes (i.e., anions and cations), this restriction occurs
when the pore sizes between individual clay particles
are sufficiently small such that electrostatic repulsion
of the ions results from the interaction of electric
fields associated with adjacent clay particles (e.g.,
Fritz 1986). Such solute restriction also leads to the
process known as chemico-osmosis, whereby water
(H2 O) migrates from a location of lower solute concentration (higher water activity) to a location of higher
solute concentration (lower water activity) (Fritz 1986,
Shackelford et al. 2003). Neutral (uncharged) solutes
also may be restricted from migrating through clays if
the solute exhibits polar charge character despite being
neutral (e.g., carbon tetrachloride), or if the physical structure of the chemical molecule is simply too
large to fit through the pores. This latter mechanism
of solute restriction is referred to as steric hindrance,
and is expected to be more prevalent in the case of
large-chain organic molecules.
Clay membrane behavior is quantified in terms of a
membrane efficiency coefficient that represents the
relative extent of solute restriction. In the geotechnical engineering literature, this efficiency coefficient
is designated by the symbol and commonly referred
to as the chemico-osmotic efficiency coefficient, or
simply the osmotic efficiency coefficient. In the science literature, the membrane efficiency coefficient
often is referred to as the reflection coefficient and
designated by the symbol, . The symbol is preferred in the engineering literature because the symbol
commonly represents applied or total stress in the
engineering literature, and the terminology for of
chemico-osmotic efficiency coefficient is preferred
because other osmotic phenomena can exist in clays,
including electro-osmosis and thermo-osmosis
(e.g., Mitchell & Soga 2005, Medved & Cern

2013). The writer prefers using the term membrane
2.2 Motivation
In terms of motivation for understanding the extent
and magnitude of semipermeable membrane behavior
in engineered bentonite-based containment barriers,
consider the simplified horizontal and vertical barrier
containment scenarios illustrated in Figs. 1a and 1b,
respectively. The two barrier scenarios assume a constant source concentration of a given contaminant, Co ,
at the upgradient side of the barrier (point A), and initial concentrations of the same contaminant of zero
both within and downgradient of the barrier. Also, for
the sake of simplicity, the hydraulic liquid flux through
the barrier, qh , is assumed to be constant at times
greater than zero with respect to solute transport, and
also equal to the product of the hydraulic conductivity, kh , and hydraulic gradient, ih , in accordance with
Darcys law (i.e., qh = kh ih ). As will be shown subsequently in this paper, such a simple expression for
the hydraulic liquid flux actually is not accurate in the
case where the barrier behaves as a membrane. Finally,
the contaminant may be subjected to sorption onto the
surface of the clay particles, e.g., via cation exchange
for metals or hydrophobic partitioning for nonpolar
organic compounds, but otherwise is assumed to be
conservative (e.g., not subject to mass transformation
or degradation due to biological activity). Under these
conditions, the possible breakthrough curves (BTCs)
representing the ratio of the concentration of the contaminant at the downgradient side of the barrier (point
B) as a function of time, C(L,t), relative to Co are
shown schematically in Fig. 1c as a function of kh
and .
In the case of a barrier with a relatively high kh
and no evidence of membrane behavior (i.e., = 0),
the BTC reflects advective (hydraulic) dominated
transport conditions with some solute dispersion due
to mechanical processes (e.g., variations in pore water
46
velocity through the barrier), such that solute breakthrough as defined by the center of solute mass, or
C(L, t)/Co = 0.5, occurs at time t1 (see Fig. 1c). However, as the kh of the barrier is reduced, solute transport
through the barrier becomes increasingly dominated
by diffusion, resulting in an overall greater degree of
solute dispersion and an increase in the time required
to achieve breakthrough, such as reflected by time t2
in Fig. 1c. This increase in containment time from t1
to t2 is the primary reason for using clays with low kh
as engineered containment barriers. Furthermore, if
the low-kh barrier also exhibits semipermeable membrane behavior (i.e., 0 < 1), then solute restriction
will reduce the maximum possible value of C(L, t) at
steady-state transport relative to the case where = 0,
such that C(L, t)/Co 0 as 1. Thus, as illustrated schematically in Fig. 1, the primary reason for
considering engineered clay containment barriers with
membrane behavior is that the containment function of
the barrier can be improved significantly if the barrier
exhibits membrane behavior.
Although no direct correlation between kh and
has been found, there is a general expectation that
will be greater than zero only in the case where kh of
a soil is low, primarily because small pore sizes are
required for both low kh and > 0. This is the reason
why membrane behavior generally is relevant only in
the case of clays. Furthermore, for the pore sizes to be
sufficiently small to restrict the migration of dissolved
chemical species, the clay particle sizes must be relatively small, which is a reason why membrane behavior
generally is substantially greater in bentonites with
smaller particle sizes relative to other clays, such as
kaolin (Shackelford et al. 2003).
3
3.1
Figure 1. Engineered barrier containment scenarios:

(a) horizontal barrier; (b) vertical barrier; (c) barrier
solute breakthrough curves [kh = hydraulic conductivity;
= membrane efficiency coefficient].
LIQUID & CHEMICAL FLUXES

Liquid fluxes
direction or outward (i.e., from A to B), whereas

q is directed in the negative-x direction or inward
(i.e., from B to A). For this reason, q often is
referred to as the chemico-osmotic counter flow (e.g.,
Malusis & Shackelford 2001, Malusis et al. 2001a,
2003, Shackelford et al. 2001). Note that, for the containment scenarios depicted in Fig. 1, the direction
of q always will be inward, whereas the direction
of qh can change depending on the direction of the
hydraulic gradient. For example, drawing the contaminated groundwater level behind the vertical cutoff wall
in Fig. 1b down via pumping to a level lower than
that outside the wall will reverse the hydraulic gradient such that both qh and q will be directed inward,
i.e., from B to A (e.g., see Kang & Shackelford 2011
and Yeo et al. 2005 for examples pertaining to Figs. 1a
and 1b, respectively). Also, note that q qh as 0,
such that Eq. 1 reduces to Darcys law in the limit when
= 0.
The total liquid flux through a barrier that behaves

as a semipermeable membrane, q, at steady state
includes a hydraulic component of liquid flux, qh ,
in response to the difference in hydraulic head, and
a chemico-osmotic component of liquid flux, q , in
response to a difference in solute concentration (e.g.,
Barbour & Fredlund 1989, Malusis et al. 2001a, 2003,
Shackelford et al. 2001), or
where k = the chemico-osmotic permeability (=kh ),

i = the chemico-osmotic gradient, h = the head loss
across the barrier, L = the thickness of the barrier,
w = the density of water (1,000 kg/m3 ), i.e., assuming dilute solutions, g = acceleration due to gravity
(9.81 m/s2 ), and = the theoretical difference in
chemico-osmotic pressure resulting from the difference in solute concentrations across the barrier. With
respect to Fig. 1, qh is directed in the positive-x
3.2 Chemical fluxes

In general, for low-kh containment barriers, the total
chemical or solute mass flux, J [units of ML2T1 ,
47
where M = mass, L = length, and T = time] across the

barrier will be the sum of three components, viz.,
always be directed outward (i.e., the positive-x direction in Fig. 1). In addition, Jd also is a function of the
term (1 ), such that Jd 0 as 1 (Malusis &
Shackelford 2002b, Manassero & Dominijanni 2003,
Malusis et al. 2013, Shackelford & Moore 2013).Thus,
the general breakthrough curves scenarios shown in
Fig. 1c are consistent with the fact that all three chemical flux terms comprising J in Eq. 2 are functions of
the term (1 ), i.e., J (=Ja + J + Jd ) 0 as 1.
where Ja = the advective (hydraulic) mass flux,

J = the chemico-osmotic mass flux, and Jd = the
diffusive mass flux. Note that Eq. 2 excludes any component of mass flux due to mechanical dispersion,
which is the common assumption in the case of transport across containment barriers due to the low kh
and the typically short distances of transport for such
barriers (e.g., 1 m) (e.g., Sleep et al. 2006).
The first term in Eq. 2, Ja , is the traditional advective transport term that results from the presence of
the applied hydraulic gradient. As a result, Ja will
occur in the same direction as qh , i.e., in the positive-x
direction or outward (from A to B) for the containment scenarios depicted in Fig. 1. However, in the case
where the barrier behaves as a semipermeable membrane (0 < 1), this traditional component of solute
mass flux is reduced by a factor of (1 ), such that
Ja 0 as 1 (e.g., see Manassero & Dominijanni
2003, Malusis et al. 2012). In physical terms, the factor
(1 ) is considered to represent the process of hyperfiltration, whereby solutes are filtered out of chemical
solution as the solution passes through the membrane
under an applied hydraulic gradient. For this reason, Ja
also has been referred to as the hyperfiltrated advective flux (e.g., Malusis et al. 2001a, 2003, Shackelford
et al. 2001).
The second term in Eq. 2, J , represents the advective transport of solutes due to chemico-osmotic liquid
flux, q . Thus, J is directed in the same direction as
q , i.e., in the negative-x direction or inward (from
B to A) for the containment scenarios depicted in
Fig. 1. For this reason, J also has been referred to as
the counter advective component of solute transport
(e.g., Malusis et al. 2001a, 2003, Shackelford et al.
2001). The form of J has been shown to differ depending on the assumptions inherent within the theoretical
development of the chemical flux terms associated
with semipermeable membranes (Malusis et al. 2012).
However, for the assumption of salt (mutual) diffusion, whereby all chemical species are diffusing in the
same direction, such as depicted in Fig. 1, Manassero
& Dominijanni (2003) proposed including the term
(1 ) in the formulation of J , such that the sum of
the two terms Ja and J (i.e., Ja + J ) in Eq. 2 represents the net advective flux resulting from the sum
of liquid fluxes qh and q in Eq. 1 (i.e., qh + q ). Of
course, for the scenario where qh is directed outward
(i.e., from A to B in Figs. 1a,b), the effect of q is
to reduce the outward mass flux of chemical species
relative to that which would occur in the absence
of q (i.e., Ja + J < Ja ).
The third term in Eq. 2, Jd , represents solute diffusion through soil in the form of Ficks first law. In
general, diffusion occurs in the direction from higher
solute concentration to lower solute concentration,
such that Jd will always occur in the direction opposite
to J (and q ) and, for any containment scenario, will
EXPERIMENTAL CONSIDERATIONS
4.1 Type of system

Membrane behavior in clays has been measured
using both open and closed hydraulic control systems (Shackelford & Lee 2003, Shackelford 2011).
The difference between these two systems is illustrated schematically in Fig. 2. In an open system
(Fig. 2a), the clay specimen to be tested is bounded
by two reservoirs, both of which are open to the
atmosphere and there is no applied hydraulic gradient (liquid levels in both reservoirs are the same).
However, the reservoirs are chemical solutions with
different concentrations of the same solutes in order
to establish the requisite concentration difference,
C, across the specimen. Although not necessarily
required, the typical experimental setup is as indicated
in Fig. 2a, whereby one reservoir contains a chemical solution with solutes at concentrations greater
than zero (C > 0), and the other reservoir contains
de-ionized water (DIW) such that C 0. If the clay
behaves as a semipermeable membrane, then this
established C will result in q occurring across
the specimen, from lower solute concentration (right
reservoir) to the higher solute concentration (left reservoir). In general, the migration of liquid from the right
to the left reservoirs would cause overflow of liquid
from the left reservoir, such that right reservoir would
require replenishment of fresh DIW to maintain the
same liquid levels in each reservoir. However, replenishment of DIW may not be required in cases where
the reservoir volumes are much larger than the size of
the specimen and/or the testing durations are relatively
short, as the magnitudes of q typically are low, such
that the decrease in the liquid level in the right reservoir
in Fig. 2a may be imperceptible.
In the case where the clay specimen behaves as a
perfect membrane (i.e., = 1), the aforementioned
process would continue indefinitely, i.e., as long as
DIW was replenished in the right reservoir. However,
in the case where the clay specimen behaves as an
imperfect membrane such that 0 < < 1, then some
solute diffusion (Jd ) also will occur simultaneously,
from higher solute concentration (left reservoir) to
lower solute concentration (right reservoir), or in the
direction opposite to q . In this case, diffusion of
the solute from left to right will cause a decrease in
the solute concentration in the left reservoir (i.e., dilution) and an increase in the solute concentration in the
right reservoir. Thus, unless solute is added to the left
48
not replenished with DIW, then the height of rise of

the liquid in the standpipe would represent the equilibrium value of the chemico-osmotic pressure head,
h , which is equivalent to the value of the chemicoosmotic pressure, P , expressed in terms of pressure
head (i.e., h = P /w g). As will be shown subsequently, the magnitude of P and, therefore, h , is
directly proportion to the magnitudes of and C,
with the maximum values of P or h for a given
C corresponding to the case of a perfect membrane
( = 1). Of course, liquid will not rise above the standpipe (i.e., overflow) as long as the standpipe is of
sufficient height to accommodate h (i.e., H > h in
Fig. 2b).
In the closed system illustrated in Fig. 2c, q is
prevented, such that P develops across the specimen
to counteract the tendency for q . This development of
P can be measured using a pressure transducer (e.g.,
Malusis et al. 2001b). However, since solute diffusion
is independent of q , solute diffusion through the clay
specimen still will occur, i.e., provided < 1.
As illustrated in Fig. 2d, the occurrence of q in an
open system can affect the steady-state solute concentration profile across the specimen relative to that in
a closed system where q is absent. In an open system, the effect of q on the solutes within the pores of
the clay membrane is counter to that of Jd , such that
the actual steady-state concentration profile within the
clay membrane is nonlinear (e.g., see Dutt & Low
1962, Quigley et al. 1987, Manassero & Dominijanni
2003). In contrast, in a closed system, the solute concentration profile within the clay membrane at steady
state will be linear, since there is no effect of q on the
diffusion of the solutes (e.g., Malusis et al. 2012). Note
that, for simplicity, the comparison shown in Fig. 2d
assumes the concentrations at each end of the clay
membrane are the same, whereas this situation may not
be the case in reality. Also, any potential discontinuity
in concentrations across the reservoir-clay boundaries
at each end of the specimen due to ion exclusion are
not explicitly illustrated in Fig. 2d (e.g., see Dutt &
Low 1962, Quigley et al. 1987).
There are at least three advantages for measurement
of membrane behavior using a closed system relative
to an open system (e.g., Shackelford & Lee 2003).
First, measurement of P using a digital transducer
is vastly less complicated and far more accurate than
measuring the small quantities of liquid flux resulting from q in an open system. Second, control of the
boundary conditions in open systems is far more difficult than control of boundary conditions in closed
systems. Third, the simplicity of the resulting linear
concentration profile at steady state in a closed system
greatly reduces the complexity associated with evaluating diffusion coefficients relative to the analysis that
would be required in an open system with a nonlinear
concentration profile (Malusis et al. 2012).
As an example of this third advantage of closed systems, Dutt & Low (1962) noted that their bentonite
specimens tested in open systems restricted solute
migration, but interpreted their measured nonlinear
Figure 2. Open versus closed systems for clay membrane

testing: (a) open system with replenishment; (b) open system without replenishment; (c) closed system; (d) relative
difference in concentration profiles within clay membrane at
equilibrium or steady state (i.e., tss = time at steady state).
reservoir and removed from the right reservoir to maintain the initially established value of C, this diffusion
eventually will equilibrate the solute concentrations
in the two reservoirs, such that the concentration difference will be destroyed and the flow of liquid and
diffusion of solute will cease to exist (i.e., q & Jd 0
as C 0).
As shown in Fig. 2b, if the left reservoir is reconfigured with a standpipe, such that any liquid flowing into
the left reservoir would rise in the standpipe instead of
overflowing the reservoir, and the right reservoir is
49
concentration profiles at steady state, which required

182 d for 70-mm-length specimens, only on the basis
of Ficks first law for Jd to indicate that concentrationdependent diffusion coefficients existed. However,
Manassero & Dominijanni (2003) have shown that
such nonlinear concentration profiles can result from
a single, concentration-independent diffusion coefficient provided that all the chemical flux processes
occurring in the system (e.g., both Jd and J ) are taken
into account in the analysis. Thus, the diffusion coefficients determined by Dutt & Low (1962) may not
be correct, but rather may have resulted from the use
of an open system and failure to recognize that J also
was operative in their system. Of course, the analysis
conducted by Dutt & Low (1962) would have been
appropriate had they conducted their experiments in a
closed system, in which case they likely would have
measured a linear concentration profile at steady state.
The study by Dutt & Low (1962) also illustrates that
failure to recognize all of the processes involved in a
test system (i.e., q and J in the case of Dutt & Low
1962) may result in a misinterpretation of test results
(i.e., concentration-dependent diffusion coefficients).
Figure 3. Circulation boundary concentrations (C) and

pressures (P) for a clay membrane specimen contained in a
rigid-wall cell between two porous disks in a closed system.
during this circulation stage of the test. As a result,

there is no volume change in the system during this
circulation stage, i.e., the system is closed, such that
q is zero. In order to establish the requisite concentration difference, C, the concentration(s) of the
solute(s) in the solution being circulated across the top
boundary, or Cot , must be different than that (those)
being circulated across the bottom boundary, or Cob .
Typically, Cot > Cob , and DIW is circulated across the
bottom boundary (i.e., Cob = 0), such that C based
on the source solutions, or Co , is represented as
Cob Cot = Cot < 0. For this case, and assuming the
specimen is not a perfect membrane (i.e., 0 < < 1),
some solute diffusion (Jd ) will occur from the top circulation boundary into the specimen and eventually
all the way through the specimen, emanating from the
bottom of the specimen into the bottom circulation
system (e.g., see Malusis & Shackelford 2002b). As a
result, the concentration(s) of the solute(s) in the circulation outflow across the top, Ct , will be lower than
that (those) in the circulation inflow across the top
(i.e., Ct < Cot ), whereas the concentration(s) of the
solute(s) in the circulation outflow across the bottom,
Cb , will be greater than that (those) in the circulation
inflow across the bottom (i.e., Ct < Cot ).
As previously noted, if the specimen behaves as
a semipermeable membrane, then a chemico-osmotic
pressure, P , will be generated in the top reservoir to
counteract the tendency to develop q from bottom
to top. This P is designated Pt in Fig. 3. Also, since
Cb > Cob , the concentration(s) of solute (s) in the bottom porous disk cannot be assumed to be zero, and
the actual C across the specimen will less than Co .
Thus, the pressure at the bottom of the specimen, Pb ,
also is measured as a reference pressure, such that the
actual chemico-osmotic pressure across the specimen
is represented by P = Pb Pt < 0.
Given the testing scenario illustrated in Fig. 3 and
the associated testing procedure and conditions, several different characteristic concentrations have been
defined for the purpose of analyzing and presenting the results of the experiments. In general, three
average concentrations (Cave ) have been identified,
including the average of the source concentrations
4.2 Type of cell

Either rigid-wall or flexible-wall cells can be used to
measure membrane behavior, although the vast majority of membrane testing of clays has involved the use
of rigid-wall cells, whereby the volume of the specimen is maintained constant. A detailed example of a
rigid-wall cell is given in Malusis et al. (2001b).
In some studies, flexible-wall cells that provide
control of the state of stress in the specimen have
been used, including both open-systems where volume change is allowed during the test (e.g., Keijzer
et al. 1999), and closed systems where volume change
is prevented during the membrane measurement stage
of the test, but can occur between successive testing stages of a single membrane test (e.g., Kang and
Shackelford 2009).
4.3
Definitions of concentrations
Given the advantages of closed systems relative to

open systems, the remaining discussion on experimental considerations will be limited to those existing
for closed systems. A detailed description of such
a closed-system testing apparatus is presented in
Malusis et al. (2001b). A more general description
based on the simplified schematic scenario shown in
Fig. 3 is given herein.
In a closed system, membrane testing commences
once chemical (source) solutions containing the same
solute or solutes are circulated across both the top and
bottom boundaries. The circulation rates are controlled
to be the same via a flow-pump hydraulic system, and
both the top and bottom circulation systems are closed
loops, such that there is no volume change in either circulation system during circulation (see Malusis et al.
2001b for details on how this is accomplished). The
volume of the specimen also is maintained constant
50
porous disks, and the disks have identical hydraulic

properties, then PDIW should be zero. However, in
most cases, some small, nonzero value of PDIW has
been observed. This nonzero PDIW generally has
been attributed to three possible causes (e.g., Malusis
et al. 2001b): (1) the existence of remnant or residual salts or other chemical species stored within the
porous disks or system plumbing, e.g., if the disks
and system plumbing previously have been used and
not completely cleaned prior to reuse; (2) slightly different circulation rates between the top and bottom
resulting in minor imperfections in the machining of
the actuators and plungers used to displace the circulation liquids; and/or (3) slightly different hydraulic
properties of the porous disks.
After establishing the steady-state magnitude of
PDIW , C is established across the specimen by
changing the circulation liquid across the top of the
specimen from DIW to some chemical solution, typically a salt solution, while maintaining DIW as the
circulation liquid across the bottom of the specimen.
This circulation proceeds until a new steady-state value
of P, or PSol , has been established and measured.
If desired, this process can be repeated sequentially
using chemical solutions with progressively higher
concentrations of the specified solutes to establish
increasingly higher values of C and, therefore, measurement of different values of as a function of C
for the same specimen. Tests that involve this sequential circulation of chemical solutions with increasingly
higher concentrations of the specified solutes are
referred to as multi-stage tests, with the results from
each different chemical solution representing a separate stage of the testing procedure (e.g., see Malusis &
Shackelford 2002a). An alternative type of multi-stage
test involving circulating the same chemical solution
during all stages, but doing so on a single specimen
that was progressively compressed or consolidated to
achieve progressively lower values of the specimen
void ratio, e, also has been conducted (Yeo et al. 2005).
being circulated across the bottom and the top of the

specimen, Co,ave , and the individual averages of the
concentrations in the top and bottom circulating liquids, or Ct,ave and Cb,ave , respectively, as follows (e.g.,
Malusis et al. 2001, Malusis & Shackelford 2002a,
Kang & Shackelford 2009):
Three differences in concentrations (C) also have

been defined as follows (e.g., Malusis et al. 2001,
Malusis & Shackelford 2002a, Kang & Shackelford
2009):
Note that, for the sign convention defined in Fig. 3,

the magnitudes of all definitions of C given by
Eqs. 57 will be negative (<0).
4.4 Preconditioning of specimens and circulation
of liquids
In several studies (e.g., Malusis & Shackelford 2002a,
Shackelford & Lee 2003, Yeo et al. 2005, Kang &
Shackelford 2009, 2010, 2011), the specimens have
been permeated prior to membrane testing with DIW
to leach or flush soluble salts from the pores of the
specimen for the purpose of enhancing the likelihood of observing significant membrane behavior.
The durations of this preconditioning by permeation
can be quite lengthy, e.g., on the order of several
months, even using relatively high hydraulic gradients
(ih > 100), primarily because of the low values of kh
for bentonites based on permeation with DIW, e.g.,
kh < 2 1011 m/s (Daniel et al. 1997). More importantly, since the bentonite used in the actual barriers in
the field is unlikely to be preconditioned in a similar
manner, the results of studies involving such preconditioned specimens probably cannot be extrapolated
directly to field applications. In this case, some assessment of the effect of this specimen preconditioning on
field performance should be undertaken.
Following preconditioning, the membrane testing
stage begins, typically by circulating DIW across both
the top and bottom boundaries of the specimen under
a closed condition to establish a baseline or reference value of P with respect to DIW, or PDIW ,
if any. Conceptually, if DIW is circulated simultaneously at the same rate through both the top and bottom
4.5 Calculating membrane efficiency coefficients

For closed systems, is defined as follows
(Groenevelt & Elrick 1976, Olsen et al. 2000, Malusis
et al. 2001a):
where (<0) is the theoretical maximum value of

P that would result across a perfect semipermeable
membrane (i.e., if = 1). The actual value of P to
be used with Eq. 8 is the effective or net pressure difference, Pe , which is equal to the difference between
Psol and PDIW (i.e., Pe = Psol PDIW ). The
value of in Eq. 8 is calculated on the basis of
the difference in solute concentration, C, across the
specimen in accordance with the vant Hoff expression
as follows (e.g., Malusis et al. 2001b):
51
cations, or diffuse double layers (DDLs), surrounding

individual clay particles with a concomitant increase in
the sizes of pore channels available to solute migration
and, therefore, a decrease in the solute restrictive capability of the specimen (e.g., Malusis & Shackelford
2002a). This effect of the higher solute concentrations on the post-peak degradation of P also is
clearly evident from the results of the multi-stage test
shown for the same GCL in Fig. 4b, where increases
in Cot (= Co ) from 3.9 mM KCl to 47 mM
KCl resulted in not only progressively higher peak
values of P, but also progressively greater
magnitudes in post-peak degradation of P.
The results shown in Fig. 4c pertain to a membrane
test conducted in a rigid-wall cell on a specimen of
a sand-bentonite backfill representative of that for a
vertical cutoff (slurry) wall as reported by Yeo et al.
(2005). The test was a multi-stage test conducted in
a rigid-wall cell, except each stage was represented
by a difference in the void ratio, e, of the specimen
instead of a difference in Cot (= Co ), which was
maintained constant at 3.88 mM KCl. Relative to the
test results for the GCL shown in Figs. 4a & 4b, the
results for the backfill in Fig. 4c illustrate much lower
peak and steady-state values of P. This difference
was attributed by Yeo et al. (2005) primarily to the
difference in bentonite contents in the two different
types of barrier materials, with the GCL essentially
containing 100% bentonite as the hydraulic/transport
resistance component of the barrier, and the sandbentonite backfill containing only 7.2% bentonite.
Also, the fact that the sand-bentonite backfill exhibited
the same trend in post-peak degradation in P as that
for the GCL, but at a much lower concentration than
for the GCL, also was attributed to the significantly
lower bentonite content of the backfill. That is, the
much higher bentonite content of the GCL relative to
the sand-bentonite backfill resulted in the GCL being
more resistant to physico-chemical interactions due to
diffusion of invading salt cations (K+ ) at the lower
concentration of these salt cations. Finally, despite the
relatively low magnitudes of P associated with the
sand-bentonite backfill, there is a distinct, albeit slight,
increase in the peak and steady-state values of P
with decreasing e, as expected.
The results shown in Fig. 4d are from Kang &
Shackelford (2010) for a single-stage membrane test
conducted on a compacted specimen of a local, natural clay known as Nelson Farm Clay (NFC) containing
89% fines but no high activity clay minerals amended
with 5% (dry weight) sodium bentonite.
The bentonite addition was deemed necessary
because a separate specimen of the NFC without such
bentonite amendment did not exhibit any significant
membrane behavior, despite measurement of a suitably low kh (i.e., <109 m/s). The kh of the compacted
specimen of bentonite amended clay was lower than
that of the unamended specimen and, therefore, also
suitably low. The membrane test was conducted in a
flexible-wall cell on the compacted specimen that had
been back-pressure saturated, consolidated to an initial
where is the number of ions per molecule of the salt,

R is the universal gas constant [8.314 J mol1 K1 ],
T is the absolute temperature (K), and C is the
solute concentration (M). For example, for 1:1 electrolyte solutions (e.g., NaCl, KCl), = 2, whereas for
2:1 electrolyte solutions (e.g., CaCl2 ), = 3. Use of
Eq. 9 assumes dilute chemical solutions and fully
dissociating chemical compounds.
In general, can be calculated on the basis of the
difference in the source solute concentrations of the
circulating liquids, designated as o , as follows:
where Co is given by Eq. 5. Alternatively, can be

calculated on the basis of the difference in the average solute concentrations across the top and bottom,
designated as ave , as follows:
where Cave is given by Eqs. 4 and 7. Since

Ct,ave < Cot and Cb,ave Cob , the magnitude of o
will be greater than that of ave , such that o < ave
for the same measured value of Pe . Thus, values of
o are more conservative (lower) than values of ave
(Malusis et al. 2001b, Kang & Shackelford 2009).
However, in the limit as the membrane efficiency
approaches 100 percent, solutes can neither enter nor
exit the specimen, such that Ct,ave Cot , Cb,ave
Cob , and ave o (Malusis et al. 2001b, Kang &
Shackelford 2009).
4.6 Typical results for closed systems
Examples of chemico-osmotic pressure responses
(i.e., P > 0) resulting from membrane testing
under closed-system conditions for specimens of different types of bentonite-based barriers are shown
in Fig. 4. The pressure responses from Malusis &
Shackelford (2002a) for both single-stage and multistage tests conducted on specimens of the same GCL
tested in a rigid-wall cell are shown in Figs. 4a and 4b,
respectively. For the single-stage test (Fig. 4a), simultaneous circulation of DIW across both the top and bottom boundaries of the specimen resulted in a baseline
pressure difference, PDIW (>0), of 4.0 kPa. Upon
replacing the DIW circulating across the top boundary
of the specimen with 47 mM KCl (= Cot = Co ),
the pressure difference increased relatively rapidly to
a peak value of 41.4 kPa before diminishing to a final
steady-state value of 32.0 kPa.This post-peak degradation of P commonly occurs at higher solute concentrations in the case of imperfect clay membranes
(0 < < 1), and is attributed to diffusion of the solutes
into the pores of the specimen resulting in progressively greater compression of the adsorbed layers of
52
represents a continuous record of P, and the daily

decreases and increases in the pressure response resulting from refilling the circulating liquids within the
system are shown. The pressure response record shown
in Fig. 4d is otherwise similar to those shown in Figs.
4a-c, except the response during the period from about
21 to 38 d is relatively erratic. This erratic response
behavior was attributed, in part, to a less uniform consistency in the natural clay-bentonite mixture and the
likely more complex nature of the network of pores,
both of which can be expected when non-processed
natural clays with varied mineralogy are evaluated for
membrane behavior.
The time-dependent values of the membrane efficiency coefficients based on the difference in source
KCl concentrations across the specimens, o , for the
chemico-osmotic pressure records in Fig. 4 are shown
in Fig. 5. All of the values for o shown in Fig. 5 are
based on the effective pressure differences, Pe .
The results for both the single-stage and the multistage tests (Figs. 4a and 4b, respectively) are shown
together in Fig. 5a for the sake of comparison. At least
two observations are readily apparent. First, in contrast to the values of P in Fig. 4b, which increased
with increasing Cot (= Co ) from 3.9 mM KCl to
47 mM KCl, the corresponding values of o in Fig. 5a
(closed data symbols) decrease with increasing Cot
(= Co ) from 3.9 mM KCl to 47 mM KCl. This
trend may seem counterintuitive, given the basic definition for represented by Eq. 8, where is shown
to be directly proportional to P. However, the reason for decreasing o with increasing Cot (= Co )
is related to the term in the denominator of Eq. 8
(or o in Eq. 10), which is directly proportional to
Cot (= Co ), as shown in Fig. 6 for KCl. Thus,
the increase in P with increasing Cot (= Co )
is offset by a greater increase in (or o ) with
increasing Cot (= Co ), such that the overall effect
from increasing Cot (= Co ) is a decrease in o .
Second, the temporal trends in o and ultimately the
steady-state value of o for the single-stage test based
on Cot (= Co ) of 47 mM KCl (open data symbols)
are similar to those for last stage of the multi-stage test
also corresponding to Cot (= Co ) of 47 mM KCl.
This similarity in o trends and values suggests that
the results were affected little by the type of test (single stage or multi-stage), i.e., all other factors being
the same.
The temporal trend in o corresponding to the values of P for the sand-bentonte backfill specimen
in Fig. 4c are shown in Fig. 5b. Three observations
are apparent. First, the range in the vast majority of
o values for the sand-bentonite backfill specimen in
Fig. 5b of 0.19 o 0.36 is significantly narrower
than that for the GCL specimen shown in Fig. 5a of
0 o 0.67 for similar values of Cot (= Co ) (i.e.,
3.88 mM KCl in Fig. 5b vs. 3.9 mM KCl in Fig. 5a).
This difference again can be attributed, in part, to
the different amounts of bentonite in the two types
of barriers (i.e., 100% for the GCL vs. 7.2% for the
sand-bentonite backfill). Second, despite the relatively
Figure 4. Chemico-osmotic pressure responses (P)

under closed-system conditions: (a) & (b) single-stage and
multi-stage tests for a GCL; (c) multi-stage test for a
sand-bentonite backfill; (d) single-stage test for bentonite
amended compacted clay (data from Malusis & Shackelford
2002a, Yeo et al. 2005, and Kang & Shackelford 2010).
effective stress, , of 34.5 kPa, and then permeated

with DIW to flush soluble salts prior to membrane
testing under closed-system conditions. Unlike the
responses shown in Figs. 4ac, the response in Fig. 4d
53
Figure 6. Correlation between maximum chemico-osmotic

pressure difference, o , and source KCl concentration
based on the vant Hoff expression (Eq. 9).
(= Co ) of 3.9 mM KCl or the sand-bentonite backfill in Fig. 5c subjected to Cot (= Co ) of 3.88 mM

KCl, despite the latter two specimens having greater
bentonite contents. This comparison illustrates the
ability of bentonite to impart a significant membrane
behavior to a soil that would otherwise exhibit little or
no membrane behavior, especially when the soil being
amended is predominantly fine-grained soil (i.e., clay)
as opposed to a coarse-grained soil (i.e., sand).
4.7 Limitation in calculated values
Neuzil & Prost (2009) noted that Eq. 8 is obtained by
assuming does not vary across the specimen, which
is not accurate as is known to be strongly affected by
solute concentration, which varies across the specimen
as depicted in Fig. 2d. Thus, in reality, generally
also will vary spatially within the membrane, such that
values of calculated in accordance with Eq. 8 are
actually apparent values that lie somewhere between
the maximum and minimum in the specimen during the experiment. For this reason, Neuzil & Prost
(2009) indicate that, although based on Eq. 8 has a
simple physical interpretation and is easily calculated,
accurate predictions of chemico-osmotic pressures or
fluxes under conditions differing from those of the
test generally are not possible using such values.
Thus, prediction of field performance using values of
based on Eq. 8 for systems unlike those upon which
the values were determined should be interpreted
with caution and, at best, should be assumed to provide
only approximate results.
Figure 5. Time-dependent measured membrane efficiency

coefficients, o , based on results from Fig. 4: (a) single-stage
(Fig. 4a) and multi-stage (Fig. 4b) results for a GCL; (b)
multi-stage results for a sand-bentonite backfill (Fig. 4c);
(c) single-stage results for a bentonite amended compacted
clay (Fig. 4d).
low bentonite content for the sand-bentonite backfill,

these values of o still may be considered as substantial. Third, there is a slight trend of increasing o with
decreasing e, as expected. For example, the steadystate values of o for void ratios of 1.212, 1.012, and
0.812 were 0.12, 0.14, and 0.16, respectively.
Finally, the temporal trend in o for the compacted
clay specimen amended with 5 % bentonite corresponding with the values of P in Fig. 4d are shown
in Fig. 5c. For this test (i.e., Cot = Co = 3.9 mM
KCl), Kang & Shackelford (2010) reported a steadystate value of o of 0.76, which is significantly higher
than the steady-state value of o of 0.01 reported
for the unamended compacted clay subjected to the
same Cot (= Co ) of 3.9 mM KCl, and higher than
that for either the GCL in Fig. 5a subjected to Cot
EFFECTS OF INDEPENDENT VARIABLES
5.1 Effect of salt concentration

As previously noted, membrane behavior in clays is
known to be a function of salt concentration. For example, consider the results from Malusis & Shackelford
(2002a) and Kang & Shackelford (2010) shown in
Fig. 7, where the measured values of o at steady state
for the aforementioned GCL and bentonite amended
54
compacted clay are plotted as a function of the Co,ave

(Eq. 3) for KCl solutions. Also, the results from a test
conducted on a specimen of the unamended compacted
clay are included in Fig. 7 for comparison.
In general, o decreases approximately semi-log
linearly with increasing Co,ave , although the slope of
the decrease in o with increasing logarithm of Co,ave
varies as a function of the barrier material. As previously noted, the decrease in o with increasing Co,ave is
attributed to physico-chemical interactions, whereby
an increase in salt concentration in the pores of the
soil decreases the thicknesses of the adsorbed layers
of cations (or DDLs) surrounding individual bentonite
particles, resulting in progressively larger pore spaces
available for ion migration (Shackelford 2011, 2012).
The maximum values of Co,ave above which membrane behavior no longer exists (o = 0) are referred to
as threshold concentrations (e.g., Shackelford et al.
2003). Conservative (low) estimates of these threshold concentrations can be obtained by extrapolating
the semi-log linear trends.
As indicated in Fig. 7, the membrane behavior of
the unamended compacted clay was negligible over
the entire range of Co,ave (1.95 Co,ave 23.5 mM).
However, significant membrane behavior was evident
in both the bentonite amended compacted clay and
the GCL for values of Co,ave less than about 9.5 and
47 mM KCl, respectively. In addition, the slope of the
regressed relationship between o and log Co,ave for
the bentonite amended compacted clay was about 2.7
times greater than that for the GCL. Thus, o for the
GCL is greater than o for the bentonite amended compacted clay for 3.7 Co,ave 47 mM KCl, whereas o
for the GCL is less than o for the bentonite amended
compacted clay for Co,ave 3.7 mM KCl. This difference has been attributed, in part, to the difference
in bentonite contents of the bentonite amended compacted clay versus the GCL, i.e., 5% vs. 100% (Kang
& Shackelford 2010, Shackelford 2012). That is, the
membrane efficiency of a barrier with a lower bentonite content generally will be more sensitive to an
increase in salt concentration than that with a higher
bentonite content, as the greater the amount of bentonite, the more resilient the barrier against chemical
attack.
5.2
Figure 7. Membrane efficiency coefficients for

bentonite-based containment barriers as a function of
the average of the source KCl concentration across the
barrier (replotted after Shackelford 2012).
(leach) soluble salts from the materials for the purpose

of enhancing the likelihood of significant membrane
behavior. Finally, the tests for the GCL specimens and
the backfill specimens were conducted using rigidwall cells, with variations in o at a given bentonite
content resulting primarily from different initial void
ratios of the specimens (eo ) and some variability
resulting from conducting single-stage versus multistage tests, whereas the tests for the unamended and
bentonite amended compacted clay specimens were
conducted in flexible-wall cells at an initial effective
stress, , of 34.5 kPa.
As indicated by Fig. 8, except for the bentonite
amended compacted clay, there is a general trend of
increasing membrane behavior with increasing bentonite content, with the steady-state values of o
increasing from virtually nil (o = 0.01) for the unamended compacted clay to 0.48 o 0.68 for the
GCL containing 100% sodium bentonite. The paucity
of data for bentonite contents between 7.2% and 100%
can be attributed to bentonite contents for containment
barrier materials greater than about 10% generally
being considered cost prohibitive, except in the case
of manufactured GCLs (Shackelford 2012).
Shackelford (2012) also noted two additional interesting aspects of the results for the bentonite amended
compacted clay shown in Fig. 8. First, since the kh
values of both the unamended compacted clay and
the bentonite amended compacted clay were less than
109 m/s, low kh apparently is a necessary but not
sufficient condition for the existence of significant
membrane behavior. Second, the membrane behavior of the bentonite amended compacted clay (i.e.,
0.73 o 0.76) was greater than that of the GCL,
which contained 100 % bentonite. Although the reasons for this apparent discrepancy are not entirely
Effect of bentonite content
The overall effect of the amount of bentonite on the

magnitude of the membrane behavior existing in various containment barriers is illustrated in Fig. 8. In
addition to the aforementioned barrier materials, the
results from Yeo et al. (2005) for a soil-bentonite
backfill consisting of the aforementioned NFC natural clay and mixed with 5% bentonite slurry to achieve
a slump of 100 mm prior to testing (i.e., a total bentonite content of 2.12% by dry weight) are shown. In
general, all of the results in Fig. 8 pertain to steadystate o values resulting from tests conducted using
Cot (= Co ) of 3.9 mM KCl (3.88 mM KCl for the
backfill specimens), and all of the specimens were permeated with DIW prior to membrane testing to flush
55
Figure 8. Correlation between bentonite content and

membrane efficiency coefficient for various bentonite-based
containment barriers subjected to 3.9 mM KCl (replotted
after Shackelford 2012, Shackelford & Moore 2013).
Figure 9. Effect of valence of salt cation on the temporal

membrane efficiency of a GCL (replotted after Shackelford
2012).
maintained constant during the test, and the GCLs

were permeated with DIW prior to membrane testing
to reduce the soluble salts contents within the bentonite
of the GCL.
As shown in Fig. 9, steady-state values for o of
0.55 and 0.28 resulted for the tests using 3.9-mM and
20-mM KCl solutions, respectively, with the lower
o for the higher KCl solution being consistent with
the aforementioned effect of increasing salt concentration on clay o . However, for the test involving
the 5 mM CaCl2 solution, complete destruction of
the initially observed membrane behavior eventually
occurred despite this test specimen being only half the
thickness of the specimens tested using KCl solutions
(i.e., denser). This destruction of membrane behavior upon exposure to Ca2+ has been attributed to
progressive collapse of DDLs and the possibility of
particle aggregation resulting from diffusion of Ca2+
(Shackelford & Lee 2003, Di Emidio 2010, Mazzieri
et al. 2010, Shackelford 2012). These results suggest
that the effect of containment of liquids comprised of
multivalent cations on the membrane behavior of a
bentonite-based containment barrier should be evaluated prior to relying on any such membrane behavior
in design of the containment system.
clear, one factor likely contributing to this difference

is the relatively high fines content of the natural clay
(i.e., 89% fines), which likely enhanced the ability of the clay upon bentonite amendment to exhibit
significant membrane behavior. Another possible factor contributing to the difference is that the tests with
the GCL specimens were conducted using rigid-wall
cells where the stress conditions in the specimens
were unknown, whereas the tests with the specimens
of bentonite amended compacted clay was conducted
using a flexible-wall cell at an initial average effective
stress of 34.5 kPa. Thus, the different stress conditions
in these two types of cells may have contributed, in
part, to the difference in the test results for the two
types of bentonite-based containment barriers.
5.3 Effect of valence of salt cation
The use of salt solutions containing multivalent
cations (e.g., Ca2+ ) versus those containing monovalent cations (e.g., K+ ) is known to result in more
compressed DDLs, larger pores, and increased kh of
bentonite based materials (e.g., Shackelford 1994,
Gleason et al. 1997, Petrov & Rowe 1997, Petrov et al.
1997a,b, Stern & Shackelford 1997, Shackelford et al.
2000, Kolstad et al. 2004, Jo et al. 2005). Given that
larger pores should correlate with lower solute restriction, an increase in salt cation valence also should be
expected to result in a decrease in membrane behavior.
For example, consider the results in Fig. 9 showing
the temporal membrane behavior of two specimens of
a GCL using KCl solutions as reported by reported
by Malusis & Shackelford (2002a) versus the temporal behavior for a specimen of the same GCL using
a 5 mM CaCl2 solution as reported by Shackelford &
Lee (2003). All the tests were conducted using rigidwall cells where the thickness of the specimen was
5.4 Effect of effective confining stress

An increase in effective confining stress, , of clay is
known to result in lower kh values based on permeation
with water. This effect generally is attributed to consolidation of the clay, resulting in a lower void ratio,
e, with increasing , which in turn results in smaller
pores and lower kh . This same effect should be evident
with respect to membrane behavior, i.e., since smaller
pores should be more restrictive to solute migration.
For example, consider the results shown in Fig. 10
from Shackelford (2012) illustrating the trends in o
56
been shown with respect to the membrane behavior of

a GCL (Kang & Shackelford 2011).
For example, consider the results in Fig. 11 where
the ratio of o at of 241 kPa (35 psi) relative to that at
of 34.5 kPa (5 psi), R , increased from 1.6 at Co,ave
of 1.95 mM KCl (i.e., Cot = 3.9 mM KCl) to 7.9 at
Co,ave of 23.5 KCl (i.e., Cot = 47 mM KCl) for the same
GCL as previously described and tested in a flexiblewall cell. Although these results are limited to a single
GCL and a simple salt solution, the results illustrate
the potentially beneficial effects of increasing on
the membrane behavior. However, additional results
for different bentonite-based containment barriers and
more complex chemical solutions clearly are required
before more general conclusions can be drawn in this
regard.
6
FUTURE DIRECTIONS
As documented in this state-of-the-art review, significant research pertaining to the existence and
magnitude of membrane behavior in bentonite-based
containment barriers has been conducted in the past
approximate decade. Although the results of this
research have answered many questions regarding
such behavior and, therefore, have advanced the state
of the art, the research also has resulted in raising
questions that still need to be answered.
In particular, future research should focus at evaluating ways to overcome or control the deleterious
impact of more realistic and more aggressive chemical
solutions on the existence and magnitude of membrane
behavior in bentonite-based containment barriers than
has heretofore been used. Some research along these
lines recently has been conducted in terms of evaluating the potential of polymer modified bentonites for
resisting the deleterious impacts resulting from chemical attack (e.g., Di Emidio 2010, Scalia et al. 2011,
Bohnhoff 2012, Scalia 2012, Bohnhoff et al. 2013),
but given the number of potential independent variables affecting such behavior, including the myriad of
possible polymer and other chemical amendments to
bentonites, considerable additional research is needed.
Also, given the results of previous research indicating that the deleterious impacts of chemical attack on
the hydraulic conductivity of bentonites can be mitigated to some extent via increases in effective stress,
and the potential benefit of increased effective confining stress on the membrane behavior (Fig. 11), more
research should be conducted to evaluate the potential benefit gained from consolidating bentonite-based
containment barriers to higher effective stress in the
presence of more aggressive chemical solutions.
Finally, all of the experimental studies to date
have focused on the membrane behavior of bentonites under fully or nearly fully saturated conditions. Given that some applications may involve the
use of bentonite-based containment barriers under
unsaturated conditions, such as those used in cover
systems or as buffers for deep, underground disposal
of high-level radioactive wastes located above the
Figure 10. Effect of initial void ratio on the membrane

efficiency of three bentonite-based barriers (replotted after
Shackelford 2012).
Figure 11. Ratio of the membrane efficiency coefficient,

o , at 241 kPa versus that at 34.5 kPa as a function of the
average KCl concentration for specimens of a geosynthetic
clay liner (replotted after Shackelford 2012).
with the initial void ratio, eo , for three of the bentonite

based barriers previously considered. As expected, o
generally decreases with increasing eo , although the
trends for the two, less compressible bentonite-based
backfills are not nearly as evident as the trend for the
more compressible GCL.
An increase in also is known to provide increased
resistance of clays to the detrimental effects on kh
resulting from permeation with liquids other than
water (e.g. Fernandez & Quigley 1991, Shackelford
et al. 2000). This same beneficial effect recently has
57
(Politecnico di Torino), and Mike Malusis (Bucknell

University), for their careful reviews of drafts of this
paper.
water table, research should be undertaken to evaluate the potential effectiveness of such barriers under
unsaturated conditions.
7
SUMMARY AND CONCLUSIONS
REFERENCES
The results of extensive research performed over the

past approximate decade have indicated that the containment function of engineered clay barriers may
be enhanced by the ability of the barrier to exhibit
semipermeable membrane behavior. This enhanced
function results from solute restriction leading to a
reduction in both advective (hydraulic) and diffusive
(concentration) driven contaminant transport. Such
semipermeable membrane behavior has been found to
exist in engineered clay barriers comprised wholly or
partially of sodium bentonite, including geosynthetic
clay liners (GCLs), soil-bentonite (SB) backfills for
vertical cutoff walls, and bentonite amended natural
clays used as compacted clay liners (CCLs).
In general, the magnitude of membrane behavior
increases with the amount of bentonite in the barrier,
from virtually no membrane behavior in unamended
CCL to substantial membrane behavior in a bentonite
amended CCL and GCLs. The magnitude of any such
membrane behavior tends to increase with decreasing average salt concentration, Cave , across the barrier
and/or increasing effective confining stress. However,
results also indicate that chemical solutions containing
multivalent cations, such as Ca2+ , can destroy the existence of any membrane behavior. Thus, despite significant advances in our understanding of clay membrane
behavior, additional studies for further advancement
are required before membrane behavior can be relied
upon to enhance the containment function of engineered clay barriers. In particular, future studies aimed
at evaluating ways to overcome or control the potentially deleterious impact on the membrane behavior of
bentonite-based containment barriers resulting from
exposure to more realistic and more aggressive chemical solutions, such as landfill leachates, are needed.
Barbour, S.L. & Fredlund, D.G. 1989. Mechanisms of

osmotic flow and volume change in clay soils. Canadian
Geotechnical Journal, 26(4): 551562.
Bohnhoff, G.L. 2012. Membrane Behavior, Diffusion, and
Compatibility of a Polymerized Bentonite for Containment
Barrier Applications. PhD Dissertation, Colorado State
University, Fort Collins, Colorado, USA.
Bohnhoff, G.L., Shackelford, C.D., Malusis, M.A., Scalia,
J., Benson, C.H., Edil, T.B., Di Emiddio, G., Katsumi. T.,
& Mazzieri, F. 2013. Novel bentonites for containment
barrier applications. 18th International Conference on
Soil Mechanics and Geotechnical Engineering, Paris,
France, Sept. 26, 2013, in press.
Daniel, D.E., Bowders, J.J., & Gilbert, R.B. 1997. Laboratory hydraulic conductivity testing of GCLs in flexible
wall permeameters. Testing and Acceptance Criteria for
Geosynthetic Clay Liners, ASTM STP 1308, L.W. Well,
Ed., ASTM, West Conshohoken, Pennsylvania, USA,
208226.
Di Emidio, G., 2010. Hydraulic and Chemico-Osmotic Performance of Polymer Treated Clays. PhD Dissertation,
Ghent University, Ghent, Belgium.
Dominijanni,A., 2005. Osmotic Properties of Clay Soils. PhD
Dissertation, Politecnico di Torino, Torino, Italy.
Dominijanni, A. & Manassero, M., 2003. Modelling advective, diffusive and osmotic transport with reference to the
molecular scale. Proceedings of the XIII European Conference on Soil Mechanics and Geotechnical Engineering,
August 2528, 2003, Prague, Czech Republic, Vancek,
I., et al., Eds., Czech Geotechnical Society, CICE, Vol. 1,
327331.
Dominijanni, A. & Manassero, M., 2005a. Modelling contaminant transport through clay membrane barriers.
Proceedings of the 16th International Conference on
Soil Mechanics and Geotechnical Engineering, Osaka,
Japan, September 1215, 2005, Millpress, Rotterdam,
22372241.
Dominijanni, A. & Manassero, M., 2005b. Modelling osmosis and solute transport through clay membrane barriers,
Waste Containment and Remediation, Alshawabkleh, A.,
Benson, C.H., Culligan, P.J., Evans, J.C., Gross, B.A.,
Narejo, D., Reddy, K.R., Shackelford, C.D., Zornberg,
J.G.,Eds, ASCE, Reston, Virgina, USA.
Dominijanni, A. & Manassero, M., 2008. Influence of
membrane behavior on contaminant transport through
geosynthetic clay liners. Characterization, Monitoring,
and Modeling of GeoSystems, Alshawabkeh, A.N., Reddy,
K.R., and Khire, M.V., Eds., GeoCongress 2008, ASCE,
Reston, Virginia, USA, 814821.
Dominijanni, A. & Manassero, M., 2012a. Modelling the
swelling and osmotic properties of clay soils. Part I:
The phenomenological approach. International Journal
of Engineering Science, 51: 3250.
Dominijanni, A. & Manassero, M. 2012b. Modelling the
swelling and osmotic properties of clay soils. Part II: The
physical approach. International Journal of Engineering
Science, 51: 5173.
Dominijanni, A., Manassero, M., & Puma, S. 2013. Coupled
chemical-hydraulic-mechanical behaviour of bentonites.
Gotechnique, 63(3): 191205.
ACKNOWLEDGEMENTS
Financial support for the results of research reported
in this paper was provided by the U. S. National
Science Foundation (NSF), Arlington, VA under
Grants CMS-9616854, Coupled Solute Migration
through Clay Barrier Materials, CMS-0099430,
Membrane Behavior of Clay Soil Barrier Materials,
CMS-0624104, Enhanced Clay Membrane Barriers
for Sustainable Waste Containment, and CMMI0757815, GOALI: Bentonite-Polymer Nanocomposites for Geoenvironmental Applications. The
opinions expressed in this paper are solely those of the
writer and are not necessarily consistent with the policies or opinions of the NSF. The writer also is indebted
to his graduate students, Catherine Hong and Kristin
Sample-Lord, and Drs. Gretchen Bohnhoff (University of Wisconsin-Platteville), Andrea Dominijanni
58
Dominijanni, A., Manassero, M., & Vanni, D., 2006.

Micro/macro modeling of electrolyte transport through
semipermeable bentonite layers. Proceedings of the 5th
International Congress on Environmental Geotechnics,
Thomas, H.R., Ed., Cardiff, Wales, UK, June 2630, 2006,
Thomas Telford, London, Volume II, 11231130.
Dutt, G.R. & Low, P.F. 1962. Diffusion of alkali chlorides in
clay-water systems. Soil Science, 93(4): 233240.
Evans, J.C., Shackelford, C.D., Yeo, S.-S., & Henning,
J. (2008). Membrane behavior of soil-bentonite slurrytrench cutoff walls. Soil and Sediment Contamination
An International Journal, 17(4): 316322
Fernandez, F. & Quigley, R.M. 1991. Controlling the destructive effects of clay-organic liquid interactions by application of effective stresses. Canadian Geotechnical Journal,
28(3): 388398.
Fritz, S.J., 1986. Ideality of clay membranes in osmotic processes: A review. Clays and Clay Minerals, 34: 214223.
Garavito, A.M., 2005. Chemical Osmosis in Clayey Sediments. PhD Dissertation, Vrije Universiteit, Amsterdam,
The Netherlands.
Gleason, M.H., Daniel, D.E., & Eykholt, G.R. 1997. Calcium
and sodium bentonite for hydraulic containment applications. Journal of Geotechnical and Geoenvironmental
Engineering, 123(5): 438445.
Greenberg, J., Mitchell, J., & Witherspoon, P. 1973. Coupled
salt and water flows in a groundwater basin. Journal of
Geophysical Research, 78(27): 63416353.
Groenevelt, P.H. & Elrick, D.E., 1976. Coupling phenomena
in saturated homo-ionic montmorillonite: II. Theoretical.
Soil Science Society of America, Journal, 40: 820823.
Henning, J.T. 2004. Chemico-Osmotic Efficiency of Two Real
World SB Slurry Trench Cutoff Wall Backfills. MS Thesis,
Bucknell University, Lewisburg, Pennsylvania, USA.
Henning, J., Evans, J.C., & Shackelford, C.D. 2006. Membrane behavior of two backfills from field-constructed
soil-bentonite cutoff walls. Journal of Geotechnical and
Geoenvironmental Engineering, 132(10): 243249.
Jo, H., Benson, C.H, Shackelford, C.D., Lee, J.M., &
Edil, T.B. 2005. Long-term hydraulic conductivity of a
non-prehydrated geosynthetic clay liner permeated with
inorganic salt solutions. Journal of Geotechnical and
Kang, J.-B., 2008. Membrane Behavior of Clay Liner
Materials. PhD Dissertation, Colorado State University,
Fort Collins, Colorado, USA.
Kang, J.-B. & Shackelford, C.D. 2009. Clay membrane testing
using a flexible-wall cell under closed-system boundary
conditions. Applied Clay Science, 44(12): 4358.
Kang, J.-B. & Shackelford, C.D. 2010. Membrane behavior
of compacted clay liners. Journal of Geotechnical and
Kang, J.-B. & Shackelford, C.D. 2011. Consolidation
enhanced membrane behavior of a geosynthetic clay liner.
Geotextiles and Geomembranes, 29(6): 544556.
Keijzer, T.J.S., 2000. Chemical Osmosis in Natural Clayey
Materials. PhD Dissertation, University of Utrecht,
Utrecht, The Netherlands.
Keijzer, T.J.S., Kleingeld, P.J., & Loch, J.P.G. 1999. Chemical
osmosis in compacted clayey material and the prediction
of water transport. Engineering Geology, 53(2): 151159.
Kolstad, D.C., Benson, C.H., & Edil, T.B. 2004. Hydraulic
conductivity and swell of nonprehydrated GCLs permeated with multi-species inorganic solutions. Journal
of Geotechnical and Geoenvironmental Engineering,
130(12): 12361249.
Lu, N., Olsen, H.W., & Likos, W.L., 2004. Appropriate
material properties for advective-diffusive solute flux in
membrane soil. Journal of Geotechnical and Geoenvironmental Engineering, 130(12): 13411346.

Malusis, M.A., 2001. Membrane Behavior and Coupled
Solute Transport through a Geosynthetic Clay Liner.
PhD Dissertation, Colorado State University, Fort Collins,
Colorado, USA.
Malusis, M.A., Kang, J.-B., & Shackelford, C.D. 2013. Influence of membrane behavior on solute diffusion through
GCLs. Coupled Phenomena in Environmental Geotechnics (CPEG 2013), July 13, 2013, Torino, Italy, CRC
Press/Balkema, Leiden, The Netherlands.
Malusis, M.A. & Shackelford, C.D. 2001. Modeling
contaminant transport through clay membrane barriers.
Proceedings, 2001 International Containment and Remediation Technology Conference and Exhibition, Orlando,
FL, June 1013, Florida State University, Tallahassee
(www.containment.fsu.edu/cd/DoubleClickToStart.htm),
146149.
Malusis, M.A. & Shackelford, C.D. 2002a. Chemico-osmotic
efficiency of a geosynthetic clay liner. Journal of Geotechnical and Geoenvironmental Engineering, 128(2):
97106.
Malusis, M.A. & Shackelford, C.D. 2002b. Coupling effects
during steady-state solute diffusion through a semipermeable clay membrane. Environmental Science and Technology, 36(6): 13121319.
Malusis, M.A. & Shackelford, C.D. 2002c. Theory for reactive solute transport through clay membrane barriers.
Journal of Contaminant Hydrology, 59(34): 291316.
Malusis, M.A. & Shackelford, C.D. 2004a. Explicit and
implicit coupling during solute transport through clay
membrane barriers. Journal of Contaminant Hydrology
72(14): 259285.
Malusis, M.A. & Shackelford, C.D. 2004b. Predicting solute
flux through a clay membrane barrier. Journal of Geotechnical and Geoenvironmental Engineering, 130(5):
477487.
Malusis, M.A., Shackelford, C.D., & Maneval, H.W. 2012.
Critical review of coupled flux formulations for clay
membranes based on nonequilibrium thermodynamics.
Journal of Contaminant Hydrology, 138139: 4059.
Malusis, M.A., Shackelford, C.D., & Olsen, H.W. 2001a.
Flow and transport through clay membrane barriers. Geoenvironmental Engineering, Geoenvironmental
Impact Management, R.N. Yong & H.R. Thomas, Eds.,
Thomas Telford Publ., London, UK, 334341.
Malusis, M.A., Shackelford, C.D., & Olsen, H.W. 2001b. A
laboratory apparatus to measure chemico-osmotic efficiency coefficients for clay soils. Geotechnical Testing
Journal, 24(3): 229242.
Malusis, M.A., Shackelford, C.D., & Olsen, H.W. 2003.
Flow and transport through clay membrane barriers.
Engineering Geology, 70(23): 235248.
Manassero, M., & Dominijanni, A., 2003. Modelling the
osmosis effect on solute migration through porous media,
Mazzieri, F., Di Emidio, G., & Van Impe, P.O. 2010. Diffusion
of calcium chloride in a modified bentonite: Impact on
osmotic efficiency and hydraulic conductivity. Clays and
Clay Minerals, 58(3): 351363.
Mazzieri, F., Van Impe, P.O., & Di Emidio, G. 2005. Chemicoosmotic behavior of modified Multiswellable bentonite.
16th International Conference on Soil Mechanics and
Geotechnical Engineering, Osaka, Japan, Sept. 1216,
2005, Millpress, Rotterdam, The Netherlands, Vol. 4,
22972300.
Mazzieri, F., Van Impe, P.O., Van Impe, W.F., & Constales,
D. 2003. Measurement of chemico-osmosis parameters of
59
clayey soils. 13th European Conference on Soil Mechanics and Geotechnical Engineering, Balkema, Rotterdam,
Vol. 1, 433438.
Medved, I. & Cern,

R. 2013. Osmosis in porous media: A
review of recent studies. Microporous and Mesoporous
Materials. 170: 299317.
Mitchell, J.K., Greenberg, J.A., & Witherspoon, P.A., 1973.
Chemico-osmotic effects in fine-grained soils. Journal
of the Soil Mechanics and Foundations Division, 99:
307322.
Mitchell, J.K. & Soga, K., 2005. Fundamentals of Soil
Behavior, 3rd ed., John Wiley & Sons, New York.
Neuzil, C.E., & Provost, A.M., 2009. Recent experimental
data may point to a greater role for osmotic pressures in
the subsurface. Water Resources Research, 45: 114.
Olsen, H.W., Gui, S., & Lu, N., 2000. Critical review of
coupled flow theories for clay barriers. Transportation
Reseach Record 1714, Recycled and Secondary Materials, Soil Remediation, and In Situ Testing, Transportation Research Board-National Research Council, National
Academy Press, Washington, DC, USA, 5764.
Olsen, H.W., Yearsley, E.N., & Nelson, K.R. 1990.
Chemico-osmosis versus diffusion-osmosis. Transportation Research Record 1288, Geotechnical Engineering
1990, Transportation Research Board-National Research
Council, National Academy Press, Washington, DC, USA,
1522.
Petrov, R.J. & Rowe, R.K. 1997. Geosynthetic clay liner
(GCL) chemical compatibility by hydraulic conductivity
testing and factors impacting its performance. Canadian
Geotechnical Journal, 34(6): 863885.
Petrov, R J., Rowe, R.K., & Quigley, R.M. 1997a. Comparison of laboratory-measured GCL hydraulic conductivity
based on three permeameter types. Geotechnical Testing
Journal, 20(1): 4962.
Petrov, R.J., Rowe, R.K., & Quigley, R.M. 1997b. Selected
factors influencing GCL hydraulic conductivity. Journal
123(8): 683695.
Quigley, R.M., Yanful, E.K. & Fernandez, F. 1987. Ion transfer by diffusion through clayey barriers. Geotechnical
Practice for Waste Disposal 87, R.D. Woods, Ed., ASCE,
Reston, Virginia, USA, 137158.
Scalia, J. 2012. Bentonite-Polymer Composites for Containment Applications. PhD Dissertation, University of
Wisconsin, Madison, Wisconsin, USA.
Scalia, J. Benson, C.H, Edil, T.B., Bohnhoff, G.L., & Shackelford, C.D. 2011. Geosynthetic clay liners containing
bentonite polymer nanocomposite. Geo-Frontiers 2011:
Advances in Geotechnical Engineering, J. Han and D.E.
Alzamora, Eds., Geotechnical Special Publication 211,
Dallas, TX, March 1316, 2011, (CD-ROM) ASCE,
Reston, VA, 20012009.
Shackelford, C.D. 1994. Waste-soil interactions that alter
hydraulic conductivity. Hydraulic Conductivity and Waste
Contaminant Transport in Soil, ASTM STP 1142, D.E.
Daniel and S.J. Trautwein, Eds., ASTM, West Conshohocken, PA, 111168.
Shackelford, C.D. 2011. Membrane behavior in geosynthetic
clay liners. Geo-Frontiers 2011, Dallas, TX, March 13-16,
2011, ASCE, Reston, Virginia, USA, 19611970.
Shackelford, C.D. 2012. Membrane behavior of engineered clay barriers for geoenvironmental containment: State of the art. GeoCongress 2012-State of the
Art and Practice in Geotechnical Engineering, R.D.
Hryciw, A. Athanasopoulos-Zekkos, and N. Yesiller,
Eds., Geotechnical Special Publication 225, Oakland,
CA, March 2529, 2012 ASCE, Reston, Virginia, USA,

34193428.
Shackelford, C.D, Benson, C.H., Katsumi, T., Edil, T.B., &
Lin, L. 2000. Evaluating the hydraulic conductivity of
GCLs permeated with non-standard liquids. Geotextiles &
Geomembranes, 18(24): 133161.
Shackelford, C.D. & Lee, J.-M. 2003. The destructive role
of diffusion on clay membrane behavior. Clays and Clay
Minerals, 51(2): 187197.
Shackelford, C.D. & Malusis, M.A. 2002. Clay membrane behavior and coupled solute diffusion. ChemicoMechanical Coupling in Clays: From Nano-Scale to
Engineering Applications, C. Di Maio, T. Hueckel, and
B. Loret, Eds., Maratea, Italy, June 2830, 2001, Swets &
Zeitlinger B. V., Lisse, The Netherlands, 289297.
Shackelford, C.D., Malusis, M.A., & Olsen, H.W. 2001. Clay
membrane barriers for waste containment. Geotechnical
News, Bi-Tech Publishers, Vancouver, B.C., 19(2):
3943.
Shackelford, C.D., Malusis, M.A., & Olsen, H.W. 2003. Clay
membrane behavior for geoenvironmental containment.
Soil and Rock America Conference 2003 (Proceedings of
the joint 12th Panamerican Conference on Soil Mechanics and Geotechnical Engineering and the 39th U. S.
Rock Mechanics Symposium), P.J. Culligan, H.H. Einstein,
and A.J. Whittle, Eds., Verlag Glckauf GMBH, Essen,
Germany, Vol. 1, 767774.
Shackelford, C.D. & Moore, S.M. 2013. Fickian diffusion of
radionuclides for engineered containment barriers: Diffusion coefficients, porosities, and complicating issues.
Sleep, B.E., Shackelford, C.D., & Parker, J.C. 2006. Modeling of fluid transport through barriers (Chapter 2).
Barrier Systems for Environmental Contaminant Containment and Treatment, C.C. Chien, H.I. Inyang, and L.G.
Everett, Eds., CRC Press, Taylor and Francis Group, LLC,
Boca Raton, Florida, USA, 71141.
Stern, R.T. & Shackelford, C.D. 1998. Permeation of sandprocessed clay mixtures with calcium chloride solutions.
Journal of Geotechnical and Geoenvironmental Engineering, 124(3): 231241.
Van Impe, P.O., 2002. Consolidation, Contaminant Transport
and Chemico-Osmotic Effects in Liner Materials. PhD
Dissertation, University of Ancona, Ancona, Italy.
Van Impe, P.O, Mazzieri F., Van Impe W.F., & Constales,
D. 2003. Coupled flow model for three-ion advectivediffusive-reactive transport in consolidating clay liners. 13th European Conference on Soil Mechanics and
Geotechnical Engineering, Balkema, Rotterdam, Vol. 3,
227232.
Van Impe, P.O., Van Impe, W.F., & Mazzieri, F., 2005. Impact
of osmotic efficiency on contaminant transport parameters. 16th International Conference on Soil Mechanics
and Geotechnical Engineering, Osaka, Japan, Sept. 12
16, 2005, Millpress, Rotterdam, The Netherlands, Vol. 4,
23432346.
Whitworth, T.M. & Ghazifard, A. 2009. Membrane effects in
clay-lined inward gradient landfills.Applied Clay Science,
43(2): 248252.
Yeo, S.-S. 2003. Hydraulic Conductivity, Consolidation, and
Membrane Behavior of Model Backfill-Slurry Mixtures
for Vertical Cutoff Walls. MS Thesis, Colorado State
University, Fort Collins, Colorado, USA.
Yeo, S.-S., Shackelford, C.D., & Evans, J.C. 2005. Membrane
behavior of model soil-bentonite backfill mixtures. Journal of Geotechnical and Geoenvironmental Engineering,
131(4): 418429.
60

Liquid and gas flow in municipal solid waste

W. Powrie, R.P. Beaven, D.S. Holmes & A.P. Hudson
University of Southampton Faculty of Engineering and the Environment, Southampton, UK
ABSTRACT: Knowledge of the hydraulic conductivity of waste is essential to the efficient and effective
management of leachate in landfills. Gas permeability is important if landfill aeration is adopted as part of a
remediation strategy. In both cases, the relevant permeability (to liquid or gas) may be significantly affected by
the presence (in terms of both quantity and distribution) of the other phase. This paper reports on the current
state of understanding of the factors influencing the liquid and gas permeabilities of municipal solid wastes,
with particular emphasis on the interactions between them.
INTRODUCTION
Landfill has been the primary means of solid waste

disposal in many countries for the past hundred years
or more. Sites containing both old and new wastes will
often require active pumping of leachate for level control or treatment purposes, and/or to achieve long-term
stabilization through biodegradation and flushing. For
these reasons, knowledge of the hydraulic conductivity of domestic wastes and an understanding of the
factors that control it are essential.
More recently, landfill aeration has been proposed
and is being increasingly adopted as part of a final
remediation strategy. In designing a landfill aeration
scheme, it is clearly important to have knowledge of
the gas permeability of the waste.
The flow characteristics for one phase (i.e., gas or
liquid) will be affected by the presence and distribution of the other. Both flows may be affected by the
structure of the waste, which may result in the presence of preferential flow paths and/or a dual porosity,
clod and matrix effect.
The presentation will discuss the factors affecting
both the liquid and gas permeabilities of municipal solid wastes, with reference to large and smaller
scale laboratory tests and field experience. Particular
emphasis will be placed on the interactions between
the two phases.
Figure 1. Pitsea compression cell.
representative results to be obtained from samples of

generally highly heterogeneous wastes.
Overburden pressures are simulated by applying a
vertical stress via hydraulic rams acting on a platen
on top of the waste. Typically, the applied stress is
increased in five or six stages to a maximum of
600 kPa, representing landfill depths of up to 60 m. At
the end of each compression stage, the bulk density,
drainable porosity and saturated hydraulic conductivity of the waste are determined.
2 APPARATUS AND EXPERIMENTAL

PROCEDURES
2.1 Large scale compression cell
Large scale tests were carried out in the Pitsea compression cell (Figure 1), a purpose-built apparatus
accommodating a sample of waste 2 m in diameter
and up to 2.5 m high. The size of the apparatus enables
61
load, the position of the top platen was fixed. This was
necessary because movement of the top platen during subsequent hydraulic conductivity testing (which
occurred over a period of several days or weeks),
would have ruptured the inflatable seals between the
top platen and the cylinder side wall. Consequently
all hydraulic conductivity data were determined on
a waste sample held at constant volume, following
compression to a given initial effective stress.
The hydraulic conductivity of the waste at each
volume and with different degrees of gas accumulation was measured in a series of constant head flow
tests. Water from header tanks was allowed to flow
upward through the refuse initially at a fast rate so
as to purge the sample as far as possible of air and
gas. The hydraulic gradient during testing was determined by means of piezometers inserted through ports
in the side of the column. Piezometers at the same horizon indicating the same hydraulic head confirmed that
flow was vertical and approximately uniform.The flow
rate was measured using electromagnetic flowmeters,
except at low flow rates when direct measurement of
the (small) fall in water level in the header tanks with
time was found to be more reliable.
Each hydraulic conductivity test ran continuously
for several weeks, during which time gas generated
naturally by waste degradation, although able to vent to
atmosphere through the top platen, began to accumulate within the sample. The volume of gas accumulated
within the sample was deduced from the reduction in
weight indicated by the load cells mounted under the
compression cell framework. Each test was concluded
when no further gas accumulation was apparent. For
each volume, hydraulic conductivity tests were carried out at two different average pore water pressures
of about 25 kPa and 60 kPa. The change in pore
water pressure was effected by switching between two
different sets of header tanks located at different levels.
Once testing at both pore water pressures over the
range of gas accumulation conditions at each volume
had been completed, the top platen seal was deflated,
the sample was compressed to an increased vertical
stress, and the cycle of measurements was repeated.
Detailed descriptions of the compression cell,

including a correction to the applied stress to account
for side friction are given by Powrie & Beaven (1999)
and Beaven (2000). The general testing procedure
adopted for the current tests was as follows. The waste
sample was loaded into the cylinder, and its as-placed
weight determined by means of the load cells on which
the apparatus is mounted. Waste composition was
determined by sorting sub-samples typically 200 kg in
mass. As-placed water content was determined from
the weight loss of 1520 kg oven dried samples. When
compression in response to the application of a vertical stress had substantially ceased (i.e. the sample
thickness was changing at a rate of less than 1% in 24
hours), the waste was saturated by allowing water to
flow into the sample through the lower platen. After the
refuse has been saturated it was allowed to drain under
gravity to field capacity (defined as the water content
of the waste in conditions of free downward gravity
drainage), and the drainable porosity was calculated
from the volume of leachate drained, per unit total
volume. The bulk density could be calculated from
the known mass and volume of the waste at any stage.
The hydraulic conductivity of the refuse at each
vertical load was measured in a constant head flow
test. Water from the header tanks was allowed to flow
upward through the refuse. The hydraulic gradient was
determined by means of piezometers inserted through
ports in the side of the column. Piezometers at the same
horizon indicating the same hydraulic head confirmed
that flow was vertical and approximately uniform.
The flow rate was measured using electromagnetic
flowmeters, except at low flow rates when direct measurement of the (small) fall in water level in the header
tanks with time was found to be more reliable. The
refuse was then drained, the applied stress increased
and the cycle of operation and measurement repeated.
Basic hydraulic conductivity tests were carried out
on four different samples of domestic waste (DM3,
PV1, DN1 and AG1) to investigate the effects of particle size reduction, degradation and compression on the
bulk vertical hydraulic conductivity. DM3 was fresh,
unprocessed waste; PV1 was fresh waste that had been
pulverized and passed through a 150 mm screen and
heavy fines (including some putrescibles) removed;
DN1 was fresh waste that had been partly sorted and
tumbled in a drum using the Dano system; and AG1
which was a 25 year old partly degraded waste containing a mixture of soil, crude waste and pulverised waste
that had been recovered from a depth of less than 5 m
from a landfill site. Full characterization analyses are
given by Powrie & Beaven (1999) for sample DM3,
by Hudson et al (2001) for waste DN1, and by Beaven
(2000) for wastes PV1 and AG1.
Further tests were carried out on sample DN1, and
on a sample of shredded, fresh MSW denoted SW1, to
investigate the effects of partial saturation and gassing.
A full characterization of the waste SW1 is given by
Powrie et al. (2008).
The procedure for the further tests was as follows.
Following initial compression under a given vertical
2.2
Small scale tests to investigate the gas

permeability of mechanically-biologically
treated (MBT) wastes
A series of air permeability tests was carried out on

two MBT wastes, one of which had a maximum particle size of 10 mm and the second 9 mm. The second
MBT appeared more finely processed and had a visibly higher organic content. A characterization of the
10 mm MBT was carried out by Velkushanova et al.
(2009).
A standard permeameter 81 mm in diameter and
200 mm high was used (Figure 2). The lower portion
was filled with gravel to hold the specimen in place
and ensure equal distribution of air across the specimen. A highly permeable geotextile provided a level
62
Measurements of mass were taken after placement of

each layer to ensure knowledge of the mass of specimen being tested and its volume. The other parameters
of the specimen could then be calculated from the
volume, mass and gravimetric moisture content.
At the end of each test, the permeameter was
emptied of MBT, which was then mixed back into
the specimen. The geotextile filter was removed and
brushed clean. If any MBT had spilled into the gravel,
the gravel was removed, washed and dried before it
was reused in the permeameter.
Figure 2. Schematic diagram of the permeameter. The specimen layers, permeameter and flowmeter and are to scale.
The other elements, including the distances between each,
are not.
3
3.1
RESULTS
Saturated hydraulic conductivity of MSW
Raw data of hydraulic conductivity, drainable porosity and density at various vertical effective stresses
are given for samples DM3, PV1 and AG1 by Beaven
(2000), and for sample DN1 by Hudson et al. (2001)
tested in the Pitsea compression cell. Figure 3 shows
the permeability in notionally saturated conditions for
all four wastes plotted as functions of (a) vertical
effective stress; (b) density and (c) porosity.
The data presented in Figure 3 show that
base for the specimen and stopped mixing of the gravel

and MBT.
The specimen was compressed into the volume
needed to achieve the density desired for a particular
test using a flat-ended metal rod.
The delivery of dry air was controlled by pressure
regulators in combination with a U-tube manometer
allowing the pressure to be controlled to 10 Pa. The
U-tube manometer was filled with 40 ml of water and
had sufficient vertical length to measure a head of
600 mm of water, or 5880 Pa.
The air flow rate was measured using a thermal mass
flow meter, which had maximum reading of 5 litres per
minute at atmospheric pressure and a stated accuracy
of 0.3% of the full-scale reading.
A measured amount of water was then added and
mixed into a dry specimen first with a heavy-duty spatula, and then by rubbing the MBT by hand wearing
needle-proof gloves. The specimen was judged to be
well mixed when it was a consistent colour. It was
noted however that both MBT wastes had a tendency to
form discrete clumps in which fines were held together
by moisture. A sub-specimen of approximately 40 g
was taken for oven drying for accurate determination
of the moisture content.
When the specimen had to be left overnight, or
for a period of more than 30 minutes, it was placed
in a sealed box to prevent moisture loss though
evaporation.
To achieve a constant density in the specimen, a
predetermined mass of MBT was placed into the permeameter, appropriate to the density required for the
test. This was then slowly compacted with the compaction tool to a 2.5 cm depth. Each layer of MBT
placed in this way had a total emplacement volume
of 12.1 cm3 . The top of each layer was then disturbed
slightly with a thin metal wire to ensure good contact
with the base of the next layer.
Air was then passed though the specimen. The
flowrate and pressure head were recorded to determine
the intrinsic air permeability.
The whole procedure, including specimen placement, was repeated three times for each density, with
four different flowrates and pressure drops each time.
1. There is a single correlation for all samples between

vertical hydraulic conductivity and vertical effective stress in first loading. Differences in hydraulic
conductivity resulting from particle size reduction
and waste degradation are essentially second order,
but appear to become more significant at higher
vertical effective stresses (with a spread of just over
one order of magnitude in hydraulic conductivity at
a vertical effective stress of 500 kPa).
2. There are individual correlations between vertical
hydraulic conductivity and density for each waste
type, with an essentially linear relationship between
the logarithm of the vertical hydraulic conductivity
and the dry density.
3. There is a single correlation between the vertical
hydraulic conductivity and the drainable porosity
of the waste. This is not surprising, as the drainable porosity represents a measure of the size and
degree of connectivity of the voids, both of which
will have a major influence on the bulk hydraulic
conductivity. However, unlike the vertical effective
stress, the drainable porosity is a difficult parameter
to estimate in advance for design purposes, so the
correlation between vertical hydraulic permeability
and vertical effective stress is of more practical use.
3.2
Effect of gas accumulation on the hydraulic

conductivity of MSW
Figure 4 shows the effect on hydraulic conductivity

of gas accumulation in the Dano processed sample
DN1combined with data from a fresh shredded domestic waste SW1 (waste compositions given in Hudson
et al. 2001 and Powrie et al. 2008 respectively) for
stresses between 40 and 228 kPa (representing landfill
63
Figure 4. Hydraulic conductivity for fresh processed MSW

in different conditions of gas accumulation and pore water
pressure.
Figure 5. Changes in hydraulic conductivity with time for

sample SW1 at a constant volume corresponding to an initial
applied stress of 40 kPa, with an average pore water pressure
of 60 kPa.
Figure 5 shows the changes in average bulk

hydraulic conductivity and the volume of accumulated
gas over a 27 day period for a test conducted at constant volume (corresponding to an initial applied stress
of 40 kPa) with a relatively high average pore water
pressure of 60 kPa. The reduction in hydraulic conductivity in response to the increase in the volume of
gas contained within the sample is clear.
The large reductions in hydraulic conductivity
recorded in these tests are understandable when the
volumes of leachate that must have been displaced
from the sample by gas accumulation are considered.
For example for sample SW1 compressed to an applied
effective stress of 40 kPa, the weight change from
nominally saturated to gas accumulated conditions
(recorded by the load cells on which the compression
cell is mounted) indicated that between 78% and 87%
of the drainable water content was displaced by gas
accumulation (there is some uncertainty to the exact
figure as it is not known how much gas was resident
in the gravel layers above and below the sample).
Although the tests lasted for periods of several
weeks, gas accumulation could occur very quickly. In
a test carried out at a pore water pressure of 60 kPa following compression to an applied stress of 87 kPa, 136
litres of gas accumulated within the first three hours
of the test, causing a 76% reduction in hydraulic conductivity. This test was carried out during the summer
in relatively warm conditions. Rapid gas generation
Figure 3. Vertical hydraulic conductivity against (a) the logarithm of the vertical effective stress in first loading; (b) the
drainable porosity; and (c) density, for wastes DM3, PV1,
AG1and DN1 (data from Beaven, 2000 and Hudson et al.,
2001).
depths of approximately 4 m to 23 m respectively) and

at two different pore water pressures of 30 to 40 kPa
(designated low pore water pressure) and 60 to 70 kPa
(designated high pore water pressure). Both samples
were free to vent to atmosphere during these tests. For
tests carried out at low pore water pressure these data
indicate that up to two orders of magnitude reduction in
hydraulic conductivity occurred in initially nominally
saturated waste in response to in situ gas generation.
Increasing the pore water pressure compressed the gas
and reduced its effect on hydraulic conductivity, typically resulting in a reduction in hydraulic conductivity
of one order of magnitude from nominally saturated
to gas accumulated conditions.
64
Figure 6. Intrinsic air permeability of the 10 mm MBT as

a function of liquid saturation for fixed values of void ratio.
The fixed values of void ratio are given in the legend on the
right of the figure.
Figure 7. Relative permeability (log scale) of the 10 mm

MBT as a function of liquid saturation for fixed values of
void ratio. The Parker et al. (1987) curve fit is given as a
solid line with the Brooks & Corey (1964) curve fit given as
a dashed line. The values of void ratio are given in the legend
on the right of the graph.
was not evident in earlier tests carried out at cooler

ambient temperatures.
3.3 Air permeability of MBT wastes
The data were processed to show intrinsic air permeability as a function of the degree of saturation for
given values of void ratio.
The value for intrinsic air permeability of dry MBT
was used as a baseline for a relative permeability. The
values of relative permeability were then compared to
the curve-fits of Parker et al. (1987) and Brooks &
Corey (1964). These are:
(Parker et al., 1987) and
Figure 8. Relative permeability (linear scale) of the 10 mm

MBT as a function of the liquid saturation for fixed volumes
of void ratio. The Parker et al. (1987) curve fit is given as a
a dashed line. The values of void ratio are given in the legend
on the right of the graph.
(Brooks & Corey, 1964).

Se is the effective saturation ratio and m and are
fitting parameters. Se is defined as the volume of drainable waster divided by the volume of drainable pore
volume. The parameter values for the curve-fits are
taken from White (2011).
Figure 6 shows the intrinsic air permeability of the
10 mm MBT as a function of the degree of saturation,
for given values of void ratio. The estimated value for
fully saturated intrinsic air permeability is also shown.
There is a plateau at approximately 3.5 1011 m2
for the values of intrinsic air permeability, for degrees
of liquid saturation below 0.4 for any void ratio. This
value is marked as a dashed line on the graph. The liquid volume at this degree saturation would be confined
to the smallest pores by capillary effects, and should
not create significant resistance to gas flow. The value
of 3.5 1011 m2 is therefore assumed to be equal to
the intrinsic air permeability and is used to calculate
the relative permeabilities in Figure 7 and 8.
The Brooks & Corey (1964) curve fit hugely overestimates both the rate and extent of the fall in
permeability at increasing liquid saturation ratio. The
Parker et al. (1989) curve fit is much closer and seems
to give an approximate lower bound. The degree to

which the intrinsic air permeability falls as liquid saturation is approached is reasonably closely matched,
but the rate of decline is matched only by the data
for a void ratio of 1.1. The reduction in intrinsic air
permeability with increasing liquid saturation is quite
small until the pore space is nearly saturated with liquid. This suggests that the pore distribution is such that
there is a connected pore network of relatively large
voids from which water is lost even at high saturation
ratios resulting in the rapid establishment of intrinsic
air permeability on drying.
Presenting the data on a linear scale (Figure 8)
illustrates the variability of relative permeability at
each void ratio.
Figure 9 shows the intrinsic air permeability of the
9 mm MBT as a function of the degree of saturation for
the whole range of void ratios. The estimated value for
fully saturated intrinsic air permeability is also shown.
Void ratios are not shown separately owing to the
large range of values (1.8 to 4.4), some of which provide few data points. This is a much greater range, with
higher values of void ratio, than for the 10 mm MBT.
The 9 mm MBT shows a greater variation in intrinsic
air permeability across the range of liquid saturation.
A similar plateau of intrinsic air permeability is apparent, but is slightly higher than the value for the 10 mm
65
Figure 11. Relative permeability (log scale) of the 9 mm

a dashed line. The values of void ratio are given to the right
of the graph.
Figure 9. All 9 mm data. The fully saturated intrinsic air

permeability is given as a dashed line.
Figure 10. Intrinsic air permeability of the 9 mm MBT as

a function liquid saturation for fixed values of void ratio.
The void ratios are given on the right hand side of the graph.
The dashed line represents the estimation for the saturated
intrinsic air permeability.
Figure 12. Relative permeability (linear scale) of the 9 mm

a dashed line. The values of void ratio are given to the right
of the graph.
MBT. The fully saturated intrinsic air permeability of

the 9 mm MBT is estimated to be 1 1010 m2 . This
is shown as a dashed line in Figure 9.
A clearer picture of the intrinsic air permeability
as a function of liquid saturation for the 9 mm MBT
waste is given by Figure 10, which gives intrinsic air
permeability as a function of liquid saturation for particular values of void ratio. Only void ratios which
contributed five or more points are shown for reasons
of clarity. At degrees of liquid saturation of about 0.2
to 0.4, the permeability rises slightly for void ratios of
2.2, 2.6, 2.9, 3.2 and 3.8. The void ratios are higher in
the 9 mm MBT due to its greater particle density, but
similar bulk densities to the 10 mm MBT.
The relative permeability curve fits were applied to
the 9 mm MBT data assuming a saturated ka value of
1 1010 m2 . The results are shown in Figure 11 on
a log scale and Figure 12 on a linear scale. The data
for the 9 mm MBT waste is much noisier than that for
the 10 mm MBT waste and the relative permeability
appears to be less than the Brooks & Corey (1964)
and also greater than the Parker et al. (1987) at lower
levels of liquid saturation. However, at higher levels of
liquid saturation the fall in intrinsic air permeability
is reasonably closely represented by the Parker et al.
(1987) curve fit both in extent and rate of decline.
DISCUSSION
An unsaturated waste has both air and water in its

voids. However, for either phase to be able to flow
through the waste, that phase must be continuous
within the pore space. Thus a near-saturated waste
may be impermeable to air, because the air present
does not offer a continuous pathway. Likewise, at low
water contents a waste may not be permeable to water
because the water present is confined to the smallest
pores, is not continuous within the waste and plays no
part in liquid flow.
As a porous medium desaturates, water drains from
the largest pores first. Because these are the most
important for water flow, the water permeability would
be expected to fall quite rapidly. Once the point is
reached at which the water phase is no longer continuous, or the water remaining in the waste is in pores
so small that it is effectively immobile, the water permeability (hydraulic conductivity) will be zero. The air
permeability of a saturated waste is effectively zero;
and will remain so as the waste desaturates until the air
phase has become continuous in the larger pores, when
it will rise dramatically. After that, desaturation of the
66
magnitude. At elevated pore water pressures, compression of the trapped gas will reduce its impact.
5. Air permeability remains substantially constant
with increasing degree of liquid saturation, until
full saturation is approached at which point the air
permeability declines rapidly. The value of liquid
saturation at which this takes place is dependent on
the void ratio of the specimen.
6. The relationships between increasing liquid saturation and air permeability form hyperbolae. This
suggests that providing continuity of air-filled pore
space is maintained, air permeability is unaffected.
The Parker et al. (1987) curve fit is a reasonable
approximation, but it understates the rate of decline
for air permeability, suggesting that the pore connectivity parameter is inaccurate for MBT wastes.
Figure 13. Variation of water and air permeability with

degree of soil saturation (schematic).
smaller pores will not make very much difference to

the air permeability.
This behaviour is well-known in soils, as illustrated schematically in Figure 13; permeabilities are
expressed as relative permeabilities krw and kra , where
krw = kw /ks and kra = ka /kd ; kw is the water permeability and ka the air permeability at the current water
content , ks is the (maximum) water permeability
when the waste is saturated ( = s), and kd is the
(maximum) air permeability when the waste is at the
residual water content, r .
The results reported in this paper suggest that the
water and air permeabilities of MSW and MBT waste
follow similar trends to those for soils. This is also
in accordance with the findings of previous investigations into drainable porosity and hydraulic conductivity e.g. Beaven (2000) and Stoltz et al. (2010). In this
respect, the properties of MSW and MBT wastes are
essentially similar to those of soils albeit with additional complicating factors such as potentially higher
compressibilities and degradation resulting in additional settlements and the internal generation of gas.
ACKNOWLEDGEMENTS
The work described in this paper has been supported
by grants from the UK Engineering and Physical Sciences Research Council (EPSRC), reference numbers
GR/T25194, EP/E041965 and EP/I012206, and from
Cleanaway Limited and Veolia under the Landfill Tax
Credit System.
REFERENCES
Beaven, R.P. (2000) The hydrogeological and geotechnical
properties of household waste in relation to sustainable landfilling. PhD dissertation, University of London
(Queen Mary and Westfield College)
Brooks, R.H. & Corey, A.T. (1964) Hydraulic properties
of porous media. Hydrology Papers, Colorado State
University.
Hudson, A.P., Beaven, R.P & Powrie, W. (2001) Interaction of
water and gas in saturated household waste in a large scale
compression cell. Proc. Sardinia 2001, 8th International
Landfill Symposium, Cagliari, Italy, III, 585593.
Powrie, W. & Beaven, R.P. (1999) Hydraulic properties of household waste and implications for landfills.Proceedings of the Institution of Civil Engineers
(Geotechnical Engineering) 137 (4), 235247.
Parker, J., Lenhard, R. & Kuppusamy, T. (1987) A parametric
model for constitutive properties governing multiphase
flow in porous media. Water Resources Research 23 (4),
618624.
Powrie, W., Beaven, R. P. & Hudson, A. P. (2008) The influence of landfill gas on the hydraulic conductivity of waste.
In, Geocongress 2008. The Challenge of Sustainability in
the GeoenvironmentAnnual Congress of the Geo-Institute
of ASCE, New Orleans, Louisiana, USA, 0912 Mar
2008.
Stoltz, G., Gourc, J.P. Oxarango, L. (2010) Liquid and gas
permeabilities of unsaturated municipal solid waste under
compression. Journal of Contaminant Hydrology 118 (1),
2742.
White (2011) Technical note: Interpretation of capillary
pressure and relative permeability data.
Velkushanova, K., Caicedo, D., Richards, D., & Powrie, W.
(2009) A detailed characterisation of an MBT waste. Proc.
Sardinia 2009, 12th International Landfill Symposium,
Cagliari, Italy. Paper E10.
CONCLUSIONS
1. For saturated MSW, there appears to be single

correlation between vertical hydraulic conductivity
and vertical effective stress in first loading. Differences in hydraulic conductivity resulting from
particle size reduction and waste degradation are
essentially second order, but appear to become
more significant at higher vertical effective stresses
(with a spread of just over one order of magnitude in
hydraulic conductivity at a vertical effective stress
of 500 kPa).
2. For a particular waste type, there is an individual
correlations between vertical hydraulic conductivity and density, with an essentially linear relationship between the logarithm of the vertical hydraulic
conductivity and the dry density.
3. There is a single correlation for all of the MSW
types tested between the vertical hydraulic conductivity and the drainable porosity of the waste.
4. Gas accumulation could reduce the hydraulic
conductivity by between one and two orders of
67

Experimental evidence of anisotropy in municipal solid waste

D. Zekkos
University of Michigan, Ann Arbor, USA
ABSTRACT: Municipal Solid Waste (MSW) includes variable waste constituents that have significantly different physical and mechanical characteristics. Many of these constituents (e.g., paper, plastic, wood) are fibrous
and provide a reinforcement effect. The properties of MSW vary not only with space, but also with time. Our
understanding of the geotechnical properties of Municipal Solid Waste has improved significantly in the last two
decades. Datasets of large-size laboratory and field experiments have become available shedding light on the
complexities of Municipal Solid Waste behavior. The available data indicates that MSW is highly anisotropic. Two
types of anisotropy are identified: Fabric anisotropy and stress-induced anisotropy. Observations and recent experimental evidence of the anisotropy of MSW in shear resistance, compressibility, wave propagation and hydraulic
conductivity are presented. The influence of waste anisotropy on engineering practice remains unknown, but
anisotropy should probably play a role in many aspects of landfill engineering practice. Currently, engineering
practices do not consider the influence of anisotropy on the design and performance of MSW landfills with the
exception, in some cases, of the anisotropy in hydraulic conductivity.
INTRODUCTION
considered in research or landfill engineering practice.

In landfill design for example, the shear strength, compressibility, or stiffness of MSW at various orientations
is considered identical. An exception is probably the
influence of anisotropy on hydraulic conductivity, and
by extension, on the liquid and gas flow through the
waste mass. A growing volume of the literature is recognizing that hydraulic conductivity of MSW in the
horizontal direction can be significantly higher than
the hydraulic conductivity in the vertical direction. For
example, Jain et al. (2005) assumed ratio of horizontal to vertical permeability to air that ranged from 1
to 100.
The purpose of this paper is to summarize recent
studies and experimental data that highlight that MSW
is a highly anisotropic material. The influence of
waste anisotropy in landfill design and operation has
not been assessed yet and may impact the operation
of landfills, especially recirculation and bioreactor
landfills.
Municipal Solid Waste (MSW) is an environmentallysensitive material that consists of solid waste constituents disposed of by residential communities and
commercial activities. Depending on waste management practices, MSW may be recycled, incinerated
or landfilled. Food scraps and green waste may be
composted. In many countries, including the US, the
majority of MSW is still disposed of in landfills.
The waste composition of MSW that reaches the
landfills is, however, affected by other waste management practices (e.g., recycling or composting), cultural
factors and social habits and priorities. Most commonly, MSW includes primarily food scraps, papers,
plastics, wood. Food scaps are readily biodegradable
and quickly become difficult to discern from other
waste constituents. Per modern landfill practices, these
constituents are mixed with soils, primarily because
regulations require the placement of a soil cover on
a daily basis. Occasionally, other contaminated soils,
or construction and demolition (C&D) debris may be
placed at MSW landfills.
Significant advances have been made in recent
years in characterizing the geotechnical properties of
MSW. An extensive review of recent studies and findings has been made recently (Zekkos 2011). These
studies, as well as additional studies since then, have
greatly improved our understanding of the geotechnical properties of MSW.
An important aspect of MSW mechanical behavior
that has not been explicitly addressed in the literature, despite occasional references, is the anisotropy of
MSW. The impact of waste anisotropy is not typically
SOURCES OF ANISOTROPY IN MSW
Two types of anisotropy are typically considered in

earth materials:
i Structural or fabric anisotropy; and
ii Stress-induced anisotropy.
These types of anisotropy need to also be considered
in Municipal Solid Waste. The sources of structural
and stress-induced anisotropy in MSW are presented
in the following sections.
69
soil covers consisting of clayey soils may have significantly lower hydraulic conductivity compared to the
hydraulic conductivity of the waste layer.
A second contributing factor to fabric anisotropy is
at the meso-scale (order of tens of cm) and is the presence of fibrous waste constituents in the waste mass.
These constituents consist primarily of papers, plastics, and wood pieces, but also occasionally, clothing,
textiles and other fabrics. These planar or elongated
constituents have a tendency to be horizontally oriented in landfills during compaction, although this
is unintentional. This horizontally layered structure
of MSW was also observed in the laboratory during specimen preparation (Zekkos et al. 2010, Bray
et al. 2009, Bareither et al. 2012c) and similar observations have been made earlier in specimens of fiber
reinforced soil (Michalowski and Cermak 2002, Gray
and Ohashi 1983; Jewell and Wroth 1987). Fibers in
fibrous peats are also found to be horizontally oriented
both in the field and the laboratory (Landva and La
Rochelle 1983, Boulanger et al. 1998).
Both sources of structural anisotropy likely contribute to the anisotropy of MSW in the field. Largesize laboratory testing (e.g., 300-mm in diameter or
larger) has captured the meso-scale anisotropy (due to
the fibrous nature of MSW), but probably is still too
small to capture the macro-scale anisotropy associated
with layering of soils and waste. Conventional, smallsize laboratory testing will not fully capture anisotropy
because such testing will either require removal of
the fibrous waste constituents or excessive milling
that would significantly reduce the size of the fibrous
constituents.
Figure 1. View of a 1 m deep trench in MSW at Los Reales

landfill in Tucson, Arizona. The top layer is the daily soil
cover, followed by a layer of MSW.
2.1
Structural (or fabric) anisotropy
There are numerous sources of structural anisotropy

in earth materials. Most common sources include
anisotropic consolidation, shear, mode of deposition or compaction of soils. In general terms, fabric
anisotropy results from both preferred orientation of
elongated or platy particles and stratification of soil
deposits (Mitchell and Soga 2005). Fiber reinforced
soils especially are anisotropic because of the orientation of the fibers in a preferred bedding plane
(Michalowski 2012).
There are two major sources of structural anisotropy
in MSW disposed of in landfills. At the macroscale
(order of tens of meters), the waste mass in landfills
consists of sequences of soil and waste layers. This
is a result of the placement of a soil cover at the end
of each day of waste disposal activities. In between
these two distinct layers, there is a mixing zone where
soil particles are mixed with the waste. Compaction
with sheepsfoot rollers also facilitates further mixing.
Fig. 1 is an illustration of the layered MSW system in
a 1-m deep trench at the Los Reales landfill in Tucson,
Arizona.
This sequence of layers can be particularly pronounced in modern landfills that comply with landfill
regulations. The layering geometry will vary across
regions and even within each landfill, and is a function
of the ratio of soil to waste used at each location. The
thicknesses of each waste and cover soil layer will also
vary. The presence of this layered structure gives rise
to significant macroscopic anisotropy in the landfill
waste mass.
Depending on the type of soil being used, there
can also be a significant difference in the mechanical
and the hydraulic properties between the two layers
(soil and waste). Some of these differences may be
less pronounced due to the mixing zone in between.
It is not uncommon, especially in older landfills, to
encounter perched saturated layers of waste that are
enclosed in between layers of daily soil cover. Daily
2.2 Stress-induced anisotropy

Geostatic stresses in soils are typically anisotropic,
unless the horizontal earth pressure coefficient is equal
to 1.0, which is rarely the case. Most commonly, the
at-rest earth pressure coefficient Ko in normally consolidated soils is in the order of 0.40.65 (Holtz et al.
2011, Mesri and Ajlouni 2007).
The horizontal earth pressure coefficient in MSW is
lower. Experimental data on MSW indicate that due to
the fibrous waste constituents, Poissons ratio and the
equivalent at-rest lateral earth pressure coefficient are
significantly affected by waste composition. Landva
et al. (2000) showed experimental data that indicate
that Ko reduced from 0.5 to 0.26 with increasing percentage of fibrous constituents. When fibrous waste
materials are present, Poissons ratio takes very low
values, and values as low as 0.1 has been reported for
MSW in the laboratory (Zekkos 2005). Assuming that
equation (1) is still valid for anisotropic materials, the
calculated Ko values are equal to 0.10.35:
Similarly low values of Poissons ratio and Ko for

MSW have been reported from field measurements
70
Figure 3. Effect of fibrous waste orientation with respect to

the shear surface in direct shear (Zekkos et al. 2010).
direct shear may be parallel to the alignment of fibrous

and elongated particles and as a result, the shear resistance of MSW may reach its minimum. Fresh waste
from Edmonton that consisted of significant amount of
plastic sheet waste yielded a friction angle of 24 , and
that low resistance was attributed to sliding between
aligned particles. Direct shear data reported by Bray
et al. (2009) and Zekkos et al. (2010) provided more
definitive data on the importance of orientation of
the fibrous waste constituents on the shear resistance
of MSW. Specimens prepared with identical waste
composition and compaction effort were sheared in a
square 300-mm size direct shear device. A set of specimens were directly prepared in the direct shear box
and, as discussed previously, fibrous constituents were
found to become oriented in the horizontal direction
during compaction. A second set of specimens were
prepared in a specially designed split-mold. The mold
allowed each test specimen to be rotated up to 90
before placing it in the direct shear device. As a consequence, MSW specimens were placed in the direct
shear device with fibrous constituents being oriented
perpendicular (i.e., vertical) to the horizontal shear
surface imposed by the direct shear box.
Representative results of shear stress as a function
of horizontal displacement for two pairs of direct shear
specimens with different waste composition and at two
different confining stresses are shown in Fig. 3. For
each pair of specimens that have the same testing conditions (confining stress and waste composition), but
different orientation of fibrous constituents, the shear
responses were significantly different. When shearing
engaged the fibrous waste constituents, a pronounced
upward curvature was observed in the shear stress
vs. horizontal displacement curve, as opposed to a
peak shear resistance. Specimen UP-15, with fibrous
constituents perpendicular to the horizontal shearing
plane, has a shear resistance that is more than twice
that of UP-10 at horizontal displacement of 55 mm.
Specimen UP-19 has a shear resistance that is about
Figure 2. View of MSW specimen exhibiting a pronounced

layered structure following triaxial compression.
(Carvalho and Vilar 1998, Dixon et al. 1998), although

the scatter in the data is higher. As a consequence,
MSW in the field is under a significant anisotropic
stress condition that should also result in anisotropy
in the engineering response of MSW. Note that the
relatively low values of Ko for MSW are very comparable to the low Ko values of 0.30.35 that have
been observed for normally consolidated fibrous peat
deposits (Mesri and Ajlouni 2007). Edil and Dhowian
(1981) also found Ko to be equal to 0.53 for amorphous
granular peat and equal to 0.30 for fibrous peats.
The fabric anisotropy of the MSW during placement
combined with the application of increasing vertical
stress during placement of overburden layers of MSW
in landfills, result in a more pronounced horizontal
layering of the fibrous waste constituents enhancing
the anisotropic structure of the MSW. An illustration
of that mechanism can be observed in the laboratory.
Fig. 2 illustrates the pronounced layering of a MSW
specimen following completion of a triaxial compression test. The material exhibits pronounced anisotropy,
and the tested specimen is disassembled in parallel
horizontal discs (Zekkos et al. 2012).
3
3.1
EXPERIMENTAL EVIDENCE OF MSW

ANISOTROPY
Shear resistance
There is significant evidence of waste anisotropy in

shear resistance of MSW. The pioneer work by Landva
and Clark (1986) pointed out that the shear plane in
71
Figure 4. Stress-displacement response for specimens with

plastic reinforcement at varying fiber orientation angles i
(Athanasopoulos et al. 2008).
Figure 5. Effect of fiber orientation angle (in degrees) for

all specimens at a normal stress of 50 kPa (Athanasopoulos
et al. 2008).
20% higher than that of UP-18 at horizontal displacement of 55 mm. The smaller difference in shear
resistance between the two specimens at the higher
normal stress (50 kPa) can probably be attributed to
the limitations of the device in terms of the maximum
horizontal displacement of 55 mm. This direct shear
data clearly indicate that the shear resistance of MSW
is anisotropic and significantly affected by the relative
orientation of the fibrous constituents to the horizontal
shear surface.
Tests byAthanasopoulos et al. (2008) on mixtures of
waste constituents and daily soil cover from a landfill
in Greece that were highly structured, highlighted the
significant effect of anisotropy. The shear resistance
of the specimens in direct shear was found to be significantly affected by the orientation angle, i, between
the fibrous waste and the horizontal shear surface, as
shown in Fig. 4 for plastic waste constituents. The ratio
of fibrous reinforcement was found to be affected by
the orientation of the fibrous constituents, and the type
of waste constituents. For example, when paper, plastic and wood waste constituents were tested separately,
wood fibrous constituents were found to contribute the
most on the shear resistance of the specimen, at all
orientation angles, as shown in Fig. 5.
These observations are consistent with observations
in reinforced soils where research studies have found
that the improvement of shear resistance of a fiber
reinforced soil was a function of the type of fibers
and their orientation compared to the shearing plane
(Shewbridge and Sitar 1989, Gray and Ohashi 1983).
For example, Fig. 6 shows direct shear tests of sand
reinforced with reed fibers. The increase in shear resistance in the reinforced sand as a function of fiber
orientation is very analogous, at least qualitatively, to
the shear resistance of the waste-soil mixtures shown
in Fig. 5.
The shear response of MSW has also many analogies to the shear response of fibrous peats. Mesri and
Ajlouni (2007) report that the undrained strength in
triaxial compression of fibrous peats is significantly
Figure 6. Direct shear test results on sand reinforced with

reed fibers or stiff, rough reinforcement (from Gray and
Ohashi, 1983).
higher when shearing engages the fibers. They reference Yamaguchi et al. (1985a, 1985b) who observed
that the friction angle in triaxial compression increased
from 35 when the fibers are oriented approximately
in the vertical direction (using horizontally oriented
recovered specimens), to 5155 when the fibers are
approximately oriented in the horizontal direction
(using vertically oriented recovered specimens).
3.2
Compressibility
The compressibility of MSW has also been the focus

of recent investigations (McDougall 2007, Chen et al.
72
Figure 7. Effect of fibrous waste orientation on the compressibility of MSW.
2010, Gourc et al. 2010, Babu et al. 2011, Bareither

et al. 2012a, Bareither et al. 2012b) and an extensive
and highly informative review of current understanding of waste compressibility has been made recently
by McDougall (2011). The topic of anisotropy in the
compression behavior of MSW has drawn less attention most likely because typically the compressibility
of the MSW in the vertical direction is of interest.
Ongoing research at the University of Michigan and
University of Patras indicates that waste anisotropy
isbserved in the response of MSW in one-dimensional
compression. An example of currently unpublished
data is shown in Fig. 7. Fig. 7 illustrates the compression response of specimens UP-18 and UP-19, i.e.,
the specimens also shown in Fig. 3. Both specimens
have identical waste composition and are subjected
to a vertical stress of 50 kPa. UP-18 has horizontally
oriented fibrous constituents, and UP-19 has vertically
oriented fibrous constituents. In response to the 50 kPa
vertical stress application, the vertical displacement of
UP-18 and UP-19 was 29 mm and 22.5 mm respectively at around 1000 min, indicating that the stiffness
of the waste perpendicular to the fibrous constituents is
higher than the stiffness along the plane of the fibrous
waste constituents. The ratio of waste stiffness perpendicular to the fibrous waste orientation compared
to parallel to the fibrous waste orientation is equal to
0.78. Additional data is presently being evaluated in
an effort to assess the influence of waste composition
on stiffness anisotropy.
Figure 8. Schematic of the testing setup at location #1 at the

Austin Community Landfill: (a) cross-section; (b) plan view.
An ongoing study (Zekkos et al. 2013) has generated field data on the shear and compression wave
propagation velocities of MSW using crosshole and
downhole methodologies. The experimental setup is
described in more detail in Zekkos et al. (2013) and is
also shown in Fig. 8.
The tests were performed at the Austin Community
Landfill, in Austin, Texas, U.S.A, following the basic
methodology proposed by Stokoe et al. (2006). Two
vertical arrays of three-component geophones were
embedded in the waste at four different depths up
to a maximum depth of about 1 m. A 0.91-m diameter, 0.3-m thick, unreinforced, prefabricated concrete
foundation was placed on top of the sensors. Source
rods for crosshole seismic testing were placed at a distance of 1.14 m from the first array as shown in Fig. 8.
Downhole seismic testing was also performed by striking the side (for shear, S) and top (for compression, P)
of the footing and recording arrivals of S and P waves,
respectiverly, at the geophone arrays. Mobile vibroseis
shakers owned and operated by the George E. Brown,
Jr. Network for Earthquake Engineering Simulation at
University of Texas (NEES@UT), were used to excite
the footing.
The mobile shakers were used to apply a static
vertical load increment on the foundation. The foundation settlement during static load application was
measured from spanning beams. The vertical load was
varied allowing for an in situ assessment of the effect
of vertical stress on propagation velocities of MSW.
For each vertical load increment, small-strain crosshole and downhole seismic testing was performed as
shown in Fig. 9. Dynamic testing was also performed
3.3 Wave propagation velocity

A significant amount of data has also been generated
in terms of the shear wave velocity of Municipal Solid
Waste (Kavazanjian et al. 1996, Sahadewa et al. 2011).
The majority of the data in the literature has been
generated using surface wave-based methodologies,
because these methodologies are non-invasive (i.e.,
they do not require boreholes) and efficient. However,
data on the anisotropy of MSW in terms of propagation
velocities are essentially non-existent.
73
Figure 9. Performance of crosshole testing at Lamb Canyon,

California. Vibroseis used to apply vertical load.
by applying a sinusoidal horizontal load and these

results are described in Zekkos et al. (2013).
For each vertical load increment, vertical and horizontal stress distributions were calculated using the
Foster and Ahvlin (1954) method. Geostatic vertical
stress was calculated using the in situ unit weight for
each location. Geostatic horizontal stress was calculated using a lateral earth pressure at-rest coefficient
K0 estimated from the measured Poissons ratio values
from crosshole and downhole seismic testing.
Crosshole seismic testing was performed to assess
the horizontally propagating P-wave velocity Vp-h
and the horizontally-propagating, vertically-polarized
shear wave velocity Vs-hZ . Downhole seismic testing
was performed to evaluate the vertically propagating
P-wave velocity Vp-v , the vertically propagating horizontally polarized in the x-axis S-wave Vs-vX , and the
vertically propagating, horizontally-polarized in the
y-axis S-wave Vs-vY .
Anisotropy in MSW can be evaluated by comparing wave propagation in horizontal and vertical
direction. Figure 10a shows a comparison between
compression waves propagating in horizontal (Vp-h )
and vertical (Vp-v ) direction. Data indicates that Vp-v
ranges from 0.71 to 0.94 of Vp-h and a mean value of
0.82. Figure 10b illustrates the relationship between
shear waves propagating in horizontal (Vs-hZ ) and vertical (Vs-vX and Vs-vY ) direction. Vs-vX and Vs-vX is
found to be between 0.74 and 1.29 of Vs-hZ with
an average value of 0.91. In general, the data suggests that wave propagation in vertical direction is
slower than that in the horizontal direction. The shear
wave velocity and the compression wave velocity are
associated to the small strain shear modulus and the
small strain bulk modulus, respectively, through the
following equations:
Figure 10. (a) Comparison of Vp propagating in vertical

and horizontal direction; and (b) Vs propagating in vertical
and horizontal direction.
to 0.94 with a mean value of 0.82 is equivalent to bulk

modulus differences of 0.5 to 0.9 with a mean value of
0.67. This ratio compares relatively well to the observations regarding anisotropy in the compressibility of
MSW presented earlier.
3.4 Hydraulic conductivity
The anisotropy in hydraulic conductivity is more commonly discussed in the literature, primarily as part of
laboratory testing, field testing and numerical analyses
of leachate flow in bioreactor or recirculation landfills. For example, Xu et al. (2013) discussed seepage
control strategies at bioreactor landfills to minimize
leachate seeps that are found to occur laterally from
the leachate injection locations and are manifested
on the landfill slopes. Numerous numerical modeling
efforts have been performed to assess the appropriate
setback distances as part of a leachate injection scheme
(Khire and Mukherjee 2007; Haydar and Khire 2007,
Jain et al. 2010) and these distances are generally a
function of the ratio of horizontal to vertical hydraulic
conductivity.
Tchobanoglous et al. (1993) recommended that the
hydraulic conductivity ratio in landfills ranges from
where is the density of the material. Eq. 2a and 2b

indicate that a ratio of Vp-v / Vp-h that ranges from 0.71
74
10 to 100. Beaven et al. (2011) recently summarized

the available experimental studies on anisotropy of
hydraulic conductivity. For example, Landva et al.
(1998) performed laboratory testing on waste samples from landfills in Canada to assess the hydraulic
conductivity in the horizontal and vertical direction
and concluded that the anisotropy ratio is around 8.
This value is likely a result of fabric anisotropy due
to the preferred orientation of the fibrous waste constituents. A very interesting study on the anisotropy of
hydraulic conductivity has been performed by Hudson
et al. (2009) on recovered and processed waste. The
assessment of the hydraulic conductivity involved
numerical modeling of the flow in a large cell. The
authors concluded that all tests exhibited higher horizontal hydraulic conductivity than vertical hydraulic
conductivity. The authors report that the hydraulic conductivity ratio increased with confining stress, from
57 at stress of 40 kPa to 10 at 300400 kPa stress.
This finding would be consistent with the laboratory and field observations of increased preferential
orientation of the fibrous constituents during the application of increasing stress. As discussed by Beaven
et al. (2011), this preferred orientation is affected by
waste composition and may also be affected by waste
degradation.
The high anisotropy in hydraulic conductivity measured in MSW is consistent with the high anisotropy
in hydraulic conductivity measured for peat deposits.
Specifically, Mesri and Ajlouni (2007) report ratios of
horizontal to vertical hydraulic conductivities of 3 to
5 for surficial peat deposits. This ratio was found to
increase to 10 for the Middleton peat deposit when
confined.
4
CONCLUSIONS
Recent experimental data indicates that Municipal

Solid Waste (MSW) is a highly anisotropic material. Both fabric (or structural) and stress-induced
anisotropy are identified. Observations and recent
experimental evidence of the anisotropy of MSW in
shear resistance, compressibility, wave propagation
and hydraulic conductivity are presented. The shear
resistance of MSW is found to be significantly affected
by the orientation and type of fibrous waste constituents. Compressibility as well as wave propagation
are also affected by waste anisotropy. MSW is found
to be stiffer in the plane parallel to the fibrous waste
constituents (which tends to be the horizontal plane)
than perpendicular to the fibrous waste constituents.
In landfills, this observation essentially indicates that
the vertical compressibility of MSW is higher than
the horizontal compressibility of MSW. Hydraulic
conductivity of MSW in the horizontal direction is
also significantly higher than the hydraulic conductivity in the vertical direction. The influence of waste
anisotropy on engineering practice remains unknown,
but anisotropy is expected to play a role in many
aspects of landfill engineering practice, including flow
of liquids and gases, site characterization, foundation
design, landfill stability, as well as, seismic response.
Currently, engineering practices do not consider the
CONSIDERATIONS OF ANISOTROPY IN
ENGINEERING PRACTICE
The anisotropy of MSW is not yet widely recognized

in engineering practice. This may be attributed to a
number of factors including: (a) the lack of available
data highlighting the anisotropy of MSW; (b) the fact
that anisotropy is not also commonly considered in
geotechnical practice for the majority of the soils; and
(c) the influence of waste anisotropy on engineering
practice is not yet well understood.
The data presented in this paper illustrates that
MSW is highly anisotropic, although more data will
be generated in the future and will allow an improved
assessment of the degree of anisotropy. The anisotropy
of MSW is very comparable to the anisotropy observed
for peats as indicated earlier.
Consideration of anisotropy is likely to play a role in
at least the following aspects of engineering practice:
understranding of anisotropy in hydraulic conductivity at different scales.

Site Characterization: The anisotropic nature of
MSW may affect interpretations of site characterization of MSW. For example, interpretations
of surface wave data do not typically account for
anisotropic materials. Also, anisotropy may bias
results from other seismic geophysics techniques
such as crosshole and downhole, as demonstrated
by the data presented earlier.
Foundation Design: In the design and performance
assessment of deep foundations that are going
through MSW material, the stiffness and shear resistance of MSW along the pile shaft may be different
than the commonly considered shear strength and
stiffness of the waste. This may affect the pile design
as well as estimates of MSW downdrag forces.
Landfill Stability: The significant difference in
shear resistance of MSW as a function of the relative
orientation of the fibrous waste constituents with the
shear plane may influence the stability assessment
of landfills. Presently, a uniform shear strength is
considered in analyses irrespective of the orientation of the failure plane compared to the horizontally
oriented fibrous waste constituents.
Seismic Response: Seismic wave propagation
through anisotropic materials is not commonly considered in practice. Given the high anisotropy of
MSW, consideration may be necessary to the influence of the observed anisotropy on site response and
seismic slope displacements.
Flow of liquid and gas: As energy recovery

from waste in landfills due to leachate recirculation becomes more common, an improved understanding of liquid and gas flow in landfills will
be key for optimizing the energy recovery process. Such understanding will require an improved
75
Chen, Y., Fredlund, D. G., Zhan, L., Xie, Y. 2010. Secondary compression of Municipal Solid Wastes and a
compression model for predicting settlement of Municipal Solid Waste landfills. ASCE Journal of Geotechnical
and Geoenvironmental Engineering, May 2010, 136(5).
Dixon, N., Jones, D. R. V., Whittle, R. W. 1999. Mechanical properties of household waste: In situ assessment
using pressuremeters, Proceedings Sardinia 99, Seventh
InternationalWaste Management and Landfill Symposium
(in cd-rom).
Edil, T. B., and Dhowian, A. W. 1981. At-rest lateral pressure
of peat soils. J. Geotech. Engrg. Div., 107(2): 201217.
Foster, C. R. and Ahlvin, R. G. 1954. Stresses and deflections induced by a uniform circular load. Proceedings the
Highway Research Board 1954; 33: 467470.
Gourc, J.P., Staub, M. J., Conte, M. 2010. Decoupling MSW
settlement into mechanical and biochemical processes
Modeling and validation on large-scale setups. Waste
Management, 30, pp. 15561568.
Gray, D. H., and Ohashi, H. 1983. Mechanics of fiber reinforcement in sand. J. Geotech. Engrg., ASCE, 109(3):
335353.
Haydar, M. and Khire, M. 2007. Leachate recirculation using
permeable blankets in engineered landfills. J. Geotech.
Geoenviron. Eng., ASCE, 133(4): 360371.
Holtz, R. D., Kovacs, W. D., Sheahan, T. C. 2011. An introduction to geotechnical engineering. Prentice Hall, 2nd
edition, pp. 853.
Hudson, A.P., Beaven, R.P. and Powrie, W. 2009. Assessment of vertical and horizontal hydraulic conductivities of
household waste in a large scale compression cell. In Ed by
Cossu R, Diaz LF and Stegman R (eds), Proceedings of the
12thWaste Management and Landfill Symposium Sardinia
2009, CISA, S. Margherita di Pula, Italy (in cd-rom).
Jain, P., J. Powell, et al. 2005. Air Permeability of Waste in a
Municipal Solid Waste Landfill. Journal of Environmental
Engineering, ASCE 131(11): 15651573.
Jain, P., Townsend, T.G., Tolaymat, T.M. 2010. Steady-state
design of vertical wells for liquids addition at bioreactor
landfills. Waste Management Journal, 30, 20222029.
Jewell, R. A., and Wroth, C. P. 1987. Direct shear tests on
reinforced sand. Geotechnique, 37 (1): 5368.
Kavazanjian, E., Jr., Matasovic, N., Stokoe, K.H.II, Bray,
J.D. 1996. In situ shear wave velocity of solid waste
from surface wave measurements M. Kamon (ed.),
Environmental Geotechnics, 1996 Balkema, 1, 97102.
Khire, M. V. and Mukherjee, M., 2007. Leachate injection using vertical wells in bioreactor landfills. Waste
Management Journal, 27: 12331247.
Landva, A. O., and La Rochelle, P. 1983. Compressibility and
shear characteristics of Radforth peats. Testing of peat and
organic soils, STP 820, ASTM, West Conshohocken, Pa.,
157191.
Landva, A. O., Valsangkar, A. J., Pelkey, S. G. 2000. Lateral
earth pressure at rest and compressibility of municipal
solid waste, Canadian Geotechnical Journal, 37, 1157
1165.
Landva, A. O., Pelkey, S. G., Valsangkar, A. J. 1998. Coefficient of permeability of municipal refuse. Proc., Third
Int. Congress on Environmental Geotechnics, Balkema,
Rotterdam, 163167.
Landva, A.O., Clark, J.I. 1986. Geotechnical testing of
wastefill. In: Proceedings, 39th Canadian Geotechnical
Conference, Ottawa, Ontario, 371385.
Matasovic, N., Kavazanjian, E. Jr. 1998. Cyclic characterization of OII landfill solid waste. Journal of Geotechnical
and Geoenvironmental Engineering, March 1998, 124(3):
197210.
influence of anisotropy on the design and performance

of MSW landfills with the exception, in some cases,
of the anisotropy in hydraulic conductivity.
ACKNOWLEDGMENTS
The author would like to acknowledge Athena Grizi
and Andhika Sahadewa, both graduate students at
the University of Michigan for being instrumental in
generating some of the data presented in this paper.
Specifically, Athena on the work related to compressibility of MSW and the effect of fibrous waste
orientation and Andhika on the work related to the
dynamic properties of MSW.
The data related to wave propogation velocity is
based upon field work supported by the National
Science Foundation Division of Civil and Mechanical Systems under Grant No. CMMI-1041566. Any
opinions, findings, conclusions and recommendations
expressed in this paper are those of the author and
do not necessarily reflect the views of the National
Science Foundation. Interactions of the author with
the research collaborators of this project, Drs. R. D.
Woods, N. Matasovic, and K. Stokoe, were invaluable.
REFERENCES
Athanasopoulos, G., Grizi, A., Zekkos, D., Founta, P.,
Zisimatou, E. 2008. Municipal Solid Waste as a Reinforced Soil: Investigation Using Synthetic Waste. Geocongress 2008, ASCE, GSP No. 177, 168175.
Babu, S. G.L., Reddy, K.R., and Chouskey, S.K. 2011. Parametric Study of MSW Landfill Settlement Model. Waste
Management Journal, 31 (6): 12221231.
Bareither, C., Benson, C., and Edil, T. 2012a. Compression
Behavior of Municipal Solid Waste: Immediate Compression. J. Geotech. Geoenviron. Eng., 138(9): 10471062.
doi: 10.1061/(ASCE)GT.1943-5606.0000672
Bareither, C., Benson, C., and Edil, T. 2012b. Compression of
Municipal Solid Waste in Bioreactor Landfills: Mechanical Creep and Biocompression . J. Geotech. Geoenviron.
Eng.. doi: 10.1061/(ASCE)GT.1943-5606.0000835
Bareither, C., Benson, C., and Edil, T. 2012c. Effects of Waste
Composition and Decomposition on the Shear Strength
of Municipal Solid Waste. J. Geotech. Geoenviron. Eng.,
138(10): 11611174.
Beaven, R. P., Powrie, W., Zardava, K. 2011. Chapter
1: Hydraulic Properties of Municipal Solid Waste. In
Dimitrios Zekkos (ed.), Geotechnical characterization,
Field Measurements, and Laboratory Testing of Municipal Solid Waste, ASCE Geotechnical Special Publication
No. 209, 143.
Boulanger, R. W., Arulnathan, R., Harder, L. F., Jr.,
Torres, R. A., and Driller, M. W. 1998. Dynamic properties of Sherman Island peat. J. Geotech. Geoenviron.
Eng., 124(1): 1220.
Bray, J.D., Zekkos, D., Kavazanjian Jr., E., Athanasopoulos,
G.A., Riemer, M.F. 2009. Shear strength of municipal
solid waste. Journal of Geotechnical and Geoenvironmental Engineering, ASCE 135(6): 709722.
Carvalho, M.de F., Vilar, O.M. 1998. In situ tests in urban
waste sanitary landfill, editor Seco e Pinto, Environmental
Geotechnics, Balkema, Rotterdam, 121126.
76
McDougall, J.R. 2011. Settlement The short and long of it.

In Dimitrios Zekkos (ed.), Geotechnical characterization,
field measurement, and laboratory testing of Municipal
Solid Waste, ASCE Geotechnical Special Publication No.
209, 76111.
McDougall, J. R. 2007. A hydro-bio-mechanical model
for settlement and other behaviour in landfilled waste.
Computers & Geotechnics, 34: 229246.
Mesri, G. and Ajlouni, M. 2007. Engineering Properties
of Fibrous Peats. J. Geotech. Geoenviron. Eng., 133(7):
850866.
Michalowski, R. L. Cross-Anisotropy in Fiber-Reinforced
Sand. GeoCongress 2012: 10861095.
Michalowski, R. L., and Cermak, J. 2002. Strength anisotropy
of fiber reinforced sand. Comp. Geotechn., 29: 279299.
Mitchell, J. K., and Soga, K. 2005. Fundamentals of soil
behavior, Wiley, pp. 577.
Sahadewa, A., Zekkos, D., Lobbestael, A., and Woods,
R. D. 2011. Shear wave velocity of Municipal Solid
Waste in Michigan Landfills. In 14th Pan-American
Conference on Soil Mechanics and Geotechnical Engineering and 64th Canadian Geotechnical Conference,
Geo-Innovation Addressing Global Challenges, October
26, 2011, Toronto, Ontario, Canada (in cd-rom).
Stokoe, K. H., II, Kurtulus, A., and Park, K. 2006. Development of Field Methods to Evaluate Nonlinear Shear
and Compression Moduli of Soil. In Proceedings of New
Zealand Earthquake Geotechnical Engineering Workshop, Canterbury 2006. Christchurch, New Zealand,
November, 5670.
Tchobanoglous, G., Theisen, H., Vigil, S. 1993. Integrated
Solid Waste Management. McGraw-Hill, New York, USA.
Xu, Q., Powell, J., Tolaymat, T., and Townsend, T. Seepage Control Strategies at Bioreactor Landfills. J. Hazard.
Toxic Radioact. Waste. doi: 10.1061/(ASCE)HZ.21535515.0000185.
Yamaguchi, H., Ohira, Y., Kogure, K., and Mori, S. 1985a.
Undrained shear characteristics of normally consolidated
peat under triaxial compression and extension conditions.
Japanese Society of Soil Mech., and Found. Eng., 25(3):
118.
Yamaguchi, H., Ohira, Y., Kogure, K., and Mori, S. 1985b.
Deformation and strength properties of peat. Proc., 11th
Int. Conf. on Soil Mechanics and Foundation Engineering,
4: 24612464.
Zekkos, D., Sahadewa A., Woods, R. D., Stokoe, K., II.
Matasovic, N. 2013. In Situ Assessment of the Nonlinear
Shear Modulus of Municipal Solid Waste. Inn proceedings
of the 18th International Conference on Soil Mechanics
and Geotechnical Engineering, Paris 2013 (accepted).
Zekkos, D. (ed.) 2011. Geotechnical characterization, Field
Measurements, and Laboratory Testing of Municipal Solid
Waste. ASCE Geotechnical Special Publication No. 209,
pp. 249, ISBN 978-0-7844-1146-9.
Zekkos, D., Athanasopoulos, G. A., Bray, J .D., Theodoratos,
A., and Grizi, A. 2010. Large-scale Direct Shear Testing of
Municipal Solid Waste. Waste Management Journal, 30:
15441555.
Zekkos, D. 2005. Evaluation of static and dynamic properties of municipal solid waste. Ph.D. Thesis, Department
of Civil and Environmental Engineering, University of
California at Berkeley.
77

Impact of subgrade water content on cation exchange and hydraulic

conductivity of geosynthetic clay liners in composite barriers
Craig H. Benson
Geological Engineering, University of Wisconsin-Madison, Madison, WI, USA
ABSTRACT: Laboratory hydration experiments are shown to simulate realistically the hydration and cation
exchange process occurring in GCLs used in composite barriers where the GCL is deployed on a subgrade and
covered with a geomembrane and cover soil. GCLs hydrated on subgrade soil compacted wet of optimum water
content underwent hydration and Ca-for-Na cation exchange, reached water contents near 70%, and retained
very low hydraulic conductivity. In contrast, GCLs hydrated on much drier subgrades hydrated to much lower
water content, underwent cation exchange, and had much higher hydraulic conductivity. Water content, cation
exchange, and hydraulic conductivity observed in the laboratory experiments are in remarkably good agreement
with properties of GCLs exhumed from the field. The findings confirm previous recommendations that GCLs
be placed on subgrades compacted at optimum water content or higher to ensure satisfactory hydration and low
hydraulic conductivity even if Na is replaced by divalent cations (e.g., Ca).
INTRODUCTION
hydraulic conductivity (Mesri and Olson 1971, Jo et al.

2001, 2004, Kolstad et al. 2004).
In a new GCL, Na generally is the primary interlayer cation (Jo et al. 2001, Kolstad et al. 2004). After
installation, the Na cations are susceptible to replacement by other cations present in the surrounding pore
water, and from dissolution of calcite within the GCL
itself (James et al. 1997, Guyonnet et al. 2005, Rauen
2007, Bradshaw et al. 2013). Replacement of Na by
divalent cations such as Mg and Ca is common, as the
exchange reactions are thermodynamically favorable
(Sposito 1981) and Ca and Mg are ubiquitous in the
geoenvironment. The replacement of Na by Ca and
Mg can affect swelling of the bentonite and alter the
hydraulic conductivity.
When GCLs are installed in a composite barrier system in a final cover, the GCL is deployed from a roll
onto a moist compacted subgrade followed by covering
with a geomembrane and cover soil.The geomembrane
practically ensures that the GCL will prehydrate by
moisture migrating upward from the subgrade rather
than by liquid migrating downward from the overlying
cover soils (Scalia and Benson 2011, Bradshaw et al.
2013). The prehydration water content achieved by
placing a GCL on a moist subgrade is related directly
to the water content of the subgrade, the contact time,
and the overburden stress (Daniel et al. 1993, Scalia
and Benson 2011, Bradshaw et al. 2013).
Scalia and Benson (2011) describe five case histories where GCLs were exhumed from final covers with
composite barriers. They found that Ca and Mg had
replaced some or all of the Na originally on the bentonite at each site, with nearly complete replacement of
Na at three sites (Fig. 1a). The hydraulic conductivity
Geosynthetic clay liners (GCLs) are factorymanufactured clay liners that are used widely in waste
containment facilities as hydraulic barriers. GCLs consist of a thin layer of granular or powdered bentonite
clay encased between two geotextiles or glued to a
geomembrane. Montmorillonite, the predominant clay
mineral in bentonite, is largely responsible for the low
hydraulic conductivity of GCLs. Montmorillonite has
high cation exchange capacity (CEC), large specific
surface area, and can be associated with a thick layer
of bound water that is essentially immobile (Mitchell
and Soga 2005). The hydraulic conductivity of bentonite is controlled by the relative amount of pore
water that is bound in the montmorillonite fraction;
when a larger fraction of the pore water is bound, the
flow paths are narrower and more tortuous, resulting
in lower hydraulic conductivity (Chapuis 1990).
Montmorillonites hydrate in two stages, known
as crystalline and osmotic swelling. When monovalent cations (e.g. sodium, Na) are predominant in
the exchange complex, bentonites undergo both crystalline and osmotic swelling during hydration. In
contrast, when polyvalent cations (e.g. calcium, Ca;
magnesium, Mg) are predominant in the exchange
complex, bentonites only undergo crystalline swelling
during hydration (Norrish 1954, Norrish and Quirk
1954). Crystalline and osmotic swelling both contribute to bound water. However, the volume of pore
water involved in osmotic swelling typically is much
larger than in crystalline swelling. Consequently, conditions that promote osmotic swelling result in a larger
fraction and greater amount of bound water, and lower
79
Figure 2. Hydraulic conductivity vs. exhumed water content of GCL (adapted from Scalia and Benson 2011).
Scalia and Benson (2011) also suggested that water

molecules imbibed in the interlayer of the montmorillonite during osmotic hydration are strongly bound by
electrical forces and, in the absence of a large energy
source, would remain in place even if Ca replaced
Na in the exchange complex after the GCL hydrated.
Consequently, the pores would remain very small
and the pathways tortuous, resulting in low hydraulic
conductivity even if Ca completely replaced Na in
the exchange complex provided the bentonite underwent osmotic swelling during hydration. Moreover,
because the GCL in a composite barrier is overlain
by a geomembrane, the GCL would not be exposed to
the energy associated with evapotranspiration, which
has been show to cause extensive drying of GCLs in
covers that are not covered by geomembranes (including removal of water from the interlayer) (Meer and
Benson 2007).
Photographs compiled by Scalia and Benson (2011)
supported their hypothesis. For example, the GCL
from Site A exhibited a swollen gel structure characteristic of bentonites that have undergone osmotic
hydration and swelling (Fig. 3a). In contrast, the GCL
from Site B appeared dry, and remnant granules were
present that appeared to have hydrated and swelled
incompletely (Fig. 3b).
Bradshaw et al. (2013) hydrated GCLs on four different subgrade soils varying in texture and chemical
composition of the pore water for 30 or 90 d. All
of the subgrades were prepared at 1% wet of optimum water content. Nearly all of the hydration of
the GCLs occurred within 30 d. However, during this
same period, virtually no cation exchange occurred.
A modest (<10%) increase in water content occurred
between 30 and 90 d. Cation exchange also occurred
during this period, with modest to significant replacement (5 to 50%) of Na by Ca and Mg. All of the GCLs
hydrated on subgrades had hydraulic conductivity
Figure 1. Mole fraction Na in exchange complex of GCLs

exhumed by Scalia and Benson (2011) (a) (shaded zone represents condition of new GCL) and water content of exhumed
GCLs. The GCLs were exhumed from covers with and without an overlying geomembrane (composite vs. GCL only).
Adapted from Scalia and Benson (2011).
of the GCLs varied widely and was not directly related

to the degree of replacement of Na by Ca and Mg.
For example, at Site A, the hydraulic conductivity of
the GCL ranged from 9.3 1010 to 1.6 1011 cm/s
(approximately the same as a new GCL), even though
approximately half of the Na had been replaced by
Ca. Similar conditions were observed at Site E, except
essentially all of the Na was replaced by Ca. In contrast,
at Site B, the hydraulic conductivity ranged between
1.5 107 to 4.5 106 cm/s, even though this site
had the least replacement of Na by Ca (Fig. 1a). The
GCL at Site B also had much lower water content than
the GCLs at Sites A and E (Fig. 1b).
Scalia and Benson (2011) found that the hydraulic
conductivity of the GCLs was strongly correlated
to their water content (Fig. 2), and that the GCL
water content increased as the subgrade water content
increased. They proffered that GCLs placed on wetter subgrades were more likely to undergo complete
osmotic swelling and that the hydration process would
occur more quickly than the cation exchange process
if the subgrade water content was sufficient.
80
MATERIALS AND METHODS
2.1 Subgrades
Subgrade soils from Sites B and E in Scalia and Benson
(2011) were used in the experiments. The soil at Site
B is nearly non-plastic silt with >90% fines. The soil
from Site E is silty clay with >76% fines. Subgrade
soils were air dried and then hydrated with deionized
(DI) water to the target water contents (3% for Site B
and 15% for Site E). The soil was allowed to hydrate
for at least 24 h to ensure uniformity, and then was
compacted following the procedure in ASTM D 698
(Method A).
2.2 GCL
The GCL contained granular Na-bentonite encased
by two geotextiles (slit-film woven geotextile and
non-woven staple fiber geotextile) bonded by needle
punching. The mass per unit area of the bentonite was
3.66 kg/m2 , the initial thickness of the GCL ranged
from 4.3 to 6.2 mm, and the average initial water content of the bentonite was 4.5%. The bentonite granules
were predominantly sand-sized with 50% of the granules larger than 0.2 mm. X-ray diffraction showed that
the bentonite contained 51% montmorillonite and 4%
calcite. The average CEC was 75.8 4.0 cmol+ /kg
with 55% of the CEC being satisfied by Na, 34% by
Ca, 10% by Mg, and 1% by K.
2.3 Hydration tests
The GCLs were hydrated on the subgrade soils in
flexible-wall permeameters with the effluent and influent lines dry and closed. A non-woven geotextile
(240 g/m2 ) was placed on an acrylic base plate followed by the compacted subgrade specimen, a GCL
specimen, 1.5-mm geomembrane disk, a geotextile
disk, and an acrylic top plate. The nonwoven face of
the GCL was in contact with the subgrade. A latex
membrane was placed around the entire assemblage
and sealed to the top and bottom plates with three
O-rings on each plate. The cell was filled with water
and a 10 kPa confining stress was applied to simulate
the surface layer in a final cover.
Hydration continued for up to 365 d, with duplicate
hydration experiments decommissioned periodically
to determine the water content, cation exchange complex, and hydraulic conductivity as a function of time.
One of the GCLs from the duplicate experiments was
analyzed immediately after completion of the hydration phase to determine the bound cation composition.
The other was permeated to determine the hydraulic
conductivity.
Figure 3. Photographs of cross-section of GCLs immediately after exhumation from Site A (a) and Site B (b) as
described in Scalia and Benson (2011).
indistinguishable from a new GCL, even though cation

exchange had occurred. These findings support the
supposition in Scalia and Benson (2011) that GCLs
hydrated on subgrades placed at optimum water content or higher will maintain low hydraulic conductivity
even if Ca-for-Na cation exchange occurs.
This paper describes laboratory experiments conducted to evaluate the conditions described by Scalia
and Benson (2011), and to determine if the behavior observed in the field by Scalia and Benson (2011)
could be replicated in the laboratory. Two sets of GCL
specimens were hydrated on subgrade soil from Site B
and Site E following the procedure in Bradshaw et al.
(2013). The subgrade soil from Site B was prepared
at 3% water content (15% dry of optimum water
content) and for Site E at 15% water content (2% wet
of optimum water content), as observed in the field
at both sites. The specimens were hydrated for up to
365 d. Water content, cation exchange complex, and
hydraulic conductivity of the GCLs were determined
at various hydration times, and compared to conditions
observed in the field.
2.4
Hydraulic conductivity and

exchange complex testing
Hydraulic conductivity tests were conducted in

flexible-wall permeameters using the falling-headwater constant-tail-water method described in ASTM
D 6766 at an effective stress of 10 kPa and with a
81
Figure 5. Mole fraction in exchange complex in GCL specimens in subgrade hydration experiments using subgrade soil
from Site B wet of optimum water content along with data
from steady permeation tests using 20 mM CaCl2 reported
by Jo et al. (2004).
Figure 4. Water content of GCLs in laboratory hydration

experiments and from field at Sites B and E as reported in
Scalia and Benson (2011).
hydraulic gradient of 120. A dilute calcium solution

(10 mM) similar to pore waters observed in the field
(Lin and Benson 2000, Egloffstein 2002) was used as
the permeant liquid.
Bound cations, soluble cations, and CEC were
determined according to ASTM D 7503. Soluble
and bound cation concentrations were determined for
major cations (Na, Ca, Mg, and K) by ICP-OES
following USEPA Method 6010B. CEC was determined using the procedure in ASTM D 7503 with
the extracted ammonium measured using a Spectronic
20 Genysys spectrophotometer (ThermoFisher Scientific, Waltham, Massachusetts) with the salicylate
method (Hach Method 10031).
3
3.1
3.2 Na replacement in exchange complex

Mole fraction of Na in the exchange complex of the
GCL is shown in Fig. 5 as a function of time for the
experiments conducted with subgrade soil from Site E.
Data from long-term tests conducted by Jo et al. (2004)
are also shown in Fig. 5. In these tests, GCLs were
permeated directly (no prehydration) with a dilute
(20 mM) CaCl2 solution. Specimens were periodically
decommissioned to determine how the composition
of the exchange complex was changing as the calcium solution migrated through the GCL. The tests
conducted by Jo et al. (2004) were conducted under
steady flow conditions and therefore represent conditions of full hydration through nearly the entire test
(initial hydration period at the start of permeation is
an exception). The tests conducted by Jo et al. (2004)
used bentonite from the same GCL used in this study.
The mole fraction of Na gradually diminishes with
hydration time in the subgrade hydration experiments.
By 365 d, less than 30% of the exchange complex
was comprised of Na, whereas Na comprised 55% of
the exchange complex in the new GCL. The temporal trend in the mole fraction of Na in the exchange
complex measured in the laboratory is also consistent
with the mole fraction of Na observed in the specimens
exhumed from the field when the trend is extrapolated.
Comparison with the data reported by Jo et al.
(2004) suggests that cation exchange in GCLs due
to contact with a subgrade occurs more slowly than
during direct permeation. The faster rate of exchange
during the tests by Jo et al. (2004) may be due to higher
concentrations in the permeant water used by Jo et al.
(2004) compared to the pore water in the subgrade
hydration experiments. Other factors, such as the rate
of upward diffusion of Ca from the subgrade or the
rate of dissolution of calcite within the GCL, may also
have affected the rate of exchange in the laboratory
experiments.
RESULTS
GCL water content
Water content of the GCLs is shown in Fig. 4 as function of hydration time. Water contents of the GCLs
exhumed from Sites B and E are also shown in Fig. 4.
Water content of the GCLs increases rapidly, as
previously reported by Daniel et al. (1993) and Bradshaw et al. (2013). Nearly all of the hydration occurred
within 65 d (first time when tests were decommissioned). A very small increase in water content (<5%)
occurred for both test conditions between 75 d and
the end of the laboratory hydration experiments. The
GCLs exhumed from the field at Sites B and E also
had nearly identical water contents as measured at the
end of the laboratory hydration experiments, suggesting that the laboratory procedure replicated the field
condition realistically. The similarity in the laboratory and field conditions also suggests that most GCL
hydration occurs in the field within 75 d, which is consistent with the findings reported in Bradshaw et al.
(2013).
82
higher hydraulic conductivity. Importantly, the water

content, cation exchange, and hydraulic conductivity
observed in the laboratory experiments were remarkably similar to those in the GCLs exhumed from the
field.
The outcome of these experiments reaffirms the
recommendation in Scalia and Benson (2011), that
GCLs should be placed on subgrades compacted at
optimum water content or higher to ensure satisfactory
hydration and low hydraulic conductivity of the GCL,
even if the Na is replaced by divalent cations such as
Ca. However, care should be used when selecting the
subgrade water content to prevent other unintended
consequences (e.g., inadequate shear strength of the
subgrade).
Additional testing is currently underway to understand the hydration mechanisms and cation exchange
occurring during hydration more thoroughly, and to
refine the recommendation for subgrade water content. Results from the ongoing experiments will be
reported in a subsequent publication.
Figure 6. Hydraulic conductivity of GCLs from laboratory

hydration experiments and exhumed from the field from
Sites B and E as a function of hydration time (laboratory)
or service time (field).
3.3 Hydraulic conductivity of GCLs

Hydraulic conductivity of the GCLs hydrated on subgrade soils from Sites B and E are shown as a function
of hydration time in Fig. 6. Hydraulic conductivity of
GCLs specimens exhumed from the field from Sites
B and E are also shown in Fig. 6.
For time periods in excess of 7590 d, hydraulic
conductivities of the GCLs hydrated in the laboratory
experiments are remarkably similar to the hydraulic
conductivities of the GCLs exhumed from the field,
suggesting that the laboratory experimental protocol provides a realistic simulation of the conditions
observed in the field. The slightly higher hydraulic
conductivities of the field specimens may reflect more
complete replacement of Na by Ca and Mg over the
longer time frames in the field.
4
ACKNOWLEDGEMENTS
Financial support for this study was provided by the
US Department of Energy (DOE) under cooperative
agreement DE-FC01-06EW07053 entitled Consortium for Risk Evaluation with Stakeholder Participation III. The opinions, findings, conclusions, and
recommendations expressed herein are those of the
author and do not necessarily represent the views
of or the US Department of Energy. Joseph Scalia
conducted many of the experiments described in this
paper. His contributions are gratefully acknowledged.
REFERENCES
Bradshaw, S., Benson, C., and Scalia, J. (2013). Hydration and
Cation Exchange During Subgrade Hydration and Effect
on Hydraulic Conductivity of Geosynthetic Clay Liners,
J. Geotech. and Geoenvironmental Eng., 139(4), 526538.
Chapuis, R. (1990). Sand-Bentonite Liners: Predicting Permeability from Laboratory Tests, Canadian Geotechnical
J., 27, 4757.
Daniel, D., Shan, H., and Anderson, J. (1993). Effects of
Partial Wetting on the Performance of the Bentonite Component of a Geosynthetic Clay Liner, Proc. Geosynthetics
93, Industrial Fabrics Association International, St. Paul,
MN, 14831496.
Egloffstein, T. (2002). Bentonite as sealing material in
geosynthetic clay liners influence of electrolytic concentration, the ion exchange and ion exchange with simultaneous partial desiccation on permeability, Clay Geosynthetic
Barriers, Swets & Zeitlinger, Lisse, 141153.
Guyonnet, D., Gaucher, E., Gaboriau, H., Pons, C.,
Clinard, C., Norotte, V., Didier, G. (2005). Geosynthetic Clay Liners Interaction with Leachate: Correlation
between Permeability, Microstructure, and Surface Chemistry, J. Geotech. and Geoenvironmental Eng., 131(6),
740749.
PRACTICAL IMPLICATIONS
The findings presented in this paper suggest that laboratory hydration experiments conducted using the
procedures described herein can be used to simulate
the hydration and cation exchange process occurring in
GCLs in the field that are placed on subgrades and covered with a geomembrane and cover soil. The impacts
of the hydration condition and cation exchange
on hydraulic conductivity are also represented
realistically.
The findings also suggest that the impacts of
cation exchange on hydraulic conductivity are strongly
influenced by the subgrade water content. The GCL
hydrated on subgrade soil compacted wet of optimum water content underwent hydration and cation
exchange, but had very low hydraulic conductivity.
In contrast, the GCL hydrated on a much drier subgrade underwent hydration and cation exchange, but
hydrated to much lower water content and had much
83
James, A., Fullerton, D., and Drake, R. (1997). Field

Performance of GCL under Ion Exchange Conditions,
J. Geotech. Eng., 123(10), 897901.
Jo, H., Benson, C., and Edil, T. (2004). Hydraulic Conductivity and Cation Exchange in Non-Prehydrated
and Prehydrated Bentonite Permeated with Weak Inorganic Salt Solutions, Clays and Clay Minerals, 52(6),
661679.
Jo, H., Katsumi, T., Benson, C., and Edil, T. (2001). Hydraulic
Conductivity and Swelling of Non-Prehydrated GCLs Permeated with Single Species Salt Solutions, J. of Geotech.
and Geoenvironmental Eng., 127(7), 557567.
Kolstad, D., Benson, C., and Edil, T. (2004). Hydraulic Conductivity and Swell of Nonprehydrated GCLs Permeated
with Multispecies Inorganic Solutions, J. Geotech. and
Geoenvironmental Eng., 130(12), 12361249.
Lin, L. and Benson, C. (2000). Effect of Wet-Dry Cycling
on Swelling and Hydraulic Conductivity of Geosynthetic
Clay Liners, J. Geotech. and Geoenvironmental Eng.,
126(1), 4049.
Meer, S. and Benson, C. (2007). Hydraulic Conductivity of
Geosynthetic Clay Liners Exhumed from Landfill Final
Covers, J. Geotech. and Geoenvironmental Eng., 133(5),
550563.
Mesri, G. and Olson, R. (1971). Mechanisms Controlling

the Permeability of Clays, Clays and Clay Minerals, 19,
151158.
Mitchell, J. and Soga, K. (2005). Fundamentals of Soil
Behavior, 3rd Ed., John Wiley and Sons, Inc., New York,
437 p.
Norrish, K. (1954). The Swelling of Montmorillonites,
Discussions of Faraday Society, 18, 120134.
Norrish, K. and Quirk, J. (1954). Crystalline Swelling of
Montmorillonite, Use of Electrolytes to Control Swelling.
Nature, 173, 255257.
Rauen, T. (2007). Effect of Leachate from Bioreactor and
Recirculation Landfills on the Hydraulic Conductivity
of Geosynthetic Clay Liners. MS Thesis, University of
Wisconsin, Madison, WI.
Scalia, J. and Benson, C. (2011), Hydraulic Conductivity
of Geosynthetic Clay Liners Exhumed from Landfill
Final Covers with Composite Barriers, J. Geotech. and
Geoenvironmental Eng., 137(1), 113.
Sposito, G. (1981). The Thermodynamics of Soil Solutions,
Oxford University Press, New York, New York, 223 p.
84

Advances in the analysis of thermo-active foundations

L. Laloui, T. Mimouni & F. Dupray
Laboratory of Soil Mechanics, Ecole Polytechnique Fdrale de Lausanne, Lausanne, Switzerland
ABSTRACT: Energy geostructures are foundations equipped with absorber pipes, exchanging heat with the
surrounding ground. They provide a good heat source for the heating and cooling of buildings but the temperature
variations they undergo bring new challenges. Thermal expansion or contraction of the foundations can lead
to building movements that must be kept within acceptable limits. Therefore, a thermo-mechanical design
should be adopted. This paper presents recent advances in the understanding of the behaviour of thermo-active
foundations. First, contributions of in situ experiments carried out on real-scale energy piles are reviewed.
The outcome of these tests is a unique numerical design tool, called Thermo-Pile, which is presented. Next,
an advanced constitutive model for environmental geomechanics accounting for non-isothermal conditions is
detailed as well as the experimental evidence it was based on. Examples of thermo-hydro-mechanical analyses
are presented to illustrate the challenges faced when using energy geostructures. Finally, the sustainability of
heat storage through particular thermo-active foundations and tunnel anchors, is assessed.
INTRODUCTION
energy piles. The numerical tool Thermo-Pile, which

was developed based on these in situ observations, is
then described.
The thermo-mechanical behaviour of soils is
detailed in Section 3. The heat propagation through
soils is assessed and thermal analytical models for
energy piles are presented. In situ and laboratory
methods to estimate the thermal properties of soils
are described. The thermo-mechanical couplings that
occur in soils are then assessed with a thermo-elastoplasticity model that was developed by our research
group (ACMEG-T model). Finally, a finite element
modelling of thermal piles is presented.
The last section proposes an optimization method of
heat production and injection through tunnel anchors
or nails. These optimisations are based on finite element analyses that assess the heat propagation through
soils as well as the mechanical implications of heat
storage. The method is first outlined and the results
for different configurations are discussed.
Ground Source Heat Pump systems (GSHPs) have

been developed in order to provide a reliable and highly
efficient heat source for the heating and cooling of
buildings, which represents an important part of the
annual global energy demand. GSHPs take advantage
of the relatively high and constant temperature levels found about 10 to 100 meters below the surface,
showing coefficients of performance (COP) up to 45
(Brandl 2006).
GSHPs are linked to the ground through Ground
Heat Exchangers (GHEs) and to the building through
the heating network, forming the so-called ground
source heat pump system, whose efficiency is directly
linked to the efficiency of these three components.
As suggested by Brandl (2006), improving heat
exchange with the ground can be achieved by embedding GHEs in foundation structures, enhancing the
GHE-earth contact with the good thermal properties of
concrete. Therefore, thermo-active foundations can be
created from slabs, walls or deep pile foundations by
inserting absorber pipes into conventional foundation
structures. However, thermally activated foundations
undergo thermal deformations that may compromise
their primary purpose of structural support.
This paper provides insight into the advances carried out to tackle the various challenges linked to
thermo-active foundations and their design.
Section 2 deals with the behaviour of energy piles.
Two major in situ experimental sites that were developed on the EPFL campus in the last 20 years are
presented along with their main contributions to the
understanding of the thermo-mechanical behaviour of
2 THERMO-MECHANICAL BEHAVIOUR OF
THERMAL PILES
2.1
In situ experiments
Two major real-scale experimental sites were built on

the EPFL campus in Lausanne (Switzerland) with the
goal of understanding the global thermo-mechanical
response of energy piles.
The first site was installed in 1998 below a 4-storey
building 100-m-long and 30-m-wide (Laloui et al.
2003). The energy pile is 25.8-m-long and equipped
85
Figure 2. Evolution of the degree of freedom of the pile

during the building construction (Laloui et al. 2003).
Figure 1. Stratigraphy and instrumentation of the first EPFL

test pile (Laloui et al. 2003).
with 4 U-loops connected in parallel to head collectors. Axial deformations are monitored using strain
gauges and 1-m-long optical fibres, deployed every
2 meters along the pile axis and attached to the reinforcing cage. The base reaction is measured with a load
cell placed at the toe of the reinforcing cage and movements of the pile head are evaluated by 4 strain gages
attached within the pile-head cross-section (Figure 1).
The strain gauges are equipped with thermistors for
temperature correction and temperature is therefore
monitored all along the pile axis. Radial expansion
of the pile is monitored with optical fibres deployed
around the pile perimeter at different depths (Figure 1).
In this configuration, the constraints acting on the
test pile is maximal as the other piles are static and
act against its thermal deformations through raft bending. Therefore, this site gives an upper bound of the
capping conditions that a thermal pile could encounter
below a raft or slab.
Seven tests were carried out, following the construction stages of the building. The influence of the pile
confinement on its thermo-mechanical reponse was
extensively documented and quantified by estimating
the degree of freedom n (Figure 2), defined by:
Figure 3. Vertical stress induced by mechanical and thermal

loading on the first EPFL experimental pile (Laloui et al.
2003).
Finally, the variations in the internal pile load were

estimated (Figure 3); an increase of 1 C induces an
internal load increase of about 100 kN (i.e. 165 kPa).
The second full scale in situ experimental site is
dedicated to the study of a group of energy piles. Four
28-m-long piles were instrumented in the same manner
as in the first site. The piles are 0.9 m in diameter and
are built below the corner of a raft supporting a water
retention tank. The absorber pipes are arranged in 4-U
loops connected in series and deployed all around the
pile shaft. The top of the loops is placed 4 m below the
pile head and the inlet and outlet pipes are insulated
along the first 4 meters from the top in order to avoid
thermal interactions with the building (Figure 4).
Interactions between the piles that occur through
raft bending are evaluated using strain gauges installed
within the pile head cross sections. Therefore, the
influence of each pile on the neighbouring ones can be
estimated. The first stage of the analyses was to determine the profiles of degree of freedom of the piles
when no structure was built on top of them. Next, the
same tests were carried out after the water retention
tank was built.
These stages, compared in Figure 5, provide interesting data about the evolution of the constraints acting
on the piles and the influence of their location and
loads. Indeed, one can assume that all four identical
where obs and free are the measured and free axial
strains.
The free axial strain free is obtained by multiplying the linear thermal expansion of the pile Tc by the
measured temperature variation T.
The degree of freedom reaches approximately 0.5
within the first 20 meters from the top of the pile at the
end of the building construction (Figure 2). This means
that in this section of the pile, about half of the free
thermal strain will be observed while the remaining
amount will turn into internal thermal stress.
86
Figure 6. Measured pressure and temperature at the toe of

pile #1.
boreholes drilled in between the energy piles; one

borehole is located at one edge of the domain while
the other is inside it. Finally, the soil behaviour is monitored using piezometers and borehole extensometers
deployed on an edge and inside the storage volume
(Figure 4).
These in situ experiments are a major tool for
investigating the behaviour of thermal piles and their
interactions while under real service conditions.
Figure 4. Scheme of the new experiment on the EPFL

campus.
2.2 Geotechnical design of energy piles

In order to include energy piles in project planning
it is necessary to have efficient tools that can provide good estimates of the pile behaviour within a
relatively short time. Currently, the use of finite element methods in engineering to assess behaviour under
thermo-hydro-mechanical conditions remains limited
since the method is time consuming. Therefore, there
is a need to develop easy-to-use numerical tools based
on simplified theories that can promote the use of
energy piles. In this framework, we have developed a
tool called Thermo-Pile that provides a simple method
for estimating the thermally-induced stresses, strains
and related quantities in energy piles (Knellwolf et al.
2011; Pron et al. 2011).
Thermo-Pile is based on the load transfer method.
The load-transfer curves utilized are those proposed
by Frank and Zhao (1982) and defined using an initial slope Ks and an ultimate value qs . The slope of
the linear part is related to the Menard modulus EM .
The relationships utilized in Thermo-Pile were empirically elaborated for fine-grained soils and weak rocks
(Frank & Zhao 1982; Amar et al. 1991):
Figure 5. Degree of freedom of the piles pre- and postconstruction of the raft and water retention tank.
piles had the same initial (without a structure on top)

profile of degree of freedom (Free head profile in
Figure 5).
However after the thick raft (0.9 m) and water
retention tank were built, the pile restraints diverged
(Figure 5). Obviously, the different piles carry different loads, given as follows: pile 1 carries a negligible
load, pile 2 carries 800 kN and piles 3 and 4 carry about
2100 kN each.
Figure 5 suggests that pile 1 is the most restrained
after the raft and tank construction. Thus, it can be
seen that the position of the piles is important as
the mechanical loads on this site because the two
restraints that were identified (static load and raft
capping strength) have implications of the same order
of magnitude.
The pile tip compression was measured in order
to estimate the evolution of the bearing capacities
of the piles. The correlation between the cell pressure and temperature, shown in Figure 6, indicates the
compression occurring at the pile toe is elastic.
This particular experimental site was also used to
investigate heat propagation into the soil between the
piles as well as assessing the sustainability of the
seasonal heat storage. Thermistors were deployed in
where Ks and Kb are the slopes used for the shaft friction and base reaction, respectively, and D is the pile
diameter.
The elastic branches of the load-transfer curves are
valid until the mobilized bearing capacity reaches half
of its ultimate value qs . The load-transfer curve then
follows a slope equal to the fifth of the elastic slope,
Ks /5, until the ultimate bearing capacity is reached.
Unloading is achieved according to an elastic branch
(Figure 7). The ultimate bearing capacities can be
87
strains is estimated. Next, the blocked strains are subtracted from the previously observed strains to obtain
a new strain state for the pile. This process is repeated
until the desired accuracy is reached.
Thermo-Pile is currently able to perform monotonic
temperature variations such as heating or cooling for
one pile with a circular cross-section and embedded in
a layered soil.
This tool was then validated against two major
in situ thermal pile experiments: the single pile test on
the EPFL campus in Lausanne, Switzerland (Laloui
et al. 2003) and the Lambeth college test in London,
United Kingdom (Bourne-Webb et al. 2009); an example of the validation curves is shown in Figure 8.
Figure 7. Example of load-transfer curve used in

Thermo-Pile (Knellwolf et al. 2011).
3 THERMO-MECHANICAL BEHAVIOUR
OF SOILS
3.1 Heat transport in soils
3.1.1 Theoretical background
Various heat transport phenomena can occur in soils
such as conduction, radiation, convection, latent transport through vaporization and condensation processes, ion exchange or freezing-thawing processes
(Hermansson et al. 2009). Underground water may
also have a significant impact in soils with significant
permeability. Enthalpy balance allows the derivation
of the heat equation taking into account all the abovementioned phenomena but simplifications are usually
made regarding their relative magnitudes.
Numerical methods can take into account multiple
heat transport phenomena but are still time consuming and may not be efficient for daily engineering
purposes. Therefore, analytical solutions were developed to consider the main heat transport phenomena
represented by the conductive/diffusive and advective
terms in equation (3). The advection-diffusion of heat
is therefore given by:
Figure 8. Example of curves used for the validation of

Thermo-Pile against the first EPFL experiment (Knellwolf
et al. 2011).
user-defined or linked to soil parameters according to

conventional methods (Lang & Huder 1978; Legrand
et al. 1993). The rigidity of the structure at the head
of the pile is accounted for with a linear spring Kh .
Indeed, while expanding or shrinking, the pile is pushing or pulling on the structure it is attached to (raft,
beam, wall), and thus develops a capping force.
The thermo-mechanical response of a pile is computed as follows: First, the stresses and strains induced
by the mechanical loading as well as the corresponding quantities (mobilized bearing capacities, displacements) are computed using the methods described by
Coyle and Reese (1966). Next, the thermal effects are
estimated through an iterative process. The initialisation is achieved by assuming that the piles are totally
free to move so that the initial strains are equal to
the free strains and directly related to the temperature variation through the concrete thermal expansion
coefficient. Next, the corresponding displacements are
calculated from the null point whose displacement is,
by definition, null.
Then, the introduction of the estimated displacements into the load transfer curves gives a set of
mobilized bearing capacity values corresponding to
the internal stresses.
The thermally-induced internal stresses are related
to the blocked strains so that a first set of blocked
where is the bulk thermal conductivity of the media,

U is the fluid velocity relative to the media, QT represents the volumetric heat sources and sinks, and ST is
the enthalpy of the system given by:
where c is the bulk heat capacity of the media and T0

is a reference temperature.
In soils with negligible underground water flow,
the heat transport is governed by the so-called heat
equation where the advective term is neglected:
Therefore, heat conduction/advection can be

described by differential equations with partial derivatives for which analytical solutions were found.
88
3.1.3 In situ observations

Two different types of thermal tests were carried out on
the last in situ experimental site at the EPFL campus.
The first one was undertaken before the construction
of the structure on top of the piles and consisted of
a conventional Thermal Response Test (TRT) carried
out on a single pile. Even though this type of test is
designed for geothermal probes, it may be possible to
use it to determine the thermal conductivity of the surrounding soil as well as the thermal resistance of a pile
by removing capacitive effects. Indeed, the capacitive
effect can be estimated as a lost flux that increases the
temperature of the concrete and does not go into the
ground.
The procedure is the same as for a conventional
TRT, that is to say a heater is connected to the absorber
pipes of the pile and a constant heat rate is injected.
Inlet and outlet temperatures of the pile are monitored and the thermal conductivity and pile thermal
resistance can be estimated (Mattsson et al. 2008).
The TRT gave an estimate of the thermal conductivity of around 2.5 W/(K m), which is close to
the estimate found previously on the first site, i.e.
2.4 W/m. The thermal resistance of the pile was estimated at 0.09 (m K)/W, in agreement with values
provided by the Swiss documentation SIA-D0-190
(2005).
Another test was conducted using a periodic injection of heat similar to periodic water pumping tests
(Renner & Messar 2006). The temperature variations
of one pile and at one meter removed from it were
observed and compared in order to estimate the thermal diffusivity of the soil, analysing the time lag and
attenuation between the source (pile) and the monitoring point. The large diameter of the piles and the high
thermal inertia of the concrete required long periods
of testing over several weeks for small temperature
variations observed in the soil. Nevertheless, this test,
whose theoretical bases are developed in the following section, gave rough estimates of the soil thermal
diffusivity of between 107 m2 /s and 106 m2 /s.
In conclusion, several in situ and laboratory tests
are available to determine the parameters that describe
the thermal interactions between thermo-active foundations and their environment.
3.1.2 Analytical solutions

Analytical solutions were developed to give reliable
results for heat propagation through soils from its surface or from embedded heat exchangers for rather
simple configurations.
The annual temperature wave propagation through
the ground is obtained by solving the transient onedimensional heat equation (5) in a forced periodic
regime within a semi-infinite body. Assuming that the
soil surface temperature TS follows a sine oscillation
around a mean value Tm with a period 0 of one year
and amplitude T0 :
the temperature evolution in the ground is delayed and

attenuated according to:
where z is the increasing depth.

Solutions were found for particular types of heat
exchangers under certain assumptions. First, the infinite line source (ILS) theory was developed for
geothermal boreholes (Mogensen 1983; Eskilson
1987). This model considers an infinite line source
within an infinite domain so that the heat flow occurs
radially. In situ thermal response tests employed to
estimate the thermal conductivity of soils are based on
this theory (Mattsson et al. 2008).
A more realistic model considers a Finite Line
Source (FLS) and is based on the solution provided
in chapter 10 of Carslaw & Jaeger (1947) for a single point. The solution for the borehole is found by
integrating this particular solution along the borehole
length (Eskilson 1987). Later, Bandos et al. (2009)
included the geothermal gradient into the FLS.
The Infinite Cylindrical Source (ICS) model considers an infinite cylinder and includes a radius to the
source so that the heat flux is applied on the cylinder shaft. The solution provided by Ingersoll & Plass
(1948) was later transformed into a more user-friendly
form by Baudoin (1988). Philippe et al. (2009) compared those three models and determined that the ICS
was suitable for short-term impact investigations (less
than 1 day) while the advantages of the FLS are mainly
on the long-term effects (more than 1000 days) and the
ILS is well suited to medium-term operations (1 day
to 1 year).
More recently, complex solutions dedicated to thermal piles were developed in order to take into account
the absorber pipe geometry. Cui et al. (2011) gave an
analytical solution based on Greens functions for a
heat exchanger with spiral coils.
Finally, Diao et al. (2004) gave an analytical solution for the temperature on the horizontal plane in the
presence of groundwater advection. This solution considers boreholes as ILSs so that the depth dependence
is removed and the solution is found in horizontal soil
slices parallel to the soil surface.
3.1.4 Periodic pumping laboratory experiment

A laboratory test was performed in order to evaluate the
potential of using periodic pumping tests to estimate
the soil thermal diffusivity through GHEs. This study
was based on a procedure utilised to estimate the
hydraulic diffusivity of soils using periodic well injections and extractions (Renner & Messar 2006). A
periodic-like temperature variation was achieved at the
borehole wall by the intermittent operation of a heater
and the induced temperature variations in the soil were
recorded at various distances from the heat source.
When the forced periodic regime is reached, a
constant delay between the injection wheel and a
monitoring point is observed as well as a constant
attenuation of the temperature variation (Figure 9).
89
Figure 9. Example of temperature shift and amplitude

attenuation between the source and thermocouple #1.
Figure 10. Diagram of the laboratory experiment used for

the thermal periodic pumping test.
The analyses of these attenuations and delays allows

for the estimation of the diffusivity of the media.
If several oscillation techniques had already been
used on the soil samples, the temperature variations
were generally applied from the external faces of the
sample.
In situ measurements were achieved by recording
the air and ground temperatures and analysing them
using the forced periodic and one-dimensional propagation of heat through a semi-infinite domain (Forbes
1849; Hinkel 1997). However, this method assumes
that the annual variation in temperature at the ground
surface follows a sinusoidal wave and that the soil does
not experience any alteration (Scott 1964). Laboratory
tests were performed on a cylindrical specimen but
the temperature variations were applied at the specimen periphery (Hoekstra et al. 1973) which may not
be applicable to in situ testing.
The presented method was applied to Bioley silt
(Peron et al. 2009) contained in a thick cylindrical
tank. Two geothermal loop-like copper tubes were
installed in the middle of the tank before placing the
over-saturated silt layer by layer.
Even if this procedure is not truly representative
of a real geothermal loop or pile installation, it was
chosen to ensure optimal soil-absorber pipe contact.
The copper tubes cross the tank from top to bottom so
that the heat transfer can be considered as radial.
The fluid circulating inside the tubes is a waterglycol mixture. The temperature measurements are
achieved with thermocouples installed around the
absorber pipes (Figure 10). Inlet and outlet temperatures of the fluid are also monitored to keep track of
the quantity of heat injected into the ground sample.
The heat source, represented by the copper tubes, is
assumed to induce sinusoidal temperature variations
whose amplitude is T0 (in K) and the pulse is (in
s1 ). The heat transport is assumed to be governed by
the transient heat equation in cylindrical coordinates
given by:
Figure 11. Redrawn from (Renner & Messar 2006), graphical illustration of the interference method.
where DT (in m2 s1 ) is the thermal diffusivity,

assumed to be isotropic, of the Bioley silt. The temperature variations at the tube walls (at r = ri ) can be
expresses as the real part of:
Solving equation (8) for the forced periodic regime

and considering that at a given distance R from the
source, the temperature is fixed to the reference temperature, the temperature in the silt is given as a linear
combination of the modified Bessel functions of order
0, K0 and I0 :
Next, if R tends towards infinity, a simplified

solution can be given by:
90
Table 1. Observed phase shifts and attenuations with the

corresponding estimated thermal diffusivities.
Sensor
TT
D (m2 /s)
TT
D (m2 /s)
1
2
3
4
5
6
0.32
0.37
0.26
0.30
0.37
0.37
7.15 107
3.83 107
8.64 107
5.78 107
3.64 107
9.71 107
0.14
0.11
0.17
0.14
0.11
0.16
3.27 107
2.44 107
3.95 107
3.00 107
2.40 107
4.14 107
From this solution, one can express the amplitude

attenuation TT and phase shift TT between the heat
source and any point within the silt as:
Figure 12. Thermo-mechanical loading paths; p the
effective mean pressure, q the deviatoric stress and T the
temperature.
the principal experimental observations concerning

thermally-induced effects on soils is given. To clarify
the presentation, two types of loading paths are distinguished (Figure 12): thermal (Path 1) and isothermalmechanical (Path 2). Path 1 represents the behaviour
of a soil subjected to a temperature variation (the difference between the present and initial or reference
temperatures) at constant stress. Path 2 corresponds to
mechanical loading at a given constant temperature.
Let N be equal
to1the ratio r/ri . Therefore, selecting the term
D ri as the main variable, the
=
f
(
D1 ri ) can be plotted and TT =
curves
TT
1
f ( D ri ).Next, it may be possible to retrieve
the value of D1 ri for the observed attenuation and phase shift and, therefore, the diffusivity DT
(Figure 11).
The temperature time series were smoothed with a
convolution vector of 100 elements and the recording frequency is 1/60 (i.e. 1 measurement per minute).
The period of the power variations is about 24 hours.
The phase shift and attenuation estimations are
given for the last of the four cycles of the the experiment in order to be as close to the forced periodic
regime as possible. The results from the test are shown
in Table 1. The average value of the retrieved thermal
diffusivity is about 4.83 107 m2 /s with a standard
deviation of 2.34 107 m2 /s.
The values determined for thermocouples 2 and 5
are closest to the real value since the assumption that
r/R goes to infinity is the most satisfied in these cases.
In conclusion, the periodic pumping test was shown
to be capable of estimating the thermal diffusivity
of the sample but assessing the boundary conditions
remains to be determined. This method requires few
mathematical tools as it does not utilize regression or
curve fitting and no assumption is made for any physical parameter. Finally, if in situ tests are carried out
using this method, only the conventional material utilized for a Thermal Response Tests plus a temperature
probe (buried or plugged thermistor) are required.
3.2.1.1 Thermal behaviour (Path 1)

Saturated soil is a two-phase material made up of a
solid part (a skeleton of grains or particles surrounded
by adsorbed water, for clays) and a fluid part (free
water) in the inter-aggregate space (or voids).
When a soil is heated, all of the constituents dilate.
In the case of clayey soils, this dilation produces a
decrease in the strength of the adsorbed layers and
a modification of the distance between the clay particles (Fleureau 1979). This changes the equilibrium
between the Van der Waals attractive forces and the
electrostatic repulsive forces, which results in one of
the most characteristic thermal behaviours of clays;
in normally consolidated conditions (NC), where the
effect of the stress is less important (than in overconsolidated conditions (OC)), the clay contracts when it
is heated and a significant part of this deformation is
irreversible upon cooling. This thermal contraction is
an unusual behaviour for any material. Figure 13 illustrates the response of a sample of saturated, drained
clay to a thermal heating-cooling cycle at constant
isotropic stress (Laloui & Cekerevac 2008). Under
heating, a NC clay sample will settle with a non-linear
volume variation. Upon cooling the volume increases.
The behaviour over the whole cycle indicates the
irreversibility of strain due to thermal loading, which
is representative of thermal hardening.
Even though there has been no physical change in
the effective stresses, this behaviour can be interpreted
as the soil undergoing densification, i.e. overconsolidated behaviour. The highly OC state mainly produces
3.2 Thermo-mechanical behaviour of soils

3.2.1 Experimental results
Temperature variation has a significant effect on the
behaviour of soils. In this section, a summary of
91
Figure 15. Isotropic consolidation of Illite samples at three

different temperatures, from Campanella & Mitchell (1968).
Figure 13. Thermal volumetric strains of Kaolin clay during
three thermal cycles (22 C 90 C 22 C) applied in NC
and OC (OCR = 12) states, from Laloui & Cekerevac (2008).
Figure 16. Influence of temperature on preconsolidation

pressure from Laloui & Cekerevac (2003).
Figure 14. Influence of overconsolidation ratio on the thermal volumetric strain of clays, from Cekerevac & Laloui
(2004).
reversible dilation. Between these states, an intermediate one with a low overconsolidation ratio OCR,
first produces dilation and then has a tendency toward
contraction. The intensity of the reversible/irreversible
parts of the deformations due to temperature cycling
depends upon the soil type and plasticity, in addition
to the stress level measured in terms of OCR. This is
illustrated in Figure 14, where the influence of OCR on
the thermal behaviour of several soils is shown. It can
be observed that, for a given increase in temperature,
compaction is smaller for higher overconsolidation
ratios and then tends to dilation.
Figure 17. Effect of temperature on friction angle at Critical

State from Cekerevac & Laloui (2004).
3.2.1.2 Mechanical behaviour (Path 2)

Under isotropic stress conditions, the slope of the consolidation line is independent of temperature in the
(e ln p ) plane, with e being the void ratio and p
being the mean effective pressure.
Experiments on saturated Illite at three different
temperatures also show that heating applied prior to
loading produces a densification of the sample at constant isotropic pressure (Figure 15). To analyse the
change in the pre-consolidation pressure with respect
to temperature, tests were conducted by heating the soil
under stress loads smaller than the pre-consolidation
value and then applying a mechanical load under a
constant temperature.
In Figure 16 the results of tests on several clays are
summarized such results for several clays (Laloui &
Cekerevac, 2003). The pre-consolidation pressure

decreases non-linearly with an increase in temperature.
This decrease means that the yield limits in an isotropic
plane decrease with a temperature increase. This
phenomenon is independent of any viscous effects, as
was shown by Boudali et al. (1994).
The thermal effect on strength is a complex process
as it combines the effect of the resultant plastic strains
with the intrinsic thermal one. Some experimental
results are summarised in Figure 17. A comprehensive analysis of this aspect is presented by Pron et al.
(2009).
Under undrained conditions, the behaviour of a soil
subjected to thermo-mechanical loading is strongly
affected. Heating results in a significant pore pressure
increase (Plum & Esrig 1969).
92
where pcyc is the mean effective stress at the last change

p,cyc,iso
is the plastic volof direction of stress and
umetric strain produced by the isotropic mechanism
since the last change of direction of the stress. The
preconsolidation pressure pc depends on the
volumetric plastic strain as:
At a constant total stress difference in a triaxial test,

the pore pressure growth can induce failure in the sample. Experimental results of tests on two clays reported
by Hueckel & Pellegrini (1989) have shown that that
clays under undrained conditions fail at temperatures
of between 70 C to 90 C. It can be seen that this failure is due to an increase in pore pressure that leads
to an effective main stress drop under constant total
stress conditions until the critical state line is reached.
The higher the initial deviatoric stress the faster the
occurrence of failure.
where is the compressibility modulus and pco the

initial preconsolidation pressure at temperature T .
The thermal effects on the yield limit are accounted
for through the temperature dependency of pco as
(Laloui & Cekerevac 2003):
3.2.2 ACMEG-T a thermo-plastic constitutive

model
The ACMEG-T model (Laloui & Franois 2009) is
developed to take into account the thermo-mechanical
responses of soils observed in the previous section, and
includes the thermal effects of both isotropic and deviatoric mechanisms implicated in soil plasticity. The
elasto-plasticity principle allows for the total strain
increment, dij , to be split between thermo-elastic deij
p
and thermoplastic dij components:
where pcoTo is the preconsolidation pressure at T0 and

T is a material parameter. Finally, the yield limit of
the isotropic mechanism is given under non-isothermal
conditions by:
The elastic part of the strain is modelled by

the conventional thermo-elasticity theory (DuhamelNeumann equation). The main advantage of the
present model with respect to the original isothermal
model (Hujeux 1979) is the addition of non-isothermal
mechanisms in the plastic component, which can be
divided into an isotropic mechanism whose yield limit
is fiso , and a deviatoric mechanism whose yield limit
is fdev .
3.2.2.2 Plastic deviatoric component

The yield limit associated to the deviatoric plastic
mechanism is given by:
where M is the slope of the critical state line in the

(q p ) plane and d is the ratio between the preconsolidation pressure and the critical pressure. M is
given by:
3.2.2.1 Plastic isotropic component

The yield limit fiso associated with the isotropic mechanism is given by:
where pc is the preconsolidation pressure and riso is the

degree of plastification of the isotropic mechanisms
allowing progressive evolution of the isotropic yield
limit during loading and partial reversal of the limit
during unloading. During loading, riso is an hyperbolic
function of the plastic volumetric strain induced by the
p,iso
isotropic mechanism :
where is the friction angle at critical state. As shown

by Cekerevac & Laloui (2004) and Hueckel et al.
(2011), the friction angle may depend on temperature. So, the following expression is proposed (Laloui
1993):
where M0 = slope of the critical state line at ambient

temperature T0 and g = average slope of variation of
friction angle at critical state with temperature.
rdev can be calculated in the same way as for the
isotropic mechanism, and represents the degree of
plastification of the deviatoric mechanism enabling
a progressive evolution of the deviatoric yield limit
during loading:
where c is a material parameter. During unloading, riso

follows the decrease of the effective mean pressure, p ,
and at reloading, it is adjusted to keep a defined elastic
e
nuclei (riso
):
93
Figure 19. 3D foundation and representation of the equivalent 2D foundation.
purpose of using realistic numbers, it has been decided

to use a theoretical storage volume of 104 000 m3 .
This foundation, shown in Figure 19, would consist of 105 piles (7 15), each with a diameter of
80 cm and a length of 20 meters, spaced 7 meters
apart in a square pattern. The foundation is considered to function as a piled raft and is located on deep,
low-permeability, clayey soil, which is assumed to be
homogeneous and to behave thermo-elastically in the
range of temperatures considered.
Interactions between the thermal, mechanical and
hydraulic responses are the main source of uncertainty
in the design of a thermal energy storage geostructure.
Heat exchange is the driving physical factor in this
problem, and it creates thermal strains and thermal
diffusion. As such, heat exchange also has hydraulic
and mechanical consequences. The numerical simulations are run using the FEM code Lagamine (Charlier
1987; Collin et al. 2002). The finite element mesh
(Figure 19) is composed of 3863 nodes and 1240
8-node elements.
Figure 18. Coupled thermoplasticity yield limits (Laloui &

Franois 2009).
e
where rdev
and a are material parameters defining the
size of the elastic nuclei of the deviatoric mechanism
p
and the evolution of rdev , respectively, while d is the
deviatoric plastic strain.
Through a combination of the previous equations,
the deviatoric yield surface becomes, under nonisothermal conditions:
3.3.1.2
Concepts for the transition from a 3D to a

2D configuration
From a mechanical point of view, the transition from
3D to 2D for a piled raft foundation can be considered
by modifying theYoungs modulus of the piles, as indicated in Prakoso and Kulhawy (2001). The responses
of vertically loaded pile foundations are controlled primarily by the axial stiffness of the piles. Because the
piles are simplified into strips in a plane, a row of piles
has to be simplified into an equivalent plane strain pile
with the following modified Youngs modulus:
Evolution of the isotropic and deviatoric yield limits

with temperature are shown in Figure 18. Validations
of the model against laboratory tests can be found in
(Laloui & Franois 2009).
3.3
Examples of numerical modelling of

geothermal structures
3.3.1 Identification of the thermo-mechanical

behaviour of a heat-exchanger foundation
3.3.1.1 The structure under consideration
In the previous sections, the thermo-mechanical
behaviour of a single isolated energy pile is studied
experimentally and the thermo-mechanical characteristics of the soil are presented from both experimental
and constitutive modelling points of view. In order to
gain a global understanding of the system it is necessary to simulate the response of a foundation from
both thermal and mechanical perspectives.
In this section, the behaviour of a multi-pile heatexchanger foundation subjected to thermo-mechanical
loading is examined numerically from both thermal
and mechanical perspectives.
A 2D-approach is used for this purpose. For satisfactory seasonal storage performance, a large reservoir
is required (greater than 30 000 m3 , according to Pahud
(2002)). For a conventional building in dense clayey
soil, the typical pile length is 20 m. Because the system
is analysed in 2D, the volume is infinite, but for the
where Eeq is the equivalent elastic modulus, Ep is the

concrete elastic modulus, Ap is the pile section, B is
the pile diameter, Lr is the slab length (in plane), np-row
is the number of piles in a row, and eeq is the 3D-2D
coefficient, as defined in Eq. 27.
An extension of this principle to thermo-hydromechanical conditions has been proposed by Dupray
et al. (2013), which makes use of the effective stress
framework to determine the correct representation of
the pile behaviour in this 2D situation.
The effect of water pressure on the actual stresses
should be written as follows:
94
where pw is the pore water pressure and p is the

real effective stress in the pile. This is also written
as the first line of the following set of equations, while
the second line simply uses the general form of the
effective stress definition, with eq and c as unknowns:
As the strains of the real and equivalent material are

assumed to be equal, the left terms are equal, leading
to the solution:
Thermal strains in the pile are also affected by its

representation as a wall. Thermal expansion in the
out-of-plane direction is blocked while it would be
almost free in a pile. The thermal expansion coefficient used in the model is therefore reduced to
compensate the increase by a factor (1 + ) of the
axial thermal expansion, as calculated in plane-strain
thermo-elasticity.
The transition from 3D to 2D simulation also has
consequences for the thermal aspect of the problem.
The choice is made to inject and extract from the corresponding wall the same amount of energy as in the
piles. This assumption leads to a different repartition
of temperature in the vicinity of the piles because this
energy is transferred to the soil over a larger surface
than in reality. It is also more appropriate in the long
term because both the heat storage ground volume and
the energy exchange are equal to the real values.
Figure 20. Pile foundation configuration and FE mesh.
is a typical value in temperate regions. A temperature of 15 C is applied at the slab nodes to account
for the presence of the thermally controlled building.
No exchange of water or heat is considered between
the soil surface and the environment. With respect to
the mechanical boundary conditions, restrictions are
applied to both the vertical and horizontal displacements at the base of the mesh and to the horizontal
displacements on the sides of the mesh. The size of
the model ensures that the outer boundaries are sufficiently far from the building. The initial stresses in
the model due to gravity are introduced by assuming
a coefficient of earth pressure at rest of K0 = 0.5.
The initial pore water pressure corresponds to the
hydrostatic profile with a water table located at the
surface. Because of the assumption of saturated conditions, the water table remains at the top of the model
during the simulation.
3.3.1.3 General THM formulation

The diffusive model that is used in this simulation
was written and implemented in the software used by
Collin (2003). Only a short description of the principles of the approach is presented here. The equilibrium
and balance equations, as well as the water and heat
flows, are expressed in the moving current configuration through a Lagrangian-updated formulation.
The model uses a phenomenological description
that averages properties of the medium (enthalpy, conductivity) depending on the spatial repartition of its
constituents. Water is considered to be compressible
and soil grains incompressible but both are affected
by thermal dilation. The properties of the soil and
concrete are presented in Table 2.
3.3.2 Boundary conditions and initial configuration
It is assumed that the energy geostructure under consideration does have any similar structures nearby. As
a consequence, a constant temperature is imposed on
the lower and lateral boundaries of the mesh, which
are at a distance from the heated volume. The axis of
symmetry is naturally an adiabatic boundary. The initial temperature of the ground is fixed at 11 C, which
3.3.3 Loading path

An evaluation of the bearing capacity of the structure was conducted according to the norm SIA-267
(2003) with assumed frictional parameters. Based on
this norm, it was decided to apply a mechanical loading of 1500 kN on each pile, which corresponds to a
95
Table 2. Thermal, hydraulic and mechanical characteristics

of clayey soil and concrete
Clayey soil
Porosity n
Intrinsic permeability
Solid thermal conductivity s
Solid specific heat cp,s
Solid specific mass s
Elastic modulus
Poissons coefficient
Volumetric thermal expansion coef.
0.39
11014 m2
2.42 W.m1 K1
732 J kg1 K1
2 700 kg m3
30 MPa
0.3
1.2 104 K1
Concrete (piles and slab)

Porosity n
Intrinsic permeability
Solid thermal conductivity s
Solid specific heat cp,s
Solid specific mass s
Elastic modulus
Poissons coefficient
3D-2D equivalent coefficient
Equivalent elastic modulus
0.12
11016 m2
1.7 W m1 K1
880 J kg1 K1
2 300 kg m3
35 000 MPa
0.25
3.6 105 K1
0.089
3 125 MPa
Water
Dynamic viscosity
Compressibility
Water thermal conductivity w
Water specific heatcp,w
Water specific mass w
f (T )
4.54 1010 Pa1
2.1 104 K1
0.57 W m1 K1
4186 J kg1 K1
1 000 kg m3
Figure 21. Loading functions as imposed on the piles.
3 MPa stress on an isolated pile. The first step consists in applying this load on top of the simulated
piles, through the raft. Sufficient time is given to allow
dissipation of additional pore pressure.
Thermal loading is then applied and cycled for
five years. The thermal load is applied directly to the
piles in the form of a uniform surface source/sink
term. Heat injection and extraction rates are estimated based on the most efficient systems in service
today, and higher rates are also studied because they
offer better volumetric efficiency. Nevertheless, sustainability and reload during the non-heating period
should be ensured (Choi et al. 2011).
In the simulation, the heat extraction and injection
rates are fixed at 90 W/m at the beginning (scenario
1, shown in Figure 21 along with other scenarios). For
comparison, in the high-rise building Palais Quartier
in Frankfurt, a total heating and cooling power of
nearly 1000 kW is made available by the use of 392
thermo-piles, each measuring 30 m (Katzenbach et al.
2009).This power corresponds to a heating and cooling
injection rate of approximately 84 W/m.
These profiles do not take into account daily variations. Heat storage comes first with a linear increase,
as would happen in May, setting day zero. Scenario 2
is used to evaluate the thermal losses, while scenario 3
aims at representing future improvements in the availability of warmer heat sources such as solar thermal
panels. Heat exchange is dissipated uniformly on the
cross-section of the pile.
Figure 22. Evolution of the temperature in the storage

system over three years for various injection/extraction ratios.
3.3.4 Thermal results

In the reference scenario, the injection and extraction
rates are 90 W/m. The total input and output are therefore 1960 GJ, or considering the theoretical size of
the system, 18 MJ/m3 on average, corresponding to
a mean temperature difference in the soil of 4.7 C. In
order to precisely characterise the temperature evolution in the storage system, node C is monitored
(Figure 20). This node is located at a depth of 10 m
beneath the slab and is equidistant between piles 3 and
4. It corresponds to a zone closer to the edge of the
storage system, where losses are more visible while
still being clearly within the storage zone. Figure 22
shows the evolution of temperature at point C over
time, and the temperature change of 5.1 C in scenario
1 indicates that point C is a good overall indicator of
the mean storage temperature.
96
In contrast to the single pile experiments, the

additional stress due to temperature is very limited,
reaching 0.25 MPa (see Figure 23) which is less than
10% of the design stress. This stress only amounts to
25 kPa/ C (primarily due to the group effect), as will
be discussed in the next section. For the extraction
period, the behaviour is slightly different because the
stress reduction appears to be slightly more dependent
on temperature, reaching 32 kPa/ C.
In the range of temperature considered, it is estimated that, for an injection rate of 90 W/m in a homogeneously used geostructure, the maximum effect of
temperature on the pile stress should not change the
design principles of such a foundation. These effects
may be taken into account either through evaluation,
such as in this simulation, or through an additional
safety coefficient. To explore the limits of acceptable
temperature increase from this point of view, simulations with higher rates of injection and extraction have
been run. Thermal scenario 3, in which the temperature in the piles evolves between 1 and 35 C, is used
for this purpose. The maximal thermal overstress value
observed is equal to 0.63 MPa, which represents 21%
of the total mechanical loading. The additional stress
in this case is 26 kPa/ C, which is similar to the value
found at lower temperatures.
When considering only one row of piles for heat
exchange, the maximum thermal overstress value
becomes equal to 1.4 MPa, or 50% of the mechanical
loading, and 62 kPa/ C, which is slightly lower than the
previously identified values which were in the range
of 100 kPa/ C (obtained in very different conditions).
The partial use of a foundation as a heat exchanger
introduces significant stresses due to the stiffness of
the rest of the geostructure. This scenario can and
should be avoided by using the entire foundation as
heat exchanger.
Figure 23. Vertical stress induced by constrained thermal

strains in the first pile.
In general, the temperature in the storage zone

varies between 6 C and 21 C during a cycle. The
temperature in this case is clearly decreasing slightly,
indicating that thermal losses are present but of a low
magnitude.
Scenario 2 serves to quantify these losses, setting
the injected heat at such a rate that temperatures in
the heat store evolve in a permanent regime over the
years. This is achieved by injecting 95 W/m instead
of 90 W/m, or 5.5% heat loss per year, in the crosssection. Doubling the losses because of the 3D effect
leads to an efficiency (ratio of extracted heat to
injected heat) of 0.89, close to the one (0.85) observed
by Lund and stman (1985) in a non-structural
system.
An increase in the heat exchange rate is shown in
scenario 3, with an extraction rate of 225 W/m and
an injection rate of 245 W/m. In line with expectations, this turns into an increase of the average storage
temperature and therefore into higher losses, with the
efficiency reaching 0.83.
3.3.5 Mechanical implications

Several experimental studies have been performed that
provide information about the mechanical response
of a single pile when thermally loaded. The main
results are from data obtained from the EPFL test
pile (Laloui et al. 2003) and the Lambeth College
test pile (Bourne-Webb et al. 2009). These studies
have enabled the calculation of a range of pile stresses
that can be created by thermal loading, the maximum
varying between 100 and 192 kPa/ C (Amatya et al.
2012).
The differences seen between a single energy pile
and a whole foundation are investigated in this section.
In order to do so, the case where the whole foundation is used as heat-exchanger is compared to the
situation where only one row of piles (the central one)
is used for heat exchange.For the reference scenario
(90 W/m), the increase in temperature in the piles is
in line with the previously mentioned experiments,
at 10 C.
ESTIMATING THE POTENTIAL OF HEAT

EXCHANGER TUNNEL ANCHORS AND
NAILS
Underground piles are not the only structures than can

be equipped with absorber pipes; structures such as
floor slabs of metro stations or tunnel walls have been
thermally activated, mainly in Austria, where different
applications of energy geostructures were tested on
tunnels. A 54-m-long portion of the Jenbach tunnel
(Austria) was equipped with a thermo-active tunnel
lining that was moulded with absorber pipes (Frodl
et al. 2010; Franzius & Pralle 2011), while thermal
geotextiles were tested on the Lainzer tunnel (Austria)
(Brandl et al. 2006).
Many new energy geostructures were tested along
the U2 metro line in Vienna (Austria). Absorber pipes
were deployed on top of the Messe-Prater station floor
slab or on the invert slab of the Traborstrasse station
tube (Adam & Markiewicz 2009).
Finally, heat exchanger anchors were tested on an
in situ experiment linked to the Lainzer tunnel but
97
Table 3. The thermal, structural and water retention parameters used to represent the porous media.
Silt
Clay
Concrete
Structural
n,k
Water retention
m,
3.43, 419, 2700

2.42, 732, 2700
1.7, 880, 2500
0.45, 1013
0.55, 1015
0.20, 1015
7/17, 50
1/3, 500
,
tunnel and with thermo-hydro-mechanical finite element analyses for the bored tunnel. The analyses are
run with the finite element code Lagamine (Charlier
1987; Collin 2003).
Different soil conditions are tested for each tunnel structure according to their respective environments. Two generic types of soils, silt and clay, are
selected in order to meet the requirement for seasonal
heat storage. Indeed, sand-like soils allow significant
underground flow that could compromise heat storage and are therefore out of the scope of the present
study. Therefore, silt is the representative soil with low
permeability and clay corresponds to an impermeable
soil.
The soil conditions are assumed to be water saturated for the bored tunnel while two different water
table levels are tested on the cut and cover tunnel (0 m
and 20 m). The different parameters used to represent the porous medium are regrouped according to
thermal, mechanical and water retention parameters,
listed in Table 3.
Finally, different heat production cycles are tested.
Based on an air temperature time series from the
Swiss meteorological service (MeteoSwiss), a simplified heat demand was established for a building in
Lausanne (Switzerland).
Next, two types of cycles were designed. The first
type, Ce, does not include heat injection, letting the
soil rest during the hot periods. The second type, Cei,
includes heat injection during the hot period and this is
split into two sub-categories: Cei,e that starts the heat
exploitation by heat extraction and Cei,i that starts with
heat injection.
Once the shape of the heat production cycles is
designed, it is optimized for each possible set of conditions in order to reach a temperature threshold at a
specific location in the ground.
The threshold is set at 273K between the anchors or
nails to prevent freezing the soil. When heat injection is
considered (i.e. for Cei cycles), it is kept constant for a
given tunnel structure in order to make the comparison
between the different configurations.
The finite element analyses are used to optimize
the heat exploitation cycles and to investigate their
mechanical impact on the bored tunnel.
Figure 24. Schematic of the modelled cut and cover tunnel

(top) and bored tunnel (bottom).
not connected to the tunnel body (Adam 2008). The

absorber pipes were inserted into the anchor bodies
forming a coaxial probe which turned out to be the
best solution for the relatively small anchor diameters
compared to thermal piles (Brandl 2006; Brandl et al.
2006).
The present study numerically investigates the
potential of using anchors and nails of shallow urban
tunnel as heat exchangers with the surrounding ground
for seasonal heat exploitation. Indeed, using heat
through tunnels crossing large and hot geological formations at great depths would be achieved with more
efficient processes than ground source heat pumps
(Brandl 2006).
4.1
Thermal
s , cs , s
Method
Two main types of urban tunnels are considered, with

a cut and cover tunnel and a bored tunnel (Figure 24),
because they present specific challenges linked to their
locations.
The cut and cover tunnel remains close enough
to the surface so that the annual temperature wave
can significantly impacts heat exploitation through the
anchors while the bored tunnel, because it is deeper, is
subjected to greater confining stresses. Therefore, two
different sets of assumptions were utilised.
The mechanical aspect of the problem was
neglected for the cut and cover tunnel because of its
shallow depth. It is assumed that the proximity of the
soil surface allows thermal expansion of the soil to
develop without inducing significant stresses within
the tunnel structure due to the buffering effect of the
backfill, which limits interactions between the walls
and tunnel. Conversely, the greater confinement of the
bored tunnel has to be taken into account and a thermoelastic approach is utilized to estimate the stress and
strain variations in the lining of the bored tunnel.
Under the assumptions described previously, the
heat exploitation is investigated by means of thermohydraulic finite element analyses for the cut and cover
4.2 Results of cycle optimization

Each configuration is given a name composed of
the soil type, cycle type and water table depth.
98
Table 4. List of the extracted (He) and injected (Hi) heat in

kWh per year and per meter of anchor.
Silt-Ce-0 m
Silt-Ce-20 m
Silt-Cei-0 m
Silt-Cei-20 m
Clay-Ce-0 m
Clay-Ce-20 m
Clay-Cei-0 m
Clay-Cei-20 m
Cut and cover

He
Hi
Bored
He
Hi
9.63
7.41
14.45
12.02
7.41
7.41
14.46
14.46
45.94
57.86
38.86
54.92
33.62
33.65
0
0
8.41
8.41
0
0
8.41
8.41
Names given to the configurations are a combination of the

soil type (clay or silt), cycle type (Ce and Cei, without and
with head injection, respectively) and water table depth (0 m
or 20 m).
Table 5. Percentage of injected heat retrieved during the

following extraction.
Clay-0 m
Silt-0 m
Silt-20 m
Cut and cover
Bored
83.8
57.4
54.8
48.3
35.4
Figure 25. Time series of stress for the bored tunnel in clay
with (top) and without (bottom) heat injection. Stress evolution are taken at locations 1, 2 and 3 which are at the tunnel
crown, middle and invert, respectively. Points 1 , 2 and 3 are
located on the intrados of tunnel lining while points 1, 2 and
3 are on the extrados. Negative stress means compression.
Names of the different configurations, combining the soil

type (Clay or silt) and the water table depth (0m or -20m).
Those names are different from the ones in Table 4 because
they regroup at the same time cycles with and without
injection.
invert and middle, were recorded and analysed to estimate the mechanical implications of heat extraction
and injection.
Points 1, 2 and 3 are on the lining extrados while
points 1 , 2 and 3 are on the lining intrados. Let
us recall that the mechanical behaviour of the tunnel structure and soil are assumed to be thermo-elastic
and that their thermal expansion coefficients are equal
(i.e. 105 C1 ).
Obviously, by inducing greater temperature variations through injecting and extracting greater amounts
of heat there is greater stress variation in the tunnel lining. Furthermore, axial stress variations on the intrados
and the extrados of the lining are in antiphase (Figure 25). When heat is extracted, the intrados stress
increases while the extrados stress decreases, meaning
that the load on the tunnel lining increases. Conversely,
when heat is injected, extrados stresses increase while
the intrados stress decreases showing that the load on
the tunnel lining is relieved.
Finally, the mechanical implications are of the order
of magnitude of 0.5 MPa when no heat injection is
considered and are 1.5 MPa when heat is injected
(Figure 25).
The configurations applied to the cut and cover tunnel are also applied to the bored tunnel except for the
unsaturated cases. The different amounts of extracted
and injected heat are summarized in Table 4.
When heat injection is ignored, the heat production
of a single meter of anchor can be about five times
greater in the bored tunnel than in the cut and cover
tunnel (Table 4). Nevertheless, the gain in extracted
energy when injection is achieved during the hot season is not as efficient for the bored tunnel as for the cut
and cover tunnel. In fact, only 48.3% of the injected
heat is retrieved during the following extraction period
in the optimum configuration for the bored tunnel
while up to 83.8% can be retrieved with the cut and
cover tunnel (Table 5).
In conclusion, the bored tunnel offers the best
means of utilizing the underground heat, while heat
injection is not necessarily required to reach significant production. The maximum production of the
bored tunnel anchors in this study reached about
3.3 MWh per year and per meter of tunnel without
heat injection and 4.2 MWh per year and per meter of
tunnel with heat injection.
CONCLUSIONS
The first thermo-active test pile built on the EPFL

campus allowed the load-transfer mechanism under
temperature changes to be observed as well as the
influence of the pile head load on the pile confinement. It was observed that the free thermal strain was
4.3 Thermo-mechanical analysis of the bored

tunnel
The variations of axial stress in the bored tunnel lining, on the intrados and extrados at the tunnel crown,
99
split into real thermal strains measured using embedded strain gauges while the remaining part was turned
into internal thermal stress. The mechanical state of the
pile after any temperature change is relevant for maintaining equilibrium between the thermal strains and
stresses and is strongly dependent on the surrounding
stratigraphy and building characteristics.
The group of test piles built close to the new
EPFL congress center gave preliminary insight into
the behaviour of a group of energy piles. It is now clear
that pile confinement below a raft depends on both the
pile load and pile position since the raft develops a
capping force as the pile expands or shrinks.
A design tool, Thermo-Pile, was developed based
on results obtained from the first experimental site
at EPFL as well as the Lambeth College test pile.
This tool was shown to be capable of reproducing
the stress and strain states of energy piles under given
temperature changes.
Next, the thermo-mechanical behaviour of soils was
assessed. Before understanding the thermal effects on
soils, heat propagation through them was investigated.
In order to retrieve the soil thermal diffusivity, periodic
pumping tests, as used in underground water investigations, were conducted on a Bioley silt sample in the
laboratory. The results suggest that the method used is
applicable to the current research but that the relatively
low value of soil thermal diffusivity (106 m2 /s)
induces errors.
A thermo-elasto-plastic model, ACMEG-T which
was developed based on the experimental results, is
described. The thermal effects on the soil response are
quantified as a variation of the preconsolidation stress,
which decreases with temperature.
Examples of thermo-hydro-mechanical analyses
are given. The behaviour of a piled raft is investigated.
Group effects are quantified by comparing the induced
stresses in piles when only one pile is heated and when
the whole foundation is heated. It is shown that heating the whole assembly of piles induces lower stresses
in the piles as the whole raft heaves. Conversely, heating a single pile maximizes its confinement by the
surrounding structure so that thermal stresses increase.
Finally, the sustainability of heat exploitation
through tunnel anchors is assessed. Two different types
of shallow tunnel structures were tested as a heat
source for the neighbouring buildings: a bored tunnel and a cut and cover tunnel. The efficiency of heat
injection during the hot season was addressed under
different service conditions. It is shown that heat injection increases the production of heat but that it is not
as efficient in the bored tunnel which benefits from
good natural heat recharge. Conversely, heat injection
is important for the cut and cover tunnel as it ensures
the sustainability of the seasonal heat storage.
The variety of the recent developments in research
about energy geostructures cater to every need, from
the engineer tasked with designing one, to future
developments in the understanding of long-term challenges. The overview proposed in this article also
highlights the maturity of the topic, making it suitable
for going forward with design regulations that would

help promote the technology.
ACKNOWLEDGEMENTS
The authors thank EOS Holding for its financial supports through the project GRETEL for the numerical
analyses of piled rafts with energy piles and the in situ
measurements, respectively.
The support of the Swiss Federal Office of Energy
(OFEN) was much appreciated during the first test
(project N 25483) and is still appreciated for the
ongoing in situ campaigns on the new site.
The investigations on heat exchanger anchors were
carried out in the framework of a preliminary study
dealing with the feasibility of the concept (FGU
2009/002), mandated by The Swiss Tunnelling Society
(FGU).
REFERENCES
Adam D. (2008). Presentation Effizienzsteigerung durch
Nutzung der Bodenspeicherung, Ringvorlesung
kologie TU Wien.
Adam D. and Markiewicz R. (2009). Energy from earthcoupled structures, foundations, tunnels and sewers.
Gotechnique 59(3): 229236.
Amar S., Clarke B. G. F., Gambin M. P. and OrrT. L. L. (1991).
The application of pressuremeter test results to foundation desing in Europe, Part 1: Predrilled pressuremeters
selt-boring pressurementers., European Soil Mechanics
and Foundation Engineering, European Regional Technical Committee N . 4. A. A. Balkema, Rotterdam, The
Netherlands.
Amatya B., Soga K., Bourne-Webb P. J., Amis T. and Laloui
L. (2012). Thermo-mechanical Behaviour of Energy Piles.
Bandos T. V., Montero ., Fernndez E., Santander J. L. G.,
Isidro J. M., Prez J., Crdoba P. J. F. d. and Urchuegua
J. F. (2009). Finite line-source model for borehole heat
exchangers: effect of vertical temperature variations.
Geothermics 38(2): 263270.
BaudoinA., 1988. Stockage intersaisonnier de chaleur dans le
sol par batterie dchangeurs baionnette verticaux: modle de prdimensionnement. Ph.D Thesis, Universit de
Reims, Reims, France.
Boudali M., Leroueil S. and Sinivasa Murthy B. R. (1994).
Viscous behaviour of natural clays. 13th International
Conference on Soil Mechanics and Foundation Engineering, New Delhi, India.
Bourne-Webb P. J., Amatya B., Soga K., Amis T., Davidson
C. and Payne P. (2009). Energy pile test at Lambeth College, London: geotechnical and thermodynamic aspects
of pile response to heat cycles. Gotechnique 59(3):
237248.
Brandl H. (2006). Energy foundations and other thermoactive ground structures. Geotechnique 56(2): 81122.
Brandl H., Adam D. and Markiewicz R. (2006). Energy
geocomposites for tunnels. 8th International Geosynthetics Conference, Rotterdam, The Netherlands, Millpress
Science Publishers.
Campanella R. G. and Mitchell J. K. (1968). Influence
of temperature variations on soil behavior. Journal of
100
soil Mechanics & Foundations Division ASCE 94(SM3):

709734.
Carslaw H. S. and Jaeger J. C. (1947). Conduction of Heat in
Solids. Oxford University Press, New York, NY, USA.
Cekerevac C. and Laloui L. (2004). Experimental study
of thermal effects on the mechanical behaviour of a
clay. International Journal for Numerical and Analytical
Methods in Geomechanics 28(3): 209228.
Charlier R., 1987. Approche unifie de quelques problmes
non linaires de mcanique des milieux continus par la
mthode des lments finis. Ph.D. Thesis, Universit de
Lige, Lige, Belgium.
Choi J. C., Lee S. R. and Lee D. S. (2011). Numerical simulation of vertical ground heat exchangers: Intermittent
operation in unsaturated soil conditions. Computers and
Geotechnics 38(8): 949958.
Collin F., Li X., Radu J. P. and Charlier R. (2002). Thermohydro-mechanical coupling in clay barriers. Engineering
Geology 64: 179193.
Collin F., 2003. Couplages thermo-hydro-mcaniques dans
les sols et les roches tendres partiellement saturs. Ph.D.
Thesis, Facult des Sciences Appliques, Universit de
Lige, Lige, Belgique.
Coyle H. M. and Reese L. C. (1966). Load transfer for axially
loaded piles in clay. Journal of the Soil Mechanics and
Foundations Division 92(2): 126.
Cui P., Li X., Man Y. and Fang Z. (2011). Heat transfer analysis of pile geothermal heat exchangers with spiral coils.
Applied Energy 88(11): 41134119.
Diao N. R., Zeng H. Y. and Fang Z. H. (2004). Improvement in modeling of heat transfer in vertical ground heat
exchangers. HVAC & R Research 10(4): 459470.
Dupray F., Laloui L. and Kazangba A. (2013). Numerical analysis of seasonal heat storage in an energy pile
foundation. Computers and Geotechnics (in revision).
Eskilson P., 1987. Thermal Analysis of Heat Extraction Boreholes. Ph.D Thesis, Department of Mathematical Physics,
University of Lund, Lund, Sweden.
Fleureau J. M., 1979. Influence dun champ thermique ou
lectirque sur les phnomnes dinteraction solide-liquide
dans les milieux poreux. Ph.D. Thesis, Ecole Central de
Paris, Paris, France.
Forbes J. D. (1849). Account of some experiments on the temperature of the Earth at different depths, and in different
soils, near Edinburgh. Transactions of the Royal Society
of Edinburgh 10: 189236.
Frank R. and Zhao S. R. (1982). Estimation par les paramtres
pressiomtriques de lenfoncement sous charge axiale de
pieux fors dans des sols fins. Bull Liaison Lab Ponts
Chausses 119: 1724.
Franzius J. N. and Pralle N. (2011). Turning segmental tunnels into sources of renewable energy. Proceedings of the
Institution of Civil Engineers-Civil Engineering 164(1):
3540.
Frodl S., Franzius J. N. and Bartl T. (2010). Design and
construction of the tunnel geothermal system in Jenbach.
Geomechanichs and Tunnelling 3(5): 658668.
Hermansson ., Charlier R., Collin F., Erlingsson S., Laloui
L. and Sren M., Sr. (2009). Heat Transfer in Soils. Water
in Road Structures. D. Andrew, Springer Netherlands. 5:
6979.
Hinkel K. M. (1997). Estimating seasonal values of thermal
diffusivity in thawed and frozen soils using temperature
time series. Cold Regions Science and Technology 26(1):
115.
Hoekstra P., Delaney A. and Atkins R. (1973). Measuring the
thermal properties of cylindrical specimens by the use of
sinusoidal temperature waves., CRREL Technical Report

244. AD 770425.
Hueckel T. and Pelligrini R. (1989). Modeling of thermal failure of saturated clays. International Symposium on Numerical Models in Geomechanics. NUMOG:
8190.
Hueckel T., Franois B. and Laloui L. (2011). Temperaturedepedent internal friction of clay in a cylindrical heat
source problem. Gotechnique 61(10): 831844.
Hujeux J. C., 1979. Calcul nemrique de problmes de consolidation lastoplastique. Ph.D. Thesis, Ecole Centrale de
Paris, Paris, France.
Ingersoll L. R. and Plass H. J. (1948). Theory of the ground
pipe heat source for the heat pump. Heating, Piping and
Air Conditioning 20(7): 119122.
Katzenbach R., Ramm H. and Waberseck T. (2009). Recent
developments in foundation and geothermal engineering.
2nd International Conference on New Developments in
Soil Mechanics and Geotechnical Engineering. Near East
University, Nicosia, North Cyprus.
Knellwolf C., Pron H. and Laloui L. (2011). Geotechnical
analysis of heat exchanger piles. Journal of Geotechnical and Geoenvironmental Engineering 137(10):
890902.
Laloui L., 1993. Modlisation du comportement thermohydro-mcanique des milieux poreux anlastiques. Ph.D
Thesis, Ecole Centrale de Paris, Paris, France.
Laloui L. and Cekerevac C. (2003). Thermo-plasticity of
clays: An isotropic yield mechanism. Computers and
Laloui L., Moreni M. and Vulliet L. (2003). Comportement
dun pieu bi-fonction, fondation et changeur de chaleur.
Laloui L. and Cekerevac C. (2008). Non-isothermal plasticiy
model for cyclic behaviour of soils. International Journal
for Numerical and Analytical Methods in Geomechanics
32: 437460.
Laloui L. and Franois B. (2009). ACMEG-T: Soil thermoplasticity model. Journal of Engineering Mechanics
135(9): 932944.
Lang H. J. and Huder J. (1978). Bodenmechanik und Grundbau: Das Verhalten von Bden und die wichtigsten
grundbaulichen Konzepte.
Legrand J., Millan A. and Renault J. Rgles techniques de
conception et de calcul des fondations des ouvrages de
gnie civil., Fascicule N 62, Titre V, M. d. l. . d. l. e. d.
transports, France, 1993.
Lund P. D. and stman M. B. (1985). A numerical model for
seasonal storage of solar heat in the ground by vertical
pipes. Solar Energy 34(45): 351366.
Mattsson N., Steinmann G. and Laloui L. (2008). Advanced
compact device for the in situ determination of geothermal characteristics of soils. Energy and Buildings 40(7):
13441352.
Mogensen P. (1983). Fluid to duct wall heat transfer in duct
system heat storages. International Conference of Subsurface Heat Storage in Theory and Pratice, Swedish
Council for Building Research.
Pahud D. (2002). Geothermal energy and heat storage,
SUPSI Laboratorio di Energia, Ecologia ed Economia,
Lugano, Switzerland.
Peron H., Hueckel T., Laloui L. and Hu L. B. (2009). Fundamentals of desiccation cracking of fine-grained soils:
experimental characterisation and mechanisms identification. Canadian Geotechnical Journal 46(10): 11771201.
Pron H., Knellwolf C. and Laloui L. (2011). A method
for the geotechnical design of heat exchanger piles.
101
Geo-Frontiers 2011 Conference, ASCE, Dallas, Texas,

USA.
Philippe M., Bernier M. and Marchio D. (2009). Validity
ranges of three analytical solutions to heat transfer in the
vicinity of single boreholes. Geothermics 38(4): 407413.
Plum L. and Esrig M. I. (1969). Some temperature effects
on soil compressibility and pore water pressure, Highway
Research Board, Special report, Report 103. Washington,
USA.
Prakoso W. A. and Kulhawy F. H. (2001). Cointribution to
piled raft foundation design. Journal of Geotechnical and
Renner J. and Messar M. (2006). Periodic pumping tests.

Geophysical Journal International 167: 479493.
Scott R. F. (1964). Heat exchange at the ground surface.
Cold Regions Science and Engineering Monograph II-D1
(AD 697136).
SIA-267 (2003): Gotechnique. Socit suisse des ingnieurs
et des architctes.
SIA-D0-190 (2005): Utilisation de la chaleur du sol par des
ouvrages de fondation et de soutnement en bton. Socit
suisse des ingnieurs et des architctes.
102

Bio-geo-chemical processes for improvement of soil engineering properties

with focus on microbially induced calcite precipitation
J.D. DeJong
Department of Civil and Environmental Engineering, University of California, One Shields Avenue, Davis, CA, USA
B.C. Martinez
Geosyntec Consultants Incorporated, Oakland, CA, USA
T.R. Ginn
Department of Civil and Environmental Engineering, University of California, One Shields Avenue, Davis, CA, USA
D.C. Nelson
Department of Microbiology, University of California, One Shields Avenue, Davis, CA, USA
ABSTRACT: The improvement of engineering soil properties through the harnessing of natural bio-geochemical processes is a novel proposition, and today represents a field within geotechnical engineering that is
rapidly expanding. Realizing the potential of this new perspective where soil is recognized as a host environment
for biological processes requires a multi-disciplinary approach that embraces biology and geochemistry. This
paper first provides a brief overview of this rapidly developing field and then examines recent experimental and
modeling work on microbial induced calcite precipitation in depth. The field of bio-mediated soil improvement
has developed primarily over the last decade, with research spanning the use of invertebrates to unicellular
organisms, mechanical reworking/recompaction to various biogeochemical processes including mineral precipitation, gas generation, biofilm formation and biopolymer generation. Geotechnical applications currently being
explored range from cementation of sands to improve liquefaction resistance, to soil erosion control, to remediation of soil and groundwater impacted by heavy metals. Microbially induced calcite precipitation (MICP) has
been the primary biogeochemical process pursued to date, with research performed by several research groups,
studies across a broad range of length scales (microscopic to field applications), and approaches ranging from
experimental to numerical modeling. Due to the geotechnical application focus emphasis has been primarily
placed on the impact of calcite precipitation on the engineering properties; less focus has been placed on a
balanced, holistic examination of the biogeochmical processes in parallel with the resulting calcite formulation.
Herein experimental and numerical modeling results that considers all of these aspects are presented to highlight
the improved understanding that can be gained from such an approach.
BIO-MEDIATED SOIL IMPROVEMENT
The use of biogeochemical processes to improve the

engineering properties of soil physical properties
(density, gradation, porosity, saturation), conduction
properties (hydraulic, electrical, thermal), mechanical properties (stiffness, dilation, compressibility,
swell/shrink, cohesion, cementation, friction angle,
erodability, and soil-water characteristic curve), and
chemical composition (buffering, reactivity, cation
exchange capacity) (DeJong et al. 2013) has developed rapidly over the past decade to a point where more
than one hundred researchers in 20+ countries are
active in the field today. Research on the topic is well
funded and identified as a research priority by national
science foundations. The emergence of the discipline stemmed largely from the review by Mitchell
and Santamarina (2005), the US National Research
Council (2006) report identifying the field as a priority for the 21st century, and several recent workshops
(e.g. DeJong et al. 2007) and conferences gathering
active researchers. Numerous papers have been published providing broad assessments of the field (Ivanov
& Chu 2008, Kavazanjian & Karatas 2008, Seagren &
Aydilek 2010, DeJong et al. 2011, Hata et al. 2011),
with the most recent (DeJong et al. 2013) summarizing the results of a recent international workshop that
assessed progress to date, identified research priorities, and forged collaborations. A schematic highlighting some of the envisioned applications is shown in
Figure 1, with more detailed assessment of specific
applications provided in DeJong et al. (2013).
The geotechnical community has long ignored biological presence in soils, with engineers often assuming that both natural soil deposits as well as graded
fills were inert, free of organisms, and in a stable state.
103
Figure 1. Schematic exemplifying possible field applications for bio-mediated soil improvement.
In fact the range across which biology can modify the

engineering properties of soil varies greatly. One perspective for understanding this is by considering the
length scale of organism, both in absolute dimension
and relative to the particle size.
Multi-cellular organisms, ranging from plant roots
down to insects and invertebrates (e.g. ants, worms),
alter soils through both mechanical and biological
processes. For example, ants are effective at soil
grading, densification, and creating preferential flow
paths (macro-pores); they also adapt and optimize
their efforts considering capillarity forces at particle
contacts (Espinosa & Santamarina 2010). Similarly,
mucous excretion from worms can strengthen soil
along tunneling paths and help bind (geotechnically
strong) fecal pellets to such an extent that the cone
penetration test (CPT) measures the strength increase
(Kuo and Bolton 2013).
Unicellular microbial organisms in soil consist primarily of bacteria and archaea (see Woese et al. 1990
for definitions of terms), which typically range in
diameter from 0.5 to 3 m. These are present in soil
either through entrapment during deposition (RebataLanda & Santamarina 2006) or migration through
pore space via hydraulic flow transport or self-motility
(Mitchell & Santamarina 2005, DeJong et al. 2010).
Unicellular activity, in general, affects soil properties indirectly through local geochemical processes.
Microbe activity creates micro-niche conditions surrounding individual cells that regulate when, where,
and at what rate geochemical processes occur. The
ability of microbes to regulate processes often stems
from the unicells containing the enzyme(s) critical to
the geochemical reaction.
Development of a bio-mediated soil improvement
technique requires an application strategy. The two primary strategies, bio-augmentation (where the required
microbes are injected into the soil) and bio-stimulation
(where natural microbes are stimulated), build on bioremediation techniques developed over the last 30+
years. While the former has been the primary strategy
used to date in exploring geotechnical applications,

the geo-environmental field is increasingly using biostimulation. Bio-augmentation is generally considered
less favorable than bio-stimulation due to the introduction of exogenous (non-native) microbes, increased
costs, difficulty of uniform application in the subsurface, etc. Bio-stimulation is generally preferable
due to the stimulation and growth of native microbes,
which are already adapted to the subsurface environment. However, many challenges exist in applying
bio-stimulation, including obtaining uniform treatment across a site and accommodating the increased
time associated with stimulation and growth.
Geochemical processes regulated through biostimulation or bio-augmentation often have multiple
products. The primary product is typically the one
required to realize the desired outcome (e.g. calcite
precipitation to bind soil particle together). The additional by-product(s) generated by the geochemical
process (e.g. ammonium ions) must also be addressed
(DeJong et al. 2013).
Bio-mediated geochemical processes that have
been identified and been the focus of research to date
include (DeJong et al. 2010, 2013):
104
bio-mineralization processes that precipitate inorganic solids (including microbially induced calcite
precipitation, or MICP), resulting in a reduction in
pore space, brittle cementation at particle contacts,
increased fines in the pore space, and increased
stiffness.
bio-film formation and production of other extracellular polymeric substances (EPS) that consist of
organic solids which occupy a portion of the pore
space with a soft, ductile, elastomeric-like material that reduces pore size, reduces rearrangement
of particles during soil deformation, and increases
ductility.
bio-gas generation from denitrification or other
biogeochemical processes that may enable longterm reduction in the degree of saturation of a
soil, thereby increasing pore space compressibility

and reducing excess pore pressure build-up during
cyclic loading.
Other approaches identified but less developed
include algal and fungal growth for near-surface soil
stabilization, bacteria and worms for methane oxidation, bio-polymers for drilling applications, organic
slurries for hydraulic control, and silicate precipitation. Research to date has investigated many of
these potential approaches, with a significant portion
of activity focused on bio-mediated cementation via
microbially induced calcite precipitation (MICP).
2
MICROBIALLY INDUCED CALCITE

PRECIPITATION
Microbial induced carbonate precipitation (MICP) is

one of the more promising bio-mediated soil improvement techniques being investigated today (DeJong
et al. 2010, 2011). MICP is a cementation process that harnesses natural subsurface bacteria using
urea hydrolysis (ureolysis) to induce calcite precipitation at particle-particle contacts in soil environments (DeJong et al. 2006, Stocks-Fisher et al. 1999,
Martinez and DeJong 2009). MICP has been extensively evaluated for geotechnical ground improvement
including increasing undrained shear strength (DeJong
et al. 2006, 2009), non-destructive geophysical monitoring (Weil et al. 2012, Al Qabany et al. 2011), and
solidification at laboratory and field scales (DeJong
et al. 2010a, Harkes et al. 2008, van Paasen et al.
2009, Whiffin 2004, Whiffin et al. 2007, Montoya et al.
2013). MICP has also been identified as an engineering solution for a wide-range of disciplines including
mineral plugging (Ferris et al. 1996), environmental remediation of heavy metals (Fujita et al. 2004),
structural concrete repair (Ramakrishnan et al. 1998,
Ramachandran et al. 2001, van Tittelboom et al. 2010),
and carbon sequestration (Renforth et al. 2009).
For practical application in soil, MICP requires the
existence of ureolytic bacteria and urea- and calciumrich solutions to drive the MICP bio-geo-chemical
reaction. However, most of this research has been
performed at the laboratory scale; few studies have
been performed at the meter scale (e.g. Whiffin et al.
2004, van Paassen et al. 2009). As a result, one of
the universal research priorities is to better understand
and control the MICP bio-geo-chemical process under
conditions that increasingly resemble natural systems.
Modeling of MICP is necessary to complement
and extend laboratory results to a more generalized
framework where the controlling parameters can be
investigated and where prediction of performance
for field scenarios can be considered. To date available mathematical modeling studies have addressed
particular aspects of MICP including mechanical,
hydro-geological, biological, and chemical processes.
Many of the early studies focused particular attention
to capturing the appropriate kinetic rates of ureolysis induced by single species and native microbes
(Ferris et al. 2004, Hammes et al. 2003, Mitchell

and Ferris 2005, Mitchell and Ferris 2006a,b, Fujita
et al. 2000, 2004, 2008) as well as cell-free enzymes
(Fidaleo and Levecchia 2003), and calcite precipitation (Nillson and Sternbeck 1999). Later studies evaluated the development of multi-component reactive
transport by coupling similar ureolysis rate expressions to calcite precipitation kinetics and fluid transport in one- and two-dimensions (Dupraz et al. 2009,
van Paasen et al. 2008, Booster et al. 2008, Harkes
et al. 2010, Barkouki et al. 2011). A few studies
have also examined the effects of reactive transport to
mechanical changes in soil studies through finite element analysis (Fauriel and Laloui 2012, Fauriel 2012).
Most of the models have been able to capture reaction rates for a variety of biological and chemical
conditions within small-scale batch experiments via
kinetic expressions accounting for chemical variants.
There is a need to improve the quantitative modeling of
MICP toward the coupling of constitutive theory from
every involved discipline, and to improve the individual components for macro-scale reactive transport in
porous media.
The study presented herein provides an overview of
an integrated experimental and modeling research program that evaluated controls on MICP for achieving
spatial uniformity of cementation in one-dimensional
flow using half meter scale column experiments. The
experimental study addressed several factors for controlling MICP, including the identification of urease
activity, achieving and monitoring uniform spatial
nutrient distributions, monitoring bio-geo-chemical
changes during treatment, and monitoring geotechnical improvements. The numerical work presented
herein builds from Barkouki et al. (2011), implementing MICP into a bio-geo-chemical reactive transport
code (TOUGHREACT; Xu et al. 2006) and calibrating unknown parameters (UCODE-2005; Poeter et al.
2005) to verify the model and develop a procedure for
predicting MICP in one-dimensional flow.
3
MICP EXPERIMENTAL PROGRAM
3.1 Experimental materials & methods

Acrylic cylindrical columns (5 cm 50 cm) having
four instrumentation locations with piezo-electric bender elements and sampling septa ports were constructed (Figure 2). Piezo-electric bender elements
(Montoya et al. 2012) for measuring shear wave velocity were placed in silicon molds and sealed at four ports
along the column length with vacuum grease. Shear
wave velocity measurements capture the cementation
level of MICP non-destructively (DeJong et al. 2006,
2009, 2010, Weil et al. 2011, Al Qabany et al. 2010).
Septa ports made from Swagelok fittings were also
fitted to the column for obtaining aqueous samples.
The columns were oriented vertically with acrylic
top and bottom caps sealed with O-rings. Soil filters
consisting of 2.5 cm layers of gravel (D50 = 5 mm)
and Ottawa 2030 sand (D50 = 0.72 mm) were placed
105
Figure 2. Schematic diagram of (a) half-meter column experiments, (b) picture of column, (c) microbial/optical density of
unattached microbes along column in Test 3A, and (d) temporal change in pH over the retention period following a ureolytic
potential injection in Test 3a.
Table 1.
Geotechnical parameters and improvements for each column experiment.
adjacent to the top and bottom caps to prevent soil

migration out of the columns. Ottawa 5070 sand
(D50 = 0.21 mm, emin = 0.55, emax = 0.87) was air
pluviated into the columns to a medium to dense condition (resulting in a DR ranging from 80%100% and
porosity, n, ranging from 0.34 to 0.37). The columns
were loaded vertically with a 100 kPa stress applied
to the top cap. Table 1 describes the setup conditions
for each column including soil height, relative density,
porosity, initial shear wave velocity distribution (Vs ),
and initial permeability (k).
Each sample was initially saturated before treatment began to ensure a relatively controlled flow field
during injection and monitoring. Pressure head was
continuously controlled at each end of the column to
ensure maximum saturation in various flow directions.
Top and bottom ports were fitted with interchangeable tubing and valves to maintain full control of water
conditions during the experiment.
A two-phase injection scheme using peristaltic
pumps was used. In the first (biological) phase, the
bacterium S. pasteurii was introduced into the column
by injection of the biological solution (Table 2)
into the columns from a single source at varying volumes and flow rates (Table 3). The bacteria were
then retained in the columns for a retention period
to allow attachment to particles (Table 3). In the second (cementation) phase, the cementation solution
(Table 2) was introduced into the columns from a single source at varying volumes and flow rates (Table
3). A third solution (the cementation solution but
without the calcium chloride, Table 3) was developed
to strictly measure ureolysis rates while monitoring pH increase, urea consumption, and ammonium
production. This amendment is called the ureolytic
potential amendment and is used to evaluate the spatial variances in ureolytic activity. This allows pH to
increase based entirely on the hydrolysis of urea for
simple lab measurements. Monitoring MICP included
biological, chemical, geo-physical, and geotechnical
analysis techniques before, during, and after treatment.
Measurements included optical density, pH, aqueous
calcium, aqueous ammonium, calcium carbonate content, shear wave velocity, and bulk permeability. Optical density measurements with a spectrophotometer
(600 nm wavelength) were performed during testing
106
Table 2.
Injection recipe.
Table 3. Treatment schemes applied to each column.
Table 4.
Sporoscarcina pasteurii amendment strategy and column-wise number balances.
to measure the temporal and spatial distributions of

bacteria suspended in the pore fluid in the columns.
Chemical analyses were performed during and after
treatment. Ammonium was analyzed on samples collected from port and effluent locations stored at 4 C
after collection using a flow injection method (Mansell
et al. 2000). Calcium was measured on samples stored
at 4 C after being collected and filtered from port
and effluent locations using an Agilent 7500i inductively coupled mass spectrometer (ICP-MS). Samples
were diluted to scale sample concentrations to values
between 0 and 100 mg/L calcium and referenced to
analytic grade calcium chloride. Calcium carbonate
was also measured on discrete column intervals after
test completion by mixing 0.5 N hydrochloric acid to
samples proportioned from the column, rinsing thoroughly, and measuring the mass difference before and
after. Shear wave velocity measurements were taken
using piezo-electric bender elements installed along
the column to non-destructively monitor the spatial
and temporal change in stiffness due to calcium carbonate precipitation during testing (Weil et al. 2010,
Al Qabany et al. 2010). Hydraulic conductivity was
also analyzed during the course of treatment using a
falling head test setup.
3.2 Test program

A set of experiments investigated a variety of injection
aspects that affect MICP reaction extents including
injection methods, sources, and recipes (Martinez
2012). Herein the results of one pair of experiments,
Tests 3A and 3B are presented (summarized in Tables
13). Test 3 examined the effect of cementation
amendment recipe on stopped-flow and continuousflow treatment schemes. An equimolar chemical formulation was designed to minimize the production of
ammonia during MICP (Table 2). Biological amendments were administered from the column tops at
10 mL/min followed by an 8 hour retention period,
designed to induce irreversible bacterial attachment
to surface. In the stopped-flow injection method,
cementation amendments (second treatment phase)
are administered by subsequent pulses (bottom to top)
at 10 mL/min for varying pore volumes (Table 3, 4)
followed by a retention period of approximately two
hours for each test. In comparison, the continuous
flow injection method involved continuously pumping
the cementation amendment (second treatment phase,
bottom to top) at 2.2 mL/min such that the net flux
(total concentration of treatment solution injected per
unit time) was equal in both columns.
107
Table 5.
3.3
Biogeochemical summary for each column.
Results & discussion
3.3.1 Microbial (urease) distribution

Controlling MICP cementation requires spatial knowledge of the microbial (e.g. urease) activity. Subsurface
microbial populations may be put in place by injection
(augmentation, e.g. DeJong et al. 2006), or stimulated by nutrient injection (e.g. Fujita et al. 2008).
Here, augmentation is assumed in 1D flow to create
an approximately exponential distribution of attached
microbes, consistent with a first order kinetically controlled rate of removal from the aqueous phase, e.g.,
colloid filtration (Yoon et al. 2006, Ginn et al. 2002).
Tables 4 and 5 summarize the S. pasteurii amendment strategy per experiment, and the overall cell
retentions in the entire column domain (assuming no
bacterium growth or death during the experiment).
Cell number densities were measured (via optical density) for all influent and effluent samples in order to
calculate the number retained in the column. Essentially all unattached microbes were flushed out of
the columns when the first cementation (or ureolytic
potential) amendment was injected.
Microbial activity was monitored indirectly by measuring urea and ammonium. Results from Test 3A
(Figure 2) provide the rates of pH increase at four
port locations after the ureolytic potential amendment
was administered. The concentration of unattached
microbes at the four port locations was measured
immediately after injection of S. pasteurii by extracting 0.5% of the pore volume per measurement (2%
total). The data indicate that lower concentrations of
attached microbes result in lower ureolysis rates.
Figure 3 presents normalized microbial profiles
for the stopped-flow and continuous flow conditions
after the biological amendment was administered. The
concentration of unattached bacteria decreases from
the injection source.
Figure 4 presents the breakthrough curves for Test
3A and 3B on samples collected in column effluent,
beginning with the start of second phase cementation amendments (immediately after data shown in
Figure 3). For stopped-flow injection, the unattached
microbes are completely flushed out of the columns
within the first hour of the first cementation amendment (Figure 4a). For continuous flow unattached
bacteria are flushed out of the column within three
hours, with the longer time period being due to a
3x slower injection rate (Figure 4b). Calculation of
the bulk concentration of unattached bacteria flushed
out of the columns (Table 4) reveals that the average
of bacteria retained is 7080% of the initial injected
concentrations.
Figure 3. Unattached microbe distributions for (a) stopped

flow Test 3A and (b) continuous flow Test 3B. Grey arrows
indicate the biological amendment flow direction. Points for
each profile indicated at 0 and 50 cm represent the initial
concentration of microbes injected and the last sampled fluid
at the end of the biological injection. An open triangle is
marked at 0 cm because no measurements were taken on
effluent samples during the biological injection of Test 3.
The results imply kinetically controlled attachment

of microbes with slower and limited detachment rate
depend on nutrient availability when biological augmentation is administered (Ginn et al. 2002). The
spatial distribution of aqueous microbes qualitatively
relates to a spatial distribution of attached microbes
given that the concentration of attached microbes
does not change during the course of treatment. The
data support this hypothesis by showing a relationship between measurements of unattached microbes
to ureolysis rates (increases in pH) produced by
attached microbe populations (Figure 2) given that all
unattached bacteria have been previously flushed out
of the column.
3.3.2 Injected urea distribution & consumption
Achieving uniform cementation with MICP in situ
requires that urea and other nutrients are distributed
uniformly throughout the target treatment zone. Figure 5a shows the average normalized aqueous urea
distributions of the stopped-flow column (Test 3A)
at two selected time intervals within the retention
phase of stopped-flow injection. Urea is consumed
during the retention phase proportionally to inferred
attached microbial distributions (reflected in Figure
3). The influence of spatially varying bacteria distributions is apparent in Test 3A. More urea is consumed
near the biological amendment source (column base)
where microbial concentrations are higher than further
108
Figure 4. Microbial breakthrough curves measured in column effluent for (a) stopped-flow Test 3A and (b) continuous
flow Test 3B.
from the source regardless of flow direction during the

cementation phase.
Urea profiles for continuous flow treatment across
the column length and over the full treatment duration
are shown in Figure 5b. Urea profiles increase relatively uniformly across the length of column over 50
hours. The microbial distribution in Test 3B was highly
non-uniform and the cementation injection source was
opposite the biological injection source. In continuous
flow, microbes near the injection source degrade more
urea than those further from the source, precipitating
more calcium carbonate near the injection source.
Bulk ureolysis rates were computed using the urea
consumption profiles from stopped-flow and continuous flow tests. For stopped-flow column tests,
maximum ureolysis rates were calculated from the
average urea consumption profiles between times near
the beginning and end of retention as follows,
where
Figure 5. Urea consumption during (a) stopped-flow and

(b) continuous flow for Test 3. Normalized urea represents
sampled urea concentrations divided by injected urea concentration. The solid and dotted black arrows represent the
biological amendment and cementation amendment flow
directions, respectively, or indicate a trend.
where ru,avg , z, t1 , and t2 represent the volume averaged maximum urea hydrolysis rate expressed as the
molar concentration of urea per minute, the number of measurements along the depth of the column,
and measurement times, respectively. Table 5 summarizes the geochemical attributes associated with Test 3.
Ureolysis rates calculated for stopped-flow columns
are assumed to be the same for continuous flow
columns as the attached microbial profiles are similar.
The distribution of nutrients such as urea plays an
important role in achieving uniform calcium carbonate precipitation using MICP. Stopped-flow injection
allows initial uniform urea distribution, which then
is consumed according to the microbial distribution.
Continuous flow at an equivalent total flux results
109
The efficiency of calcium carbonate precipitation

was affected by a combination of ureolytic activity
and the amendment formulation. Calcium carbonate
precipitation efficiency was calculated based on the
total amount of calcium carbonate precipitated versus
the total amount of calcium injected as follows,
Figure 6. Effect of injection method on incremental changes

in biogeochemistry (a) stopped-flow Test 3A and (b) continuous flow Test 3B.
in high urea concentration near the injection source

and lower concentrations further away. Therefore,
stopped-flow is preferred over continuous flow to distribute nutrients evenly and hence precipitate calcium
carbonate more uniformly.
3.3.3
Biogeochmnical changes during treatment

intervals
The MICP reaction extent was monitored during treatment with chemical analyses. Mass balance on urea,
ammonium and calcium was tabulated (Table 5) to calculate the formation of by-products (e.g. ammonium)
and efficiency of processes associated with the two
main components of MICP; urea hydrolysis and calcium carbonate precipitation. Figure 6 presents typical
trends in aqueous urea, aqueous ammonium, aqueous, calcium, and pH for the injection method and
treatment formula used. The stopped-flow plots show
changes over an average retention period after the
cementation amendment while the continuous flow
plots show changes over 20 hours of cementation
amendment injection. For both recipes the urea is
consumed while ammonium is produced, followed by
steady decreases in calcium concentration as calcium
carbonate is precipitated.
Differences in ammonium production depend on
the microbial spatial variability. Ammonium concentration was calculated based on the urea consumption
and checked with selected ammonium analysis. The
1:1:1 [50 mM] formulation used represents the minimum ammonium production condition possible given
the ability to precipitate calcium carbonate.
where [CaCO3 (s)]T is equal to the total amount of

calcium carbonate measured after treatment over the
entire column and [CaCl 2 ] is the total amount of
calcium chloride injected into the column during
treatment. Precipitation efficiency was higher (43%
compared to 36%) for the continuous flow cases due to
more calcium carbonate precipitation occurring where
more urea consumption occurred (Table 5).
MICP was effectively optimized by reducing the
molar ratios of urea to calcium chloride (e.g. formula
1:1:1 [50 mM] used in Test 3). This formula does not
induce a change in the pH because these two processes are effectively balancing each other, lowering
the ureolysis rate and calcium carbonate precipitation efficiency. Reducing the molar ratios of urea to
calcium chloride in the formulation improves MICP
because it reduces the amount of ammonium produced during treatments while continuing to yield
precipitation at a stable background pH.
3.3.4 Changes to calcium carbonate concentration,
shear wave velocity & hydraulic conductivity
Monitoring the change in soil stiffness with shear
wave velocity measurements occurred at four bender element locations throughout treatment in all tests.
Figures 7a and 7b present typical shear wave velocity
time histories for Test 3A and 3B, respectively, at four
positions along the column length for stopped-flow
and continuous treatment. Final calcium carbonate
content was compared to the final shear wave velocity
profiles to confirm cementation gradients associated
with the test parameters examined. Figures 7c and 7d
show the calcium carbonate distribution for Test 3A
and 3B in stopped-flow and continuous flow, respectively. Calcium carbonate profiles generally follow
the trend of microbe distributions along the column
length. The 1:1:1 [50 mM] formulation under continuous flow injection (Test 3B) proved to be most sensitive
to the initial microbe distribution.
For practical purposes correlations of shear wave
velocity with physical and chemical properties such
as total calcium carbonate precipitated and hydraulic
conductivity was investigated. Figure 8a presents the
final local calcium carbonate content versus the final
shear wave velocity at every measurement location in
every experiment (Test 3A and 3B presented herein
as well as the additional tests described in Martinez
110
Figure 7. Effect of injection method on MICP improvement (a,b) Shear wave velocity time histories for stopped-flow Test
3A and continuous flow Test 3B, respectively and (c,d) end-of-test calcium carbonate profiles for stopped flow and continuous
flow, respectively. The vertical axis represents the distance from source of biological amendment (BA) injection in all cases.
2012). A linear trend exists across the range of measurements obtained (which is generally consistent with
Al-Qabany et al. 2011 and Weil et al. 2012).
Hydraulic conductivity was monitored before, during, and after treatment to investigate reduction due
to calcium carbonate precipitation. Figure 8b presents
a compilation of all bulk hydraulic conductivity measurements obtained versus the maximum shear wave
velocity in the column at the same time. The hydraulic
conductivity was significantly reduced with the continuous flow technique (open symbols) relative to the
stopped-flow technique (closed symbols).
Bulk hydraulic conductivity decreases during
MICP improvement as calcium carbonate precipitates
due to void space reduction. The bulk hydraulic conductivity in 1-D columns is controlled by the zone
where the highest calcium carbonate precipitation
occurs. Figure 8c presents all final bulk hydraulic
conductivity measurements versus the maximum
calcium carbonate concentration in the column for

all tests.
4
4.1
MICP NUMERICAL MODEL

Overview of numerical model
The MICP model presented in this study was built

in the reactive transport code, TOUGHREACT (Xu
et al. 2006) and setup for calibration utilizing
a model-independent calibration program UCODE2005 (Poeter et al. 2005) (Barkouki et al. 2011).
TOUGHREACT is a reactive transport code developed at the Lawrence Berkeley National Labs (LBNL)
that is capable of modeling non-isothermal, multiphase, three-dimensional fluid flow in heterogeneous and variably saturated porous and fractured
media coupled with multi-component transport with
equilibrium and kinetic aqueous speciation, mineral
111
ensoformis) derived enzymatic ureolysis (Fidaleo and

Lavecchia 2003) and calcite precipitation (derived in
the TOUGHREACT framework).
The ureolysis reaction and reaction rate expression are:
For ureolysis, the variables ru , max , ZH, S, P, KM ,

and KP represent the rate of ureolysis (mol/L/sec), ureolysis rate constant (1/sec), the active urease enzyme
concentration (mol/L), concentrations of substrate
(mol/L urea) and product inhibitor (mol/L ammonium), half-saturation constant for substrate (mol/L),
and inhibition constant for the product inhibitor
(mol/L), respectively. K1 and K2 (mol/L) are conventional speciation constants for the protonation/
deprotonation reactions, which are an adaptation of
the pH driven inhibition term described by Fidaleo
and Levecchia (2003) model to fit the TOUGHREACT
input structure. Values used for half-saturation, inhibition, and speciation constants follow Fidaleo and
Levecchia (2003), leaving the urease enzyme concentration the only unknown parameter in the ureolysis
expression.
The calcite precipitation reaction and rate expression are:
Figure 8. Relationships between (a) shear wave velocity and

hydraulic conductivity, (b) shear wave velocity and carbonate
precipitation, and (c) hydraulic conductivity and carbonate
precipitation.
precipitation/dissolution, gas diffusion/exsolution,

cation exchange, linear sorption/desorption, and
radioactive decay (Xu et al. 2006; Xu 2008a; Xu et al.
2008b). Reactive flow and transport are described
through continuity relationships to describe flow and
transport (e.g. Advection-Dispersion Equation).
Mixed equilibrium-kinetics chemical reactions are
solved from mass balance on a basis of components
following Parkhurst et al. (1980) and Reed (1982).
Equilibrium reactions follow mass action laws associated with aqueous complexation reactions occurring
at local equilibrium and kinetically controlled reactions are developed from relevant expressions for
enzyme kinetics and precipitation. The minteq.v4.dat
thermodynamic database was adopted and reformatted from PHREEQC (Parkhurst and Appelo 1999).
MICP is simulated in TOUGHREACT by implementing the reaction for the hydrolysis of urea (or ureolysis)
through kinetic expressions for jack bean (Canavalia
For calcite precipitation, rc , kc , Ac , and Ksp are

the rate of precipitation (mol/kgH 2O /sec), kinetic rate
constant (mol/m2 /sec), specific mineral reactive surface area (m2 /kgH 2O ), and solubility product constant
(mol/L), respectively. In order to have a positive reaction rate for mineral precipitation, the specific mineral
reactive surface area must be initially set to a positive number (Ac,min > 0). Ac,min is the only unknown
parameter in the expression.
An improved spatially distributed expression for
the unknown parameter for total immobile urease
enzyme (ZHt ) was adopted to represent bacterial
augmentation.
The parameter, ZH t , represents the total urease

enzyme present including the protonated and deprotonated forms (Z and ZH +
2 ). The expression represents a
112
Figure 9. Residual analysis for the calibration of stopped-flow Test 3A including observed versus simulated values of
(a) calcite, (b) pH, (c) calcium, (d) urea, (e) ammonium.
gamma distribution of enzyme as a function of length

simulating bio-colloid advection and filtration theory while allowing for flexibility in the shape of a
semi-log-linear distribution accounting for instances
of microbial detachment. The expression consists of
two scaling parameters and a shape parameter, a, d,
and s respectively.
The unknown parameters to be calibrated by coupling UCODE-2005 to TOUGHREACT consist of
parameters associated with ureolysis rate (a, d,and s)
and the initial reactive surface are for mineral precipitation, Ac,min . UCODE-2005 is a model-independent
parameter calibration program that performs a modified Gauss-Newton search to optimize parameter
values of a given model in order to fit observed data
and provides a suite of sensitivity analysis statistics to
evaluate model calibration and robustness (Poeter et al.
2005, Hill and Tideman 2007). An objective function is
defined and minimized until the chosen parameter(s)
yield residuals within defined constraints.
The variables S, b, y, y, , and e represent the objective function value, the vector of unknown parameters,
the vector of observations, the vector of simulated values, the vector of observation weights, and the vector
of residuals, respectively. UCODE-2005 is coupled to
TOUGHREACT such that UCODE-2005 iteratively
perturbs parameter values in search of lower residuals until convergence criteria are met. Guidelines for
effective model calibration are described by Hill and

Tiedeman (2007).
Test 3A and 3B were setup independently for calibrating the four parameters associated with the reaction rates for ureolysis and calcite precipitation. The
urease enzyme distribution was constrained to allow an
exponential distribution (s = 1) or a gamma distribution (s > 1), where a and d are allowed to range from
102 to 102 . Based on Fujita et al. (2003), Whiffin et al.
(2007), and Ferris et al. (2003) it was determined that
the initial mineral reactive surface area could be constrained within one order of magnitude of the estimated
bulk average ureolysis rate experimentally measured.
4.2 Numerical model calibration to
experimental data
For Tests 3A (stopped-flow) and 3B (continuous) the
set of tests, biological treatment was administered
from the top-down (Table 3) and cementation treatments were administered from bottom-up. For calibration, the inversion converges on parameter values
but the solution is determined to be constrained on
the initial mineral reactive surface area, Ac,min . The
calibrated urease enzyme distributions follow gamma
distributions similar in shape and magnitude with maxima located near the injection source (top of column)
of bacteria.
Residual analysis including observed versus simulated values for calcite, pH, aqueous ammonium,
calcium, and urea are shown in Figure 9 for column
Test 3A. The calibrated model tends to overestimate
113
relatively low values while underestimating relatively

high values of calcite (Fig. 9.2a), pH (Fig. 9.2b),
ammonium (Fig. 9.2c), and calcium (Fig. 9.2d). Urea
data tend to be slightly overestimated overall (Figure 9.2e). The difference reflects the observation
weight assignments favoring calcite (100% confidence) observations while devaluing calcium (60%
confidence) and pH data (50% confidence). In addition to the weighting scheme, the models enzymatic
representation of ureolysis is expected to produce
higher rates of urea hydrolysis than shown by the
Figure 10. Biogeochemical changes during the retention

of a stopped-flow Test 3A injection cycle; experimental and
calibrated numerical results.
experimental data because of other ongoing microbial

processing interfering with direct enzyme processing. This is likely to cause conditions of higher pH,
larger ammonium production, quicker reduction in
urea, and faster dissolved calcium depletion. Overall, the residual analysis reflects the assigned weights
based on qualitative assessment of data types, inherent
discrepancies between the model and the experiments,
and inaccuracies in the observations. Residual analysis of column Test 3B data showed similar trends,
with favored calcite residuals between observed and
simulated values, and the overestimation of aqueous
chemistry data.
Modeling results for aqueous chemistry time
courses during a selected retention period and location of a stopped-flow cycle for Test 3A are shown
in Figure 10. Total ammonium, urea, and calcium
are normalized by their respective initially injected
concentrations during treatment. During a single retention period of a stopped-flow cycle, it is expected
that the column reacts similar to a batch experiment (no flow condition) where urea is fully depleted
through hydrolysis, producing twice the amount of
ammonium beyond background concentrations, while
subsequently decreasing calcium as calcite precipitates. Urea hydrolysis and calcite precipitation prove
to be competing processes neutralizing pH at equal
concentration of urea and calcium. The calibration
shows reasonable agreement between simulated aqueous chemistry and experimental data albeit for a slight
overestimation in pH.
Modeling results for the final calcite distribution upon completion of treatment compared to
experimental data and calibrated enzyme profiles
are presented for both stopped-flow and continuous flow injection schemes (Test 3A and 3B) in
Figure 11. Both treatment schemes (stepped-flow
Figure 11. Calibration modeling results for(a) stopped-flow Test 3A and (b) continuous flow Test 3B.
114
versus continuous flow) produced similar final calcite profiles primarily dependent on the distribution
of bacteria in the columns. Both calibrations produce similar urease enzyme profiles in shape and
magnitude, indicating similar quantity of injected
bacteria and their attachment behavior as expected.
The effective enzyme gamma distribution produces
calibrated calcite profiles reasonably similar to the
experimental data being calibrated. It is not surprising
to achieve good calibration on the final calcite profiles
as the calcite data received a relatively larger weight
assignment compared to the aqueous chemistry data.
The enzymatic representation of ureolysis for MICP
implemented in TOUGHREACT and calibrated with
UCODE-2005 captures the general trends and final
calcite distributions although discrepancies exist as
shown in the residual analysis for the two independent
calibrations.
MICP model calibration requires careful selection
of constraints on parameters associated with reaction
rates of ureolysis and calcite precipitation. Constraining the initial mineral reactive surface area Ac,min
parameter for the calibration allowed for unique convergence of the urease enzyme distribution in all cases
indicating promise for predicting MICP described in
the following section.
By sequentially coupling UCODE-2005 to
TOUGHREACT, the model can predict MICP given
input of the spatially distributed immobile urease activity for the half-meter column experiments
described previously. The prediction procedure may
be applied to a variety of scenarios including multidimensional flow regimes and complex geochemical
conditions. The process has high potential for upscaling given the strictly enzymatic representation
of ureolysis, which could be measured in the field
with pumping tests or by extracted cores for lab
analysis.
CONCLUSIONS
The field of bio-mediated soil improvement is

advancing rapidly, and the potential demonstrated is
promising. Advancement of the field required multidisciplinary research in the academy as well as a
perspective shift in geotechnical practice towards a
broader science based view of soils. Based on progress
to date it is possible that applications of bio-mediated
soil improvement in the field may become relative
routine over the next decade.
Development of microbially induced calcite precipitation (MICP) has progressed the most rapidly to
date, and arguably still holds the most promise moving forward. A subset of research performed by the
authors has been presented herein to exemplify the unified biogeochemical-geotechnical approach that can
be performed to significantly improve understanding,
advance the field, and enable scaling towards field
applications. Experimental results clearly demonstrated the challenges of bacterial bio-augmentation
of soil, the dependence of treatment uniformity of

bacterial (ureolysis) distribution, the preference for
stopped-flow to improve uniformity, the significance
of treatment formulation on ureolysis rate and byproduct generation, and the link between calcite precipitated and the engineering parameters of small
strain stiffness and hydraulic conductivity. Numerical
modeling highlighted the challenges of biogeochemical modeling, the calibration of a model to site-specific
ureolysis activity distributions, the ability to predict
precipitated calcite distributions, and the framework
to model larger treatment scenarios.
ACKNOWLEDGEMENTS
The collaboration of Brina Montoya, Chris Hunt, Dave
Major, Burak Tanyu, and Tammer Barkouki is appreciated. Funding for the current study was provided
by Geosyntec Consultants, Inc., the UC Discovery
Grant Program, and the National Science Foundation
(#0727463 and #0628782). Any opinions, findings,
and conclusions or recommendations expressed in
this material are those of the writer(s) and do not
necessarily reflect the views of the National Science
Foundation.
REFERENCES
Al Qabany, A., Mortensen, B., Martinez, B., Soga, K., &
DeJong, J. 2011. Shear wave velocity as an indicator
for microbially induced carbonate precipitation. ASCE
Geo-Frontiers 2011.
Barkouki, T., Martinez, B., Mortensen, B., Weathers, T.,
DeJong, J., Spycher, N., Ginn, T., Smith, R., & Fujita,
Y. 2011. Forward and inverse bio-geochemical modeling
of microbially induced calcium carbonate precipitation
in half-meter column experiments. Transport in Porous
Media 90:2329.
Booster, J.L., van Meurs, G. A.M., Pruiksma, J.P., van
Paassen, L.A., Harkes, M., & Whiffin, V. 2008.
1D-modelling of microbially induced calcite precipitation
for geotechnical applications, in: 1st International Conference on Bio-Geo-Civil Engineering, June 2325, 2008,
Netherlands, 4550.
DeJong, J.T., Fritzges, M.B., & Nsslein, K. 2006. Microbial induced cementation to control sand response to
undrained shear. ASCE Journal of Geotechnical and
DeJong, J.T., Martinez, B.C., Mortensen, B.M., Nelson,
D.C., Waller, J.T., Weil, M.H., Ginn, T.R., Weathers, T.,
Barkouki, T., Fujita, Y., Redden, G., Hunt, C., Major,
D., & Tanyu, B. 2009. Upscaling of bio-mediated soil
improvement. Proceedings of the 17th International Conference on Soil Mechanics and Geotechnical Engineering,
Alexandria, Egypt, 4 pp.
DeJong, J.T., Mortensen, B, & Martinez, B. 2007. Bio-Soils
Interdisciplinary Science and Engineering Initiative, NSF
Final Report on Workshop, 84 pp.
DeJong, J.T., Mortensen, M.B., Martinez, B.C., & Nelson,
D.C. 2010. Biomediated soil improvement. Ecological
Engineering 36(2):197210.
DeJong, J.T., Soga, K., Banwart, S.A., Whalley, W.R., Ginn,
T., Nelson, D.C. Mortensen, B.M., Martinez, B.C., &
Barkouki, T. 2011. Soil engineering in-vivo: harnessing
115
natural biogeochemical systems for sustainable, multifunctional engineering solutions. J. Royal Society Interface 8:115.
DeJong, J.T., Soga, K.S., Kavazanjian, E., Burns, S., van
Paassen, L., Al Qabany, A., Aydilek, A., Bang, S.S.,
Burbank, M., Caslake, L., Chen, C.Y., Cheng, X., Chu, J.,
Ciurli, S., Fauriel, S., Filet,A.E., Hamdan, N., Hata,T., Inagaki, Y., Jefferis, S., Kuo, M., Laloui, L., Larrahondo, J.,
Manning, D.A.C., Martinez, B., Montoya, B.M., Nelson,
D.C., Palomino, A., Renforth, P., Santamarina, J.C., Seagren, E.A., Tanyu, B., Tsesarsky, M., & Weaver, T. 2013.
Biogeochemical processes and geotechnical applications:
progress, opportunities, and challenges. Geotechnique
(in press).
Dupraz, S., Parmentier, M., Mnez, B., & Guyot, F. 2009.
Experimental and numerical modeling of bacterially
induced pH increase and calcite precipitation in saline
aquifers. Chemical Geology 265(12):4453.
Espinoza, D. & Santamarina, J. C. 2010.Ant tunneling-a granular media perspective. Granular Matter, 12(6):607616.
Fauriel, S. 2012. Bio-chemo-hydro-mechanical modeling
of soils in the framework of microbial induced calcite
precipitation. PhD thesis, Lausanne, Switzerland.
Fauriel, S. & Laloui, L. 2012. A bio-chemo-hydromechanical model for microbially induced calcite precipitation in soils. Comput. Geotech. 46:104120.
Ferris, F.G., Phoenix, V., Fujita, Y., & Smith, R.W. 2004.
Kinetics of calcite precipitation induced by ureolytic bacteria at 10 to 20 C in artificial groundwater. Geochimica
et Cosmochimica Acta 68 (8):17011710.
Ferris, F.G., Stehmeier, L.G., Kantzas, A., & Mourits, F.M.
1996. Bacteriogenic mineral plugging. The Journal of
Canadian Petroleum Technology 35 (8): 5661.
Fidaleo, M. & Lavecchia, R. 2003. Kinetic study of enzymatic urea hydrolysis in the pH range 4-9. Chemical and
Biochemical Engineering 17(4):311318.
Fujita, Y., Redden, G.D., Ingram, J.C., Cortez, M.M.,
Ferris, F.G., & Smith, R.W. 2004. Strontium incorporation into calcium carbonate generated by bacterial
ureolysis. Geochimica et Cosmochimica Acta 68(15):
32613270.
Fujita, Y., Taylor, J.L., Gresham, T.T., Delwiche, M.M.,
Colwell, F., McLing, T.L., Petzke, L.M., & Smith, R.W.
2008. Stimulation of microbial urea hydrolysis in groundwater to enhance calcium carbonate precipitation. Environmental Science and Technology 42:30253032.
Fujita,Y., Ferris, F.G., Lawson, R.D., Colwell, F.S., & Smith,
R.W. 2000. Calcium carbonate precipitation by ureolytic subsurface bacteria. Geomicrobiology Journal, 17:
305318.
Ginn, T.R, Wood, B.D., Nelson, K., Scheibe, T.D.,
Murphy, E.M., & Clement, T.P. 2002. Processes in microbial transport in the natural subsurface. Advances in Water
Research 25(812):10171042.
Hammes, F., Boon, N., Clement, G., de Villiers, J.,
Siciliano, S.D., & Verstraete, W. 2003. Molecular, biochemical, and ecological characterization of bio-catalytic
calcification reactor. Applied Microbiology and Biotechnology 62: 191201.
Harkes, M.P., Booster, J.L., van Paassen, L.A., & van
Loosdrecht, M.C.M. 2008. Microbial induced carbonate
precipitation as ground improvement method bacterial
fixation and emperical correlation CsCO3 vs strength. 1st
International Conference on Bio-Geo-Civil Engineering,
Netherlands, June 2325:3741.
Harkes, M.P., van Paassen, L.A., Booster, J.L.,
Whiffin, V.S., & van Loosdrecht, M.C.M. 2010. Fixation
and distribution of bacterial activity in sand to induce
carbonate precipitation for ground reinforcement. Ecol.

Eng. 36:112117.
Hata, T., Tsukamoto, M., Mori, H., Kuwano, R., & Gourc,
J.P. 2011. Evaluation of multiple soil improvement techniques based on microbial functions ASCE GeoFrontiers
2011: Advances in Geotechnical Engineering, Geotechnical Special Publication 211, pp. 39453955.
Hill, M.C. & Tiedeman, C.R. 2007. Effective Groundwater
Model Calibration: With Analysis of Data, Sensitivities,
Predictions, and Uncertainty. Wiley-Interscience.
Ivanov V. & Chu J. 2008. Applications of microorganisms to
geotechnical engineering for bioclogging and biocementation of soil in situ. Reviews in Environmental Science
and Biotechnology 7:139153.
Kavazanjian, E., Jr. & Karatas, I. 2008. Microbiological
improvement of the physical properties of soil. Proc. 6th
International Conference on Case Histories in Geotechnical Engineering, Rolla, MO (on CD ROM).
Kuo, M.Y.H. & Bolton, M.D. 2013. The nature and origin of deep ocean clay crust from the Gulf of Guinea,
Geotechnique (in press).
Mansell, B.O., de Vellis, L., & Schroeder, E.D. 2000. Automated separation and conductimetric determination of
inorganic nitrogen. Journal of Environmental Engineering. pp. 778780.
Martinez, B. C. 2012. Experimental and numerical upscaling of MICP for soil improvement. Doctoral dissertation,
University of California, Davis, CA.
Martinez, B.C. & DeJong, J.T. 2009. Bio-mediated soil
improvement: load transfer mechanisms at micro- and
macro-scales. ASCE Proceedings of the 2009 USChina Workshop on Ground Improvement Technologies,
pp. 242251.
Mitchell, A.C. & Ferris, F.G. 2005. The coprecipitation of
Sr into calcite precipitates induced by bacterial ureolysis in artificial groundwater: Temperature and kinetic
dependence. Geochimica et Cosmochimica Acta 69(17):
41994210.
Mitchell, A.C. & Ferris, F.G. 2006a. Effect of strontium
contaminants upon the size and solubility of calcite
crystals precipitated by the bacterial hydrolysis of urea.
Environmental Science and Technology, 40:10081014.
Mitchell, A.C. & Ferris, F.G. 2006b. The influence of bacillus pasteurii on the nucleation and growth of calcium
carbonte, Geomicrobiology Journal, 23(3):213226.
Mitchell, J.K. & Santamarina, J.C. 2005. Biological considerations in geotechnical engineering. ASCE Journal
of Geotechnical and Geoenvironmental Eng. 131(10):
12221233.
Montoya, B. M., DeJong, J. T. & Boulanger, R. W. 2013. Seismic response of liquefiable sand improved by microbial
induced calcite precipitation. Geotechnique (in press).
Montoya, B.M., Gerhard, R., DeJong, J.D., Weil, M.,
Martinez, B.C., Waller, J.T., & Pederson, L. 2012. Fabrication, operation, and health monitoring of bender elements
for aggressive environments. ASTM Geotechnical Testing
Journal, 35(12):115.
Nilsson, . & Sternbeck, J. 1999. A mechanistic model for
calcite crystal growth using surface speciation. Geochimica et Cosmochimica Acta, 63(2): 217225.
Parkhurst, D.L. & Appelo, C.A.J. (1999), Users guide to
PHREEQC (Version 2)A computer program for speciation, batch-reaction, one-dimensional transport, and
inverse geochemical calculations, U.S. Geological Survey
Water-Resources Investigations Report 99-4259, 310 p.
Poeter, E. P., Hill, M. C., Banta, E. R., Mehl, S., &
Christensen, S. 2005. UCODE_2005 and Six Other
Computer Codes for Universal Sensitivity Analysis,
116
Calibration, and Uncertainty Evaluation, Techniques and

Methods 6-A11, U.S. Geological Survey, Reston, Virginia.
Ramachandran, S.K., Ramakrishnan, V., & Bang, S.S. 2001.
Remediation of concrete using micro-organisms. ACI
Material Journal 98(1):39.
Ramakrishnan, V., Bang, S.S., & Deo, K.S. 1998. A novel
technique for repairing cracks in high performance concrete using bacteria. Proceedings of the International
Conference on HPHSC, Perth, Australia: 597618.
Rebata-Landa, V. & Santamarina, J.C. 2006. Mechanical limits to microbial activity in deep sediments. Geochemistry,
Geophysics, Geosystems, 7(11):112.
Rebata-Landa, V. & Santamarina, J.C. 2012. Biogenic gas
bubbles in soils: Skemptons B parameter, P-wave velocity
and liquefaction resistance. ASCE Journal of Geotechnical and Geoenvironmental Eng. 138:127137.
Reed, M. H. 1982. Calculation of multicomponent chemical equilibria and reaction processes in systems involving
minerals, gases and aqueous phase. Geochim. Cosmochim.
Acta 46:513528.
Renforth, P., Manning, D.A., & Lopez-Capel, E. 2009.
Carbonate precipitation in artificial soils as a sink for
atmospheric carbon dioxide. Applied Geochemistry 24:
17571764.
Seagren, E.A. & Aydilek A.H. 2010. Biomediated Geomechanical Processes. Chapter 24 in Environmental Microbiology, 2nd Edition, R. Mitchell and J-D. Gu, eds., John
Wiley & Sons, Inc., Hoboken, NJ, pp.319348.
Stocks-Fischer, S., Galinat, J.K., & Bang, S.S. 1999. Microbiological precipitation of CaCO3 . Soil Biology and
Biochemistry, 31:15631571.
van Paassen, L.A., Harkes, M.P., van Zweiten, G.A, van
der Zon, W.H., van der Star, W.R.L, & van Loosdrecht,
M.C.M. 2009. Scale up of BioGrout: a biological ground
reinforcement method. Proceedings of the 17th International Conference on Soil Mechanics and Geotechnical Engineering, M. Hamza et al. (Eds.), IOS Press:
23282333.
van Paassen, L.A., J. O. Isimite, C. Picioreanu, M. C. M. & van
Loosdrecht. 2008. Computational model to study microbial carbonate precipitation in porous media at microscale. In: 1st International Conference on Bio-Geo-Civil
Engineering. Netherlands, pp. 4550, June 2325.
van Tittelboom, K., Belie, N.D., de Muynck, W., & Verstraete,

W. 2010. Use of bacteria to repair cracks in concrete.
Cement and Concrete Research, 40:157166.
Weil, M.H., DeJong, J.T., Martinez, B.C., Mortensen, B.M., &
Waller, J.T. 2012. Seismic and resistivity measurements
for real-time monitoring of microbially induced calcite
precipitation in sand. ASTM Journal of Geotechnical
Testing 35(2):112.
Whiffin, V.S. 2004. Microbial CaCO3 Precipitation for the
production of Biocement. ScD Thesis, School of Biological Sciences and Biotechnology, Murdoch University,
Western Australia, 154 p.
Whiffin, V.S., van Paassen, L.A., & Harkes, M.P. 2007.
Microbial carbonate precipitation as a soil improvement
technique. Geomicrobiology Journal, 25(5):417423.
Woese, C.R., Kandler, O., and Wheelis, M.L. 1990. Towards a
natural system of organisms: Proposal for the domains of
Archaea, Bacteria, and Eucarya. Proc. National Academy
of Sciences, USA, Vol. 87, pp. 45764579.
Xu, T, Sonnenthal, E., Spycher, N., & Pruess, K. 2008b.
Toughreact Users Guide: A simulation program for nonisothermal multiphase reachtive geochemical transport in
variably saturated geologic media. LBNL Report 554602008, Lawrence Berkeley National Laboratory, Berkeley,
California.
Xu, T. 2008a. Incorporation of aqueous reaction kinetics
and biodegradation into TOUGHREACT: Application of
a multi-region model to hydrobiogeochemical transport of
denitrification and sulfate reduction, Vadose Zone Journal
7(1), 305315.
Xu, T., Sonnenthal, E.L., Spycher, N., & Pruess, K.
2006. TOUGHREACT A simulation program for nonisothermal multiphase reactive geochemical transport
in variably saturated geologic media: Applications to
geothermal injectivity and CO2 geological sequestration,
Computers & Geosciences 32(2):145165.
Yoon, J.S., Germaine, J.T., & Culligan, P.J. 2006.Visualization
of particle behavior within a porous medium: Mechanisms for particle filtration and retardation during downward transport. Water Resources Research, 42(W06417):
116.
117

Multiphase aspects of soil contamination by immiscible

petroleum hydrocarbons
P. Delage
Ecole des ponts ParisTech, Navier/CERMES (CNRS UMR 8205 IFSTTAR), France
ABSTRACT: To better understand the phenomena occurring during infiltration of non aqueous phase liquid
(NAPL) in the vadose zone, the concepts describing the retention and transfer of water in granular and finegrained soils are first recalled. The extension of these concepts to immiscible petroleum hydrocarbon is then
dealt with by first describing macroscopical evidence of the interaction between NAPL and fine-grained soils.
Finally, the phase changes occurring during soil vapour extraction by venting are commented, with emphasis put
on the conditions under which local equilibrium at the liquid-gas interface occurs or does not occur. The mass
transfer coefficients, that allow describe some retardation processes in vapour extraction, are then introduced.
When introduce in diffusion models, they allow a better modelling of the tailing effects that often indicate a
decontamination slower than predicted based on local equilibrium hypotheses.
INTRODUCTION
Soil pollution by hydrocarbons (HC, also called NAPL

that holds for Non Aqueous Phase Liquid) is a serious
geoenvironmental problem that can affect the environmental quality of the soil, groundwater, air and
vegetation (Mercer and Cohen 1990). Accidental soil
contamination by hydrocarbons originating from the
oil industry has been a strong concern for various
decades. It often occurs from unknown or uncontrolled
leaks from hydrocarbon processing/storage/transport
facilities. Gasoline and diesel fuels are widely used
in individual and industrial transportation. They are
the main pollutants found in contaminated soils. Consequently, various remediation techniques for in-situ
decontamination of HC contaminated soils have been
developed and are commonly used. The most used one,
soil vapour extraction (SVE), is based on the phase
change properties of hydrocarbons and their volatility.
It works for light HCs that are volatile enough, but
not for heavy ones, that are fortunately not as commonly used. SVE is also quite efficient for chlorinated
solvent.
As seen in Figure 1, soil pollution by hydrocarbons
may occur due to leaks from hydrocarbon tanks and
subsequent infiltration into the upper unsaturated layer
of soil, called the vadose zone. This upper zone is
unsaturated due on the one hand to climatic effects
(evaporation) and on the second hand to weak water
retention properties of the layer, a behaviour feature
that will be described in more details later on. Note
also that vegetation may have some effects through
evapotranspiration, from trees or even grass.
Figure 1. Typical scheme of light NAPL infiltration from

leaking container.
As seen in the Figure, various multiphase processes

are involved during hydrocarbon infiltration:
119
Infiltration of the non miscible phase of the hydrocarbon in the unsaturated layer that is composed
of minerals (either hydrophilic in the clay fraction
or neutral in the silt of sand/gravel fraction), water
(either under capillary state in the non clay fraction
or in an adsorbed non free state in the clay fraction) and air (at a humidity that depends of the soil
suction, as described later)
Hydrocarbons are complex mixtures of components also including volatile phases that change into
vapour once put in contact with air and propagate
along the continuous air channels existing if the
soil is enough desaturated. The occurrence of air
continuity when a soil is desaturated depends of
the climatic conditions (duration of dry periods),
of vegetation and of the retention capability of the
soil.As a result, a cloud of vapour is progressively

generated all around the infiltrating non miscible
liquid phase as shown in Figure 1, with obvious
concerns related to the quality of the atmosphere
above
Once touching the water table, the non miscible
phase of light HCs will spread over the water
table, resulting in potential hazards due to possible
movements related to the natural water gradient
and possible downwards transport and exposition to
vulnerable target. Note that adapted and performing techniques have been developed to successfully
extract by pumping this floating phase
There still exist miscible components in HCs that
dissolve into water and may be transported forward
and affect potential vulnerable targets.
Note that the scheme of Figure 1 describes the fate
of light HCs (including diesel and gasoline), the most
common case. In the case of heavy HCs and of chlorinated solvents, there is also a penetration into the
water table with some trapping in the saturated zone
that makes further decontamination quite complex.
Whereas heavy HCs are less prone to phase changes,
this is the case of chlorinated solvents that constitute
a serious vadose zone contamination issue.
As commented by Harper et al. (2003), the efficiency of soil venting extraction depends of many
parameters including soil type, particle size distribution and clay fraction, degree of saturation, permeability, water content, organic carbon content (with
significant sorption occurring), contaminant physical
and chemical properties and air pore velocity.
Many complex multiphase mechanisms are involved
in the infiltration process described in Figure 1:
The status of the vadose zone, the water content
of which depends on climatic conditions (evaporation/wetting) and on the water retention and
transport properties of the unsaturated soil. Beside
capillary actions that take place in the granular
phase (gravel and sand), clay water interaction
intervene in fine-grained soils and add some significant difficulty to understanding the mechanisms
The phase change properties of the mixture of
different components of the HC, briefly commented above. Note also that the various chemical
additives added to HC (sulphurs, acids, nitrogen
compounds, heavy metals, . . .) obviously play a
role (Izdebska-Mucha and Korzienowska-Rejmer
2000), adding some significant chemical complexity to the process. As in the soil/water interaction,
the soil/water/oil interaction is governed by both
capillary actions and physico-chemical interactions
along the clay fraction, with possible interaction
and competition between water and other HC components that are governed by both solid/liquid capillary interface parameters and physico-chemical
phenomena.
The retention and transport properties of water in
unsaturated soils are well understood. The retention
and transport of NAPL in unsaturated soils are likely
to be governed in a similar fashion. However, further

complexity is added by this third phase that co-exists
with water (that can develop physico-chemical interactions with the clay fraction of the soil) and air in
the soil.
When considering the retention and transport properties of NAPL in an unsaturated soil, the two
fundamental characteristics to investigate are the
NAPL/soil retention curve and the NAPL permeability. As far as soil decontamination is concerned, the
former provides quantitative information on the energy
necessary to extract the NAPL from the soil, whereas
the latter allows for the prediction of movements of the
pollutant in the vadose zone. Obviously, both parameters have a fundamental importance regarding soil
clean-up techniques.
Other important parameters are related to the phase
change behaviour of the petroleum hydrocarbons.
Given the complex composition of petroleum hydrocarbons, the evaporation of hydrocarbons depends
of a lot of parameters characterizing the various
components.
In the following, the water retention and transport
properties of water in unsaturated soils are recalled
and some aspects of the infiltration of a NAPL in
unsaturated soils are presented. Recent research on soil
vapour extraction emphasizing phase changes carried
out by researchers working in the field of environment and pollution, based on some specific tools less
often used in geoenvironmental engineering will also
be described.
2 WATER IN UNSATURATED SOILS
2.1 Introduction
In this section, the status of water in unsaturated soils
(multiphase geomaterials composed of three phases,
solid, water and air) will be first recalled in terms
of retention and transport properties. This status will
afterwards be used as a basis to describe the status
of contaminated soils, multiphase geomaterials composed of 4 phases (solid, water, NAPL and air) that are
governed up to a point by comparable physical laws.
Note however a fundamental difference between the
fluids involved: whereas water is a polar liquid, NAPL
are non polar. This difference in nature explains the
non miscibility between water and NAPL fluids. As
seen later, this feature theoretically makes things simpler in the framework of capillary actions and is also
of some importance in the clay-fluid interaction.
2.2 Water retention properties in unsaturated soils
Water retention properties are characterised by the
water retention curves that give the changes in water
content (and degree of saturation) with changes
in suction. There are determined by various techniques well described in the literature (Croney 1952,
Fredlund and Rahardjo 1993) the most often used
being the axis translation method in which an air back
120
Figure 2. Water retention curves of various soils (Barbour

1998).
Figure 3. Water retention in granular soils.
pressure (ua > 0) is imposed while maintaining the

water pressure at atmospheric pressure (uw = 0),
resulting in a suction (s > 0) defined by the difference
between air pressure and water pressure:
This situation of (positive) suction is comparable

to that prevailing in unsaturated soils in which the air
pressure is equal to atmospheric and the water pressure
negative, at least when the degree of saturation is not
too close to 100% in order to ensure the continuity of
the air phase.
Figure 2 (Barbour 1998) presents the water retention curves obtained starting from an initial saturated
state along a desaturation path (drying path) by progressively increasing suction in different soils of
various plasticity index. The suction at which the saturation degree starts to decrease is called the air entry
value. Its value is of some importance since it indicates
the ability of the soil to start desaturating under given
climatic conditions.
In this regard, one can see that the sand sample
starts desaturating quite easily at a low value smaller
that 10 kPa whereas the Regina clay requires a suction of around 1 MPa, like soils of medium plasticity
(Ip 30). The Indian Head till is intermediate with
desaturation occurring close to 100 kPa whereas the
Botkin silt desaturates at around 20 kPa. As a consequence, a leak occurring over a layer of plastic soil that
will remain saturated in a temperate country will stay
retained at the surface. This is not the case in granular or silty soils that will desaturate much more easily
giving rise to the development of the vadose zone and
allowing infiltration down to the water table. Given
that industrial sites are often located along rivers on
alluvial deposits that tend to behave like sands, one
concludes from the figure that significant desaturation
should prevail, even in wet seasons. Typically, sandy or
alluvial deposits will start desaturating at an air entry
value of various tens of kPa whereas low plasticity
silty soils will do so at an air entry value of various
hundreds of kPa.
In granular soils, soil water interactions are only due
to capillary actions in the meniscus located at intergrains contacts. In a vertical capillary tube of radius r
placed in contact at its base with a liquid and submitted
to water ascension, the difference in pressure between
Figure 4. Water retention in fine-grained soils.
air (ua ) and water (uw ) at the air-water interface i.e.

the suction s (s = ua uw ) is linked to the capillary
radius according to Jurins law:
in which aw is the air-water superficial tension and

the solid liquid contact angle. In the air/water system,
aw = 72.75 103 N/m and between water and the
solid phase is equal to 180 , illustrating the wetting
behaviour of water, with the meniscus tangent to the
inner surface in the capillary tube.
In granular soils submitted to suction, the meniscus
are located in the smallest pores, i.e. at the inter-grains
contacts (Figure 3). This relation shows that the higher
the suction the smaller the meniscus radius (and the
water content).
In fine grained soils, an additional interaction is
exerted on water molecules by the clay fraction thanks
to physico-chemical interactions, as shown schematically in Figure 4. Water adsorption on clay minerals is
quite complex and still thoroughly investigated. It is
currently admitted that this attraction is mainly due to
Van der Waals links that are independent of parameters
like the pore fluid salinity or permittivity.
Repulsion phenomena between clay particles are
linked to the electrical deficiency of clay platelets
that result in an electrical field in their proximity.
Water molecules, in spite of being electrically neutral,
are dipoles that are sensitive to electrical fields. The
electrical field around clay platelets allows the development of the so-called diffuse double layer (DDL)
characterised by a distribution of the clay exchangeable cations in the electrical field and described by the
121
Figure 5. Inter-particle interactions (after Mitchell and

Soga 2005).
Figure 6. Effect of permittivity of solvent on double layer
of illite (Shang et al. 1994).
DDL theory (see for instance Mitchell and Soga 2005).

A useful parameter is the DDL equivalent thicknes
characterised by the following expression:
in which:
D is the dielectric constant (permittivity) of the
medium (D = 80 for water, 2030 for alcohols and
acetone, both polar liquids and D = 2 for aromatic
compounds. The permittivity of non polar liquid is
smaller than that of polar liquids)
K is the Boltzmann constant (k = 1.38 1023 J/K)
T is temperature
n0 the salt concentration
is the elementary electric load ( = 1.6
1019 C)
the cation valence (Na+ , K+ , Ca++ )
The DDL theory is often used when trying to explain
the interactions between clay particles and fluids or
between clay particles themselves, including in contaminated soils, according to the scheme represented
in Figure 5.
Basically, DDLs around two parallel and close clay
platelets will exert a mutual electrical repulsion due to
the negative charges of the exchangeable cations that
make impossible any overlapping. The DDL repulsion
increases with thicker DDL when the salt concentration and salt valence decreases and when the permittivity increases according to the expression above. A
thick DDL (small cation valence Na+ or K+ small
salt concentration, large permittivity, higher temperature) will enhance repulsion whereas a thin DDL (large
cation valence Ca++ large salt concentration, small
permittivity, lower temperature) enhance repulsion.
DDL repulsion is counterbalanced by the constant
Van der Waals attraction. When repulsion is dominant
(see above), particles flocculate and form aggregates
(observed in reality by SEM and other microstructure
observation techniques). When attraction predominates, a so-called dispersed structure (not really
observed on natural soils) is created.
The interest of the DDL theory is that it provides
a possible interpretation the effects of the changes of
both salt the concentration and the permittivity of the
pore fluid. The pore fluid permittivity is obviously
changed when an unsaturated soils containing water
(a polar liquid with high permittivity) is infiltrated
by NAPL (a non polar fluid with low permittivity),
resulting in stronger attraction and aggregation. As
an example, Figure 6 shows from theoretical calculations how the DDL thickness reduces in an illite
when comparing water with alcohols and aromatic
compounds. This significant reduction in thickness
enhances attraction between particles, resulting in
flocculation. Note the very small values of the permittivity of aromatic compounds (frequent in fuels) that
indeed has a significant effect on the DDL reduction.
The DDL theory is often referred to when explaining adsorption phenomena. It is a theroretical model
that is able, up to a point, to capture the effect of
some parameters on inter-particle interactions. However, water adsorption along clay platelets is indeed
a complex phenomenon that includes other physicochemical interactions (e.g. Prost and Sposito 1982).
Going back to suction and water retention in soils,
the state of desaturation of a soil profile from the surface down to the water table depends on both climatic
conditions and on the water retention properties of
the soil. At equilibrium, the suction profile is hydrostatic above the water table, with negative pressures
defined by the water column height above water table.
The suction profile resulting from the linear profile
of negative pressure is itself not linear, given the non
linearity that most often characterises water retention
curves (see Figure 2). During dry periods, the suction
profile moves towards higher values than hydrostatic
(and stronger desaturation) whereas it goes to smaller
values (and smaller desaturations) in wet seasons.
122
An important characteristic of water retention

curves is the hysteresis that makes different the drying
path from saturated state (showed in Figure 2) from
the wetting path from dry state.
Table 1.
Physical parameters of water, air and Soltrol 170.
soltrol170
soltrol170 w
(Mg/m3 ) (Mg/m3 ) (Pa s)
0,785
w
(Pa s)
a
(Pa s)
2,028 103 1,00 103 1,83 105
2.3 Water transfer in unsaturated soils

Water transfer in unsaturated soils obeys Darcys law,
with a permeability coefficient that depends on the
degree of saturation, starting from low values in dry
state to reach the highest saturated permeability value:
Table 2. Interfacial tension values in a multiphase system

composed of air, water and Soltrol (Busby et al. 1995).
Interfacial
tension
in which is the volumetric water content and the

hydraulic head h is defined by h = s/w + h.
Water transfers (and hence water infiltration) in
unsaturated are governed by Richards equation that
is obtained from combining Darcys law and the water
conservation equation, finally giving:
In this equation, the term h

is obtained from the
water retention curve.

The determination of the unsaturated permeability
to water is most often obtained in transient regimes
by back analysing infiltration through Richards
equation. A convenient method is the so-called instantaneous profile method (Daniel 1982) in which onedimensional infiltration is monitored in a column in
which a soil cylindrical sample is wetted at one end
while changes in suction with time are monitored
at various levels. The suction isochrones obtained
allow the calculation of local hydraulic gradients. The
knowledge of the water retention curve of the soil provides the water content isochrones that are deduced
from suction isochrones. Water contents isochrones
allow calculating the water flux.
3
3.1
NON MISCIBLE HYDROCARBONS

IN THE VADOSE ZONE
Interaction with unsaturated granular soils
If only capillary phenomena were to occur in a contaminated soil between the two liquids (water and NAPL),
the gaseous phase and the solid skeleton, one should
pay attention to the respective values of the interfacial
parameters described above.
Most laboratory investigations carried out on contaminated soils have been made by using a light
hydrocarbon called Soltrol 170 (Lenhardt and Parker
1998), the physical properties (density and viscosity)
of which are presented in Table 1. Together with that
of water and air. Soltrol 170 is a non aromatic non
toxic hydrocarbon composed of a mixture of C14-C16
isoalkanes specially designed to be safe when used in
laboratory investigations. It has very low volatility and
water solubility and can be considered as non miscible
Soltrol-air
sa
N/m
Soltrol-water
sw
N/m
Air-water
aw
N/m
24,24 103
44,21 103
72,75 103
Figure 7. Scheme of the mutual placement of water, air

and oil.
with water. As indicated by its density, Soltrol is a light

hydrocarbon, like all fuels.
In the multiphase system defined by the three fluids (air, water and Soltrol), three interfaces can be
defined: air-water, water-Soltrol and air-Soltrol. Table
2 presents the values of the corresponding interfacial
tension parameters.
In a porous medium only governed by capillarity,
the mutual arrangement of the three fluids is governed
by the spreading coefficient S:
The positive value of S indicates that the Soltrol will

spread between water and air, as schematically indicated in Figure 7. In other words, the Soltrol is more
wetting than air but less than water. It can hence move
air and replace it when put in contact with the water-air
system, like during infiltration in an unsaturated soil.
The scheme of Figure 7 could be applied to the configuration of the grain assembly of Figure 3, defining
the location of the NAPL along the water menisci at
inter-grains contacts.
3.2 NAPL in unsaturated fine-grained soils
This exclusive capillary mechanism would be relevant
in granular soils with no clay fraction. In fine-grained
soils containing some clay, non polar fluids should
theoretically not be sensitive to electrical fields like
water does. Pure non polar fluids should hence only
obey to capillarity laws with no effects of adsorption
and physico-chemical interactions.
123
Actually, a significant effect of a non polar fluid

(xylene) on soil microstructure was however noticed
by Fernandez & Quigley (1985) who observed a significant aggregation in a sample compacted from
a powder with xylene as a compacting fluid. This
aggregation, compatible with the DDL theory presented above, led to a drastic increase in permeability
up to 106 m/s (as compared to 1011 m/s with water
and 1010 m/s with ethanol), obviously a major concern when considering waste isolation by compacted
liners. Fernandez & Quigley emphasised the importance of the fluid dielectric constant (D = 80 for water,
2 for xylene and 32 for ethanol, a polar fluid). The role
of this parameter through the DDL theory has been
commented above (Figure 6).
Interactions between NAPL and fine-grained soils
appear to be more complex and not only capillary.
More generally, the complexity of the interactions
between crude oil and reservoir rocks has been known
for a long time in reservoir rocks as distinct as sandstones or chalk, with for instance either water-wet
(Ekofisk chalk) or oil-wet (Valhal chalk) behaviour
in the North-sea reservoir rocks. Indeed, the very long
period of time during which crude oil has been in contact at various temperatures with the grain surface of
the rocks is prone to some complex chemical interface reactions that can result in some coating of the
grain surfaces by some specific components among
the many different ones that compose crude oil. This
complex chemistry and the understanding gained in
reservoir engineering regarding the wettability properties of reservoir rocks is certainly a potential tool
to be fruitfully developed and applied to contaminated
soils. Again, significant efforts would certainly be necessary to include the effects of the clay fraction on the
interaction with oil components in fine-grained soils.
In this context, some techniques exist to turn a
water wet porous medium (hydrophilic, the natural
behaviour of porous media containing water, given the
high tension interface properties of water, a wetting
fluid) into an oil wet material (hydrophobic). As mentioned later on, they are based on the use of specific
organic components like Glassclad, a compound used
by some authors that provide a coating along the solid
surface of the porous stones and inverses its wettability
properties.
It is well known that Atterbergs limits are a good
indicator of the ability of fine-grained soils to attract
water. Indeed, the plastic index Ip (Ip = wp wl ) is a
measure of the quantity of water that is necessary to
change the soil from a solid state (w < wp ) to a liquid one (w > wl ). A high Ip hence indicates a large
amount of adsorbed water along the clay particles.
Also, the changes in liquid limit in a series of distinct
soils are satisfactory correlated with that of their specific surfaces and cation exchanges capacity (CEC).
As a consequence, it appears interesting to check possible effect of NAPL on fine grained soils, as done by
Izdebska-Mucha and Korzienowska-Rejmer (2010) on
various soils. They determined the Atterberg limits of
a series of soils of various plasticity index by carried
Figure 8. Influence of crude oil content on Atterbergs

limits and plasticity index (after Ebadi and Kermani 2010).
out standard tests (with distilled water) and also with

two petroleum substances.
The effect of the petroleum substances appeared
to be drastic: in all soils, it was impossible to determine any plastic index because the soils lost cohesion
and had no consistency at all, they were qualified of
non plastic soils. The determination of the liquid limit
with Casagrandes device showed a drastic decrease
in fluid content with highest decrease observed in the
most hydrophilic (and plastic) ones, with a spectacular
effect on the bentonite and little effect ion the glacial
till. Following Fernandez and Quigley (1985), these
authors raised the important issue of the suitability of
these soils as insulating compacted barriers in waste
disposal.
In a similar fashion, they observed significant
linear shrinkage on the Miopliocene clay with
water and much less when initially saturated with
petroleum substances. Confirming earlier findings by
Khamehchiyan et al. (2007) and Singh et al.
(2009), Izdebska-Mucha and Korzienowska-Rejmer
also observed some significant effects on the compaction properties with a 2% decrease in relative
compaction. These observations are compatible with
the effects of the dielectric constant on the DDL
equivalent thickness described in Figure 6. Indeed, by
reducing the DDL equivalent thickness, clay particles
are much less able to retain some fluid molecules in
their vicinity, resulting in smaller liquidity indexes.
Given that thin DDL favour van der Waals attraction between particles, low permittivity fluids enhance
flocculation that tend to change mixtures at low fluid
content into grains, with some difficulty to determine
the plastic limit. These changes in structure due to
NAPL are also macroscopically observed when compacting samples with some percentage of NAPL (e.g.
Manassero et al. 2006).
Ebadi and Kermani (2010) added various amounts
of crude oil to a reconstituted silty soil (Ip = 22) at a
low water content (3.9%) and determined Atterbergs
limit and plastic index of the sample at various degree
of oil contamination, as indicated in Figure 8. The Figure demonstrates a significant effect of the initial oil
content with an increase in both the plastic and liquid limits that finally results in a decrease in plastic
index. One can observe here that small initial water
124
Table 3. Liquid limits of soils contaminated by petroleum

substances (after Izdebska-Mucha and KorzienowskaRejmer 2010).
Plastic limit (%) Liquid limit (%)
Soil
Water
Water Petrol Diesel
Miopliocene clay
Glacial till
Bentonite
Illitic clay
Kaolinite
25
13
55
26
40
65
24
257
46
51
36
24
34
24
45
34
23
37
26
37
Figure 10. Intrinsic permeability of a compacted silt to

air, Soltrol and water (Cui et al. 2003).
Figure 9. Oedometer compression tests on unsaturated

compacted silt samples at a constant water content of 15%
(air), after water saturation (water) and after Soltrol saturation
(Soltrol) (Delage et al. 2000).
content allowed the determination of plastic index,

unlike in the case of Table 3. The interpretation of
these data are however complicated given that it a
combined effect of the successive introduction, in the
silt, of water (3.9%), oil (up to 12%) and, again, water
so as to carry out the index determination. The ain
trend is a significant decrease in plastic index from
20 to 9 when introducing the first 4% oil. This trend
is compatible with that observed by Izdebska-Mucha
and Korzienowska-Rejmer (2010). However, the parallel increase in wL and wP while introducing more oil
is not straightforward.
A macroscopic mechanical effect of oil infiltration in low plasticity compacted silt (Ip = 18) has
been detected by Delage et al. (2000) who compared the oedometer compression curve of compacted
samples i) at constant water content of 15% (initial
degree of saturation of 58%), ii) saturated with Soltrol
(same water content) and iii) saturated with water
(see Figure 9).
The responses in compression show that Soltrol
has a role comparable to that of water with an
initial comparable swelling upon saturation and a
larger compressibility with respect to that of the

constant water content sample. Soltrol saturation
brings the interfacial tension from the air-water situation with aw = 72.75 103 N/m down to the
Soltrol-water one with sw = 44.21 103 N/m (see
Table 2) whereas water saturation moves from
aw = 72.75 103 N/m down to ww = 0. The corresponding suction relaxation upon saturation should
be smaller, resulting in smaller swelling, which is not
observed. Indeed, the same swelling observed with
water (polar fluid) and Soltrol (non polar) is surprising.
Conversely, the intermediate compression behaviour
once saturated with Soltrol is compatible with the
gradation of interfacial tensions.
Further macroscopic evidence on the status of
Soltrol with respect to air and water can be gained
by considering the data of permeability tests (to air,
Soltrol and water) as seen in Figure 10 (Delage et al.
2000). Air permeability was obtained by using the
technique of Yoshimi and Osterberg (1962) and the
water and Soltrol permeabilities were obtained by running changing head standard tests. Data are presented
in terms of intrinsic permeability expressed in m2 .
In an inert medium with no fluid/solid interaction,
the three curves should be similar to the air permeability curve. It is not surprising that the test run with
water provides the smaller values given the significant interaction between water and the clay fraction.
The intermediate position of the oil permeability curve
shows that Soltrol is not completely inert with respect
to the soil skeleton, namely an aggregate microstructure like any dry compacted soil. Again, this indicate
that the soil/Soltrol interaction is not only governed by
capillarity with some physico-chemical effects.
Another remark is that laboratory investigations carried out with refined hydrocarbons like Soltrol will
probably not be able to capture all the features if oil/soil
interactions, given that many components (including
125
Figure 11. Cross sectional view of the three phase retention

cell developed by Lenhard and Parker (1988).
aromatic ones still present in fuels) have been extracted

for safety reasons.
These observations on the macroscopic effects of
oil/soil interactions through the changes in Atterberg
limits, plastic index or compression behaviour can
be completed in terms of oil-retention behaviour by
investigations in which the techniques used in water
retention have been adapted to the retention capacity
exerted by soils on oil compounds.
3.3
NAPL retention properties in

unsaturated soils
In the literature, the data concerning the experimental

determination of the parameters governing the retention and transport of NAPL are less frequent than for
water retention and transport (e.g. Lenhardt and Parker
1988, Busby et al. 1995, Steffy et al. 1997, Cui et al.
2002, Manassero et al. 2005, Rabozzi et al. 2006).
In the framework of capillarity, more relevant to
granular soils, some authors (Lenhard and Parker
1988, Ferrand et al. 1990 and Busby et al. 1995) aimed
at extending the findings of Leverett (1941) who proposed to extend the two phase retention properties of a
porous medium to predict the fluid retention behaviour
of a three phase system and to determine whether a
scaled retention function can be employed to describe
the water retention curve (along a drainage path) of
arbitrary two phase fluid system, according to scaling factors based on the various values of interfacial
tensions.
Lenhard and Parker (1988) developed a three phases
retention cell with specially treated porous ceramics to
control and monitor liquid movements and pressures
(Figure 11). In this system, a cylindrical sample of soil
was placed in a cell composed of two ceramic high air
entry value (HAEV) rings, respectively hydrophilic
and hydrophobic. The hydrophobic behaviour of the
porous ceramic was obtained by using a Glassclad
18 solution. Negative fluid pressures were applied by
applying vacuum, volume changes were monitored by
using burettes and the negative fluids pressures were
measured by using standard tensiometers (maximum
suction 80 kPa). The hydrophobic porous ceramic
Figure 12. Retention cell for LNAPL developed by Steffy

et al. (1997).
allowed for the NAPL pressure to be measured via a

transducer outside the cell and allowed for controlling
the movement of NAPL into and out of the sample. The
hydrophilic ceramic was used to measure water pressure and control water movements. With much smaller
pores than in the sample, the ceramics remain always
saturated.
Provided that the fluid wettability sequence follows
the order water NAPL air (see Figure 7), they
experimentally observed good agreement when water
contents in two-phase NAPL-water systems were compared to water contents in the three-phase system.They
also obtained good agreement when NAPL contents
in two phase air-NAPL systems were compared to
total liquid content in three-phase system. They however mentioned a limitation already commented that
fluid/structure interactions should not affect the soil
microstructure. This is acceptable in granular soils but
obviously not in fine-grained soils in which NAPL
infiltration is known to affect the soil microstructure
as commented above from macroscopic responses.
Steffy et al. (1997) used the cylindrical retention
cell shown in Figure 12 where two hydrophobic and
hydrophilic porous plates were placed on both sides
of the sample. Low negative fluid pressures were
applied by the difference in fluid levels, allowing
control independently each fluid negative pressure
(at low values, compatible with their investigation of
NAPL entrapment).
They devoted particular attention to the residual
NAPL saturation remaining in sands as a result of a
water table rise occurring once a sand layer has been
contaminated. They were also interested in the effect of
antecedent water saturation on residual NAPL saturation. In a carbonate sand sample containing water and
decane, they observed a residual water saturation of
20% under a tension of 1 m of water (10 kPa). In sands,
they linked NAPL entrapment with the pore space configuration, evidencing the effects of the aspect ratio
of pore throat diameter to pore body diameter and
observed little influence of the air phase that was
flushed before NAPL/water phenomena responsible
126
Figure 13. Cross sectional view of the three phase retention

cell developed by Cui et al. (2003).
for entrapment started. Note however that this investigation, comparable to that conducted in oil recovery,
only holds for sand and coarser granular soils.
Infiltration in granular soils mainly involves capillarity whereas other physico-chemical effects also
take place in fine grained soils in the clay fraction, in
the organic fraction and along the oxides components
(see for instance Lagaly 1984). As in unsaturated soils,
techniques of controlling the matrix air-oil suction account for both capillary and physico-chemical
soil-fluid interactions.
In order to provide a relevant system for fine grained
soils in which higher values of oil-air suction are
expected, Cui et al. (2003) developed an axis translation cell with hydrophilic and hydrophobic ceramic
porous stones placed on both faces of a cylindrical
sample (30 mm high and 70 mm in diameter), shown in
Figure 11. In this device, the air pressure was imposed
through the lateral connection and the oil volume
exchanges were monitored with an inclined tube.
The data obtained indicated that, due to the significant air water suction illustrating strong soil/water
attraction, water could be considered as belonging to
solid phase. The oil retention curves obtained in three
compacted samples (A, B and C) having a water content of 15.6% and densities equal to 1.3, 1.73 and
1.61 Mg/m3 respectively are presented in Figure 14.
Samples were first oil saturated and subsequently submitted to values of oil suction (so = uo uw with
uw = 0 and uo > 0) of 25, 50, 100, 200 and 300 kPa.
The loosest sample A is naturally starting from a higher
initial oil content. Data show that significant desaturation starts at an oil suction of 25 kPa, a point where
the loosest (A) and densest (B) samples behave similarly. Note that the three samples exhibit a residual oil
saturation of around 2% at 300 kPa suction.
3.4 Phase changes in the vadose zone
The physical phenomenon on which soil venting is
based is the evaporation of the volatile phases contained in fuels. In this regard, the previous investigations carried out about the ability of Soltrol, a non
aromatic non volatile petroleum hydrocarbon, to be
attracted by soils, both by capillarity and physicochemical clay-fluid interactions contributes to characterize the status of the fluid but does not provide
Figure 14. Oil retention curves of compacted samples at

w = 15.2 at various densities (A: 1.3 Mg/m3 ; B: 1.73 Mg/m3 ;
C: 1.61 Mg/m3 ).
information about phase changes and decontamination

processes.
Actually, researches about phase changes and venting techniques have been conducted in the framework
of environmental engineering, a discipline devoted to
pollution issues in general that put more emphasis
on the physico-chemical behaviour of fluids (liquid and gas) with less attention paid on the soils
characteristics. Many laboratory investigations have
been conducted on sands and that conducted on finegrained soils emphasized the added difficulty due to
the presence of clay.
A first important issue addressed is that of knowing whether local or non local equilibrium conditions
prevail at liquid interfaces where phase changes occur.
Obviously, local equilibrium is more likely at low venting rates, when the imposed air velocity along the
interface is small.
By running column tests in which a cylindrical
sand sample (6.60 cm in diameter and 13.96 cm in
height) was submitted to venting along the axial direction, Farhan et al. (2002) determined the relationship
between the decrease in contaminant content (two
chlorinated solvents, PCE and TCE, bonded by capillarity) and the related increase in gas permeability,
as shown in Figure 15. Cumulative mass lost was
determined by regularly weighing the column. Air permeability was determined based on Darcys law and
gas pressure measurements in both inlet and outlet.
Interestingly, the curves are bilinear showing the
consecutive evaporation of the two contaminants. The
larger air pore velocity (5.70 m/s) results in a lower
residual saturation.
Farhan et al. (2002) developed a soil model composed of parallel capillary tubes (in the direction of air
flow) lined with the contaminant. Each tube is divided
into segments. The evaporation in each segment is
127
Figure 16. Comparison of various modelling approaches

for binary mixture at w = 1516% (Harper et al. 2003).
Figure 15. Variation of measured mixture mass (solid symbols) and permeability (open symbols) with time versus
calculated values (lines), medium Ottawa sand (0.30.6 mm),
relative density 1 (Farhan et al. 2002).
modelled by diffusion and calculated by the following

equation (Crank 1975):
where a2n are the positive roots of Bessels function

of the first kind and order zero, D the diffusion of
the contaminant in air, Mt the mass that is diffused
into the cylinder within time t and M the final mass
that diffuses into the cylinder. The diffused contaminant is moved to the next segment by the flowing air
(advection).
Given the good agreement between the data of
the 60 experiments that they conducted (with water,
PCE and TCE as fluids) on sands samples of various
densities and the results obtained by their diffusion
model, Farhan et al. (2002) concluded that local equilibrium was ensured at the velocities considered in
their tests (from 2 to 9.2 cm/s). They emphasized the
necessity of accounting for the increase in permeability while the contaminant was removed to correctly
predict the process of venting. They confirmed the
correctness of the well known cubic law describing
the changes in permeability with the fluid degree of
saturation:
This statement obtained in somewhat ideal conditions of having a fluid (or a fluid mixture) contained
in a sand and put in direct contact with venting air

is however not confirmed by other authors who commented about some delayed decontamination and tail
effects (a too high concentration remains after some
venting period, that cannot be governed in the framework of equilibrium between phases). Among other
things, a possible reason for retardation is, in contaminated layers submitted to water table fluctuations,
the fact that a portion of free NAPL can be isolated and entrapped as blobs of residual NAPL, in
a layer called smear layer. As a result, contaminant
removal by soil venting extraction can be controlled
by slow diffusive transport through the water filled
pores isolating the residual NAPL from the air flow
pathways (Harper et al. 2003). Indeed, Harper et al.
(1998) observed that contaminant removal in a silt
loam (34% sand, 50% silt and 16% clay) was affected
by the soil water content, with local equilibrium in
dry conditions (w = 2.7%) and obvious mass transfer limitations at a water content of 22%, even on the
presence of an appreciable volume of NAPL. Also, to
get closer from complex contaminant mixtures such
as gasoline, Harper et al. (2003) carried out tests
with a quaternary mixture of petroleum hydrocarbons
components.
Non equilibrium models are based on the use of
interface mass transfer kinetics, with generally constant mass transfer coefficients (e.g. Lingineni and
Dhir 1992, Armstrong et al. 1994). In such approaches,
the evaluation of mass transfer coefficients has to
be made by curve fitting in the absence of effective correlation for soils. Some authors completed
the analysis by accounting for the structure of finegrained soils by modelling the effects of aggregate in
soils, with immobile domains consisted of water filled
micropores and mobil domains consisted of air-filled
macropores. For instance, Gierke et al. (1992) considered advection and diffusion in the air filled domain,
radial diffusion in the water filled spherical aggregates
and first order mass transfer kinetics between the two
domains.
Harper et al. (2003) proposed a model in which the
contaminant removal due to soil venting was described
128
by two equations including one for the mobile vapour

phase contaminant transport, as follows:
in which g is the air volumetric fraction, q

the superficial air velocity (m/h), Dej the effective molecular diffusion coefficient (m2 /h), Kg a the
overall air NAPL volumetric mass transfer coefficient (h1 ), Cgj the vapour phase concentration
(mol/m3 ), Kgl the air/NAPL partition coefficient, Clj
the NAPL concentration (mol/m3 ) and j the specie
index.
The second term of this balance equation illustrates
the diffusion process whereas the third term introduces the air NAPL mass transfer coefficient taken
non constant as follows:
in which ol is the initial NAPL volumetric fraction and

m (h1 ) an adjustable parameter capturing dependence
of the overall volumetric mass transfer coefficient on
the NAPL content, so as to model the decrease in
NAPL recovery with lower NAPL content.
The model also includes a second contaminant
transport equation for the NAPL, aqueous and solid
phases as follows:
in which l is the volumetric NAPL content, w the

volumetric water content and b the bulk density f
the porous medium. The NAPL/aqueous phase interaction is described by the partition parameter Kwl
(m3 /kg) and the sorption of NAPL along the solid
phase defined by the sorbed/NAPL partition parameter
Ksl (m3 /kg).
Figure 16 shows how different approaches compare,
with a two fast decreases in effluent vapour concentration (toluene and xylene) obtained with constant
air/NAPL mass transfer coefficient Kg a. Parameters
from the two equations above other than mass transfer
coefficients have been either taken from literature (in
particular sorption coefficients) or from experimental data. The curves were fitted by adopting a relevant
value of Kg a.
Figure 17 shows the behaviour of a quaternary
mixture (hexane, toluene, m-xylene and trimethyl benzene) during removal experiment under dry conditions
(Harper et al. 2003).
Indeed, the mass transfer coefficient appeared to
depend on the soils water content (higher coefficient in dry state) and of the composition of the
mixture.
Gidda et al. (2011) stressed the importance of
also considering the aqueous/air and sorbed/aqueous
mass transfer coefficients and provided some expressions. They also evidenced the significant effect of
Figure 17. Behaviour of quaternary mixture removal

experiment under dry conditions (Harper et al. 2003).
organic matter in sorption and the difficulty of properly accounting for the effect of the clay fraction to
fully explain the mass transfer resistance observed in
fine grained soils.
CONCLUDING REMARKS
Multiphase aspects constitute an essential component

in the pollution and remediation of unsaturated soils by
immiscible pollutants like petroleum hydrocarbons or
chlorinated solvents. They also play a major role in the
most common remediation technique of soil vapour
extraction by soil venting.
The extension to immiscible fluids of the knowledge and techniques of investigating the retention and
transfer of water in unsaturated soils provided some
basic tools that provided relevant data able to help
model the movements of NAPL in the vadose zone.
In granular soils and sands in which only capillary
actions prevail, the extension of data from air water
interface could be successfully extended to the third
phase constituted by the NAPL. In fine-grained soils,
it appeared that, in spite of being non polar, NAPLs
have some complex interactions with the clay fraction that is probably not fully understandable through
the diffuse double layer concepts developed for water
solutions of ions (Lagaly 1984). Indeed, some significant interactions can be macroscopically observed by
running common geotechnical tests like Atterbergs
129
limits, plastic and shrinkage indexes, compressibility

and permeability.
These interactions obviously play a role and affect
the retention and transfer behaviour. They also affect
the phase change behaviour. This complex aspect has
been faced by environmental scientists by using geochemical tools. Whereas in sands or in dry fine-grained
soils local equilibrium appeared to be respected under
common air velocities, allowing for vapour extraction calculations with diffusion laws, non equilibrium
states and retardation processes were evidenced in wetter fine-grained soils. These effects, related to sorption
effects (along clay minerals and organic matter) are
accounted for by adopting mass transfer parameters
that are to be fitted, once known all the other involved
parameters. In other words, the rate of vapour extraction governed by diffusion models is slowed down by
a limited availability of the NAPL, modelled by the
mass transfer coefficient.
Interesting results have been obtained I this framework, including with mixtures of up to 4 components.
However, these mass transfer parameters have to be
fitted once all other parameters known, given that, for
the same soil, they depend on the soil density and water
content.
The results of such advanced analysis obviously
help further understanding of the many complex phenomena interacting during soil vapour extraction,
permitting a better use of this efficient technique in
practice.
REFERENCES
Armstrong J.E., Frind E.O. and McClellend R.D. 1994. Non
equilibrium mass transfer between the vapor, aqueous and
solid phases in unsaturated soils during vapour extraction.
Water Resour. Res., 30(2), 355368.
Barbour S. L. 1998. The soil-water characteristic curve: a
historical perspective. Can. Geotech. J. 35, 873894.
Busby R.D., Lenhard R.J. and Rolston D.E. 1995. An investigation of saturation capillary pressure relations in two
and three-fluid systems for several NAPLs in different
porous media. Groundwater J., 33(4), 570578.
Crank J. 1975. The mathematics of diffusion, 2nd Ed., Oxford
University Press, Oxford, United Kingdom.
Croney D. 1952. The movement and distribution of water in
soil. Gotechnique, 3 (1): 116.
Daniel D.E. 1982. Measurement of hydraulic conductivity of
unsaturated soils with thermocouple prychrometers. Soil
Sci. Soc. Am. J., vol. 46, No. 6, 11251129.
Ebadi T. and Kermani M. 2010. The effect of crude oil contamination on geotechnical properties of fine grained soils
in the vicinity of Tehran oil refinery site. Proc. 6th Int.
Cong. Env. Geotech., 669674, New Delhi, Balkema.
Farhan S., Budiman J. and Holsen T. 2002. Experimental
investigation of the interaction of soil air permeability and
sopil vapour extraction. J. Env. Eng. 128(2), 120130.
Fernandez F. and Quigley R.M. 1985. Hydraulic conductivity of natural clays permeated with simple liquid
hydrocarbons. Can. Geotech. J. 22, 205214.
Ferrand L.A., Milly P.C.D., Pinder G.F., Turrin P.D. 1990. A
comparison of capillary pressure-saturation relations for
drainage in two- and three-fluids porous media. Advances
in Water Ressources 13 (2), 5463.
Fredlund D.G. & Rahardjo H. 1993. Soil mechanics for
unsaturated soils. John Wiley, New York.
Gierke J.S., Hutzler N.J. and McKenzie D.B. 1992. Vapor
transport in unsaturated soil columns: implication for
vapour extraction. Water Ressour. Res. 28(2), 323335.
Harper B.M., Stiver W.H. and Zytner R.G. 1998. Influence
of water content on SVE in a silt loam soil. J. Env. Eng.
124(11), 10471053.
Harper B.M., Stiver W.H. and Zytner R.G. 2003. Nonequilibrium nonaqueous phase liquid mass transfer model for soil
vapour extraction systems. J. Env. Eng. 129 (8), 745754.
Izdebska-Mucha D. and Korzienowska-Rejmer E. Selected
characteristics of clay soils polluted by petroleum substances in the context of their barrier properties. Proc. 6th
Int. Cong. Env. Geotech., 8486, New Delhi, Balkema.
Khamehchiyan M., Charkhabi A.H. and Tajik M. 2007.
Effects of crude oil contamination on geotechnical properties of clayey and sandy soils. Eng. Geol. 89 (34),
200229.
Lagaly G. (1984). Clay-organic interactions. Phil. Trans.
Royal Soc. London A 311, 315332.
Lingineni S. and Dhir V.K. 1992. Modelling of soil venting processes to remediate unsaturated soils. J. Env. Eng.
118(1), 135152.
Lenhard R.J. and Parker 1988. Experimental validation of the
theory of extending two-phase saturation pressure relations to three fluid phase systems for monotonic drainage
paths. Water Resources 24(3), 373380.
Leverett M.C. 1941. Capillary behaviour in porous solids.
Trans. Am. Inst. Min. Metall. Eng., Pet. Eng. Div. 142,
152169.
Manassero M., Musso G., Rabozzi C., Ribotta L. 2005.
Retention curves for a polluted soils. Proc. Int. Symp.
Advances Exp. Unsat. Soil Mech., Trento, Balkema,
459465.
Mercer J. W. and Cohen R.M. 1990. A review of immiscible
fluids in the subsurface: properties, models, characterization and remediation. J. Contam. Hydrol. 6, 107103.
Mitchell J.K. and Soga K. 2005. Fundamentals of soil
behavior. J. Wiley. New-York.
Prost R. and Sposito G. 1982. Structure of water adsorbed on
smectites. Chem. Rev. 82 (6) 553573.
Rabozzi C., Ribotta L., Gremigni G. 2006. Retention curves
and hydraulic properties of a soil contaminated by NAPL.
Proc. 5th Int. Cong. Env. Geotech., 12321239, Cardiff,
ICE London.
Shang J.Q., Lo K.Y. and Quigley R.M. 1994. Quantitative determination of potential distribution in Stern-Gouy
doublr-layer model. Can. Geotech. J. 31, 624636.
Singh S.K., Srivastava R.K. and John S. 2009. Studies on soil
contamination tue to used motor oil and its remediation.
Can. Geotech. J. 46, 10771083.
130

Electrokinetic remediation of soils at complex contaminated sites:

Technology status, challenges, and opportunities
Krishna R. Reddy
University of Illinois at Chicago, Chicago, Illinois, USA
ABSTRACT: Subsurface contamination has become a widespread and pervasive global problem. Heavy metals
and organic compounds often coexist, and such mixed contaminant distribution within the subsurface is highly
dependent on particle and macro-scale heterogeneities. Vast amounts of resources have been invested to develop
efficient remediation technologies, yet few have been successful. In-situ remediation is often preferred due
to lower site disturbance, safety, simplicity, and cost-effectiveness. However, the effectiveness of in-situ technologies depends largely on contaminant chemistry and subsurface heterogeneities (including particle-scale
heterogeneities, e.g. fine-grained soils, soils with reactive minerals and/or organic matter, and macro-scale
heterogeneities, e.g. irregular soil layers and/or lenses). Under complex site and contaminant conditions, electrokinetic remediation has great potential. This paper presents the status of this technology based on numerous
laboratory investigations and limited field demonstrations, the challenges in applying this technology at actual
field sites and opportunities for electrokinetics as an efficient and economical remedy for complex contaminated
sites.
Keywords: Electrokinetic Remediation; Soils; Pollution; Heavy Metals; Organic Compounds
INTRODUCTION
Polluted sites pose a serious hazard to public health

and the environment. The United States Environmental Protection Agency (USEPA) estimated that over
217,000 sites require urgent cleanup at an estimated
cost of over $187 billion (USEPA, 1997). Soil and
groundwater contamination has been a major problem
at these sites. The types of contaminants encountered at these sites include heavy metals such as lead
(Pb), chromium (Cr), nickel (Ni), cadmium (Cd) and
mercury (Hg), and organic compounds such as polycyclic aromatic hydrocarbons (PAHs) and polychlorinated biphenyls (PCBs). Heavy metals and organic
contaminants are often found to coexist at many sites.
Recently, environmental professionals have focused
on risk-based approaches to remediating polluted sites.
Risk assessment includes detailed site characterization, human and ecological risk quantification and
selection of remedial goals (Sharma and Reddy, 2004).
The end-use of the site is also taken into consideration
in the risk assessment process. If the risk posed by the
contamination is unacceptable, remedial action must
be selected and implemented to achieve the remedial
goals in an efficient and cost-effective manner. Remediation of polluted sites can also help preserve the
green lands (pristine land) from pollution due to industrial development as well as provide opportunities for
economic growth.
Several technologies have been developed to remediate contaminated sites (Sharma and Reddy, 2004).
These technologies can be grouped under ex situ or
in situ technologies. Ex situ remediation technologies involve removing the contaminated soils and/or
groundwater from the subsurface and then treating
them on-site or off-site. On the other hand, in situ remediation technologies involve treating the contaminated
soils and/or groundwater in-place without removal
from the subsurface. Often, in situ remediation technologies are preferred because they minimize site
disturbance and reduce exposure of the contaminants
to personnel and the surrounding public. In addition,
in situ technologies are often simple and less costly.
Common in situ soil remediation technologies include
soil vapor extraction, soil flushing, solidification and
stabilization, thermal desorption, vitrification, bioremediation, and phytoremediation. Common in-situ
groundwater remediation technologies include pump
and treat, air/ozone sparging, flushing, permeable
reactive barriers, immobilization, chemical oxidation,
and bioremediation.All of these remediation technologies are based on physicochemical, thermal or biological processes that aim to remove the contaminants from
the soils and groundwater or immobilize and/or detoxify the contaminants within the soils and groundwater.
The USEPA has documented inadequate performance of various remediation technologies implemented at numerous polluted sites (USEPA, 2000,
131
2006). The inadequate performance of the remediation

technologies may be attributed to complex site conditions, such as: (1) clayey soils are difficult to remediate
because of their low permeability and complex composition (mineralogy and organic content); (2) many
common technologies fail under heterogeneous subsurface conditions (e.g., clay lenses within sand formation); (3) hydrophobic organic contaminants such as
PAHs and PCBs are difficult to treat; (4) very few technologies are available to cleanup sites contaminated
with heavy metals; and (5) very few technologies exist
that can remediate sites contaminated by mixed contaminants (e.g., organic compounds combined with
heavy metals and/or radionuclides). There is an urgent
need to develop new technologies that can overcome
these challenges as well as be cost-effective (Reddy,
2010). In situ electrokinetic remediation has great
potential to fill this need.
A comprehensive electrokinetic remediation
research program has been on-going at the University
of Illinois at Chicago since 1993. The main objectives of this research program are to: (1) investigate
geochemistry/biogeochemistry during electrokinetic
remediation in different soil and contaminant environments, and (2) investigate methods to engineer
geochemistry/biogeochemistry to favor enhanced contaminant remediation. The results of this research have
provided the most comprehensive fundamental knowledge to develop electrokinetic remediation as a practical soil and groundwater remediation technology. The
research approach includes bench-scale experiments
to investigate process fundamentals and optimal operational parameters and mathematical modeling to serve
as a screening and optimization tool.
This paper provides an overview of fundamental
aspects of electrokinetic remediation and describes
different studies conducted to assess the effectiveness of electrokinetic remediation to cleanup soils
polluted by various types of contaminants in clayey
soils and/or heterogeneous soils. The challenges and
opportunities of applying electrokinetic remediation
in the field are highlighted. Attempts to integrate electrokinetic remediation with other technologies such as
chemical oxidation, bioremediation and phytoremediation, to make the technology practical, efficient and
economical are summarized.
ELECTROKINETIC REMEDIATION
TECHNOLOGY
An electrokinetic remediation technology for the

in-situ treatment of contaminated sites consists of
drilling wells (drains or trenches) in which electrodes are installed and then applying a very low
direct current electric potential. Pumping and conditioning systems may be needed at the electrodes
depending on the site conditions. A schematic of an
in situ electrokinetic remediation system is shown
in Figure 1. Similarly, electrokinetic treatment may
Figure 1. Schematic of in-situ electrokinetic remediation

system.
be accomplished ex-situ by using specially designed

above-ground reactors. Generally, the contaminants
accumulated at the electrodes are removed by either
adsorption onto the electrodes or withdrawal followed
by treatment. Electrokinetic remediation offers the
following advantages as compared to conventional
remediation methods: (1) simplicity requires simple
equipment; (2) safety the personnel or the public in the vicinity are not exposed to contaminants;
(3) wide range of contaminated media can be used
for soils, sludges, sediments, and groundwater (particularly well-suited for low-permeability clays and
heterogeneous soil deposits within the vadose zone
where conventional remedial methods have proven
to be ineffective or expensive); (4) wide range of
contaminants can be used for metals, organic compounds, radionuclides, or a combination of these
contaminants; (5) flexibility can be used as an insitu or ex-situ remediation system, and it can be easily
integrated with other remediation technologies such
as bioremediation; and (6) cost-effectiveness- requires
low electrical energy (relative to other thermal technologies) leading to lower overall cost, with costs
ranging from $20 to $225 per cubic yard depending
on the site-specific conditions.
The successful implementation of electrokinetic
remediation requires a thorough understanding of
the transport mechanisms and physico-chemical processes that affect the transport and fate of contaminants
under an induced electric potential. The major contaminant transport mechanisms under an induced electric
potential are: (1) electroosmosis bulk movement
of pore fluid through the electrical double layer in
clayey soils, generally occurring from anode to cathode; (2) electromigration transport of ions and ion
complexes within the pore fluid towards oppositely
charged electrodes; (3) electrophoresis transport
of charged colloids, micelles, bacterial cells, etc.
within the pore fluid towards oppositely charged electrodes; and (4) diffusion transport of chemicals due
132
to concentration gradients (Acar and Alshawabkeh

1993). Electroosmosis is the major transport process
for non-polar organic compounds, while electromigration is the dominant transport process for ionic
compounds. These two processes govern the overall
contaminant migration in compact soil systems; and
the roles of the other two processes, electrophoresis
and diffusion, are often negligible. However, electrophoresis may be a significant contaminant transport
process in contaminated sludge and sediments.
Research conducted to date shows that the electrochemical processes are quite complex and are
influenced by local geochemistry. As a result of the
induced electric potential, electrolysis of water occurs
at the electrodes. The electrolysis reactions generate
H+ ions and O2 gas at the anode and OH ions and
H2 gas at the cathode. The gases may be allowed
to escape into the atmosphere, while the H+ ions
migrate towards the cathode and OH ions migrate
towards the anode. Depending on the extent of migration of H+ and OH ions, pH changes occur across
the soil. Generally, low pH (acidic) conditions exist
near anode and high pH (basic) conditions exist near
cathode. The pH changes in the soil will affect the
geochemical processes, namely adsorption and desorption, precipitation and dissolution and oxidation
and reduction.
The lower soil pH near the anode causes desorption and solubilization of cationic (negatively charged)
metals, such as Pb, Ni and Cd, enhancing their electromigration towards the cathode. However, the higher
pH near the cathode causes these metals to adsorb
and/or precipitate, hindering electromigration and
removal at the cathode. The change in pH also affects
the surface charge of soil particles. If the pH is less than
the point of zero charge (PZC), the soil surfaces are
positively charged. However, if the pH is greater than
the PZC, the soil surfaces will be negatively charged.
PZC is the pH at which the net charge on the particle
surfaces is zero. The changes in surface charge will
affect the diffuse double layer, consequently affecting
the electroosmotic flow and adsorption of contaminants. Thus, changes in surface charge will affect
the migration and removal of both organic and metal
contaminants. Therefore, the effects of pH and other
geochemical (redox potential, electrolyte concentration, etc.) changes under applied electric potential on
the site-specific soil and contaminant conditions and
remediation should be assessed.
For electrokinetic remediation to be feasible, contaminants must be desorbed and/or solubilized in the
soil. H+ transport causes desorption or solubilization
of cationic metals, while OH transport causes desorption of anionic metals. Electromigration of cations and
anions occurs towards the cathode and anode, respectively. Electroosmotic advection also causes ions and
non-polar organics to transport towards the electrodes.
The contaminants are then captured in the electrolyte
at electrodes and treated above-ground using common
wastewater treatment technologies.
Figure 2. Schematic of typical bench-scale electrokinetic

test setup.
ELECTROKINETIC REMOVAL
OF HEAVY METALS
3.1 Systems effects

Several series of bench-scale experiments have provided valuable information on the geochemistry and
transport of heavy metals under applied electric
potential. Figure 2 shows the typical bench-scale
electrokinetic test setup used. Experiments have been
conducted using kaolin and glacial till, spiked with
typical heavy metals such as Cr, Ni, Cd, and Hg, individually and in combinations. Kaolin is a clayey soil
with homogeneous mineralogy, while glacial till represents a field soil with complex mineralogy. Typical
properties of these two soils are summarized in Table 1.
The simultaneous testing on these two soils under identical contaminant and treatment conditions allowed the
investigation of soil compositional effects on the extent
of heavy metal removal. Cr, Ni and Cd were chosen as
typical heavy metals as they are commonly prevalent,
toxic and need to be remediated (e.g., electroplating
sites). In addition to the spiked soils, experiments were
also conducted on highly variable actual field soils
obtained from sites contaminated with multiple heavy
metals. Most of the experiments were conducted under
an electric potential of 1 VDC/cm, as this voltage gradient was found to be optimal and practical (Reddy
and Shirani 1997).
The removal of heavy metals when they exist individually and in combination in soils was investigated
using the soils that were contaminated with Cr(VI)
only, with Ni(II) only, and with Cr(VI), Ni(II), and
Cd(II) combined (Reddy et al. 2001; Reddy and
Parupudi 1997; Reddy et al. 1997). Typical results
shown in Figure 3(a) demonstrate that cationic metals migrate towards the cathode, but their migration is
retarded by high pH near cathode. In contrast, as shown
in Figure 3(b), anionic metals migrate towards anode,
but their migration is slowed by low pH near anode.
133
Table 1.
Properties of kaolin and glacial till soils.
Property
Kaolin
Glacial till
Mineralogy
Kaolinite:
100%
Muscovite:
trace
Illite: trace
Particle size distribution

(ASTM D422)
Gravel
Sand
Silt
Clay
Atterberg limits
(ASTM D2487)
Liquid limit
Plastic limit
Plasticity index
Specific gravity
(ASTM D854)
Hydraulic conductivity
(cm/s) (ASTM D5084)
pH (ASTM D4972)
Cation exchange capacity
(meq/100 g)
(ASTM D9081)
Organic content
(ASTM D2974)
USCS classification
(ASTM D2487)
0%
Quartz: 31%
Feldspar: 13%
Carbonate: 35%
Illite: 15%
Chlorite: 46%
Vermiculite: 0.5%
Smectite: trace
0%
4%
18%
78%
20%
44%
36%
50%
27.4%
22.6%
21.7%
11.7%
10.0%
2.6
Figure 3b. Migration and removal of Cr(VI) in kaolin due

to electrokinetic treatment.
2.71
8
1.0 10
4.1 108
4.9
11.6
8.2
1318
2.8%
CL
CL
Figure 3c. Migration and removal of Ni(II) in glacial till
due to electrokinetic treatment.
Figure 3d. Migration and removal of Cr(VI) in glacial till

due to electrokinetic treatment.
Figure 3a. Migration and removal of Ni(II) in kaolin due to
electrokinetic treatment.
Electromigration is the most dominant transport process for heavy metals. The different compositions of
soils lead to different contaminant migration behavior.
For example, glacial till soil possesses high acid buffering capacity due to its high carbonate content and the
soil remains alkaline even after the application of electric potential. The high soil pH hinders the migration of
cationic metals and enhances the migration of anionic
metals (Figures 3(c) and 3(d)).
It was found that in kaolin, a significant pH variation occurred due to electric potential application,
affecting the adsorption-desorption and dissolutionprecipitation, as well as the extent of migration of the
contaminants. In glacial till, however, pH changes were
not affected significantly. In both kaolin and glacial
till, the migration of Cr(VI) and Ni(II) was higher when
they were present individually compared to when they
existed together with Cd(II). Cr(VI) migration as single or combined contaminant was lower in kaolin as
compared to that in glacial till. This result was due to
134
the low pH conditions created near the anode in kaolin

that led to high Cr(VI) adsorption to the clay surfaces.
In glacial till, however, Ni precipitated with or without
the presence of co-contaminants due to high pH conditions in the soil. Overall, this study demonstrated that
adsorption, precipitation and reduction are the significant hindering mechanisms for the removal of heavy
metals using electrokinetic remediation. The direction
of the contaminant migration and overall removal efficiency depend on the polarity of the contaminant, the
presence of co-contaminants and the geochemistry and
composition of the soil.
The transient behavior of Cr, Ni and Cd during
electrokinetic remediation was assessed by performing electrokinetic experiments with 1, 2, 4, and 10 d
of treatment time (Reddy and Al-Hamdan 2008). In all
tests, the contaminants were Cr(VI), Ni(II) and Cd(II)
combined in the soil. Results showed that in kaolin, the
extent of Ni(II) and Cd(II) migration towards the cathode increased as the treatment time increased. Unlike
kaolin, in glacial till the treatment time had no effect
on Ni and Cd migration because of its high buffering
capacity. In both kaolin and glacial till, the extent of
Cr(VI) migration towards the anode increased as the
treatment time increased. However, Cr(VI) migration
was higher in glacial till as compared to kaolin due to
the higher pH. In all tests, some Cr(VI) was reduced to
Cr(III), and the Cr(VI) reduction rate to Cr(III) as well
as the Cr(III) migration were significantly affected by
the treatment time.
The initial moisture content of the soils as well as
initial contaminant concentrations can also influence
the electrokinetic remediation (Reddy et al. 2002).
These aspects were investigated by conducting experiments with soils at different moisture contents (30,
50 and 70%) with same heavy metal contaminants,
Cr(III), Ni(II) and Cd(II), at concentrations of 1000,
500 and 250 mg/kg, respectively. Then, two additional
experiments were conducted at lower Cr(III), Ni(II)
and Cd(II) concentrations of 500, 250 and 125 mg/kg
and 250, 125 and 62.5 mg/kg, respectively. These
tests all used the same moisture content of 30%.
The results revealed that the electrical current and
electroosmotic flow increased considerably when the
soil contained higher moisture content (70%), but
the moisture content effect on heavy metal contaminant migration and removal appeared to be minimal. Furthermore, these experiments indicated that
the moisture content remains nearly the same during the electrokinetic process. The results showed
that soils possessing higher contaminant concentrations could have a lower percentage of contaminant
migration and removal. Although similar amounts of
removal occurred for the Cr(III) contaminant, Ni(II)
and Cd(II) migration was evidently inhibited when the
three co-contaminants were present in the test using the
highest concentrations of 1000, 500 and 250 mg/kg,
respectively.
Some heavy metals can have several valence states
and the electrokinetic removal can depend on the contaminant valence state (Reddy and Chinthamreddy
2003). For example, Cr can exist as Cr(III) or Cr(VI).

In order to investigate this aspect, electrokinetic experiments were conducted with soils containing Cr(III)
alone, Cr(VI) alone or a combination of Cr(III) and
Cr(VI), all with the same total Cr concentration of
1000 mg/kg. Ni(II) and Cd(II) in concentrations of
500 mg/kg and 250 mg/kg, respectively, were also
introduced into the soils as co-contaminants to simulate typical electroplating waste constituents. The test
results showed that Cr migration was highest when it
was present in kaolin in the Cr(III) form and in glacial
till in the Cr(VI) form. When Cr was present in Cr(III)
form, migration occurred towards the cathode due to
the existence of Cr(III) as cation and cationic hydroxide complexes. Cr(III) migration was not observed
in glacial till because of precipitation that resulted
from the high pH conditions that existed throughout the glacial till. However, when Cr was present
in Cr(VI) form, the migration occurred towards the
anode, due to the existence of Cr(VI) as soluble oxyanions. The migration of Cr(VI) was higher in glacial till
as compared to kaolin due to alkaline conditions that
existed in the glacial till, resulting in negligible Cr(VI)
adsorption to soil solids. When Cr was present as a
combination of Cr(VI) and Cr(III), Cr(VI) migrated
towards the anode, while Cr(III) migrated towards
the cathode. For these cases, the total Cr migration
was lower than the migration observed when only
Cr(III) was present in kaolin or when only Cr(VI)
was present in glacial till. No migration was observed
for the co-contaminants Ni(II) and Cd(II) in glacial
till due to precipitation as a result of alkaline conditions. In kaolin, however, Ni(II) and Cd(II) migrated
towards the cathode. The test results showed that significant removal of contaminants from the soils was
not achieved for the processing periods utilized.
The presence of natural oxidizing or reducing
agents in the soil can influence speciation of the
contaminants and affect their electrokinetic removal
(Reddy and Chinthamreddy 1999; Chinthamreddy and
Reddy 1999). To investigate this, bench-scale electrokinetic experiments were conducted using the soils
with and without a reducing agent. The reducing agent
used was either humic acid, ferrous iron or sulfide,
at a concentration of 1000 mg/kg. The soils were then
spiked with Cr(VI), Ni(II) and Cd(II) at concentrations
of 1000, 500 and 250 mg/kg, respectively, and tested
under an induced electric potential of 1 VDC/cm for
a duration of over 200 h. The reduction of Cr from
Cr(VI) to Cr(III) occurred prior to electrokinetic treatment. The extent of this Cr(VI) reduction was found
to be dependent on the type and amount of reducing
agents present in the soil. The maximum reduction
occurred in the presence of sulfides, while the minimum reduction occurred in the presence of humic
acid. The concentration profiles in both soils following
electrokinetic treatment showed that Cr(VI) migration
was retarded significantly in the presence of sulfides
due to the reduction of Cr(VI) to Cr(III) as well as
an increase in soil pH. This low migration of Cr is
attributed to: (1) migration of Cr(VI) and the reduced
135
Cr(III) fraction in opposite directions, (2) low Cr(III)

migration due to adsorption and precipitation in high
pH regions near the cathode in kaolin and throughout
the glacial till, and (3) low Cr(VI) migration due to
adsorption in low pH regions near the anode in both
soils. Ni(II) and Cd(II) migrated towards the cathode in
kaolin; however, the migration was significantly lessened in the presence of sulfides due to increased pH
through most of the soil. Initial high pH conditions
within the glacial till resulted in Ni(II) and Cd(II)
precipitation, so the effects of reducing agents were
inconsequential. Therefore, one must characterize the
soils for the presence of the reducing agents, particularly sulfides, in soils that may affect redox chemistry
and soil pH, ultimately affecting the electrokinetic
remediation process. Similarly, the effects of naturally
occurring oxidizing agents should also be carefully
assessed.
3.2
Geochemical assessment
In order to better understand speciation and distribution of heavy metals before and after electrokinetic
remediation, various geochemical analyses including
transmission electron microscopy (TEM), energy dispersive X-ray spectroscopy (EDX) and X-ray diffraction (XRD) and sequential chemical extractions as well
as geochemical modeling were performed.
TEM, EDX and XRD: A detailed investigation of
physical distribution of heavy metals in kaolin soil and
the chemical and structural changes in kaolinite minerals that result from electrokinetic remediation was
performed by TEM, EDX and XRD analyses on the
soil samples before and after electrokinetic remediation (Roach et al. 2009). Results showed that the heavy
metal contaminant distribution in the soil samples was
not observable using TEM and EDX. X-ray diffraction patterns showed a decrease in peak height with
decreasing soil pH value, which indicates the possible
dissolution of kaolinite minerals during electrokinetic
remediation. The changes in particle morphology were
found to be insignificant, but a relationship was found
between the crystallinity of kaolin and the pH changes
induced by the applied electric potential.
Sequential Extractions: Sequential extractions
were performed on the contaminated soils before and
after electrokinetic treatment to provide an understanding of the distribution of the contaminants in
the soils (Reddy et al. 2001). The speciation of contaminants after electrokinetic treatment showed that a
significant change in exchangeable and soluble fractions occurred. In kaolin, exchangeable and soluble
Cr(III), Ni(II) and Cd(II) decreased near the anode and
increased near the cathode, whereas exchangeable
and soluble Cr(VI) decreased near the cathode and
increased near the anode. In glacial till, exchangeable and soluble Cr(III), Ni(II) and Cd(II) were low
even before electrokinetic treatment and no significant changes were observed after the electrokinetic
treatment. However, significant exchangeable and soluble Cr(VI) that was present in glacial till prior to
electrokinetic treatment decreased to non-detectable

levels near the cathode and increased significantly near
the anode. In both kaolin and glacial till, low migration
rates occurred as a result of contaminants existing as
immobile complexes and precipitates.
Geochemical Modeling: The speciation of the
metals was predicted using the chemical speciation
program MINEQL+. The results showed that the speciation and distribution of cationic metals [Ni, Cd
and Cr(III)] in glacial till soil remain unaffected or
slightly affected during electrokinetics (Al-Hamdan
and Reddy 2008). This is attributed to the high pH and
buffering capacity of the glacial till, leading the metals
to precipitate in the soil prior to and after electrokinetics. This study showed that during electrokinetics,
Cr(VI) exists as anionic complex and migrates towards
the anode and the migration is greatest in case of a
single-contaminant system.The study also showed that
near the anode in the absence of any reducing and oxidizing agent, Cr(VI) mostly adsorbed, and some of
Cr(VI) reduced to Cr(III) and migrated towards the
cathode and finally precipitated due to high pH conditions. Ni and Cd remain adsorbed or precipitated due
to the high pH conditions throughout the soil.
A geochemical assessment of how the contaminants are held within the kaolin soil under induced
electric potential was also made by using the equilibrium model MINEQL+ (Al-Hamdan and Reddy
2006). As the soil pH changes due to applied electric potential, the adsorption of the Cr(VI), Ni(II)
and Cd(II) by kaolin was implemented as an electrostatic behavior. The FITEQL 4.0 model was used
to determine the equilibrium constants of the electrostatic adsorption model of kaolin for the metals by
optimizing the experimental titration and adsorption
data of kaolin. The adsorbed and precipitated forms of
Cr(III), Ni(II) and Cd(II) increased near the cathode
and decreased near the anode, whereas the adsorbed
form of Cr(VI) increased near the anode as well as
in the middle region. However, the precipitated form
of Cr(III), Ni(II) and Cd(II) as Cr2 O3 or Cr(OH)3 ,
Ni(OH)2 and Cd(OH)2 , respectively, dominated over
their adsorbed form as they got closer to the cathode.
A one-dimensional transport model was developed
to predict the transient transport and speciation of
heavy metals in soil during electrokinetic remediation
as a function of time and space (Al-Hamdan and Reddy
2008). The model assumes: (1) pH-dependent adsorption of contaminants to the soil surface; (2) sensitivity
of soil surface potential and electroosmotic flow to
the pore water properties; and (3) synergistic effects
of multiple chemical species on electrokinetic remediation. The model considers that: (1) electrical potential
in the soil is constant with time; (2) surface complexation reactions are applicable in the highly concentrated
clay suspensions; (3) the effect of temperature is
negligible; and (4) dissolution of soil constituents
is negligible. The predicted pH profiles, electroosmotic flow and transport of Cr, Ni, and Cd in kaolin
soil during electrokinetic remediation were found to
reasonably agree with the bench-scale electrokinetic
136
experimental results. The predicted contaminant

speciation and distribution (aqueous, adsorbed and
precipitated) allowed for an understanding of the transport processes and chemical reactions that control
electrokinetic remediation.
3.3 Enhancement strategies
Several investigations have demonstrated that using
water as an electrolyte does not advance the removal of
heavy metals from the soils. Therefore, different strategies are needed to enhance removal efficiency. These
strategies include: (1) increasing treatment duration,
(2) increasing electric potential gradient, (3) applying
electric potential in different modes (e.g., continuous, periodic, AC/DC combinations, etc.), (4) using
cation/anion exchange membranes in the electrodes,
(5) circulating electrolytes, and (6) using enhancement
(electrode conditioning) solutions such as organic
acids (e.g. acetic acid, citric acid) and chelating agents
(e.g. EDTA, DTPA). The use of enhancement solutions
has been found to be the most promising strategy to
achieve high removal efficiency.
Batch and electrokinetic experiments were conducted to investigate the removal of different heavy
metals from kaolin soil by using ethylenediamine
tetraacetic acid (EDTA) as a complexing agent (Reddy
et al. 2004). Experiments were conducted on kaolin
spiked with Cr(VI), Ni(II) and Cd(II) at concentrations of 1000, 500 and 250 mg/kg, respectively, which
simulate typical electroplating waste contamination.
The batch experiments revealed that highly successful
removal of these heavy metals (62100%) was possible
by using either a 0.1 M or 0.2 M EDTA solution over
a wide range of pH conditions (210). However, the
results of the electrokinetic experiments using EDTA
at the cathode showed low heavy metal removal efficiency. Using EDTA at the cathode along with the
pH control (NaOH) at the anode increased the pH
throughout the soil and achieved high (95%) Cr(VI)
removal, but the removal of Ni(II) and Cd(II) was limited due to the precipitation of these metals near the
cathode. Apparently, the low mobility of EDTA and
its migration direction, which opposed electroosmotic
flow, prevented EDTA complexation from occurring.
Subsequently, several other purging solutions were
evaluated to enhance the removal of the heavy metals
from kaolin (Reddy and Chinthamreddy 2003). The
removal of heavy metals from the soil using tap water
as the purging solution was very low. When 1 M acetic
acid was used as the purging solution in the cathode,
the removal of Cr, Ni and Cd was increased to 20, 19
and 13%, respectively. Using 0.1 M EDTA as the purging solution in the cathode, 83% of the initial Cr was
removed; however, the Ni and Cd removal was very
low. A sequentially enhanced electrokinetic remediation approach involving the use of water as a purging
solution at both the anode and cathode initially, followed by the use of acetic acid as the cathode purging
solution and a NaOH alkaline solution as the anode
purging solution, was tested. This sequential approach
resulted in a maximum removal of Cr(VI), Ni(II) and

Cd(II) of 6871, 7173 and 8794%, respectively. This
study shows that the sequential use of appropriate electrode purging solutions, rather than a single electrode
purging solution, is necessary to remediate multiple
heavy metals in soils using electrokinetics.
A similar investigation was performed on glacial till
soil (Reddy and Chinthamreddy 2004). The electrolyte
solutions tested were 0.1 M EDTA, 1.0 M acetic acid,
1.0 M citric acid, 0.1 M NaCl/0.1 M EDTA, and 0.05 M
sulfuric acid/0.5 M sulfuric acid. The results showed
that 4682% of the Cr(VI) was removed from the soil,
depending on the purging solution used. The highest
removal of Ni(II) and Cd(II) was 48 and 26%, respectively, achieved by using 1.0 M acetic acid. Although
cationic contaminant removal was low, the use of 0.1 M
NaCl as an anode purging solution and 0.1 M EDTA as
a cathode purging solution resulted in significant contaminant migration towards the soil regions adjacent
to the electrodes. Compared to low buffering capacity
soils, such as kaolin, the removal of heavy metals from
the glacial till soil was low, likely due to the complex
composition of glacial till.
Besides Cr(VI), Ni(II) and Cd(II), another toxic
metal of great concern is mercury (Hg) and its
removal from soils using electrokinetics was investigated (Reddy et al. 2003). Initially, batch tests were
performed with soils spiked with Hg(II) to investigate Hg desorption and complexation under different pH environments (pH 212). The complexing
agents included Na-EDTA, potassium iodide (KI), and
sodium chloride (NaCl), and these solutions were used
at a concentration of 0.1 M. In addition, deionized
water was used for comparison purposes. Based on
the batch tests, Na-EDTA and KI were identified as
the complexing agents with the greatest potential. The
removal efficiency of these complexing agents was
then examined by conducting electrokinetic experiments that employed the same solution concentration
(0.1 M) and voltage gradient (1.0 VDC/cm) conditions. These tests indicated that for both soils, KI was
a more effective complexing agent than Na-EDTA
under electrokinetics. For the kaolin soil, the electrokinetic treatment using KI removed approximately
97% of the initial contaminant present (500 mg/kg of
Hg(II)), leaving a residual concentration of 16 mg/kg
of Hg in the soil, whereas on the glacial till soil, KI
removed only 56% of the initial contaminant present
(500 mg/kg of Hg(II)), leaving a residual concentration of 220 mg/kg of Hg in the soil. The lower Hg
removal from glacial till is attributed to the presence
of organic matter, which increased Hg adsorption or
the formation of insoluble Hg complexes.
Further studies investigated optimizing the iodideenhanced electrokinetic treatment for the removal
of Hg from soils (Reddy et al. 2003). Experiments
were conducted at two voltage gradients (1.0 or
1.5 VDC/cm) to evaluate the effect of the voltage gradient when employing a 0.1 M KI solution. Additional
experiments were performed to assess the effect of
using a higher iodide concentration (0.5 M KI) when
137
using a 1.5 VDC/cm voltage gradient. The tests conducted on the kaolin soil showed that when the 0.1 M
KI concentration was employed with the 1.0 VDC/cm
voltage gradient, approximately 97% of the Hg was
removed, leaving a residual concentration of 16 mg/kg
in the soil. The tests conducted on glacial till indicated that it was beneficial to use the higher (0.5 M
KI) iodide concentration and the higher (1.5 VDC/cm)
voltage gradient to enhance Hg removal, because,
under these conditions, a maximum of 77% of the
Hg was removed from the glacial till, leaving a residual concentration of 116 mg/kg in soil after treatment.
Compared to kaolin, the lower Hg removal from the
glacial till soil is attributed to its more complex soil
composition, such as the presence of carbonates and
organic matter, which caused Hg(II) to adsorb to the
soil and/or exist as an immobile chemical species.
The above studies were conducted on soils spiked
uniformly with heavy metals; however, soils at contaminated sites often possess variable compositions
and multiple metals (Reddy and Ala 2005). In order
to investigate the effects of field conditions, a clayey
soil contaminated with multiple heavy metals was
obtained from an actual contaminated site. In addition to the contaminants present on-site, additional
Pb and Hg were added to soils in order to simulate
the source zone contamination. Bench-scale electrokinetic experiments were conducted under a voltage
gradient of 2 VDC/cm and a hydraulic gradient of 1.4
using four different extracting solutions (0.2 M EDTA,
0.2 M DTPA, 0.2 M KI and 10% HPCD). EDTA and
KI were found to be efficient in the removal of Pb
and Hg, respectively. On a mass-efficiency basis, the
EDTA-enhanced system is found to be more effective
for the simultaneous removal of a variety of metals,
while KI was found to be effective for the selective
removal of Hg from the field soil.
4
4.1
OF ORGANIC CONTAMINANTS
System effects
Many sites around the globe are contaminated with

polycyclic aromatic hydrocarbons (PAHs) such as
phenanthrene and chlorinated aromatic and nitroaromatic compounds such as pentachlorophenol (PCP),
2,4-dinitrotoluene (2,4-DNT), and lindane. There are
several Superfund sites (surface and subsurface soils)
contaminated with these toxins, which are considered priority pollutants by the USEPA. Remediation
of these contaminants is more complicated in low
permeability and/or heterogeneous soils. These contaminants are difficult to remove from soils due to
their low solubility and strong adsorption to soil surfaces and organic matter. Electrokinetically enhanced
in situ flushing using solubilizing agents, (i.e. surfactants, cosolvents and cyclodextrins) has the potential
to remove these contaminants from low permeability
clay soils. A solubilizing agent is used to increase
contaminant desorption and solubilization, and an
applied electric potential facilitates contaminant transport primarily by electroosmosis and/or electrophoresis (Saicheck and Reddy 2005; Cameselle and Reddy
2012). Electroosmotic flow generally improves soil
solutioncontaminant interaction, especially in zones
of high organic content and/or low hydraulic conductivity. The electrokinetically enhanced remediation
process is fairly simple to implement and operate, but
the fundamental reactions that govern the remediation
method are complex. An adequate knowledge of the
contaminant transport mechanisms and the physical,
chemical and electrochemical processes is essential in
order to optimize system performance.
4.2
Remediation of PAHs
An investigation was conducted to evaluate electrokinetically enhanced flushing for the removal of PAHs
from kaolin and glacial till using different flushing
solutions: deionized water, a surfactant, or a cosolvent (Reddy and Saichek 2003). The results indicated
that the contaminant was more strongly bound to the
glacial till than the kaolin, and this was attributed to its
higher-organic content. The glacial till also generated
a greater electrical current and electro-osmotic flow,
and this was probably a result of its higher-carbonate
content and more diverse mineralogy. Based on the
contaminant mass remaining in the soil, it was apparent that the surfactant or cosolvent solution caused
contaminant desorption, solubilization, and/or migration in both soils, but PAH removal efficiency was low.
Additional evaluation of surfactants to remove PAHs
during electrokinetics was performed by Saichek and
Reddy (2003).
Different cosolvents (n-butylamine, tetrahydrofuran or acetone) were further investigated on glacial
till with phenanthrene (Li et al. 2000). Desorption
equilibrium was investigated by batch tests. Water or
20% (v/v) cosolvent solution was constantly supplied
at the anode. The concentration of phenanthrene in the
effluent collected at the cathode was monitored for
100 to 145 days. Results showed that the presence of
n-butylamine significantly enhanced the desorption
and electrokinetic transport of phenanthrene; about
43% of the phenanthrene was removed after 127 days
or 9 pore volumes. The effect of acetone was not
as significant as butylamine. The effluent flow in
the tetrahydrofuran experiments was minimal, and
phenanthrene was not detected in the effluent. The
use of water as the conducting solution did not cause
observable phenanthrene migration.
It is proposed that in soils with low buffering capacity (e.g., kaolin), controlling the pH at the anode to
counteract the electrolysis reaction and prevent low
pH conditions, which are responsible for low electroosmotic flow, will improve contaminant removal
(Saicheck and Reddy 2003). Three different flushing
solutions [deionized water, a surfactant or a cosolvent with and without a 0.01 M NaOH solution] at
the anode to control the pH were tested. The test using
deionized water with pH control generated a higher
electroosmotic flow than the equivalent test performed
138
Figure 4b. Electroosmotic flow in kaolin during pulsed

(periodic) electrokinetic treatment.
Figure 4a. Migration and removal of phenanthrene in kaolin

during pulsed (periodic) electrokinetic treatment.
without pH control, but the electroosmotic flow difference between the surfactant and cosolvent tests with
and without pH control was minor compared to that
observed with the deionized water tests. Controlling
the pH was beneficial for increasing contaminant solubilization and migration from the soil region adjacent
to the anode, but the high contaminant concentrations
that resulted in the middle or cathode soil regions
indicated that subsequent changes in the soil and/or
solution chemistry caused contaminant deposition and
low overall contaminant removal efficiency.
To improve removal efficiency, pulsed electrokinetics was investigated. It consists of a periodic voltage
application on a 7-day cycle of 5 days of continuous application and 2 days of down time, when the
voltage was not applied (Reddy and Saicheck 2004).
The periodic voltage effects were evaluated by performing four different bench-scale electrokinetic tests
with the voltage gradient applied continuously or periodically, under relatively low voltage (1.0 VDC/cm)
and high anode buffering (0.1 M NaOH) as well as
high voltage (2.0 VDC/cm) and low anode buffering
(0.01 M NaOH) conditions. For all the tests, kaolin
clay soil was spiked with 500 mg/kg phenanthrene. A
nonionic polyoxyethylene surfactant, Igepal CA 720,
was used as the flushing solution in all tests.The results
of these experiments show that considerable contaminant removal can be achieved by employing a high,
2.0 VDC/cm, voltage gradient along with a periodic
mode of voltage application (Figure 4a). The increased
removal was attributed to increased phenanthrene solubilization and mass transfer due to the reduced flow
of the bulk solution during the down time as well as to
the pulsed electroosmotic flow that improved flushing
action (Figure 4(b)). Overall, such studies have shown
that electrokinetic remediation is a viable technique to
remove hydrophobic organic compounds from soils.
4.3 Remediation of chlorinated aromatic and
nitroaromatic contaminants
An investigation was undertaken to determine if low
permeability soils contaminated with PCP can be
remediated using the electrokinetic technique (Reddy

et al. 2011). A total of six tests were performed using
kaolin soil spiked with 100 mg/kg of PCP. The first
three tests were performed with an applied voltage gradient of 1 VDC/cm, where each test employed one of
the three different flushing solutions: deionized water,
electrolyte or buffered electrolyte. The other three tests
were performed using the same electrolyte solution,
but each employed a voltage gradient of 2 VDC/cm
under constant and periodic application modes and
constant voltage gradient with electrolyte liquid recirculation. The results showed that PCP can be degraded
in an electrokinetic system via direct electrochemical reduction at the electrodes. The degradation of
PCP ranged from 52% to about 78% depending on
the conditions present. As the applied voltage gradient across the PCP-contaminated soil increased, the
electroosmotic flow and PCP migration toward the
cathode increased, resulting in higher PCP degradation. In the tests with electrolyte flushing solution, PCP
degradation was about 58% and 65% under the applied
voltage gradients of 1 and 2 VDC/cm, respectively.The
mode of application of voltage potential across the
PCP-contaminated soil showed a noticeable effect on
the system pH and electroosmotic flow and hence the
PCP degradation. The highest PCP degradation (i.e.,
78%) was in the test with constant 2 VDC/cm voltage gradient and recirculation application. Overall,
this study demonstrated that electrokinetic technology
has the potential to remediate PCP-contaminated clay
soils by the direct reductive process. The electroosmotic flow and the degree of PCP degradation during
electrokinetics were dependent on the applied potential gradient and properties of the aqueous phase such
as pH, ionic strength and presence of carbonates.
Nitroaromatic compounds have also been successfully degraded via electrokinetics. Khodadoust et al.
2006 assessed the removal of 2,4-dinitrotoluene (2,4DNT), a munitions waste constituent and an industrial
intermediate, from contaminated soils using enhanced
electrokinetic remediation. Kaolin and glacial till soils
were spiked with 480 mg of 2,4-DNT/kg of dry soil
and deionized (DI) water and cyclodextrin solutions
were used as purging solutions. Cyclodextrin was
139
selected as a nonhazardous solubility enhancer for

enhancing the desorption and removal of 2,4-DNT
from soils in EK remediation. Two aqueous solutions
of hydroxypropyl-cyclodextrin (HPCD) at concentrations of 1 and 2% were selected for kaolin and glacial
till, respectively, based on results for batch extraction of 2,4-DNT from the same soils. Less 2,4-DNT
remained in the kaolin soil (up to 94% transformed)
than in the glacial till soil (20% transformed) due to
strong retention of 2,4-DNT by the soil organic matter in glacial till. For kaolin, less 2,4-DNT remained in
the soil using HPCD solutions than using DI water. For
glacial till, comparable levels of 2,4-DNT remained in
the soil for both EK solutions. Since no 2,4-DNT was
detected in the effluents from the EK cells, the decrease
in 2,4-DNT concentration in the kaolin and glacial till
soils was attributed to electrochemical transformation
of 2,4-DNT to other species.
A comparative assessment on the removal of PCP,
lindane and DNT using electrokinetic technique was
performed (Reddy et al. 2011).Three bench-scale electrokinetic experiments were conducted on kaolin soil.
In each test, the soil was spiked with one of the following contaminants: PCP, lindane and DNT at target
concentrations of 100, 500 and 1000 mg/kg, respectively. The applied voltage gradient across the soil
in each test was 1 volt direct current (VDC)/cm for
about 500 h. The study demonstrated that degradation
of PCP, lindane and DNT occurred during electrokinetics without using any solubilizing or reducing
agents in the flushing solution. Results showed that
PCP, lindane and DNT were degraded by a direct
electrochemical reduction at the electrodes. Contaminant degradation across the soil ranged from 40%
to 95%, 28% to 80%, and 26% to 60% for PCP, lindane and DNT, respectively. The effectiveness of the
electrochemical reduction of the contaminant during
electrokinetics depended on the contaminant solubility
and soil pH. Overall, this study demonstrated that electrokinetic technology has the potential to remediate
chlorinated aromatic and nitroaromatic contaminants
in low permeability soils via direct reduction.
5
OF MIXED CONTAMINANTS
Previous studies have shown that electrokinetic remediation has potential to remove heavy metals and
organic compounds when they exist individually in
low permeability soils. Mixed contaminants, a combination of heavy metals and organic contaminants, are
often encountered at contaminated sites (e.g., manufactured gas plant sites). Electrokinetic remediation
can induce substantial and uniform electroosmotic
flow through low permeability and heterogeneous
soils, and remove both metals and organic contaminants by electromigration, electroosmosis and electrophoresis (Reddy et al. 1999).
The feasibility of using surfactants in electrokinetic
remediation was evaluated to remove PAHs in the
Figure 5. Simultaneous removal of phenanthrene and nickel

during electrokinetic remediation using surfactant (Igepal).
presence of heavy metals from clayey soils (Maturi

et al. 2009). Kaolin was selected as a model clayey
soil and it was spiked with phenanthrene and Ni at a
concentration of 500 mg/kg-dry wt. each to simulate
typical mixed contamination. Bench-scale electrokinetic tests were performed using deionized water and
two different surfactants, Igepal CA-720 and Tween
80, at 3% and 5% (w/w) each at the anode. These solutions at the anode were circulated and were buffered
with 0.01 M NaOH to maintain neutral pH conditions. A periodic voltage gradient of 2 V/cm (with
5 days on and 2 days off cycles) was applied for
all the tests. There was a significant migration of
phenanthrene towards cathode in all the tests. Among
all the extractant solutions used, complete removal
of phenanthrene was observed using 5% Igepal
CA-720 (Figure 5). In case of Tween 80, the low electroosmotic flow limited the delivery of Tween 80 into
the soil and therefore, limited phenanthrene solubilization and transport. Nickel electromigrated as a cation
towards the cathode and precipitated in the soil near
the cathode due to the high pH in all tests. Overall, Ni
migration pattern was not affected by the presence of
phenanthrene and the extracting solutions used in this
study.
Instead of surfactants, the feasibility of using
cyclodextrins was evaluated for the same soil and contaminant conditions (Maturi and Reddy 2006). Benchscale electrokinetic experiments were conducted using
hydroxypropyl b-cyclodextrin (HPCD) at low (1%)
and high (10%) concentrations and using deionized
water in control test. A periodic voltage gradient
of 2 VDC/cm (with 5 days on and 2 days off) was
applied to all the tests, and 0.01 M NaOH was added
during the experiments to maintain neutral pH conditions at anode. In all tests, Ni migrated as Ni2+ ions
towards the cathode and most of it was precipitated
as Ni(OH)2 within the soil close to the cathode due
to high pH condition generated by electrolysis reaction. Even though high flow was generated in tests
using deionized water and 1% HPCD, migration and
removal of phenanthrene was low due to low solubility
of phenanthrene in these solutions. The test with 10%
140
HPCD solution showed higher solubility of phenanthrene, which caused it migrate towards the cathode,
but further migration and removal was retarded due
to reduced electric current and electroosmotic flow.
Approximately one pore volume of flushing resulted
in about 50% removal of phenanthrene from the soil
near the anode.
Subsequently, the feasibility of using cosolvents to
enhance the electrokinetic removal of PAHs was investigated (Maturi and Reddy 2008). Experiments were
performed using n-butylamine (cosolvent) at concentrations of 10 and 20% and deionized water, each
mixed with 0.01 M NaOH solution and circulated at the
anode to maintain alkaline conditions. A periodic voltage gradient of 2 VDC/cm in cycles of 5 days on and
2 days off was applied in all the tests. During the initial
stages when the soil pH was low, Ni existed as a cation
and electromigrated towards the cathode. However, as
the soil pH increased due to hydroxyl ions generated at
the cathode and also flushing of high pH n-butylamine
solution from the anode, Ni precipitated with no further migration. Phenanthrene was found to migrate
towards the cathode in proportion to the concentration
of n-butylamine. The extent of phenanthrene removal
was found to depend on both the electroosmotic flow
and the concentration of n-butylamine, but the presence of Ni did not influence the transport and removal
of phenanthrene.
In the above cases, heavy metals were not effectively removed from the soil. Therefore, the feasibility
of using surfactants and organic acids sequentially
and vice versa during electrokinetic remediation was
evaluated for the removal of both PAHs and heavy metals from the soils (Reddy et al. 2009). Bench-scale
electrokinetic experiments were performed with the
sequential anode conditioning using (1) 1 M citric acid
followed by 5% Igepal CA-720; (2) 1 M citric acid followed by 5% Tween 80; and (3) 5% Igepal CA-720
followed by 1 M citric acid. A periodic voltage gradient of 2 V/cm with 5 days on and 2 days off cycles
was applied in all the tests. A removal of about 96% of
phenanthrene was observed in the test with 5% Igepal
CA-720 followed by 1 M citric acid sequence. Most of
the Ni (90%) migrated from anode to cathode in this
test; however, it precipitated in the section very close
to the cathode due to the high pH conditions. Conversely, the removal efficiency of Ni was about 96 and
88% in the tests with 1 M citric acid followed by 5%
Igepal CA-720 sequence and 1 M citric acid followed
by 5% Tween 80 sequence, respectively. However, the
migration and removal efficiency of phenanthrene in
both of these tests were very low. Overall, it can be
concluded that the sequential use of 5% Igepal CA720 followed by 1 M citric acid may be an effective
remedial strategy to remove coexisting heavy metals
and PAHs from clayey soils.
The above investigations were conducted using
kaolin or glacial till spiked with phenanthrene and
Ni. At actual contaminated sites, the soil composition can be varied and multiple metals and PAHs
could be present. To investigate the effectiveness of
electrokinetics on field soils, a manufactured gas plant

(MGP) soil contaminated with PAHs and heavy metals was used in laboratory tests (Reddy et al. 2006).
Four flushing agents, which included two surfactants
(3% Tween 80, and 5% Igepal CA-720), one cosolvent (20% n-Butylamine) and one cyclodextrin (10%
hydroxypropyl--cyclodextrin or HPCD) were examined to enhance the solubilization of PAHs in the
soil. Four electrokinetic experiments were conducted
at 2.0 VDC/cm voltage gradient and 1.4 hydraulic gradient. It was found that the cosolvent increased the soil
pH, while the surfactants and HPCD did not induce
substantial change. Electroosmotic flow was the highest with the cosolvent, while the lowest flow was
observed with Tween 80 surfactant. Igepal CA-720
surfactant yielded the highest removal efficiency due
to the partial solubilization of PAHs, causing some
PAHs to migrate towards the cathode. Heavy metals
were found to be strongly adsorbed/precipitated and
showed negligible migration behavior in all the tests.
Based on the contaminant mass remaining in the soil,
it is apparent that further optimization of the electrokinetic system is required to improve PAH removal
efficiency for the MGP soil.
A similar study was conducted on contaminated
sediment obtained from Indiana Harbor (Reddy and
Ala 2005). The sediment is a fine-grained material
with high moisture (78%) and organic (19%) contents and is contaminated with a wide range of PAHs
and heavy metals (Reddy et al. 2010). Four benchscale electrokinetic experiments were conducted at
2.0 VDC/cm of constant voltage gradient using two
surfactants: 5% Igepal CA-720 and 3% Tween 80,
a cosolvent: 20% n-butylamine and a cyclodextrins:
10% HPCD as flushing solutions for the solubilization/desorption of PAHs. The mobility and removal
of heavy metals was also investigated while flushing
these solutions. The experimental results show that
20% n-butylamine and 5% Igepal 720 systems are
effective for the partial solubilization of the PAHs
in the sediments; however, both 3% Tween 80 and
10% HPCD systems are ineffective for desorption/
solubilization of PAHs in the sediment. None of the
selected flushing agents is effective for the removal
of heavy metals from the contaminated sediment. The
low removal of PAHs and heavy metals is attributed
to the high organic content and high acid buffering
capacity of the sediment.
Electrokinetic remediation is generally applied to
low permeability soils, but it can also be applied to
sandy soils to solubilize and enhance removal of the
contaminants. An integrated use of hydraulic flushing
and electrokinetic treatment was investigated for the
remediation of silty sand contaminated by both PAHs
and heavy metals. The soil was collected from another
polluted former manufactured gas plant (MGP). Four
bench-scale experiments were conducted to analyze
the ability of the combined hydraulic flushing and the
electrokinetic treatment for the simultaneous removal
of PAHs and heavy metals. Sequential flushing with
EDTA or Igepal CA-720 was tested with or without the
141
Figure 6a. Flow and mass of phenanthrene removal from

sand during surfactant flushing.
Figure 6c. Flow and mass of phenanthrene removal from
soil profile with top clay layer and bottom sand layer during
electrokinetically enhanced surfactant flushing.
Figure 6b. Flow and mass of phenanthrene removal from

kaolin during electrokinetically enhanced surfactant flushing.
simultaneous application of a low intensity direct electric field (1 VDC/cm). The best results were obtained
with 0.2 M EDTA flushing in two stages (without
and with voltage gradient, 1 VDC/cm), followed by
5% Igepal flushing in two stages (without and with
1 VDC/cm). Heavy metals were removed mainly during the EDTA flushing, with removal efficiencies of
about 60% for Zn, 80% for Pb and 30% for Cu. During Igepal flushing, no heavy metals were removed,
but PAHs were removed, including 40% phenanthrene,
30% pyrene and 20% benzo[a]pyrene. Combining
electrokinetics with hydraulic flushing did not improve
contaminant removal from the soil.
6
ELECTROKINETIC REMEDIATION IN
HETEROGENEOUS SUBSURFACE
Generally, subsurface conditions are heterogeneous

in nature consisting of clay layers and sand layers
and/or clay layers interbedded in sand formations.
Common remediation techniques based on hydraulic
flushing can only remediate sand formations and the
contamination associated with clay layers and lenses
cannot be remediated. An investigation was undertaken to determine the applicability of electrokinetic
remediation technology to treat contaminated soils
Figure 6d. Flow and mass of phenanthrene removal from

sand layer with interbedded clay lenses during electrokinetically enhanced surfactant flushing.
under soil heterogeneities such as layers, lenses,

and mixtures of different soils (Saicheck and Reddy
2005). Specifically, this study evaluated surfactantenhanced electrokinetic remediation of PAHs under
heterogeneous soil conditions. A series of benchscale experiments was conducted using two soils
(sand and kaolin) spiked with a representative PAH
compound (phenanthrene) in a two-dimensional electrokinetic test apparatus under various layered, lens,
or mixed soil configurations (Figure 6). In addition,
the homogeneous sand and kaolin soils were each
tested alone for comparison purposes. All the experiments employed the same nonionic surfactant (5%
Igepal CA-720) flushing solution and a low (0.05)
hydraulic gradient. The results showed that the surfactant flushing under the low hydraulic gradient alone
was sufficient for complete removal of the contaminant from the homogeneous sand profile (Figure
6(a)), whereas the electroosmotic flow generated by
the application of a DC 2.0 V/cm electric potential in
a periodic mode considerably enhanced the removal
efficiency for the homogeneous and heterogeneous
soil profiles containing kaolin (Figures 6(b), 6(c)
142
and 6(d)). The voltage gradient varied spatially and

temporally through the soil profiles and affected the
electroosmotic flow and contaminant removal.
7
CHALLENGES AND OPPORTUNITIES
7.1 Challenges
Electrokinetic remediation is uniquely applicable to
remediate complex sites containing low permeability and heterogeneous soils contaminated by heavy
metals, organic contaminants or both. However, many
practical challenges need to be overcome to make
this technology feasible for field application. Some
of these challenges include:
The effectiveness of the technology is sensitive to

variable in situ geochemistry. Hence, laboratory test
results may not be reflective of expected variable
geochemical conditions in the field.
The complex field soil and aged multiple contaminants lead to low remedial efficiency as compared to laboratory demonstrations using spiked,
homogeneous soils.
The remediation time can be longer and difficult to predict, making the technology unattractive
if the site needs to be closed quickly for future
development.
Except for the in situ electrochemical reduction
of some organic contaminants, the contaminants
are transported into the electrodes wells/trenches;
effluent needs to be pumped out and treated above
ground either on-site or off-site. The treatment of
such effluent requires regulatory permits and incurs
additional treatment costs.
The technology cost is high due to high cost of materials such as electrodes and electrode conditioning
solutions and electrical power.
It may be difficult to get regulatory approval to
inject electrode conditioning solutions (e.g., EDTA,
surfactants, etc.) into the ground as the fate and
risk of their residual presence in the subsurface is
unknown.
Many investigations are based on spiked soils and
very limited information is available on field contaminated sites. The well documented field pilot
studies are extremely scarce.
Full-scale applications are very limited in the US.
Some field applications such as Lasagnaprocess
are well documented.
Technology developers information on pilot or
full-scale field applications is often incomplete.
For example, Lageman (2005) reported several
successful full-scale field applications dealing
with inorganic contaminants in the Netherlands,
but information available on design, implementation and monitoring is limited. US-based ElectroPetroleum, Inc. claims completion of several
projects, but it has not published detailed technical
information.
As the technology is material- and energy-intensive,
it may not be as sustainable as other technologies.
7.2 Opportunities
Implementation of electrokinetic remediation as a
stand-alone technology to remove contaminants from
soils may be inefficient, uneconomical and unsustainable. Nevertheless, electrokinetic remediation is
uniquely applicable for complex sites with low permeability/heterogeneous soils and mixed contaminants.
For such sites, the common remediation technologies
(e.g. pump and treat, soil flushing, permeable reactive barriers, chemical stabilization/oxidation, in situ
bioremediation, and soil heating) may not be effective or efficient. Integrating electrokinetic remediation
judiciously with common technologies to overcome
the limitations of the common technologies may provide an effective and efficient remedial strategy for
complex sites. For example, electrokinetic remediation may be integrated with common technologies
such as chemical oxidation/reduction, bioremediation
and phytoremediation to remediate sites with the following advantages: (1) organic contaminants can be
detoxified within the soil, hence no effluent treatment
is required; (2) removes metals from soils, providing
a long-term, effective solution; (3) remediates mixed
contaminants (both metals and organic contaminants);
(4) cost-effective; (5) practical; and (6) green and sustainable with a renewable energy source (e.g., solar
power). Some examples of integrated technologies are
presented below.
7.3 Coupled electrokinetic-chemical oxidation
technology
Fenton Oxidation of PAHs: An integrated electrochemical oxidation process that utilizes electrokinetics
(EK) to deliver the oxidant (510% hydrogen peroxide, H2 O2 ) and chelant (40 mM EDTA or DTPA)
or iron chelate (1.4 mM Fe-EDTA or Fe-DTPA) to
oxidize PAHs in soils was investigated (Reddy and
Chandhuri 2009). Batch and bench-scale EK experiments were conducted using: (a) kaolin spiked with
phenanthrene at 500 mg/kg and (b) former manufactured gas plant (MGP) soil, a high buffering silty
soil, contaminated by a variety of PAHs (1493 mg/kg).
Batch experiments showed that chelant solutions dissolve native iron minerals to form soluble Fe-chelates
that remain available even at higher pH conditions of
soil for the Fenton-like oxidation of the PAHs. In EK
experiments, a 510% H2 O2 solution was delivered
from the anode and a chelant solution or iron-chelate
was delivered from the cathode. Preflushing of soil
with 5% ethanol and ferrous sulfate (1.4 mM) prior
to oxidant delivery was also investigated. An electric
potential of 2 VDC/cm was applied in all tests to induce
electroosmotic flow for 58 days for kaolin and 25 days
for the MGP field soil. In the absence of any chelating agent, phenanthrene oxidation was catalyzed by
native iron present in kaolin soil, and 49.882.3% of
phenanthrene was oxidized by increasing H2 O2 concentration from 510%. At 5% H2 O2 concentration,
phenanthrene oxidation was not increased by using
40 mM EDTA, 40 mM DTPA or 1.4 mM Fe-DTPA,
143
Figure 7a. Residual phenanthrene distribution after integrated electrokinetic Fenton-like oxidation treatment.
Figure 7b. Residual nickel distribution after integrated

electrokinetic Fenton-like oxidation treatment.
but it increased to 70% using 1.4 mM Fe-EDTA. Maximum phenanthrene oxidation (90.5%) was observed
by 5% ethanol preflushing and then treating with 5%
H2 O2 at the anode and 1.4 mM Fe-EDTA at the cathode. However, preflushing with 1.4 mM ferrous sulfate
did not improve phenanthrene oxidation. The findings of the MGP field soil test indicated that delivery
of 5% H2 O2 alone resulted in oxidation of 39.8% of
total PAHs (especially 2- and 3-ring PAHs). The use of
EDTA and Fe-EDTA did not increase PAHs oxidation
in this soil. Overall, an optimized in situ combined
technology of EK and Fenton-like process has the
potential to oxidize PAHs in low permeability and/or
high buffering soils.
Fenton-Oxidation of PAHs and Simultaneous
Removal of Heavy Metals: The coupled Fentonlike oxidation and electrokinetic remediation was also
investigated for mixed contaminants (combination of
heavy metals and PAHs). This remediation process
aims at oxidation of organic contaminants and simultaneous removal of heavy metals. Fentons reagent,
consisting of H2 O2 and native iron catalyst, is utilized for chemical oxidation. Laboratory batch and
electrokinetic experiments were performed on kaolin
spiked with Ni and phenanthrene each at a concentration of 500 mg/kg of dry soil to represent typical
heavy metal and PAH contaminants found at contaminated sites. Experiments were conducted using H2 O2
solution in 5%, 10%, 20% and 30% concentrations and
also using deionized (DI) water as control (Reddy and
Karri 2008). For electrokinetic experiments, a voltage
gradient of 1 VDC/cm was applied and H2 O2 solution
was introduced at the anode for a total duration of four
weeks. Batch tests showed that phenanthrene oxidation
increases from 76% to 87% when the H2 O2 concentration increases from 5% to 30%. The electrokinetic
experiments showed substantial electroosmotic flow in
all the tests. Oxidation of phenanthrene increased with
increasing concentration of H2 O2 ; a maximum of 56%
oxidation was observed with 30% H2 O2 (Figure 7a).
Nickel migrated from anode to cathode.This migration
was more pronounced in the H2 O2 tests as compared
to the DI baseline test. Nickel precipitated in all the

tests near the cathode due to high pH conditions (Figure 7b). These results emphasize that the optimization
of H2 O2 /catalyst concentration and voltage gradient as
well as the control of soil pH are required to increase
the removal of Ni and the oxidation of phenanthrene.
In a supplemental study, the effect of voltage gradient on the remedial efficiency was investigated
by testing two voltage gradients, 1 and 2 VDC/cm
(Reddy and Karri 2006). The H2 O2 solution in two
different concentrations at 5% and 10% was introduced at the anode, and each experiment was conducted for a total duration of four weeks. The results
showed that increasing the voltage gradient from
1 VDC/cm to 2 VDC/cm did not increase the electroosmotic delivery of H2 O2 significantly. About 28%
and 34% of the phenanthrene were oxidized within
the soil in the 5% and 10% H2 O2 experiments, respectively, under 1 VDC/cm. The phenanthrene oxidation
increased to about 32% and 42% using the 5% and 10%
H2 O2 concentrations, respectively, under 2 VDC/cm.
Ni migrated towards the cathode and then precipitated close to the cathode, due to high pH conditions
in all the experiments. The Ni migration was slightly
higher in the case of 2 VDC/cm than in the case of
1 VDC/cm, due to greater migration of the acidic pH
front towards the cathode under 2 VDC/cm. Overall,
the results showed that an increase in the voltage gradient from 1 VDC/cm to 2 VDC/cm improved overall
remedial performance slightly.
Persulfate Oxidation of PCBs: Persulfate is
another oxidant used for remediation purposes. Integrating electrokinetic remediation with persulfate has
the potential to facilitate the delivery of persulfate in
low permeability soils as well as activate oxidizing radicals and simultaneously induce oxidative/reductive
reactions directly in the soil. In order to investigate
this, a series of laboratory batch and bench-scale electrokinetic experiments were conducted using kaolin
spiked with 50 mg/kg of 2,20,3,50 tetrachlorobiphenyl
(PCB 44), a representative PCB (Yukselen-Aksoy and
Reddy 2013). Persulfate oxidation activators [elevated
144
temperature (45 C) and high pH (at the cathode)]

were investigated to maximize the PCB degradation.
In addition, the effect of oxidant dosage on PCB
degradation was investigated. The electrokinetically
enhanced temperature-only activated persulfate oxidation test resulted in better PCB 44 remediation
(77.9%) than the temperature and high-pH activated
persulfate oxidation (76.2%) over a 7-day period. The
optimal dosage for effective remediation was 30%
Na-persulfate (76.2%) because a 20% concentration of
the oxidant yielded a lower rate of degradation (55.2%)
of PCB 44. These results are encouraging for the use
of electrokinetically enhanced persulfate oxidation for
the effective remediation of PCBs in soils.
In a supplemental study, the glacial till soil
was investigated to determine the effects of soil
type on electrokinetic-persulfate treatment (YukselenAksoy and Reddy 2012). The activation of persulfate
improved the oxidation of PCB in kaolin, but the effect
in the glacial till was insignificant. The final pH values
of the kaolin were around 1, while they were around 7
for glacial till because of the high buffering capacity of
glacial till. Low pH conditions are also shown to activate persulfate, leading to higher PCB oxidation. The
degradation of PCB 44 was significantly higher for the
kaolin than the glacial till. In kaolin, the highest level
of PCB oxidation, 77.9%, was achieved with temperature activated persulfate in 7 days. However, in the
glacial till the highest PCB oxidation was 14.4% with
30% Na-persulfate concentration without any activation. The high buffering capacity, nonhomogeneous
mineral content and high organic content of glacial till
may have contributed to the low persulfate oxidation
of PCB.
Permanganate Oxidant Delivery: Potassium permanganate (KMnO4 ) is another common oxidant used
for remediation purposes. A bench-scale study was
conducted to investigate the feasibility of using electrokinetics as a method of delivering KMnO4 into
kaolin. KMnO4 was introduced at the anode, the cathode and both electrodes to investigate (1) the extent
of oxidant transport via electroosmosis and electromigration and (2) to investigate the stability of the
oxidant at each electrode location. Separate tests were
performed under different voltage gradients to explore
the effect of electric potential on oxidant transport
and stability. Results indicate that (1) electromigration is not as effective as electroosmotic advection for
the transport of permanganate into the soil and (2)
permanganate may not be stable under the pH-redox
conditions that are imposed by the application of electric potential. Additional research is being conducted
to explore the stability of permanganate under electrokinetic environments and to determine whether a
substantial amount of permanganate can be injected
into low-permeability soils.
7.4
Coupled electrokinetic-chemical reduction

technology
Zero-valent iron (ZVI) has been demonstrated to

effectively dehalogenate organic contaminants in the
subsurface. ZVI is also used commonly in permeable

reactive barriers for treating groundwater. Recent synthesis of nanoscale iron particles (NIP) allow them to
be injected into contaminated soil zones for reduction of the organic contaminants. However, bare NIP
were found to be unstable, while lactate modified NIP
(LMNIP) was found to be stable and ideal for transporting through the subsurface soils. The challenge
is delivery of the LMNIP in low permeability soils;
hence electrokinetics was proposed to be used to inject
LMNIP into the contaminated soils. This approach
was evaluated in bench-scale electrokinetic experiments performed at constant voltage gradient with
DNT spiked kaolinite at a concentration of 920 mg/kg
(Reddy et al. 2011). Bare NIP or LMNIP (4 g/L)
was injected at some distance away from the anode.
Aluminum lactate 10% (w/w) was used as modifier for
LMNIP. The results showed 4165% of DNT degradation in the soil near the anode, while it was lower at
3034% near the cathode. The highest DNT degradation was achieved using LMNIP. The total degradation
of DNT was attributed to both NIP and electrochemical process. Overall, it was found that electrokinetic
system can enhance the delivery of LMNIP in low
permeability soils for the degradation of energetic
organic contaminants such as DNT.
7.5
Coupled electrokinetics-bioremediation
technology
Electrokinetics can serve as an effective nutrient delivery system for bioremediation of contaminants in low
permeability soils. Bioremediation can include degradation of organic contaminants or immobilization of
heavy metals. A preliminary laboratory investigation
was performed wherein electrokinetics was used for
the delivery of nutrients to metal-reducing microorganisms in a low permeability clayey soil (Reddy
et al. 2003). In particular, the microorganisms were
used to reduce a toxic and mobile Cr(VI) to a less toxic
and immobile form Cr(III). Three bench-scale electrokinetic experiments were conducted using kaolin
artificially contaminated with Cr(VI) at an initial
concentration of 1000 mg/kg. All the experiments
included a control test without micro-organisms or
nutrients, a test with microorganisms but without nutrients and a test with microorganisms and supplemental
nutrients, specifically acetate, phosphate and ammonium. The results showed that acetate and phosphate
amendment by electrokinetics was effective because
both nutrients electromigrated into the soil. Moreover,
the results indicate that employing the microorganism
cultures improved Cr(VI) reduction. These results suggest that nutrient amendment by electrokinetics for the
bioremediation of heavy metals has great potential;
however, the microbial strains responsible for Cr(VI)
reduction must be identified so the electrokinetic system can be engineered to provide the optimal nutrient,
pH and environmental conditions for these strains.
Several other studies have reported enhanced bioremediation of organic contaminants using electrokinetics
(Reddy and Cameselle 2009).
145
7.6
Coupled electrokinetics-phytoremediation
technology
Phytoremediation involves the removal, stabilization or degradation of contaminants in soils and

groundwater by plants. Both organic and inorganic
contaminants can be accumulated in plant tissues
or degraded via enzymatic activity. Alternatively,
contaminants can be transformed to other chemical species less bioavailable in the soil. The different mechanisms involved in phytoremediation
include: rhizofiltration, phytodegradation, phytoaccumulation (or phytoextraction), phytostabilization, and
rhizodegradation (also called phytostimulation). The
combination of phytoremediation and electrokinetic
remediation has been proposed in an attempt to avoid
the limitations of phytoremediation. The electric field
may enhance the removal of the contaminants by
increasing the bioavailability of the contaminants by
desorption and improved mobility of polar species.
Some significant variables that affect the coupled technology are: the use of AC or DC current, the voltage
level, the mode of voltage application (continuous or
periodic), the change in soil pH caused by the electrolysis of water at the electrodes, and the possible
addition of facilitating agents to enhance the mobility and bioavailability of the contaminants. Limited
laboratory studies are reported on combined electrokinetics and phytoremediation (Reddy and Cameselle
2009), but the most critical issue is on how it can be
readily implemented in the field.
8
CONCLUSION
Although many technologies have been developed to

remediate polluted sites, a substantial number of them
are ineffective or inefficient for complex sites containing low permeability soils, heterogeneous soils or
mixed contaminants. Electrokinetic remediation technology has great potential for in-situ remediation of
low permeability and/or heterogeneous soils that have
been contaminated by organics, heavy metals or a combination of these contaminants. However, electrokinetic remediation is highly dependent on site-specific
geochemical conditions such as the soil composition,
native electrolytes, contaminant aging, and contaminant mixtures. Several research studies have been
undertaken to develop fundamental geochemical characterization and enhance geochemistry for effective
electrokinetic remediation. Electrokinetic remediation
can be easily integrated with conventional remedial
systems to enhance remedial efficiency and decrease
overall remedial cost. Field pilot tests and full-scale
studies are limited, and many practical issues on implementing the technology at actual sites are yet to be
resolved.
REFERENCES
Acar, Y.B., and Alshawabkeh, A.N. 1993. Principles
of Electrokinetic Remediation. Environ. Sci. Technol.,
27(13):26382647
Al-Hamdan, A.Z., and Reddy, K.R. 2008. Geochemical

Assessment of Metal Transport in Glacial Till During Electrokinetic Remediation. Environ. Monit. Assess.
139(13):137149.
Al-Hamdan, A.Z., and Reddy, K.R. 2008. Transient Behavior
of Heavy Metals in Soils During Electrokinetic Remediation. Chemosphere, 71(5):860871.
Al-Hamdan, A.Z., and Reddy, K.R. 2008. Electrokinetic Remediation Modeling Incorporating Geochemical Effects. J. Geotech. Geoenviron. Eng-ASCE, 134(1):
91105.
Al-Hamdan, A.Z., and Reddy, K.R. 2006. Geochemical
Reconnaissance of Heavy Metals in Kaolin after Electrokinetic Remediation. J. Environ. Sci. Heal. A41(1).
Cameselle, C., and Reddy, K.R. 2012. Development and
Enhancement of Electro-osmotic Flow for the Removal of
Contaminants from Soils. Electrochim. Acta. 86:1012.
Chinthamreddy, S., and Reddy, K.R. 1999. Oxidation and
Mobility of Trivalent Chromium in Manganese Enriched
Clays During Electrokinetic Remediation. Journal of Soil
Contamination, 8(2), 1999, pp. 197216.
Khodadoust, A.P., Reddy, K.R., and Narla, O. 2006.
Cyclodextrin-Enhanced Electrokinetic Remediation of
Soils Contaminated with 2,4-Dinitrotoluene. J. Environ.
Eng-ASCE. 132(9):10431050.
Li, A., Cheung, K.A., and Reddy, K.R. 2000. Cosolvent
Enhanced Electrokinetic Remediation of Soils Contaminated with Phenanthrene. J. Environ. Eng-ASCE,
126(6):527533.
Maturi, K., Reddy, K.R., and Cameselle, C. 2009. Surfactantenhanced Electrokinetic Remediation of Mixed Contamination in Low Permeability Soils. Separation Science &
Technology, 44(10):23852409.
Maturi, K., and Reddy, K.R. 2008. Cosolvent-enhanced desorption and transport of organic and metal contaminants
in soils during electrokinetic remediation. Water, Air, and
Soil Pollution, 189(14):199211.
Maturi, K., and Reddy, K.R. 2006. Simultaneous Removal of
Heavy Metals and Organic Contaminants from Soils by
Electrokinetics using a Modified Cyclodextrin. Chemosphere, 63(6):10221031.
Reddy, K.R., Darko-Kagya, K. and Al-Hamdan, A.Z. 2011.
Electrokinetic Remediation of Pentachlorophenol Contaminated Clay Soil. Water, Air and Soil Pollution,
221(14):3544.
Reddy, K.R., Darko-Kagya, K., and Al-Hamdan, A.Z. 2011.
Electrokinetic Remediation of Chlorinated Aromatic and
Nitroaromatic Organic Contaminants in Clay Soil. Environ. Eng. Sci., 28(6):405413.
Reddy, K.R., Darko-Kagya, K., and Cameselle, C. 2011.
Electrokinetic-enhanced Transport of Lactate-Modified
Nanoscale Iron Particles for Degradation of Dinitrotoluene in Clayey Soils. Sep. Purif. Technol. 79(2):
230237.
Reddy, K.R., Cameselle, C., and Ala, P.R. 2010. Integrated
electrokinetic-flushing to remove mixed organic and metal
contaminants. J. App. Electrochem. 40(6):12691279.
Reddy, K.R., and Cameselle C. 2009. Electrochemical remediation technologies for polluted soils, sediments and
groundwater. Wiley, New York, USA.
Reddy, K.R., 2010. Technical Challenges to In-Situ Remediation of Polluted Sites. Geotech. Geol. Eng. J. 28(3):
211221.
Reddy, K.R., and Chandhuri, K.S. 2009. Fenton-like Oxidation of PAHs in Clayey Soils Using Electrokinetics.
J. Geotech. Geoenviron. Eng.-ASCE. 135(10):14291439.
Reddy, K.R., Maturi, K., and Cameselle, C. 2009. Sequential Electrokinetic Remediation of Mixed Contaminants
146
in Low Permeability Soils. J. Environ. Eng-ASCE.

135(10):989998.
Reddy, K.R., and Karri, M.R. 2008. Effect of Oxidant
Dosage on Integrated Electrochemical Remediation of
Contaminant Mixtures in Soils. J. Environ. Sci. & Heal.,
43(8):881893.
Reddy, K.R., and Karri, M.R. 2006. Effect of Voltage
Gradient on Integrated Electrochemical Remediation
of Contaminant Mixtures. Land Contam. Reclamation.
14(3):685698.
Reddy, K.R., Ala, P.R., Sharma, S., and Kumar, S.N.
2006. Enhanced Electrokinetic Remediation of Contaminated Manufactured Gas Plant Soil. Eng. Geol. 85(12):
132146.
Reddy, K.R., and Ala, P.R. 2006. Electrokinetic Remediation of Contaminated Dredged Sediment. J. ASTM Inter.
3(6):254267.
Reddy, K.R., and Ala, P.R. 2005. Electrokinetic Remediation of Metal-contaminated Field Soil. Separ. Sci. Technol.
40(8):17011720.
Reddy, K.R., Danda, S., and Saichek, R.E. 2004. Complicated Factors of using Ethylenediamine Tetracetic
Acid to Enhance Electrokinetic Remediation of Multiple Heavy Metals in Clayey Soils. J. Environ. Eng-ASCE,
130(11):13571366.
Reddy, K.R., and Saichek, R.E. 2004 Enhanced Electrokinetic Removal of Phenanthrene from Clay Soil by Periodic
Electric Potential Application. J. Environ. Sci. Heal. A.,
39(5):11891212.
Reddy, K.R., and Chinthamreddy, S. 2004. Enhanced Electrokinetic Remediation of Heavy Metals in Glacial Till
Soils Using Different Electrolyte Solutions. J. Environ.
Eng-ASCE, 130(4):442455.
Reddy, K.R., Chaparro, C., and Saichek, R.E. 2003.
Iodide-Enhanced Electrokinetic Remediation of MercuryContaminated Soils. J. Environ. Eng-ASCE, 129(12):
11371148.
Reddy, K.R., Chinthamreddy, S., Saichek, R.E., and Cutright,
T.J. 2003. Nutrient Amendment for the Bioremediation
of a Chromium Contaminated Soil by Electrokinetics.
Energy Sources, 25(9):931943.
Reddy, K.R., and Saichek, R.E. 2003. Effect of Soil Type on
the Electrokinetic Removal of Phenanthrene using Surfactants and Cosolvents. J. Environ. Eng-ASCE. 129(4):
336346.
Reddy, K.R., Chaparro, C., and Saichek, R.E., 2003. Removal
of Mercury from Clayey Soils Using Electrokinetics. J.
Environ. Sci. Heal-A., 38(2):307338.
Reddy, K.R., and Chinthamreddy, S. 2003. Sequentially
Enhanced Electrokinetic Remediation of Heavy Metals
in Low Buffering Clayey Soils. J. Geotech. Geoenviron.
Eng-ASCE, 129(3):263277.
Reddy, K.R., and Chinthamreddy, S. 2003. Effect of Initial Form of Chromium on Electrokinetic Remediation in
Clays. Adv. Environ. Res., 7:353365.
Reddy, K.R., Saichek, R.E., Maturi, K., and Ala, P. 2002.
Effects of Soil Moisture and Heavy Metal Concentrations on Electrokinetic Remediation. Indian Geotechnical
Journal, 32(3):258288.
Reddy, K.R., Chinthamreddy, S., and Al-Hamdan, A.
2001. Synergistic Effects of Multiple Metal Contaminants During Electrokinetic Remediation. Remediation:
The Journal of Environmental Cleanup Costs, Technologies & Techniques, John Wiley Publishers, 11(3):
85109.
Reddy, K.R., Xu, C.Y., and Chinthamreddy, S. 2001. Assessment of Electrokinetic Removal of Heavy Metals from
Soils by Sequential Extraction Analysis. J. Hazard. Mater.
84(23):85109.
Reddy, K.R., Donahue, M.J., Saichek, R.E., and Saasoka, R.
1999. Preliminary Assessment of Electrokinetic Remediation of Soil and Sludge Contaminated with Mixed Waste.
J. Air and Waste Management Association, 49:174181.
Reddy, K.R., and Chinthamreddy, S. 1999. Electrokinetic
Remediation of Heavy Metal Contaminated Soils under
Reducing Environments. Waste Manage., 19(4):269282.
Reddy, K.R., Parupudi, U.S., Devulapalli, S.N., and Xu,
C.Y. 1997. Effects of Soil Composition on Removal of
Chromium by Electrokinetics. J. Hazard Mater. 55(13):
135158.
Reddy, K.R., and Parupudi, U.S. 1997. Removal of
Chromium, Nickel and Cadmium from Clays by Insitu Electrokinetic Remediation. J. Soil Contam. 6(4):
391407.
Reddy, K.R., and Shirani, A.B. 1997. Electrokinetic Remediation of Metal Contaminated Glacial Tills. Geotech. Geol.
Eng. 15(1):329.
Roach, N., Reddy, K.R., and Al-Hamdan, A.Z. 2009. Particle
Morphology and Mineral Structure of Kaolin Contaminated with Heavy Metals Before and After Electrokinetic
Remediation. J. Hazard. Mater., 165(13):548557.
Roach, N., and Reddy, K.R. 2006. Electrokinetic Delivery of
Permanganate into Low Permeability Soils. International
J. Environ. Waste Manage. 1(1):419.
Saichek, R.E., and Reddy, K.R. 2003. Effects of System
Variables on Surfactant Enhanced Electrokinetic Removal
of Polycyclic Aromatic Hydrocarbons from Clayey Soils.
Environ. Technol., 24:503515.
Saichek, R.E., and Reddy, K.R. 2005. Surfactant-enhanced
Electrokinetic Remediation of Polycyclic Aromatic
Hydrocarbons in Heterogeneous Subsurface Environments. J. Environ. Eng. Sci., 4(5):327339.
Saichek, R.E., and Reddy, K.R. 2005. Electrokinetically
Enhanced Remediation of Hydrophobic Organic Compounds in Soils: A Review. Crit. Rev. Environ. Sci.
Technol., 35(2):115192.
Saicheck, R.E., and Reddy, K.R. 2003. Effect of pH Control at
the Anode on the Electrokinetic Removal of Phenanthrene
from Kaolin Soil. Chemosphere, 51:273287.
Sharma, H.D., and Reddy, K.R. 2004. Geoenvironmental
Engineering: Site Remediation, Waste Containment, and
Emerging Waste Management Technologies. John Wiley &
Sons, Inc., Hoboken, NJ.
United States Environmental Protection Agency (USEPA).
1997. Cleaning Up the Nations Waste Sites: Markets and
Technology Trends. EPA 542-R-96-005, Office of Solid
Waste and Emergency Response, Washington, DC.
USEPA. 2000. Innovative Remediation Technologies: FieldScale Demonstration Projects in North America, 2nd
Edition, EPA 542-B-00-004, Washington, D.C.
USEPA. 2006. Abstracts of Remediation Case Studies,
Vol. 10, EPA 542-R-06-002, Washington, D.C
Yukselen-Aksoy, Y., and Reddy, K.R. 2013. Delivery and
Activation of Persulfate for Oxidation of PCBs in Clayey
Soils by Electrokinetics. J. Geotech. Geoenviron. EngASCE, 139(1):175184.
Yukselen-Aksoy, Y., and Reddy, K.R. 2012. Effect of Soil
Composition on Electrokinetically Enhanced Persulfate
Oxidation of Polychlorobiphenyls. Electrochim. Acta,
86:164169.
147

Cement bentonite cutoff walls for polluted sites

K. Soga
University of Cambridge, Cambridge, UK
K. Joshi
Formerly at University of Cambridge, Cambridge, UK
J.C. Evans
Bucknell University, Lewisburg, PA, USA
ABSTRACT: Cement-bentonite (CB) cutoff walls have long been used to control ground water flow and
contaminant migration at polluted sites. Hydraulic conductivity and unconfined compressive strength are two
short-term properties often used by industry and owners in CB specification and are important parameters
discussed in this paper. For polluted sites, long-term compatibility is also an important issue. These properties
are coupled to a number of external factors including the mix design, construction sequence, presence/absence
of contaminants at the site. Additional short-term properties for engineering assessment include the stressstrain characteristics in both drained and undrained shear in both with and without confinement as well as
one-dimensional consolidation properties. Long term CB properties are affected by aging, reaction chemistry,
drying, in situ stress state, and interaction with the polluted environment.
INTRODUCTION
Slurry trench cutoff walls were originally developed

to control ground water flow beneath dams and levees and for temporary excavations below the ground
water table. In Europe, the technique of choice was
the cement-bentonite (CB) slurry wall originally composed of a mixture of bentonite-water slurry and
ordinary Portland Cement (OPC). As the technology
evolved the CB mixtures included granulated ground
blast furnace slag (GGBFS) as a replacement for 70 to
80% of the OPC (Jefferis 1981).The resulting mixtures
were found to produce laboratory values of hydraulic
conductivity (k) less than 1 109 m/s and unconfined compressive strength (UCS) values of 200 kPa
or more after one month of curing. With time, society
became aware of subsurface contamination from past
industrial practices and the need to mitigate the negative consequences of this pollution was recognized.
Slurry trench cutoff walls became a technology of
choice where remedial measures called for control of
polluted ground water. This use imparted the need for
a better understanding of CB cutoff walls in both the
short-term and long-term and in a contaminated environment. This paper presents a general report on slurry
walls for polluted sites with a focus on CB technology.
The paper draws on the published literature and on site
studies over time of a CB wall built at a disused gasworks site and studied over a period of nearly 12 years
(Joshi 2009).
MIX DESIGNS
The mix design for CB walls varies from contractor to

contractor and from project to project although generally within a reasonable narrow range. Bentonite-water
slurry is prepared with 3 to 6% (typical) bentonite and
clean (often potable) water available near the site. This
bentonite-water slurry is often hydrated for a period of
24 hours or more and, once the bentonite hydration is
complete, will typically have a viscosity, as measured
in a Marsh Funnel, of 32 to 40 seconds. The quality
of the slurry is evaluated in a filtrate loss test and filtrate loss is generally limited to 25 mL at 695 kPa for
30 min. To this slurry, the binder is added usually in
the range of 10 to 35% (or water to cement ratio of
2.87 to 10). The binder may be 100% OPC but more
often is a blend of OPC and GGBFS.
The most common (and recommended) blend would
include at least 60% GGBFS and the optimum performance (lowest k and highest strength) is often at about
80% GGBFS in the blend.
It is important to understand that the primary role
of the bentonite in CB mixtures is a short-term one.
That is, the viscosity imparted to the slurry by the bentonite keeps the binder particles in suspension during
the time elapsed between mixing and initial set. A measure of the stability of the slurry is bleed, which is the
excess water in the slurry as binder particles settle prior
to initial set. Bleed is commonly measured by pouring freshly mixed slurry into a 1 L graduated cylinder
149
Figure 1. Effect of slag content on hydraulic conductivity

(after Opdyke and Evans 2006).
Figure 2. Curing time effects on hydraulic conductivity

(after Opdyke and Evans 2006).
and observing the accumulation of the water at the

top of the cylinder. A balanced, stable slurry mixture
may exhibit zero bleed even though for most applications, up to 2% is acceptable (Jefferis, 2012). CB
mixtures without GGBFS exhibit significantly greater
bleed than those with GGBFS.
3
3.1
HYDRAULIC BEHAVIOUR
Laboratory hydraulic conductivity
For polluted sites, the purpose of a CB cutoff wall is

typically to minimize contaminant transport by minimizing advective flow. The controlling mechanism
of contaminant transport is predominantly diffusion
at k values less than about 1 109 m/s and typical
field gradients. Little has been done to investigate
reductions diffusive transport in CB cutoffs.
The effect of the GGBFS binder on the k is shown
on Fig. 1. For these laboratory mixtures, the lowest k
was found for a binder with 80% GGBFS and 20%
OPC. Further, mixtures of 50/50 which are typically
readily available from cement suppliers, show no benefit to using the GGBFS compared with 100% OPC.
Also notice there is not a clear trend between binder
content and k as was found by when combining data
from many projects.
The hydration process is time-dependent and therefore the k is time dependent. For CB mixtures a more
realistic measure of the k requires many months or
more of curing and, during this time, the k is dropping. Shown on Fig. 2 are medium term data for the
20% binder content mixtures shown on Fig. 1. The data
in Fig. 2 clearly show that, over the one year duration
the k decreases about two orders of magnitude. Interestingly, the mixtures at 80% and 90% slag content in
the binder had the lowest values of k at one-month but
not at one-year. While the mixtures with 60% and 70%
were the most permeable at one month of curing, they
Figure 3. Effect of time and sample type on Hydraulic

conductivity (after Joshi, 2009).
were the least permeable at one year of curing. Clearly

comparisons of k between mixtures must be coupled
with consideration of curing time.
The k is not only coupled with time but is coupled
with sample origin. Samples may come from the laboratory, from the field mixer, from the field trench
or from block samples. In situ measurements of k can
also be made. Field mixed samples will inherently have
greater variability than laboratory mixed samples and
sample storage and handling conditions may affect the
results. Shown on Fig. 3 is the relationship between k
and curing time for various sample types all from a
disused gasworks site in the UK.
Several trends are apparent from the data on Fig. 3.
First, a reduction in k occurs throughout the 11-year
curing period. Second, there is considerable variability
in results for all data types. Third, the mixer samples
show the least variability and the block samples show
the most. Fourth, at 11 years of curing, the measured
values of k was the lowest in the mixer cast samples,
then trench cast samples and then the highest values
in the block field samples. Taken together these data
emphasize the need to gather several types of samples and to test replicates to be able to examine the
variability in results.
150
Figure 4. Hydraulic conductivity with depth.
Figure 5. Pore pressure dissipation test results.
The mixer cast, trench cast and block samples all

came from the same disused gasworks site but would
have had different contaminant exposure levels. Contamination in the CB will occur due to mixing with the
surrounding soil during construction and due to exposure to contaminants in the adjacent soil and groundwater. This no doubt introduces variability compared
to laboratory or other samples.At the disused gasworks
site, laboratory permeability tests were conducted on
three types of samples (block, mixer-cast and trenchcast). The block samples were observed to have stones,
mud pockets, foul lime, spent oxides, air cavities and
fissures. As compared to laboratory prepared mixercastsamples, the blockfield samples have additional
Quartz mineral due to presence of sand/stone. Furthermore, the block field obtained from heavily sulphate
contaminated area showed Ettringite and quartz mineral in XRD examination. ESEM tests on heavily
contaminated block sample confirmed the presence
of Ettringite, a reaction product of sulphates with the
hydrated slag-cement-bentonite matrix.
For this particular site with its site specific chemistry, the effect of long-term exposure was not detrimental in terms of meeting site k requirements
(less than 1 109 m/s). At the age of 11 years,
the hydraulic conductivity measured by laboratory
tests of mixer-cast was the lowest (1.1 1011 to
8 1011 m/s), followed by the trench-cast samples (5.8 1011 to 1.7 1010 m/s) and the block
field samples (2.4 1011 to 5.6 109 m/s). Moreover, the hydraulic conductivities of the block field
samples exhibit larger variability compared with the
mixer-cast and the trench-cast samples. Mixer cast
samples are contaminant free, trench cast samples have
in situ soils and ground water mixed in prior to curing
and block samples have in situ soils and ground water
mixed in prior to curing as well as long-term exposure
to the site contaminants.
Another long-term performance issue relates to the
position of the CB barrier with respect to the water
table. It is obvious from observations in the laboratory
that CB, if allowed to dry, will exhibit degradation in
k. Researchers often need to decide between curing
samples at 100% humidity or under water. Data from
the disused gasworks site are useful in this regard and
are shown in Fig. 4.
The data on Fig. 4 are for block samples tested in

the laboratory in triaxial k tests at a confining stress of
100 kPa. These data support the conclusion that, for the
site specific meteorological and geologic conditions,
there is no pronounced effect on k of samples above the
water table after 11 years. It is relevant to note that the
site is generally fine grained and the water table is
relatively shallow in the UK. Therefore caution must
be taken in extrapolating these findings to sites with
a more coarse-grained geology, with a deeper water
table or with dryer/hotter climates.
When carving laboratory samples from block samples, obvious cracks and defects are avoided so that
the results represent the intact CB material and do not
show the influence of in situ defects that would influence the overall bulk k of the wall. Furthermore, the k
is often measured under an isotropic stress state which
may not be representative of the in situ stress state and
which may provide for the closure of micro-cracks.
Data on the influence of water table as measured
from in situ tests is discussed in Section 3.2. Addition
insight into long-term behavior as influenced by the
polluted environment is discussed in Section 4.
3.2
In situ measures of hydraulic conductivity
As an alternative to laboratory tests on field samples,

in situ testing may provide additional insight into the k
of CB cutoff walls. Using a piezocone, pore pressure
dissipation tests can provide data to calculate the k in
the vicinity of the cone tip. From the time it takes to
dissipate 50% of the induced excess pore water pressure (t 50 ) and the methods of analysis presented by
Parez and Fauriel (1988) and by Manassero (1994),
the in situ k can be estimated.
Pore pressure dissipation test data from the CB cutoff wall at the disused gasworks site after 7.4 years of
curing are shown on Fig. 5 and the computed values
of k for the two methods are shown on Fig. 6.
Several observations are in order with respect to
this in situ data. First, the dissipation is much faster
than that found in any low permeability natural soil
materials, indicating the CB material is a very stiff
material. Parez and Fauriel (1988) assume typical values of soil stiffness to obtain their relationship between
k and t 50 , whereas Manaserros method require an
151
Figure 6. Hydraulic conductivity from pore pressure dissipation tests.
input of a rigidity index (i.e. stiffness-strength ratio),

which in this case was obtained from the laboratory
test data. Hence, the use of Parez and Fauriels method
is not appropriate for this material. Second, even using
Manaserros method, the rate of pore pressure dissipation at tests shallower than 2.5 m was much greater, and
hence the computed values of k much higher (around
106 m/sec) than what was measured by other methods (less than 108 m/sec, see next). The very act of
pushing the CPT into a brittle material introduces both
disturbance (such smearing and cracking) and volume
change. The large values of k at shallow depths (less
than 2.7 m) may be due to cracking, which in turn
accelerates the dissipation process along the cone and
material interface. The k values at two greater depths
(2.8 m and 3 m) are considered to be more reliable as
they are similar to the values measured in other field
and laboratory tests. The large confinement conditions
may have prevented the CB material to crack during
cone penetration.
In addition to laboratory testing on field mixed samples and the use of piezocone pore pressure dissipation
tests to determine k, in situ borehole permeability tests
are quite common. Borehole permeability tests may be
in an open borehole or in a specific range by using a
packer using constant flow or falling head. For the
disused gasworks site, tests were also made with a
Cambridge self-boring pressuremeter (SBP). Comparisons between the SBP and borehole tests with packers
were made in the local Lias clay to verify comparable precision of the SBP with more widely employed
borehole tests (Joshi, et al. 2010). All of these data are
presented on Fig. 7.
The data in Fig. 7 show a difference in measured
values in k in boreholes using packers and with the selfboring pressuremeter (SBM). The falling head tests
(PFH) produced, on average, lower values of k than
those from constant flow (PCF) tests. Three mechanisms are postulated to explain the difference. First,
the PCF test was a closed system such that entrapped
air may have influenced the pressure response leading to higher computed values of k. Secondly, the
injection pressures in the PCF test, which are often
greater than those in the PFH test, may have produced
leakage between the inflated packer membrane and the
Figure 7. Hydraulic conductivity measured in situ in borehole tests using packers.
Figure 8. Effect of pocket length on hydraulic conductivity.
sidewalls of the borehole. Finally, injection pressures

may have induces microfractures increasing the k in
the test range. Note that the self-boring pressuremeter
gave results consistent with the PFH tests.
Scale effects have long been an issue in laboratory and field measurements of k on compacted clays
(Benson et al. 1997) and for soil-bentonite slurry
trench cutoff walls (Filz et al. 2003 and Britton et al.
2004). To determine if a similar scale effect exists for
CB barriers, the k was measured using different pocket
lengths in PCF, PFH and SBM tests. These results
are shown on Fig. 8 and illustrate the scale effect in
CB wall permeability testing. The increase in pocket
length reflects an increase in the L/D ratio from 11 to
21, 31 and 41 for the packer tests and from 2.7 to 11
for the SBP tests.
The foregoing detailed analysis of CB cutoff walls
illustrate: a) the changing value of k with time for
periods of a year or more, b) the variation of k with
mix proportions, particularly GGBFS content, c) the
dependence of k upon test method and d) the effect of
sample size (scale) on k values. While it may be intuitive that in situ testing is the best way to measure k for
a completed barrier, examination of the results from
various in situ methods reveal the difficulties with
this approach. In summary, carefully conducted laboratory tests on field grab samples in combination with
borehole permeability tests at low hydraulic pressures
may provide the best understanding of the hydraulic
conductivity of the as-built CB cutoff wall.
152
Figure 9. Variation of fracture pressure with confining pressure in cavity expansion test and leak-off test of laboratory
cast cement bentonite (i = 0.1 cm3 /min).
3.3 Hydraulic fracturing and stress state
Figure 10. Variation of peak cavity pressure with injection

rate in cavity expansion test and leak-off test of laboratory
cast cement bentonite ( 3 = 75 kPa).
Hydraulic fracturing was first described for field permeability testing over 40 years ago by Bjerrum, et al.
(1972). For the conditions of tensile failure in terms
of major principal stress 1 , a value of u/ 1 between
0.5 and 1.0 was found to be critical. Bjerrum et al. also
point out that, for the case of a compressible material
in a narrow trench, the overburden pressure is, in part,
carried by arching and fracturing has been observed
simply by filling a borehole casing with water while
drilling. Hence the potential for hydraulic fracturing is
inextricably linked to the stress state in the CB wall.
Fundamentally, for a material without tensile capacity, tensile cracking occurs when the minor principal
effective stress 3 (compression positive) becomes
zero. For soils with tensile capacity t (tension positive) such as CB cutoff walls, fracturing will occur
(e.g. Mitchell and Soga, 2005) when:
Viewing hydraulic fracturing in borehole in a CB

cutoff wall as cavity expansion, hydraulic fracturing can occur as the radial total stress in the cavity
increases and the circumferential stress decreases in
the linear elastic range. Radial cracks develop when the
effective stress becomes equal to the tensile strength of
the CB material. For these conditions including knowledge of the initial minor principal effective stress 3i
and pore pressure u0 , the fluid pressure P f that caused
the CB to fracture (Mitchell and Soga, 2005) is:
In a series of laboratory experiments on CB, hollow

cylinder fracturing tests were carried out by Ng (2008).
Cavity expansion tests were conducted to model membrane inflation during packer testing and leak-off tests
to model field permeability testing. Tests were conducted by varying confining pressure and rate of
injection. Fig. 9 shows the effect of confining pressure
on the fracture pressure (P b is the back pressure, which
can be taken as groundwater pressure in the field). The
fracture pressure increased with confining pressure in
both tests, but the rate of increase diminished in higher
confining pressure. The fracture pressure in leak-off
Figure 11. Examples of cracks from the hydraulic fracturing

tests (Ng, 2008) and borehole cracks due to drilling fluid
pressures (courtesy of Keller Hayward Baker).
test is much lower than that in cavity expansion test

due to the seepage of water into microcracks the material, which affects the effective stress distribution and
hence the materials tendency to tensile failure.
The effect of injection rate on fracture pressure in
both tests is shown in Fig. 10. The injection rate did
not have significant effect on the fracture pressure in
the cavity expansion tests. For the leak-off tests, the
fracture pressure of the tests with an injection rate of
0.1 cm3 /min is 30% higher than that of the tests with an
injection rate of 0.03 cm3 /min. However, the difference
between the tests of injection rate = 0.1 cm3 /min and
the tests of injection rate = 0.3 cm3 /min is negligible.
This could be due to the tendency of nearly undrained
behaviour in higher injection rates.
Figure 11a&b show examples of cracks (highlighted in red) after the hydraulic fracture tests.
153
Figure 12. Influence

Conductivity.
of
Cracking
on
Hydraulic
It is clear from the discussion of hydraulic fracturing that the state of stress and tensile strength of the CB
materials are important contributors to CBs the resistance to hydraulic fracturing. The tensile strength will
be discussed in Section 5 of this paper but a brief discussion of stress state is important (see also Eq. (1)).As
indicated above, placement of a compressible material
in a narrow trench leads to arching and a reduction in
vertical stresses. Transverse stresses are balanced by
the lateral earth pressures as discussed in Filz (1996)
and Ruffing et al. (2010). Longitudinal stresses are
the resultant of the vertical and transverse stresses and
are the lowest of the three directions as evidenced by
the location of fracture planes in field photographs of
cracked bore holes. The stress state is further complicated by the volume changes that occur after the CB
has initially set. Additional shrinkage will occur due
to cement hydration processes (chemical shrinkage)
and cooling after exothermic processes are complete
(thermal cracking).
4
Borehole cracks due to excess pore water pressures

during core drilling are show on the borehole photograph presented as Fig. 11c (Cermak et al., 2012).
It was found that the peak pressure (that initiating
fracturing) was significantly larger in the cavity expansion tests using a membrane than for the leak-off tests
where the water was in direct contact with the CB material. It is believed that where water is in direct contact
with the CB, the microstructual characteristics offer
crack initiation locations permitting the water to penetrate and further crack the material. For example, using
the data presented in Fig. 9, when an insitu permeability test is conducted at a depth of 5 m (i.e. the effective
confining pressure of 45 kPa), the water injection pressure should be less than 90 kPa + u0 (where u0 is the
initial groundwater pressure, t = 0 in Eq. (1)). Otherwise, the CB will likely to fracture. In contrast,
when the membrane is expanded, it spans over small
microstructural features and cannot capitalize on these
features to initiate cracking.
The importance of hydraulic fracture cracking in
field permeability testing is illustrated in Fig. 12. For
an in situ mixes wall made with CB slurry, downhole borehole falling head permeability tests were
conducted after coring and the hydraulic conductivity
determined. Borehole photographs, such as that shown
as Fig. 11c were then used to determine the crack
length. Using 15 point running average, the relationship between crack length and k is shown on Fig. 12.
As shown, the k values for the uncracked boreholes as
approximately 2 1010 m/s, whereas the computed k
values increase rapidly with cracking. It can be argued
that the computed values of k in a cracked borehole
are not values of k for the material but some composite between the material and crack permeability. Hence
realistic in situ borehole testing requires care to avoid
hydraulically fracturing the material both during core
hole drilling and during testing.
DURABILITY
4.1 General
Durability is a term used to describe the long-term
behavior effects on the barrier material due to the site
pollution or other environmental stresses such as wetting and drying. While much has been written about
chemical effects of compacted clays and soil-bentonite
mixtures, little is available for CB materials. Similarly,
effects of cyclic wetting and drying have not received
much attention for CB materials. This portion of the
paper examines the work of others and the results of
long-term site and laboratory studies for the CB wall at
the disused gasworks site described earlier. Compatibility is frequently evaluated in long-term permeability
tests and immersion tests using the contaminants of
concern in the permeant or immersion bath. Durability in cyclic wetting and drying is also done and may
include strength and/or permeability tests as well as
visual observations.
4.2 Chemical effects
Permeation with organic contaminants has been shown
to have a detrimental effect on compacted clays
(Brown and Anderson, 1983) and Soil Bentonite (SB)
slurry trench cutoff wall materials (Evans et al. 1985).
This poor performance can be explained by examining
the pore fluid interactions with the colloidal clay minerals in the system. CB cutoff wall materials formed
from hydration products of the ingredients and would
be expected to perform much differently than clayey
systems. In fact, XRD studies of CB pastes over time
show that the bentonite in the mixture is incorporated
into the hydration product matrix such that bentonite,
as a distinct clay mineral, is no longer found after
hydration is complete (Yeboah et al., 2013). Thus,
colloidal models of clay behavior in the presence of
various pore fluids cannot be extended to hydrated CB
cutoff wall materials.
154
Figure 14. Immersion test experiment on CB material:

(left) confined specimen before immersion in plastic mould
and cable ties, and (right) duplicate specimen immersed in
magnesium sulphate solution.
Figure 13. Long term permeation of CB with Aniline.
Long-term triaxial permeability compatibility studies reveal CB material to be more resistant to degradation in the presence of organic contaminants. This
has been demonstrated for organic contaminants such
as ethanol and aniline (Evans and Opdyke, 2006). For
example, after three months of permeation of duplicate CB samples with concentrated (pure) aniline, the
k is essentially constant as shown on Fig. 13. In addition, immersion tests on duplicate samples earned a
rating of very good using the methods developed by
Gavin and Hayes (1999). Note that the data in Fig. 13
presented in terms k rather than intrinsic permeability
which factors out the influence of the pore fluid density and viscosity. The pore fluid density and viscosity
changes from that of water to that of aniline as the test
progresses and the time rate of change is not known.
However, comparing the initial and final conditions, a
35% increase in intrinsic permeability was found due
to permeation with aniline.
Aniline has been shown to cause rapid increases of
up to four orders of magnitude in k in both compacted
clays (Anderson and Brown, 1983) and Soil Bentonite
(SB) backfill (Evans et al., 1985). Comparisons of
these findings indicate the potential for better field
performance from CB barriers than from SB barriers.
Durability evaluations in long term permeability
tests may lead to different conclusions from those
determined from immersion tests. For example, in
one study of the interaction of CB with an acidic
sulphate leachate, long-term permeability tests (>2
years) showed good resistance to degradation whereas
immersion tests indicated some cracking was observed
(Fratalocchi et al., 2006). The beneficial effect of
confinement is absent from immersion tests such
that volume changes due to ettringite formation may
degrade the material.
Immersion tests with sodium sulphate solutions
where conducted for samples of various ages from
young to 11 year block samples from the disused gasworks site as shown in Fig. 14 (Joshi 2009). As a means
Figure 15. Immersion tests (a) 6 week old mixer sample

having different surface and core colour after immersion,
(b) 1 year old mixer sample having intact solid core after
immersion, (c) 11 years old mixer sample after immersion
and (d) block sample after immersion.
to evaluate the impacts, samples were visually examined and measured. As shown in Fig. 15, unaffected
materials had a deep blue solid core where as fully
affected samples had turned gray indicative of oxidation. The results of the visual examinations revealed a
coupled impact of contamination exposure and sample
age as shown on Fig. 16.
Since immersion tests can give a rapid indication
of the potential for damage due to incompatibility,
these tests make a excellent screening, or indicator, test
where potential effects are unknown or not predictable
(Jefferis, 2012). It is recommended that the immersion fluid be changed regularly and that weight/volume
data be recorded over time. If possible, analysis of the
immersion water will provide added insight into any
chemical reactions that may be occurring.
155
Figure 16. Effect of contaminant exposure on CB.
Figure 18. Effect of cyclic wetting and drying (a) sample

retrieval after the test, (b) changes in hydraulic conductivity.
Figure 17. Air dried CB.
Long-term permeability tests are recommended but

are time consuming. A pore volume displacement of
two or three, usually sufficient for compacted clay
or SB, may be insufficient to accommodate the time
required for the reaction front to fully migrate through
the sample. While it may not be possible to conduct the
test until the reaction front is through the sample, the
time rate of the advancement of the front can be scaled
up to see if the CB will perform for the design life of
the cutoff wall.
4.3 Wetting and drying effects

Durability under wetting and drying must also be considered for CB cutoff walls. CB materials are largely
water and are significantly affected by drying. For
example, shown on Fig. 17 is a series of photographs
of a CB sample undergoing air drying in the laboratory (Joshi, 2009). Notice cracking beginning within
8 hours and the sample is completely cracked within
24 hours. The moisture content reduced from about
376% to 10% in the 8 day drying period and significant
cracking occurred by the time the moisture content had
dropped to 275%.
While air drying produces dramatic results, a more
representative (for the UK) experiment where the CB
is buried in sand produced very different results. In this
study, the sand was cyclically saturated for three days

and drained for two days and subsequently left in a
drained condition for 13 days (Joshi 2009). After 12
cycles (just over 7 months) the CB was excavated from
the sand bed and found to be intact (see Fig. 18a). Permeability tests before and after showed in all samples
(three different ages) there was a slight decline in the
hydraulic conductivity as shown in Fig. 18b.
There are no apparent reasons for reduction in
hydraulic conductivity due to wetting-drying test,
except for hydration and pozzolanic reaction in 90-day
mixer sample. It can be expected that the pozzolanic
reaction in the 90 days mixer-cast sample would continue during 7.25 months and may reduce its hydraulic
conductivity. However, it should be noted that the
reduction in hydraulic conductivity is within the possible variation of the measurement for the heterogeneous
material like CB. The error bars in Fig. 18 shows
the variation of more than one order of magnitude
in the measurements for the different specimens of
the same sample. Thus, it is assumed that wettingdrying does not affect the hydraulic conductivity of
CB samples and the change in their value can be
due to variation in measurements because of inherent
heterogeneous nature of the material.
It is important that studies of the effect of cyclic wetting and drying reflect the expected suction pressures
during drying under field conditions. For the case of
the disused gasworks site, the laboratory data are consistent with field observations. Shown on Fig. 19 is a
photo of a portion of the wall above the water table
at the disused gasworks site taken 8 years after wall
156
Figure 19. Photo of 8 year old CB wall above the water

table.
installation. As the photo shows, the wall is intact and

has not suffered degradation.
MECHANICAL BEHAVIOUR
5.1 General
The performance specification (ICE, 1999) for hardened slurry requires unconfined compressive strength
(UCS) larger than 100 kPa at 28 days. It has been the
UK practice to specify that the hardened cement bentonite should reach an axial strain of greater than 5%
prior to failure in a consolidated drained triaxial compression test (Tedd, 2005). Therefore in practice, both
UCS and drained triaxial tests are carried out on hardened CB sample. Tedd (2005) further mentions that
the strain criterion is specified because of a perceived
need for a deformable plastic cut off wall that will not
crack when subjected to movement. Similarly, if the
CB wall is robust, then it can allow construction and
site regeneration activities in its vicinity. The understanding of mechanical properties is therefore useful
to determine the future use of the contained land.
Before discussing the details of mechanical
behaviour of CB material, it would be useful to develop
a conceptual mechanical model. The most unique
property of CB material is that the slurry contains
large quantity of water and much of the water remains
available in the voids. By general convention of soil
mechanics, this free pore water can be treated as void
volume. But the water is integral part of the hardened
set slurry and loss of this water causes permanent
shrinkage of the material. The dried material cannot
be categorized as the real CB wall material and therefore the calculation of void ratio and specific gravity
obtained from the dried CB wall material may not be
correct. But having said that, CB material possesses a
porous and rigid structure as shown in Fig. 20a.
As described in later sections, the mechanical
behaviour of CB is similar to that of metallic foam
as shown in Fig. 20b. This foam has network of alloy
forming closed pores, thus it has low density, high
Figure 20. Two materials that have high strength, high

porosity, low density, low permeability and zero poissons
ratio (a) CB material, and (b) metallic foam.
porosity but very low permeability. Similarly the CB

material on micro level, it has a network of cemented
solids that contributes to high void ratio, high strength
and low hydraulic conductivity simultaneously. The
structure is rigid when the applied load is small due
to the network. However, once the load exceeds its
yield stress, the porous network rapidly collapses,
generating large volumetric compression.
5.2 Compressibility
In case of CB material, the yield stress in compression
represents a threshold of pressure value beyond which
the cementation bond would break and voids will start
closing rapidly. Manassero et al. (1995) and Opdyke
and Evans (2006) conducted oedometer tests on CB
wall material to study its consolidation behaviour.
Opdyke and Evans (2006) reported compression index
(Cc) of 0.97 and a recompression index (Cr) of 0.10
and their data is shown in Fig. 21. The preconsolidation
pressure was interpreted to be in the range from 100 to
200 kPa. Notice the high void ratio reported between
9.9 and 11.1, which imply very low solids content.
157
triaxial tests. The yield stress found in all four cases is

between 300 and 320 kPa, which is between the values
reported by Manassero et al. (1995) and Opdyke and
Evans (2006).
The practical implication of these results is that the
mean effective stress of the CB wall should be less than
the yield stress. If not, very large settlements would
occur. This further illustrates the need to understand
the stress state in a completed wall.
5.3
Figure 21. One-dimensional consolidation of CB material

after 15 months of curing (Opdyke and Evans, 2006).
Figure 22. Volume change in triaxial specimen after applying effective minor principal stress in both drained and
undrained triaxial compression test a) 35 day old mixer-cast
sample, b) 90 day old mixer-castsample, c) 3.25 to 4.24 year
old mixer-cast sample, and d) 11.5 to 12.25 year old block
sample.
Manassero et al. (1995) found that the yielding

stress (or preconsolidation pressure) as 700 kPa, which
is about 3 to 7 times higher than that measured by
Opdyke and Evans (2006). They reported void ratio
between 4.2 and 5.2 for isotropic consolidation under
similar magnitude of pressure, which is again half than
reported by Opdyke and Evans (2006). This suggests
that the mix of Manassero et al. (1995) is less compressible (smaller void ratio, higher yielding stress
and larger coefficient of consolidation) and this can be
due to double amount of cement in its mix with similar amount of bentonite and slag contents compared
to Opdyke and Evans (2006). Also such difference in
measured void ratio could be due to difference in definition of voids and assumed specific gravity of the
material, but the papers have not reported the way to
determine these quantities.
Joshi (2009) performed a series of drained and
undrained compression tests on isotropically consolidated CB samples. The volume change occurred
in specimen at the consolidation stage is plotted in
Fig. 22. It shows volume change data from drained and
undrained tests for all four type of specimens used in
Drained behaviour
Drained triaxial tests were performed the CB samples

from the disused gas work site at different ages. Tests
were conducted at two different confining stresses
(100 kPa and 500 kPa). Figure 23 shows deviator
stress versus axial strain curve. The 100 kPa tests,
in which the specimens were consolidated below the
compression yield stress, showed drained strength as
660850 kPa and peak axial strain at 510%. The specimens exhibited ductile behaviour. While the 500 kPa
test, in which the specimens were consolidated above
the compression yield stress, did not reach a failure
and exhibited significant strain hardening behaviour.
The test was stopped at 2025% axial strain to avoid
damage to the triaxial instrument.
Figure 24 shows the volumetric strain versus axial
strain graph for the 100 and 500 kPa tests for all four
samples. Both 100 and 500 kPa tests showed contractive behaviour with a large amount of volume
change. The 500 kPa test displayed almost linear relation between axial and volumetric strain and both of
them increased at nearly 1:1 ratio. For example, the
35 days old sample showed the axial strain 21.895% at
the end of test while the volumetric strain was 22.02%.
This implies that volumetric strain was predominant
affected by axial strain and there was negligible radial
strain during the compression. This zero-Poisson ratio
behaviour is a typical characteristic of a porous metal
form introduced in the previous section.
The 100 kPa test initially has a straight relation
between axial and volumetric strain until the deviator
stress reaches the peak value. The volumetric strain
drastically increased from 12 to 15% axial strain till
the end of test. This is because the stress path of the
100 kPa drained test crossed the compression yield
stress value at this stage. Figure 25 shows that the
CB specimen showed apparently no volume change
till the mean effective stress p value reached 300
400 kPa, which correspond to the compression yield
stress. When the stress path goes beyond the yield
stress, the specimens started to exhibit large volume
change. After the failure, the p value drops and the
structure of the specimen collapses. The specimen
displays a large volume change.
Based on other drained test data in addition to
those shown above, Joshi (2009) proposed Figure 26
that shows three modes of failure in drained loading depending upon age and confining pressure of
samples. The first mode of failure is called as strainhardening, where a sample undergoes continuous
compression and strength increase due to large void
158
Figure 24. Volumetric strain versus axial strain in drained

triaxial test. (a) 35 day old mixer-cast sample, (b) 90 day old
mixer-castsample, (c) 4.25 year old mixer-castsample and
(d) 12.25 year old block field sample.
Figure 23. Deviator stress versus axial strain in drained triaxial test. (a) 35 day old mixer-cast sample, (b) 90 day old
mixer-castsample, (c) 4.25 year old mixer-castsample and
(d) 12.25 year old block field sample.
reduction. Such failure was prominent at 500 kPa confining pressure for all the tests. Additionally, strain
hardening was also found in 90 days old sample at
200 kPa pressure and in 28 day old sample at 100 and
200 kPa pressure. It is likely that the young samples
allow easy breakage of bond and voids closure as a
159
Figure 25. p versus volume change plot for drained triaxial tests conducted at 100 kPa confining pressure on all four
samples.
Figure 26. Proposed model for strain hardening, ductile

and brittle failure mode depending up on age and confining
pressure under drained triaxial loading.
Figure 27. Deviator stress versus axial strain of 3.25 year

old mixer-cast sample in undrained triaxial test.
Figure 28. Excess pore water pressure versus axial strain of

3.25 year old mixer-cast sample in undrained triaxial test.
result of their incomplete cement hydration. The moderate confining pressure is enough to maintain their
internal stability and therefore they do not reach a
failure point. The specimens lost significant volume.
5.4
Undrained behaviour
Joshi (2009) carried out isotropically consolidated

undrained triaxial compression tests on young 35 day
old mixer-cast samples to 11.5 year old block field
samples. These samples were tested under different
confining pressures to determine their behaviour with
respect to deformation.
Figures 27 and 28 show the change in deviator
stress and pore water pressure change with axial
strain, respectively. The undrained strength is mostly
unchanged with the change in confining pressure and
the pore water pressure is very high. The undrained
strength ranged between 535 to 745 kPa, at a range
of axial strain between 0.5 and 2%. The 900 kPa test
showed the maximum strength as the minor effective
principal stress was greater than zero. While for all
the tests having 3 = 200 kPa, the excess pore water
pressure increased high enough to cause the minor
effective principal stress to become equal to zero at
time of failure, indicating that the specimens have
failed in tension. Beyond the failure, the excess pore
Figure 29. p q stress path of 3.25 year old mixer-cast

sample in undrained triaxial test.
water pressure reduced and the minor effective principal stress became more than positive. Noticeably the
excess pore water pressure reached 30.7 kPa towards
the end of 0 kPa test.
The difference between 3 = 200 kPa and 3 >
200 kPa can be better visualised by the effective stress
paths in the qp space, as shown in Figure 29. The
stress paths of samples subjected to 3 = 200 kPa first
behaved elastically by moving up vertically and later
160
Figure 30. p q stress path for undrained triaxial test

conducted at different age under 50 and 200 kPa confining
pressure.
Figure 31. p q stress path for undrained triaxial test

conducted at different age under 500 kPa confining pressure.
they reached the 3 = 0 line (or q = 3p ) to fail in

tension. It should be noted that this point did not
necessarily correspond to the stress state when the
first tensile crack happened in the specimen. This is
because the pore pressure change at which the tensile crack just initiated might not be detected since
the measurement was only located at the base of the
specimen. The samples then mobilised the remaining
strengths and travelled along the 3 = 0 line. Finally,
tensile cracks are significantly developed to induce
large strains and the specimens reached their maximum deviator stresses. Subsequently, the stress paths
of various tests fell onto the same line towards the end.
On the other hand, the stress paths of the specimens
subjected to 3 > 500 kPa started moving upwards
and failed before hitting the 3 = 0 line. The pore
water pressures at maximum and the residual strength
have not exceeded the effective confining pressure
as observed in the previous cases. Thus the specimen has failed under shear but not in tension. Beyond
the peak strength, the stress-path moved towards the
3 = 0 line and showed residual strength similar to the
other cases. The 900 kPa stress path touched the shear
failure envelope while approaching the peak strength
and continued strain softening after the failure.
Figures 30 and 31 shows stress path in p q plane
for 50, 200 and 500 kPa tests conducted on 35 days,
90 days, 3.25 years old mixer-cast and 11.5 year old
Figure 32. Shear and tension failure mode depending up on

age and confining pressure under undrained triaxial loading.
Figure 33. Test specimens displaying tensile splits and

crumbled blocks after compression at low confining pressure:
35 day old mixer-cast sample at 200 kPa confining pressure (left), 90 day old mixer-cast sample at 0 kPa confining
pressure (right).
block samples. Their comparison can highlight the

change in triaxial properties due to change in age of
CB material at the same confining pressure. The 90
days and 3.25 years sample show similar strength.
This implies that the cement-slag-bentonite hydrates
sufficiently to achieve a good strength by 90 days
and mechanical properties remain stable thereafter.
The graph shows the minimum strength for 35 day
old mixer-cast sample while this was maximum for
11.5 year old block.
Based on the undrained compression data of CB
materials, Joshi (2009) proposed Figure 32, which
illustrates that the failure mode is not only dependent
on the confining pressure but also the age (cement
reaction) of specimen. The confining pressure dictates shear or tensile mode, but these failure modes
are possibly affected by amount of strength gained by
a specimen due to cement reaction till that age.
Most specimens that were consolidated below
200 kPa prior to undrained shearing failed in tension.
The maximum undrained strength would depend on
ability of specimen to withstand the tensile stress or
resistance to split in to individual blocks (see Fig. 33).
Such failure patterns were also observed by Manassero
161
et al. (1995), where they reported low or negative effective minor principal stress for 3 < 200 kPa and all
peak strengths lied on the 3 = 0 line for low confining pressure cases. The slope of stress path in any
of the undrained tests did not exceed 3 and always
followed on the 3 = 0 line. However, it should not
be deduced that the CB samples do not have tensile
strength to cross the 3 = 0 line. The reason for the
q = 3p slope is due to inherent nature of the triaxial
tests, where negative value of the minor principal stress
cannot be detected in this type of tests. The maximum
possible measurable limit for the pore water pressure
is equal to the cell pressure because the pore water
pressure is measured by the system of radial drainage
and porous stones at the boundary of specimen. As
soon as pore water pressure exceeds the cell pressure
at the boundary of specimen, the pore fluid has liberty to expand and the pressure drops to the level of
confining cell pressure. In reality, the pore pressure
might have exceeded the cell pressure in the interior
of specimen, while experiencing tensile stresses. For
the same reason Manassero et al. (1995) proposed that
the real stress path in the field could be on the left
of the 3 = 0 line.
For the higher confining pressure tests like 500 and
900 kPa (3 > 200 kPa), the excess pore water pressure
never exceeded the initial effective confining pressure
and stress path did not reached the tension cut-off line.
In particular, the 900 kPa tests stress path touched
shear failure line and softened further past its peak
strength. However, the critical state line is not identified due to some abruptness in stress-strain curve and
many of the tests has been terminated just past the
peak stress (approximately at 6% axial strain) because
of the limitation of the triaxial set up.
The 500 and 900 kPa tests samples have been subjected to the minor effective principal stress more
than zero and thus showed slightly higher undrained
strength than all other tests at low confining pressure. Skempton parameter at failure (Af ) is greater than
unity and the material structure has collapsed due to
large water content. The crack pattern on the specimen after test were very different than that observed
at the tests of 3 = 200 kPa and this also confirmed
shear failure, as shown in Figure 34. Manassero et al.
(1995) observed positive pore pressure (Af > 1) for
3 > 400 kPa and inclined cracks on specimen due to
shear failure.
It is likely that the 200 kPa test would have failed in
shear, if CU triaxial tests have been conducted on very
young sample (say 7 day old). A younger sample at low
confining pressure may fail in shear before reaching
the 3 = 0 line, as the cement reaction has not provided
enough strength.
5.5
Unconfined compressive strength
Unconfined compression strength (UCS) tests give the

ultimate strength at failure under compressive loading.
The strength of CB wall material improves as it ages.
Manassero et al. (1995) reported summary of the UCS
of CB wall material as a function of curing time as
Figure 34. Test specimens displaying a well defined shear

failure plane after compression at high confining pressure:
90 day old mixer-cast sample at 900 kPa confining pressure
(left) and 11.5 year old block sample at 500 kPa confining
pressure (right).
Figure 35. The increase in UCS with curing time for CB

wall material (Manassero et al., 1995).
shown in Figure 35. This phenomenon is similar to the

commonly observed behaviour of cement concrete and
therefore the ICE (1999) specification recommends
conducting the UCS test at the age of 28 days with
a requirement of UCS greater than 100 kPa. For CB
material, UCS tests are essentially equivalent to the
undrained compression tests with zero confinement
that were shown in Section 5.4. Hence, the failure
mode is tensile. The failure pattern for many samples
in the UCS tests was by a number of vertical cracks in a
brittle manner, while occasionally samples had incline
cracks (see Fig. 36). Unfortunately the latter case gives
false impression that CB materials fail in shear, which
may lead to assigning UCS as shear strength in engineering analysis. More appropriate analysis is to check
the minor principle stress does not exceed the tensile
strength of the material (see next Section).
Figure 37 shows UCS value plotted as a function
of the age of the samples obtained from the disused gas work site. These samples display a definite
162
Figure 38. Change in UCS with sample age for the

mixer-cast, trench-cast and block samples.
Figure 36. Photograph of vertical tensile spilt type failure
(left) and inclined shear type failure (right) of UCS test
samples.
Figure 37. Change in UCS of mixer-cast sample with age.
increase in strength with age. For example, the mean

UCS at 28 days of uncontaminated mixer samples
was approximately 360 kPa, whereas that at 90 days
had increased to 890 kPa. After 90 days, there was
relatively no improvement in strength of the uncontaminated slurry. Similar continuous increase in strength
up to 8 months is reported by Manassero et al. (1995).
The figure also presents that all the specimen after 7
days of curing comply with the ICE (1999) specification of minimum 100 kPa UCS (pink line in the figure).
There are variations in the range of UCS results, irrespective of the samples age. This variation in UCS
has no correlation with bulk density or moisture content. Also the amount of strength depends upon the
nature of cracking (tensile splits) in the specimen in
the unconfined condition.
The hydraulic conductivity reduces at slower rate
from 90 days till 3 years of period, as reported in
Figure 3. This could be because the cement-slag reaction beyond 90 days imparts negligible improvement
in the cemented bond strength, but their reaction
might be closing pores of the hydrated gel to reduce
the hydraulic conductivity. The CB material contains porous and amorphous gel of Calcium-SilicateHydrate (CSH) since 21 days of age. There is a
possibility that the hydrated slag (or hydrated cement)
would continue to react with other hydration product
after 90 days, where the newer reaction products would

occupy the vacant void space in the CB wall material rather than strengthening the contacts between the
porous network of the CHS gel.
Figure 38 show a comparison between the mixercast samples and the trench-cast samples at various
ages. The average strength of both sample types shows
a difference of 100 kPa at age of 28 days and the difference increases more than 300 kPa at 90 days of age.
These results indicate that the presence of contaminants hamper the strength gain of the CB wall material.
In all the cases, the variation in strength of the mixer
samples was narrow than that of the trench samples,
which means that the mixer samples are relatively
less heterogeneous. Also most of the samples have
strength above 100 kPa after age of 28 days i.e. more
than the specified minimum strength requirement.
Figure 38 also displays UCS of block samples
along with mixerand trenchsample against the age.
This comparison shows that the block sample at age
of 11 years has vast range of UCS from 170 to 790 kPa.
In all the cases the scatter of results for the mixer-cast
samples was smaller than the trench-cast samples,
followed by the block samples. The high variation in
results of the field blocksamples is because of higher
exposure to contamination and presence of impurities,
which leads to more heterogeneous behaviour.
The block samples have many pockets of
entrapped soil/stone, air voids and contaminants.
These impurities are relatively soft compared to the
set CB slurry, which resulted in poor bonding between
them causing variable UCS. Further level of heterogeneity may be introduced by aggressive environment
like chemical attack, weathering action at ground
surface and wetting-drying cycles due groundwater
fluctuation. This can cause the behaviour of the block
samples different to that of the laboratory cured samples. Therefore, in situ testing can be a way forward
to characterize heterogeneity of CB cut-off walls at
contaminated sites (Soga et al. 2005).
The trench-cast samples gain strength at slightly
slower rate until an initial 90 days compared to mixercast, but both of them achieved the target UCS value
of 100 kPa by the age of 28 days. The range of UCS
values at the age of 11 years for the mixer-cast,
163
occur to generate suction in the specimen. Therefore,

the tensile strength is believed to be contributed by the
cementation rather than suction due to negative pore
pressure.
The tensile strength of CB material is about 7.8%
of the average UCS. This is less than cement concrete,
which has tensile strength nearly 10% to 15% of the
compressive strength (ACI 318, 2008). Namikawa and
Koseki (2007) reported the tensile strength measured
by the Brazilian split method for cemented soils as 8
to 13% of the UCS, which is near the lower range for
the CB material.
6
Figure 39. Stress-strain relation in Brazilian tensile test of
cement bentonite samples.
trench-cast and the block field samples were 660

to 910 kPa, 670 to 1064 kPa and 170 to 790 kPa,
respectively. This implies that the block field samples have inferior strengths and hydraulic conductivity
along with large variability because of heterogeneity caused by aggressive environment and impurities.
These heterogeneities introduce weaker zones, which
would encourage stress localisation and progressive
collapse, thereby exhibiting failure at high axial strain
during the UCS tests.
5.6 Tensile strength
The previous sections discussed that the CB specimen may possess tensile strength, though none of the
stress path cross tension cut-off line. Therefore, it is
considered to find out the tensile strength of CB material. Brazilian test was performed on two 10 year old
mixer-cast and trench-cast samples. The test was
conducted in accordance with BS/EN-12390 part 6
(2000), which is one of the standard ways for evaluating tensile splitting strength for concrete specimens.
The test involved compressing a 100 mm diameter by
150 mm long specimen along its circumference. The
force can split apart the cylinder along its diameter and
the tensile strength can be calculated. Figure 39 shows
the stress-strain behaviour exhibited by the samples. A
prominent vertical crack was developed in the samples
as they fail in tension and the average tensile strength
determined is 65 kPa. This implies that the CB material
will start developing tensile cracks and fail in triaxial
tests with low confining pressure as soon as it exceeds
the tensile strength.
This test is a total stress test and measure of pore
water pressure is unavailable. One of the assumptions
is that the effective stress may be zero and the negative excess pore pressure would be holding the cracks.
Such stress condition is similar to undrained triaxial
test at zero confining pressure near the end of experiment. But prior to failure, the dilative behaviour of CB
material has not been noticed in any of the previous
experiments and pore volume expansion is unlikely to
CONCLUSIONS
The behavior of CB cutoff walls for polluted sites

has been presented above. The hydraulic conductivity, compatibility and strength are all highly dependent
upon stress state, sample age, mix formulation, and
chemical effects. Importantly, the properties measured
are affected by the measurement methods.
Hydraulic conductivity decreases with increasing
curing time and is lowest when the CB mixture is
approximately 80%/20% GGBFS and OPC. Laboratory tests on mixer samples yielded the lowest values
of hydraulic conductivity while borehole constant flow
tests produced the highest. Field testing demonstrated
that increasing the length of the increment tested
generally increased the measure value of hydraulic
conductivity. Hydraulic fracturing in borehole tests has
been shown to occur under some field conditions.
CB appears more resistant to degradation in the
presence of organic pollutants compared to soil bentonite materials. Under UK climate conditions, CB
does not appear to degrade due to seasonal (or
simulated) wetting and drying. However, the effect
should be examined for other climate conditions. CB
exhibits compressibility behavior similar to soils with
yielding stresses (preconsolidation pressures) in the
range of 100 to 700 kPa, depending upon mix design.
All strength tests show an increase in strength with
increasing sample age. In drained triaxial shear testing,
at low stress level, the material exhibits stiff behaviour
at the beginning. However, when the mean effective
stress in the drained stress path reaches the yield stress,
the open structure collapses and the stiffness suddenly
reduces. At high stress level, the structure continuously collapses from the beginning of shearing and
large volumetric strain is generated. The behaviour is
similar to that of metal foam material. In undrained
triaxial shear testing, at low stress levels, the material fails in tension. At higher stresses, the material
exhibits large excess pore pressures as the open structure collapses. Examination of failed samples show
that those from UCS or triaxial tests at low confining
pressure fail in tension whereas those at higher confining stress fail in shear. Such failure mechanisms need
to be beard in mind when conducting engineering analysis of CB walls. Further studies of the in situ stress
state are recommended since CB properties exhibit
strong dependence on the state of stress.
164
ACKNOWLEDGEMENTS
The writers would like to thank the UK EPSRC (Engineering and Physical Sciences Research Council), the
Hinduja Foundation, the Cambridge Commonwealth
Trust, St. Edmunds College, the Engineering Department, and the Cambridge Philosophical Society for
their financial support of K. Joshi during his graduate studies. The writers would also like to thank
Cambridges Churchill College and Bucknell University for their financial contributions to the J. Evans
sabbatical at Cambridge University.
REFERENCES
ACI Committee 318 (2008) ACI 318-08: Building Code
Requirements for Structural Concrete and Commentary.
American Concrete Institute. ISBN 0870312642.
Benson, C. H., Gunter, J. A., Boutwell, G. P., Trautwein,
S.J., and Berzanskis, P. H. (1997) Comparison of Four
Methodsto Assess Hydraulic Conductivity J. Geotechnical and Geoenvironmental Engineering, ASCE, 123(10),
929937.
Bjerrum, L., Nash, J.K.T.L., Kennard, R.M. and Gibson, G.E.
(1972) Hydraulic fracturing in field permeability testing
Gotechnique 22(2), 319332.
Britton, J. P., Filz, G. M. and Herring, W. E. (2004) Measuring the Hydraulic Conductivity of Soil-Bentonite Backfill J. Geotechnical and Geoenvironmental Engineering,
ASCE, 130(12), 12501258.
Brown, K. W. and Anderson, D. C. (1983) Effect of Organic
Solvents on the Permeability of Clay Soils EPA-600/
2-83-016, Mar., 153 pp.
Cermak, J., Evans, J.C. and Tamaro, G.J. (2012) Evaluation
of Soil-Cement-Bentonite Wall Performance Effects of
Backfill Shrinkage DFI 4th International Conference on
Grouting and Deep Mixing, New Orleans, LA.
Evans, J. C., Fang, H. Y., and Kugelman, I. J. (1985) Organic
Fluid Effects on the Permeability of Soil-Bentonite
Slurry Walls, Proceedings of the National Conference on
Hazardous Wastes and Environmental Emergencies,
Cincinnati, OH, May, 267271.
Evans, J. C. and Opdyke, S. M. (2006) Strength, Permeability, and Compatibility of Slag-Cement-Bentonite
Slurry Wall Mixtures for Constructing Vertical Barriers, Proceedings of the 5th International Conference on
Environmental Geotechnics, Cardiff, Wales, June 2630,
Thomas Telford Publishing, UK.
Filz, G.M. (1996) Consolidation stresses in soil-bentonite
back-filled trenches. Proc., 2nd Int. Congress on Env.
Geotechnics, M. Kamon, Ed., Osaka/Japan, 497502.
Filz, G. M., Evans, J.C. and Britton, J.P. (2003) Soilbentonite hydraulic conductivity: measurement and variability Soil and Rock America Conf. 2003: Proc. of the
Joint 12th Panamerican Conf. on Soil Mechanics and
Geotechnical Engineering, Vol. 2., 13231328.
Fratalocchi, E., Pasqualini, E. and Balboni, P. (2006) Performance of a cement-bentonite cut-off wall in an acidic
sulphate environment. Proc. 5th Int. Conf. Environmental
Geotechnics, Cardiff, Vol. 1, 133139.
Garvin, S. L. and Hayes, C. S. (1999) The chemical
compatibility of cement-bentonite cut-off wall material.
Construction and Building Materials, Vol. 13, 329341.
Institute of Civil Engineers (1999) Specification for the construction of slurry trench cut-off walls as barriers to pollution migration, Thomas Telford Publishing, London, UK.
Jefferis, S. A. (1981) Bentonite-Cement Slurries for

Hydraulic Cut Offs Proceedings of the Tenth International Conference on Soil Mechanics and Foundation
Engineering, Stockholm, June 1519, A.A. Balkema,
Rotterdam, 435440.
Jefferis, S. A. (1997) The origins of the slurry trench cut-off
and a review of cement-bentonite cutoff walls in the UK
First International Containment Technology Conference
and Exhibition, Florida, 5261.
Jefferis, S. A. (2012) Cement-Bentonite Slurry Systems
ASCE Proceedings of the 4th International Conference on
Grouting and Deep Mixing GSP 228, 124.
Joshi, K. (2009) Long-term Engineering Performance and
In-Situ Assessment of Cement-Bentonite Cut-off Walls,
Dissertation submitted for the degree of Doctor of Philosophy, University of Cambridge.
Joshi, K., Kechavarzi, C., Sutherland, K., Ng, M. Y. A.,
Soga, K., &Tedd, P. (2010) Laboratory and in situ tests for
long-term hydraulic conductivity of a cement-bentonite
cutoff wall J. Geotechnical and Geoenvironmental
Engineering, 136(4), 562572.
Khoury, M. A., Fayad, P. H., & Ladd, R. S. (1992) Design,
construction and performance of a soil-bentonite cutoff wall constructed in two stages. Slurry wall: Design,
construction, and quality control (STP 1129), 289308.
Manassero, M. (1994) Hydraulic conductivity assessment
of slurry wall using piezocone test J. Geotechnical
and Geoenvironmental Engineering, ASCE, 120(10),
17251747.
Manassero, M., Fratalocchi, E., Pasqualini, E., Claudia, S.
and Verga, F. (1995). Containment with vetical cutoff walls. Proc. Geoenvironment 2000, Geotechnical
special publication 46, ASCE, New Orleans, 11421172.
Mitchell, J. K. and Soga, K. (2005) Fundamentals of Soil
Behavior. 3rd edi., John Wiley & Sons, USA.
Namikawa, T. and Koseki, J. (2007) Evaluation of tensile strength of cemented-treated sand based on several
types of laboratory tests, Soils and Foundations, 47(4),
657674.
Ng, M. Y. A. (2008) Modeling of Hydraulic Fracturing in
Cement Bentonite Geomaterials Dissertation submitted
for the degree of Doctor of Philosophy, University of
Cambridge.
Opdyke, S. M. and Evans, J. C. (2006) Slag-CementBentonite Slurry Walls, ASCE J. of Geotechnical and
Geoenvironmental Engineering, 131(6), 673681
Parez, L., Fauriel, R. (1988) Le pizocne amliorations
apportes la reconnaissance de sols Rvue Franaise
de Gotechnique, 44, 1327
Ruffing, D. G., Evans, J. C., and Malusis, M. A., (2010)
Prediction of Earth Pressures in Soil-Bentonite Cutoff
Walls, ASCE GeoFlorida 2010 Advances in Analysis,
Modeling and Design GSP 199, 2416-2425.
Soga, K., Sutherland, K. J., Kechavarzi, C. and Whittle,
R. W. (2005) In situ Permeability Measurement of a
Contaminant Containment Wall. The 16th Int. Conf. on
Soil Mechanics and Geotechnical Eng., Osaka, Vol. 4,
24452448.
Tedd, P. (2005) Field Validation Of Cement Bentonite
Slurry Cut-Off Walls In Chemically Aggressive Ground
At Gloucester Gasworks BRE Client report number
221476, Building Research Establishment Ltd.
Yeboah, N., Evans, J. C. and Burns, S. E. (2013) The Role
of Bentonite in Slag-Cement-Bentonite Slurry Wall Performance: in preparation for ASCE J. of Geotechnical and
Geoenvironmental Engineering.
165
Specific lectures

Osmotic phenomena in bentonites

A. Dominijanni, M. Manassero & S. Puma
Politecnico di Torino, Torino, Italy
ABSTRACT: A theoretical approach has been proposed in order to derive constitutive equations for the coupled
chemical-hydraulic-mechanical behaviour of bentonites, which are clay soils characterized by a high specific
surface and a permanent negative electric charge on their solid skeleton. The phenomenological parameters
that govern the transport of electrolyte solutions through bentonites, i.e. the reflection coefficient, which is also
called the chemico-osmotic efficiency coefficient, and the osmotic effective diffusion coefficient, have been
measured through laboratory tests on a sodium bentonite with porosity of 0.81, over a range of sodium chloride
concentrations in the pore solution that varied from 5 mM to 100 mM. The relevance of the osmotic phenomena
has been shown to decrease when the salt concentration increases for this bentonite. The reflection coefficient
has been measured also on a calcium bentonite, over a range of calcium chloride concentrations in the pore
solution that varied from 5 mM to 10 mM: in this case, the osmotic behaviour has resulted to be negligible. The
obtained results have been interpreted by assuming that the microscopic deviations of the pore solution state
variables from their average values are negligible. In this way, it has been possible to interpret the macroscopic
behaviour on the basis of the physical and chemical properties of the bentonite mineralogical components.
INTRODUCTION
The term bentonite is commonly used to indicate a

clay soil with a high content (>70%) of montmorillonite, a mineral of the smectite group. Montmorillonite particles are thin lamellae that are characterized
by a high specific surface (defined as the surface per
unit weight) and a permanent negative electric charge.
Bentonite is used in hydraulic and contaminant barriers, because of its low hydraulic conductivity, k, to
permeation with water and dilute aqueous solutions
(k typically 3 1011 m/s). Geosynthetic clay liners
(GCLs), which consist of a thin layer of bentonite (5to 10-mm thick) sandwiched between two geotextiles,
are examples of such barriers. GCLs are currently used
in bottom and cover landfill barriers to limit water
infiltration and contaminant migration.
Owing to its physical and chemical properties,
the macroscopic mechanical behaviour and transport
properties of bentonite cannot be modelled through
classical soil mechanics approaches. In fact, bentonite
swells or shrinks in response to changes in the chemical composition of the pore solution. Moreover, when
a hydraulic pressure gradient, or a solute concentration
gradient, is applied to it, both a volumetric flux and a
diffusive solute mass flux occur, as in semipermeable
membranes. For this reason, a theoretical approach
that accounts for the electric interaction that occurs
between the solid skeleton of the bentonite and the ions
contained in the pore solution is proposed in the first
part of this paper. The experimental determination of
the swelling pressure and the transport properties of

both a natural sodium bentonite and a calcium bentonite is described in the second part. The obtained
results have been interpreted with the proposed theoretical model, in which the microscopic properties
of the bentonite have been linked to the observed
macroscopic behaviour.
2 THEORY
Montmorillonite lamellae are characterized by a negative electric charge, due to the isomorphic substitution
of a portion of the tetravalent silicon (Si4+ ) and the
trivalent aluminium (Al3+ ) in their crystalline structure, with metals, such as magnesium (Mg2+ ), which
have a lower valence. This electric charge per unit solid
volume can be expressed as F c sk,0 , where F is Faradays constant (96,485 C mol1 ) and c sk,0 is the molar
concentration per unit solid volume of the solid skeleton electric charge, which is assumed to have unit
valence (i.e. zsk = 1). c sk,0 represents the moles of
solid skeleton electric charge per volume of solids and,
in order to be compared with the ion concentrations of
the pore solution, it needs to be divided by the void
ratio, e, which represents the pore volume per volume
of solids. Dominijanni and Manassero (2012b) have
shown that c sk,0 is proportional to the effective specific surface of the solid particles and decreases when
the montmorillonite lamellae aggregate to form the
so-called tactoids.
169
the hydraulic pressure (referenced to the atmospheric

pressure, as is usual in soil mechanics), u, and the
salt concentration, cs . The corresponding variables of
the pore solution can be evaluated from the following
conditions:
If the pore solution contains a single salt that is completely dissociated with the following stoichiometric
reaction:
where z1 and z2 are the electrochemical valences of

the cation and the anion, and 1 and 2 are the stoichiometric coefficients of the cation and the anion,
respectively, the following condition has to be satisfied in order to preserve electroneutrality within a
saturated porous medium, even in the presence of the
solid skeleton electric charge:
where c 1 and c 2 are the molar concentrations of the

cation and the anion, respectively, and e is the void
ratio.
As a consequence, the solid skeleton electric charge
influences the distribution of the ions contained in the
pore solution. This phenomenon is known as the ionpartition effect and is expected to be more relevant for
porous media characterized by higher solid skeleton
charge concentrations.
2.1
where T and T are the absolute temperature in the pore

solution and in the external bulk solution, respectively;
w and w are the water chemical potential in the pore

solution and in the external bulk solution, respectively;
ec
ec
i and i are the electrochemical potentials of the
i-th ion in the pore solution and in the external bulk
solution, respectively.
The water chemical potential, w , and the ion electrochemical potentials, ec
i , of the external solution
can be related to the hydraulic pressure, u, and the
salt concentration, cs , for a dilute solution, as follows (Katchalsky and Curran, 1965; Dominijanni and
Manassero, 2012a):
Equilibrium conditions
When an electrically charged porous medium is placed

in contact with an external bulk solution that contains
the same ions that are present in the pore solution, a
thermodynamic equilibrium condition is reached, after
a certain period of time, in which the water chemical potential and the ion electrochemical potentials
between the two solutions are equal. The external bulk
solution can be considered as a chemical thermometer in order to evaluate the equilibrium conditions of
the porous medium (Coussy, 2004). The electroneutrality condition in the external solution is given by:
where c1 and c2 are the molar concentrations of the

cation and the anion that are contained in the bulk
solution.
It is convenient to define the salt concentration, cs ,
of the external solution as follows:
Using Eq. (3), the electroneutrality condition,

Eq. (2), provides the relation between the electrochemical valences and the stoichiometric coefficients:
where 0w and 0i are integration constants that only

depend on the absolute temperature T; cw is the water
2

molar concentration; = RT i cs is the osmotic
i=1
pressure; i is the chemical potential of the i-th ion;

R is the universal gas constant (8.314 J mol1 K1 );
and is the electric potential.
The state variables in the external bulk solution can
be measured easily, whereas it is very difficult to determine the corresponding variables in the pore solution.
Moreover, the relations obtained by linking the chemical potentials to the state variables of the pore solution
are more uncertain, due to the interaction with the
solid skeleton charge, which alters the ion concentration distribution near the solid particles. The simplest
assumption that can be adopted involves using analogous relations to Eqs. (8) and (9), as they are also
considered valid for the pore solution. This assumption, which was first proposed by Donnan (1911),
neglects the microscopic deviations of the ion concentrations from their average values that are induced
by the electric potential distribution within the pores.
If this approximation is accepted, the water chemical
potential,
w , and the ion electrochemical potentials,
ec
,
of
the
pore solution can be expressed as follows:
i
As a result, the equilibrium condition can be characterized by the following state variables of the
external bulk solution: the absolute temperature, T,
170
On the basis of Eqs. (17) and (18), the swelling

pressure can be expressed as follows:
where
0w and
0i are integration constants that only
depend on the absolute temperature T; c w is the molar
concentration of the water in the pore solution, which
can be taken equal to the molar concentration of
the water in the external bulk solution, i.e. c w
= cw ;
2

= RT c i is the osmotic pressure of the pore solu
i=1
tion;
i is the chemical potential of the i-th ion in the
pore solution; and is the electric potential in the pore
solution.
The hydraulic pressure of the pore solution, u ,
and the ion partition factors, i , defined as the ratio
between the ion concentration of the pore solution and
the ion concentration of the external bulk solution, can
therefore be expressed on the basis of Eqs. (6) and (7),
and using Eqs. (8)-(11), as follows:
= is the electric potential of the porous

where
medium, which is also called Donnans potential.
On the basis of this approach, the hydraulic pressure of the pore solution is different from the hydraulic
pressure of the external solution that is in equilibrium
with it. The pressure difference between the pore solution and the external solution is called the swelling
pressure, usw , and is given by:
2.2 Transport equations

The most general approach for modelling coupled
fluxes is to invoke phenomenological equations by
applying the formalism of the Thermodynamics
of Irreversible Processes (Katchalsky and Curran,
1965;Yaroshchuk, 1995; Dominijanni and Manassero,
2012a,b). The main advantage of this approach is to
avoid any specification of physical properties of the
membrane, maintaining the model as general as possible. Using such a formalism, Dominijanni et al. (2013)
derived the following equations for the volumetric
flux, q, and the salt flux, Js , for a semipermeable porous
medium permeated by a solution containing a single
salt (e.g. NaCl or CaCl2 ):
where
Eqs. (12) and (13) for i = 1, 2, together with Eq. (1),

constitute a set of four equations that can be solved to
find the four unknown variables: the hydraulic pressure, u , the ion concentrations, c i for i = 1, 2, and the
electric potential, .
When the ion electrochemical valences are both
unitary, such as for NaCl, Eq. (13) implies that
Inserting Eq. (15) into Eq. (1) results in the following equation:
which has a positive solution of the following form:
In Eqs. (20)(25), k is the hydraulic conductivity,

n is the porosity, w is the water unit weight, is the
hydraulic friction coefficient, is the reflection coefficient, Di is the macroscopic diffusion coefficient of
the i-th ion and D is the osmotic effective diffusion
coefficient.
Dominijanni and Manassero (2012b) have demonstrated that, if the microscopic deviations of the variables from their average values are assumed to be
negligible, the macroscopic ion diffusion coefficients,
Di , result to be equal to the ion effective diffusion
coefficients, Di :
171
corresponding global values: in fact, inserting Eq. (24)

into Eq. (29) with Ds = Ds leads to:
where m is the dimensionless matrix tortuosity factor

that accounts for the tortuous nature of the actual diffusive pathway through the porous medium (Malusis and
Shackelford, 2002b), Di,0 is the free (aqueous) solution diffusion coefficient of the i-th ion, Ds is the salt
effective diffusion coefficient and Ds,0 is the free solution diffusion coefficient of the salt. When the solid
skeleton electric charge is equal to zero, the ion partition coefficients, i , are equal to 1 and Eqs. (20) and
(21) reduce to the Darcy equation and the classical
advective-diffusion equation, respectively.
The osmotic effective diffusion coefficient, D ,
results to be related to the reflection coefficient, ,
through Eq. (24), so that D = 0 when = 1. As a
result, the condition = 1 implies a null salt flux
through the porous medium, which, in this case, can
The coefficient k can be measured, under steady
state conditions, using traditional permeameters.
Malusis et al. (2001) developed a testing apparatus to
determine and D . This apparatus is able to impose
the condition of no-volumetric flux (q = 0) through
a soil sample in contact with two external solutions,
maintained at constant salt concentrations, so that the
global or averaged values of the coefficients can be
measured. The global values of and D are defined
as follows:
where ct and cb represent the salt concentration at

the top and the bottom boundaries of the clay sample, respectively, and cs = ct cb is their difference.
These coefficients can be determined by means of the
following relations under steady state conditions:
where u = ut ub and = t b represent the

differences between the hydraulic pressure and the
osmotic pressure at the boundaries of the clay sample,
and L is the length of the sample.
It is interesting to observe that the relationship
between D and is also maintained between their
In the case of a salt constituted by monovalent ions,

inserting Eq. (23) into Eq. (28) and using Eqs. (17)
and (18), the following expression of g is obtained:
where
3.1 Materials
The powdered bentonite tested in this study is an
Indian sodium bentonite that is used for the production
of a needle-punched GCL. The bentonite is characterized by a cation exchange capacity (CEC, measured using the methylene blue adsorption method) of
105 meq/100 g. The mineralogical composition, evaluated through x-ray diffraction analysis, indicates a bentonite that is primarily composed of smectite (>98%)
with traces of calcite, quartz, mica and gypsum.
The bentonite is characterized by a liquid limit (LL)
of 525% and a hydraulic conductivity of 8 1012 m/s,
measured at a 27.5 kPa confining effective stress using
de-ionized water as the permeant liquid.
Sodium solutions were prepared with sodium chloride (ACS reagent, purity = 99%) and de-ionized
water (DW). The sodium solutions were prepared at
different molarity values, in the 5 mM to 100 mM
range, with the aim of investigating the effect of
the monovalent cations on the osmotic behaviour
of the bentonite. The DW (pH = 6.95; EC at
20 C = 0.6 mS/m) consisted of tap water processed
through a series of activated carbon filters, a reverse
osmosis process and, finally, a UV lamp (Elix Water
Purification system). Moreover, the DW was deaerated
prior to use. The electrical conductivity (EC) measured
at 20 C for the NaCl solutions ranged from 60.5 mS/m
to 1.1 mS/m.
3.1.1 Calcium bentonite preparation
Calcium bentonite has been obtained through an accelerated degradation of sodium bentonite that was promoted by the cation exchange of calcium for sodium.
172
Sodium bentonite was kept in contact with a highly

concentrated (1 M) calcium chloride solution for a
week. In this way, sodium bentonite exchanged sodium
cations with calcium cations present in the equilibrium solution. Excess soluble calcium ions, contained
in the exchanged calcium bentonite were successively
removed by a series of hydration with de-ionized water
and consolidation cycles. Calcium bentonite was left
to consolidate in de-ionized water and, when complete
settlement was reached, excess water was removed and
the material was hydrated once again. This procedure
was stopped when the electrical conductivity of the
equilibrium solution was less than 70 mS/m.
3.1.2 Salt removal pretreatment
Prior to the osmotic property determination, the bentonite was submitted to a process with the aim of
removing the soluble salts, mainly sodium, which
are naturally present inside the material, due to its
marine origin. The treatment prevents soluble salts
from interfering with the determination of the osmotic
properties.
Previous studies (Malusis et al., 2001; Malusis
and Shackelford, 2002a, 2002b; Shackelford and Lee,
2003; Yeo et al., 2005; Kang and Shackelford, 2009;
Di Emidio 2010) have used the flushing method to
remove soluble salts. This method consists of an initial permeation phase, performed under back pressure,
which requires a long period of time (i.e. from months
to a year), because of the low bentonite hydraulic
conductivity.
In this study, the squeezing method has been used
with the aim of reducing the salt removal time. The
squeezing method consists of a series of consecutive phases of powder bentonite hydration with DW,
at a higher water content than the liquid limit, and
drained consolidation, performed in a consolidometer under a maximum load of 500 kPa. Moreover, the
drained solution is sampled daily and the EC is monitored to evaluate the soluble salt concentration in the
bentonite pore water. After the squeezing process,
the material is oven dried at 105 C and pulverized
once again. When a 5 L consolidometer is used, the
above procedure can produce about 500 g of squeezed
dry powder bentonite, characterized by a lower EC
value than 50 mS/m, in 4050 days. The specimens
for the chemico-osmotic test are prepared by rehydrating the squeezed bentonite with DW at a lower water
content than the liquid limit value and then by statically
compacting the material in a compaction mould, while
allowing the excess water to be released. Dry powder
bentonite is required for the swelling pressure test.
3.2 Testing apparatus and procedure
3.2.1 Chemico-osmotic test
The testing apparatus used to measure the global
reflection coefficient and the global osmotic effective
diffusion coefficient is described in detail in Malusis
et al. (2001). The main components of the apparatus
include the osmotic cell, the flow-pump system, the
pressure transducer, which is used to measure the differential pressure that develops across the specimen
during the test, and the data acquisition system.
The cell consists of a modified rigid wall permeameter, in which the top piston and the bottom pedestal
are equipped with three ports each: two enable the
different solutions to circulate through the top (electrolyte solution) and the bottom (DW) porous stones
with the aim of establishing a constant concentration
gradient across the specimen. The third port is installed
in both the top piston and the bottom pedestal to allow
the differential pressure across the specimen to be
measured.
The flow-pump system, which consists of a dualcarriage syringe pump and two stainless steel accumulators (Model 33 Twin syringe pump, produced
by Harvard, Holliston, MA), prevents the volumetric
flux through the specimen by simultaneously injecting
into and withdrawing from the porous stones the same
volume of solution. In order to obtain this result, the
syringes have to move at the same rate.
The test was performed according to the procedure
proposed by Malusius et al. (2001): a solution containing a known electrolyte concentration was circulated
in the top porous stone, while DW was circulated in
the bottom porous stone. The concentration difference
across the specimen was maintained constant by continuously infusing the two liquids at the boundaries of
the specimen.
Since the specimen was preliminary squeezed with
DW to remove the soluble salts, the EC of the electrolyte solutions in the flux exiting from the porous
stones at the steady state was induced solely by the
contributions of Cl and Na+ ions.
A calibration was performed and showed that the
relation between the EC and solution molarity was
linear over the examined concentration range both for
NaCl and CaCl2 solutions. As a consequence, the EC
of the withdrawn fluxes (i.e. from the top and bottom
porous stones, respectively) was monitored by sampling the solution contained in the pistons, and the
electrolyte molar concentration was derived using a
linear relation. Since the volumetric flux through the
specimen was hindered, the global reflection coefficient could be calculated using Eq. (33).
The diffusive solute flux through the specimen was
calculated for the n-th sampling interval as follows:
where cm
s is the solute molar concentration measured
by sampling the solution coming out from the bottom
porous stone, Vm is the volume of the solution circulating in the porous stones in the tm interval, AS
is the cross-section of the specimen and Qn is the
cumulative salt molar mass per unit area that passed
through the specimen. The global osmotic effective
173
diffusion coefficient, Dg , is calculated at the steady

state as follows:
where ct,avg and cb,avg are the average top and bottom
salt concentrations, respectively.
3.2.2 Swelling pressure test
The swelling pressure apparatus primarily consists of a
stainless steel oedometer cell, a NaCl solution supply
tank that is placed above the pressure panel, a displacement transducer connected to the cell top piston, which
is used to measure the axial strains of the specimen, a
load cell and a data acquisition system.
The swelling pressure apparatus consists of a rigid
cell that confines the sample (i.e. the oedometer),
which allows access to the water through both porous
stones. The cell is connected to a pressure panel that
allows the specimen to be back-pressurized. The rigid
piston above the upper porous stone is connected to
the load cell, which measures the pressure that has to
be applied in order to hinder the axial strain of the
specimen.
The test procedure requires a known amount of dry
material to be dusted inside the oedometer ring, the cell
to be assembled and a NaCl solution to be supplied.
The specimen, which is characterized by an initial dry
height of 5 mm, is allowed to swell to 10 mm. The
piston is then blocked, the sample is back-pressured
and the steady state swelling pressure is recorded after
a short transitional phase.
Since the bentonite that is initially dusted inside the
oedometer is dry, the pressure increases for a number
of days, during the hydration phase, and the steady
state swelling pressure is reached when hydration has
been completed.
4
4.1
RESULTS AND INTERPRETATION

Chemico-osmotic test on sodium bentonite
The first chemico-osmotic test was performed using

the oven dried squeezed sodium bentonite, rehydrated
with DW and then statically compacted, in a drained
compaction mould, at a porosity, n, equal to 0.81
(e = 4.26). After the preparation phase, the 17 mm
thick specimen was transferred to the cell for the
osmotic test.
After assembling the cell, DW was circulated
through the top piston and the bottom pedestal for two
weeks in order to establish a steady baseline differential pressure, before a concentration gradient was
applied to the specimen. A source concentration of
NaCl then was injected into the top porous stone,
while DW was continuously circulated in the bottom
porous stone.
A multiple-stage chemico-osmotic test was performed by sequential circulation of chemical solutions
containing 5.16, 10.27, 20.24, 51.94 and 109.31 mM

NaCl concentrations at a constant flow rate of
0.05 mL/min.
The EC values of the salt mass fluxes withdrawn
from the top and the bottom porous stones, measured
during the testing stages, are shown in Fig. 1.
The measured values depend on the NaCl concentrations imposed at the boundaries of the specimen: the
EC values progressively increase during the test as the
NaCl concentration of the injected solution in the top
porous stone rises. The trends of the electrical conductivity of the flux withdrawn from the top porous stone,
ECt,exit, and the electrical conductivity of the flux
withdrawn from the bottom porous stone, ECb,exit,
both show that a steady state has been reached for
each stage.
Moreover, the difference between the EC values
measured in the flux withdrawn from the top porous
stone (ECt,exit) and the EC values of the solutions
injected into the same stone (ECt,ref) is due to the loss
in NaCl mass induced by the diffusion through the
bentonite from the top to the bottom boundary.
The global reflection coefficient values, g ,
obtained during the multiple-stage chemico-osmotic
test, are shown in Fig. 2 as a function of time. The g
values are determined using Eq. (33), on the basis of
the differential pressure, u, measured during the test
with a time step of 10 min, and the osmotic pressure,
, calculated from the average of the top and bottom
NaCl concentrations. The steady state values of the
variables are reported in Table 1 for each concentration
stage.
As far as the EC measurements are concerned, the
trend of the global reflection coefficient shows that
a steady state has been reached for each stage. The
steady state g values tend to decrease as the salt
concentration in the top porous stone increases. The
recorded values range from 68%, for a 5.16 mM NaCl
source concentration, to 5%, for a 109.31 mM NaCl
source concentration. The global reflection coefficient
can be assumed approximately null for higher molarities. The cumulative molar mass per unit area, Q, of the
NaCl that migrated through the specimen during the
multiple-stage test is reported in Fig. 3. The values of
the global osmotic effective diffusion coefficient, Dg ,
which have been obtained from the Q measurements
shown in Fig. 3, are reported in Table 1.
4.2
Chemico-osmotic test on calcium bentonite
The second chemico-osmotic test was performed on

a calcium bentonite specimen, in order to analyse the
osmotic behaviour of bentonite in long term landfill
condition, i.e. when the cation exchange phenomenon
has been developed completely and the bentonite
exchange sites are entirely saturated with calcium ions.
The chemico-osmotic test was performed using the
oven dried squeezed material, rehydrated with DW
and then statically compacted, in a drained compaction
mould, at a porosity, n, equal to 0.67. After assembling
the cell, DW was circulated through the top piston
and the bottom pedestal for two weeks in order to
174
Figure 1. Electrical conductivity of the salt flux withdrawn from the top porous stone (a) and the bottom porous stone (b) as
a function of time during the multiple-stage chemico-osmotic test on sodium bentonite.
establish a steady baseline differential pressure, before

a concentration gradient was applied to the specimen.
A source concentration of CaCl2 then was injected
into the top porous stone, while DW was continuously
circulated in the bottom porous stone.
A multiple-stage chemico-osmotic test was performed by sequential circulation of chemical solutions
containing 5.37 and 10.40 mM CaCl2 concentrations
at a constant flow rate of 0.05 mL/min.
The global reflection coefficient values, g ,
obtained during the multiple-stage chemico-osmotic
test, are shown in Fig. 4 as a function of time. The g
values are determined using Eq. (33), on the basis of
the differential pressure, u, measured during the test
with a time step of 10 min, and the osmotic pressure,
, calculated from the average of the top and bottom
CaCl2 concentrations.
The trend of the global reflection coefficient shows

that a steady state has been reached for each stage.
The steady state g values tend to decrease as the salt
concentration in the top porous stone increases. The
recorded values are very low and range from 0.5%, for
a 5.37 mM CaCl2 source concentration, to 0.2%, for
a 10.40 mM CaCl2 source concentration. The global
reflection coefficient can be assumed completely null
for higher molarities. The results highlight that calcium bentonite shows very low osmotic behaviour at
the tested molarities values.
4.3 Swelling pressure test
The swelling pressure test was performed using dry
specimens, prepared with the squeezed oven dried bentonite and characterized by an initial dry height of
175
Figure 2. Global reflection coefficient as a function of time during the multiple-stage chemico-osmotic test.
Table 1. Steady state values of the variables involved in the multiple-stage chemico-osmotic test, where: ct,exit and cb,exit are
the NaCl concentrations of the flux withdrawn from the top and the bottom porous stones, respectively; ct,avg and cb,avg are
the NaCl average concentrations in the top and the bottom porous stones, respectively; u is the hydraulic pressure difference
between the top and the bottom specimen boundaries, measured by the differential transducer; is the osmotic pressure
difference; g is the global reflection coefficient and Dg is the global osmotic effective diffusion coefficient.
ct,ref
ct,exit
(mM)
cb,exit
(mM)
ct,avg
(mM)
cb,avg
(mM)
u
(kPa)

(kPa)
g
()
Dg
(m2 /s)
5.16 mM
10.27 mM
20.24 mM
51.94 mM
109.31 mM
5.12
9.61
18.93
47.39
97.18
0.83
0.85
1.45
4.39
9.78
5.14
9.94
19.58
49.67
103.24
0.42
0.43
0.72
2.19
4.89
15.65
26.87
30.32
32.38
23.96
23.02
46.33
91.89
231.30
479.21
0.68
0.58
0.33
0.14
0.05
2.54 1010
3.52 1010
4.19 1010
4.60 1010
5 mm, which were allowed to swell to 10 mm during

hydration. The final hydrated volume of the specimens
corresponded to n = 0.81.
The test was performed by hydrating five different
specimens with five different NaCl solutions, characterized by increasing concentrations, i.e. 5, 10, 20, 50
and 100 mM. After hydration, the specimen volume
change was inhibited and the value of the swelling
pressure was recorded after a short transitional phase.
The swelling pressure trend is reported in Fig. 5 for
each test as a function of time. Since, during the tests,
the load cell was unloaded until the specimen swelled
to 10 mm, the initial swelling/hydration phase of the
dry material (from 5 to 10 mm) was characterized by
null swelling pressure values.
All the tests show that the swelling pressure
increases for approximately 1520 h, during the controlled hydration phase (i.e. when the volumetric
strain is inhibited), and that the equilibrium swelling
pressure is reached when the hydration phase is completed. Moreover, in the tests with lower NaCl concentrations (i.e. for 5 and 10 mM NaCl equilibrium
solutions), the specimens were successively backpressurized to 300 kPa. The obtained results show
that the swelling pressure value does not change after

back-pressurization.
In the test performed using the 100 mM NaCl solution, the bentonite specimen did not rise to 10 mm,
as it stopped at a height of 9.5 mm, and the load cell
was never loaded during the test. For this reason, the
swelling pressure for this test was taken equal to zero.
4.4 Interpretation of the results

on sodium bentonite
The experimental results can be related to the physical
and chemical properties of the tested bentonite under
the assumption that the microscopic deviations of the
state variables from their average values are negligible.
In such a case, both the global reflection coefficient
and the swelling pressure depend on the solid skeleton
electric charge through Eqs. (19) and (37). Therefore,
from the best fitting of the theoretical curves with the
experimental data of both tests, a value of c sk,0 equal
to 90 mM was found. Therefore, from the best fitting
of the theoretical curves with the experimental data of
both tests, a value of c sk,0 equal to 90 mM was found.
176
Figure 3. Cumulative molar mass of NaCl per unit area as a function of time during the multiple-stage chemico-osmotic test.
(a) ct,ref = 10 mM, (b) ct,ref = 20 mM, (c) ct,ref = 50 mM, (d) ct,ref = 100 mM.
Figure 4. Global reflection coefficient as a function of time during the multiple-stage chemico-osmotic test on the calcium
bentonite specimen.
The obtained theoretical curves are reported in Figs. 6

and 7, together with the experimental data.
The salt concentration at the top boundary was
taken equal to ct,avg to determine g , while the
salt concentration at the bottom boundary was
considered equal to zero, i.e. cb
= cb,avg
= 0. The
sodium transport number was calculated from the
sodium and chloride free-solution diffusion coefficient values (Shackelford and Daniel, 1991):
DNa,0 = 13.3 1010 m2 /s, DCl,0 = 20.3 1010 m2 /s.
In Fig. 8, the experimental reflection coefficient

data were also fitted with the empirical semi-log linear curve proposed by Shackelford et al. (2003) and
Malusis et al. (2003):
where A and B are the regression parameters. The
value of the coefficient of determination R2 close to
one confirms the ability of this empirical curve to fit
the g experimental data, as it was found by Malusis
177
Figure 5. Swelling pressure as a function of time. (a) cs = 5 mM, (b) cs = 10 mM, (c) cs = 20 mM, (d) cs = 50 mM.
Figure 6. Global reflection coefficient, g , as a function of

the average NaCl concentration at the top boundary of the
bentonite specimen, with the best fitting theoretical curve,
obtained for c sk,0 = 90 mM in Eq. (33) (continuous line).
Figure 7. Swelling pressure, usw , as a function of the NaCl

equilibrium concentration, with the best fitting theoretical
curve, obtained for c sk,0 = 90 mM in Eq. (19) (continuous
line).
et al. (2003) for the Kemper and Rollins (1966) and

Malusis and Shackelford (2002a) data. However, the
regression parameters A and B should be intended as
functions of the soil porosity (Malusis et al., 2003)
and of the bottom boundary condition (i.e. cb,avg ). The
advantage of interpreting the experimental data with
the proposed theoretical model is that, when the single unknown parameter, c sk,0 , has been calibrated on a
restricted experimental data set, the global reflection

coefficient values can be estimated for different soil
porosities and boundary conditions through Eq. (37).
The tortuosity factor was determined by plotting
the measured values of Dg as a function of the corresponding values of the complement to 1 of g , i.e.
(1 g ) and finding the intercept of the linear regression with the ordinate axis at (1 g ) = 1, i.e. g = 0
178
(Fig. 9). The tortuosity factor in Eq. (36) is in fact

given by:
Figure 8. Semi-log linear regression of the measured global

reflection coefficients versus average NaCl concentration at
the top boundary of the bentonite specimen.
Figure 9. Global osmotic effective diffusion coefficient,

Dg , as a function of the complement to 1 of the global reflection coefficient, g , with the theoretical linear relation given
by Eq. (32) (continuous line).
where Ds,0 is the NaCl free solution diffusion coefficient, which is equal to (Shackelford and Daniel,
1991).
A value of m equal to 0.31 was obtained from the
data plotted in Fig. 8. The resulting theoretical curve
of Dg is reported in Fig. 9 as a function of the top
boundary salt concentration.
The obtained values of c sk,0 and m are compared in Table 2 with those derived by Dominijanni
and Manassero (2012b) from the interpretation of
the experimental results of Malusis and Shackleford
(2002a, 2002b), relative to a geosynthetic clay liner,
with a bentonite porosity, n, of 0.79 for different
concentrations of potassium chloride (KCl). The differences in the parameters obtained from the two
studies can be attributed to both the different mineralogical compositions of the tested bentonites and the
different salts contained in the pore solutions.
The theoretical linear relationship between Dg and
(1 g ) in Fig. 9 is a consequence of assuming that
the pore-scale variations in the hydraulic pressure, ion
concentrations, and water velocity within the soil are
negligible: as a result, the agreement of the experimental data with the linear relationship is an indication of
the acceptability of this assumption.
The goodness of the linear fitting shown in Fig. 9
(R2 = 0.9810) and the possibility of fitting both the
global reflection coefficient and the swelling pressure
data with a single value of c sk,0 are indications of the
ability of the proposed theoretical approach to simulate
the bentonite behaviour.
5
CONCLUSIONS
A theoretical approach that takes into account the interaction between the electric charge of the bentonite
Table 2. Comparison between the physical parameters derived from the interpretation of the experimental results in this work
and those obtained by Dominijanni and Manassero (2012b) from the interpretation of the tests carried out by Malusis and
Shackelford (2002a, 2002b).
Experimental data
Material
This study
Natural sodium bentonite
Method for removing soluble salts

Tests
Squeezing
Chemico-osmotic test and
swelling pressure test
NaCl
0.81
90
KCl
0.790.80
46
0.31
0.14
Salt in pore solution

Porosity, n ()
Solid skeleton charge concentration,
c sk,0 (mM)
Tortuosity factor, m ()
179
Malusis and Shackelford (2002a; 2002b)

Geosynthetic clay liner containing natural
sodium bentonite
Flushing
Chemico-osmotic test
solid skeleton and the ions contained in the pore

solution has been proposed. The phenomenological
parameters introduced in this theoretical approach
were measured for a bentonite specimen with porosity,
n, of 0.81, over a range in sodium chloride concentration in the pore solution varying from 5 mM to
100 mM. The global reflection coefficient, g , was
found to decrease with an increase in the salt concentration.This result is in agreement with the trends given
by the proposed theoretical model, assuming that the
microscopic deviations of the pore solution state variables from their average values are negligible. If this
assumption is accepted, the experimental data can be
used to derive the electric charge of the solid skeleton
(per unit solid volume), c sk,0 , and the tortuosity factor, m . The transport properties of bentonite can be
estimated from these physical properties to evaluate
its performance as a hydraulic and contaminant barrier in field applications. However, in order to verify
the applicability of the proposed model under different
boundary conditions and for different salts contained
in the pore solution, further experimental evaluations
must be conducted. Moreover, the results obtained for
a single salt contained in the bentonite pore solution
need to be extended to the more general problem of
a solution containing an unspecified number of salts,
in order to evaluate the performance of bentonites that
are used as contaminant barriers for real leachates.
REFERENCES
Coussy, O. (2004). Poromechanics. Chichester: Wiley.
Di Emidio, G. (2010). Hydraulic and chemico-osmotic performance of polymer treated clays. Ph. D. Thesis., Ghent:
Ghent University.
Dominijanni, A. & Manassero, M. (2012b). Modelling the
swelling and osmotic properties of clay soils. Part I:
The phenomenological approach. International Journal
of Engineering Science 51, 3250.
Dominijanni, A. & Manassero, M. (2012b). Modelling the

Science 51, 5173.
Dominijanni, A., Manassero, M. & Puma, S. (2013). Coupled
chemical-hydraulic-mechanical behaviour of bentonites:
Geotechnique 63 (3), 191205.
Donnan, F.G. (1911). Theorie der Membrangleichgewichte
und Membranpotentiale bei Vorhandensein von nicht
dialysierenden Elektrolyten. Ein Beitrag zur physikalischchemischen Physiologie, Zeitschrift fr Elektrochemie
und angewandte physikalische Chemie 17, 572581.
English translation republished in Journal of Membrane
Science 100 (1995), 4555.
Kang, J.-B. & Shackelford, C.D. (2009). Clay membrane
testing using a flexible-wall cell under closed-system
boundary conditions. Applied Clay Science 44, 4358.
Kemper, W.D. & Rollins, J.B. (1966). Osmotic efficiency
coefficients across compacted clays. Soil Science Society
of America, Proceedings 30, 529534.
Malusis, M.A., Shackelford, C.D. & Olsen, H.W. (2001). A
laboratory apparatus to measure chemico-osmotic efficiency coefficients for clay soils. Geotechnical Testing
Journal 24, 229242.
Malusis, M.A. & Shackelford, C.D. (2002a). Chemicoosmotic efficiency of a geosynthetic clay liner. Journal
of Geotechnical and Geoenvironmental Engineering 128,
No. 2, 97106.
Malusis, M.A. & Shackelford, C.D. (2002b). Coupling effects
during steady-state solute diffusion through a semipermeable clay membrane. Environmental Science and Technology 36, No. 6, 13121319.
Malusis, M.A., Shackelford, C.D. & Olsen, H.W. (2003).
Flow and transport through clay membrane barriers.
Engineering Geology 70, 235248.
Shackelford, C.D. & Daniel, D.E. (1991). Diffusion in
saturated soil: I. Background. Journal of Geotechnical
Engineering, 117, No. 3, 467484.
Shackelford, C.D. & Lee, J.-M. (2003). The destructive role
of diffusion on clay membrane behavior. Clays and Clay
Minerals 51, No. 2, 186196.
Yeo, S.-S., Shackelford, C.D. & Evans, J.C. (2005). Membrane behaviour of model soil-bentonite backfill mixtures. Journal of Geotechnical and Geoenvironmental
Engineering 131, No. 4, 418429.
180

Effect of mechanical settlement on biogas transfer in MSW landfills

L. Oxarango, J.P Gourc, G. Stoltz & A.J. Tinet
LTHE (Laboratoire dtude des Transferts en Hydrologie et Environnement), Universit Joseph Fourier,
Grenoble, France
ABSTRACT: Municipal Solid Waste (MSW) stored in landfills is a very compressible material. A dedicated
oedopermeameter apparatus is used to study the primary settlement of MSW. A complete characterization
of the material physical properties (dry density, porosity and gravimetric moisture content) is presented. The
corresponding gas permeability is measured. A correlation between the gas permeability and the volumetric
gas content is obtained. This set of results is used to simulate biogas flow around a collection well. Taking into
account the strong coupling between the mechanical state of the material and its transfer properties, a significant
effect of the vertical heterogeneity of the material on the gas flow distribution is obtained.
INTRODUCTION
Municipal Solid Waste is a very compressible porous

medium. Waste storing in a landfill cell leads to a gradient of material porosity from the top to the bottom
of the waste deposit. On the other hand the collection of biogas thanks to vertical wells is a key issue,
seeing that biogas is a greenhouse gas (environmental
issue) which can be recovered. The efficiency of the
gas collection in vertical wells rather than the leakage
through the cap cover is dependent of the gas transfer through the waste mass. Tentative modelling of
the gas flow, coupling gas permeability with compression of waste, is presented. A preliminary campaign
of laboratory tests is required to propose a relationship between compressibility and gas permeability of
the porous material.
2
CHARACTERISING WASTE MATERIAL
The LTHE developed a specific experimental method

and associated devices (Stoltz et al., 2010 WM and
2010 CH) in order to characterize the bio-hydromechanical behaviour of biodegradable waste generally called Municipal Solid Waste (MSW).
2.1
Material for compressibility measurements
Description of the experimental setup

A compression cell, called oedopermeameter, was
developed in laboratory (Fig. 1). It is capable of loading a waste sample under oedometric conditions (no
lateral strains) to a given compression stress . This
compression makes it possible to simulate the overburden stress of the overlying column of waste for
the sample placed in-situ. The vertical settlement H
Figure 1. Oedopermeameter apparatus experimental setup.
induced by applying compression stress is monitored

by a vertical displacement sensor as a function of time.
The sample is compressed between two diffusers
allowing the circulation of fluids without any significant hydraulic pressure loss. The two diffusers are
porous plates with the top one attached to a piston dedicated to transfer load and the bottom diffuser serves
as a support for the waste sample at the base. The possibility of fluid flows allows, on one hand, to drain
leachate during compression of a wet sample and, on
181
The gravimetric Liquid Content of each component (i)

is obtained after an oven drying as:
The mass fractions i are calculated with respect to

the dry mass as:
These values are useful to calculate the gravimetric

Moisture Content os the global sample w as:
Figure 2. View of the material tested.
2.2 MSW physical parameters measurement

Table 1.
Composition of the tested MSW.
Waste component
i
wi
Paper/cardboard
Plastic
Textiles
Glass
Metal
Garden/Food
Timber
Miscellaneous
0,261
0,140
0,055
0,061
0,057
0,327
0,035
0,065
0,479
0,303
0,879
0,000
0,000
0,483
0,262
0,125
0,240
0,146
0,040
0,083
0,077
0,300
0,027
0,078
Set of physical parameters

Determination of the set of physical parameters [d , w
and n] is required in order to characterize physically
the porous medium (Stoltz et al., 2010b).
The gravimetric Liquid Content w is obtained by
drying the sample of waste of total mass MT in an
oven.
the other, to measure the gas permeability of the sample without external significant head loss. In the case
of hydraulic tests corresponding to a complete saturation of the sample, the saturated liquid permeability
can be measured. The liquid permeability which is not
presented here was also studied by Powrie & Beaven
(1999).
Description of the tested waste material
The tested MSW (Fig. 2) comes from a French landfill. It is composed of a mix of household waste and
industrial waste sampled before landfilling. In order to
match the characteristic size of the oedopermeameter
cell, it is shreded to insure a maximum grain size of
70 mm. The characterization of the waste material is
done following the MODECOM procedure on a 150 kg
sample.
The waste composition is presented in Table 1.
MTi and wi correspond to the wet mass and the
gravimetric moisture content of the waste component i.
The mass fractions i of each component i are calculated with respect to the total mass of the sample (ie to
the wet mass MT ):
The temperature of drying oven is fixed at 85 C in

order to avoid the loss of solid mass that may likely
occur because of combustion of organic matter constituting a part of the MSW. Furthermore, the risk
of burning of some types of paper is also reduced.
Gravimetric Liquid Content is obtained after complete
stabilisation of solid mass MS (around 5 days). For
practical reasons, MT (around 10 kg) is assessed at the
beginning of oedopermeameter test and MS at the end.
The dry density d is obtained by measuring gravimetric Liquid Content w and total density of MSW
sample put in the oedopermeameter.
Measurement method of total porosity n of an unsaturated sample put in the oedopermeameter consists in
measuring its volumetric Liquid Content L and its
volumetric Gas Content G .
The volumetric Liquid Content is deduced from the

dry density d and gravimetric Liquid Content w by
with L the density of the liquid.
182
The volumetric Gas Content G is measured by gas

pycnometer method. This method was described in
detail by Stoltz et al. (2010a).
The gas pycnometer method allows the assessment
of the volume of the Gas Voids. The principle consists
of connecting the oedopermeameter cell to an external
chamber initially put under pressure with nitrogen N2 .
The use of nitrogen N2 as gas permits to avoid aerobic
biodegradation during the test. The volume of the Gas
Voids is determined from the BoyleMariotte relation
applied between the two conditions before and after
connection of the oedopermeameter sample and the
external chamber.
presents long term compression tests in oedometric

conditions, in order to assess the waste secondary
settlement (Conte et al., 2013).
The primary settlement can be modelled using a logarithmic dependence with the compression load following the classical law proposed by Terzaghi (1976)
for fine grained soils. One should note that in the case
of unsaturated waste this primary settlement does not
correspond to the decrease of the interstitial pressure
with time. That is to say that the total stress is an
effective stress = .
The oedometric compression test can be decomposed in two stages:
In the first stage, the material is over-consolidated.
That is to say that the imposed stress is lower
than the pre-consolidation stress ( < c ). c is the
maximum stress seen by the material before the
experiment during the filling up of the cell. In this
condition, the settlement H reads as:
Evolution of the physical parameters with

settlement in the oedometric test
Waste materials are known to present a very high
compressibility (Olivier & Gourc, 2007). Thus, the
evolution of the characteristic physical parameters
with respect to the compression load is very relevant. For a waste sample placed in an oedometric cell,
which section is constant, the initial sample height is
noted H0 . The corresponding initial physical parameters are noted d0 , n0 and w0 . The sample settlement
H is measured under a constant vertical stress (compression load) . The evolution of the characteristic
physical parameters can be calculated straightforward
using the formulae:
with e the void ratio defined from the porosity n as:
In the second stage, the material is normally consolidated. The imposed stress is lower than the preconsolidation stress ( > c ) and the settlement
H reads as:
In the case of waste material, the settlement

ofover-consolidated material should be neglected
(Cs = 0). Thus, the primary settlement can be
described using a single settlement coefficient Cc .
If the initial void ratio e0 is unknown, an overall
settlement coefficient called compression ratio Cr
should be used:
where MLout is the possible mass of liquid drained
out of the sample during the compression (this phenomenon occurs only for high initial gravimetric
moisture content).
The settlement H reads as:
2.3 Analysing the primary settlement of a MSW

Mechanical parameters
Primary settlements correspond to the period of
operation-compaction of waste and can be considered
as a purely mechanical process whereas secondary
settlement happens after closure of a cell when the
vertical load is stabilized. Secondary settlement is
interpreted as resulting from both biodegradation and
mechanical creep (Machado et al., 2008; Gourc et al.,
2010). In the present paper, only the experimental data
corresponding to the primary settlement will be used
for the modelling of the landfill behaviour. Its worth
noting that an another paper in the present conference
In the oedometric compression diagram, the preconsolidation stress c is conventionally evaluated as

the intersection of the lines corresponding respectively to the over-consolidated stage and the normally
consolidated stage.
Procedure of the experimental compression test
Samples with various initial moisture contents are
tested in the oedopermeametercell (Table 2). The
loading path, under vertical constraints , varies in
the different tests. Each loading step is applied for
183
Table 2. Values of the initial moisture content, the underconsolidate stress c , the Cc and Cr compressibility coefficients for the 10 tested samples.
w0
c
kPa
Cc
Cr
T1
T2
T3
T4
T5
T6
T7
T8
T9
T10
0.26
16
0.47
8
0.61
12
0.74
16
0.77
16
0.81
16
0.87
18
0.99
15
1.16
9
1.16
10
1,03 1,20 1,17 1,12 1,40 1,29 1,48 1,22 1,16 1,13
0,269 0,291 0,293 0,291 0,335 0,318 0,365 0,316 0,287 0,294
Figure 4. Device for vertical gas permeability tests with the

Oedopermeameter cell.
Figure 3. Compressibility of the waste samples.
24 hours. A complete test lasts up to a maximum of

7 days. For each compression step, the settlement H
and the potential released leachate mass (MLout ) are
recorded. The waste physical parameters could then be
computed for the initial state using eq. (6), (7) and (8)
and using eq. (9), (10) and (11) for each compression
step.
Determination of the compression ratio for the
waste material
The compressibility of the waste samples is analyzed
using the relationship between the void ratio and the
logarithm of stress (eq. (14)). At the initial state, the
waste sample is characterized by its initial void ratio
e0 under a stress corresponding to the oedopermeameterpiston weight only. For graphical convenience,
this stress is set arbitrarily to 1 kPa on Fig. 3. The
waste sample is slightly over-consolidated due to its
placement in the cell. The settlement model (eq. (16))
neglects the material compressibility in the overconsolidation phase. Under this assumption, this phase
should be interpreted using the pre-consolidation
stress c only. The settlement curves exhibit a linear
trend for every test for stresses higher than 20 kPa.
It shows than the pre-consolidation stress c must be
close to 20 kPa. The compression index Cc is evaluated
by a linear regression on the normally-consolidated
curve for c > 20 kPa.
The pre-consolidation stress c , the compression
index Cc and the compression ratio Cr (calculated from
equation (14)) are summarized in Table 2.
The average value of the compression index Cc is

Cc = 1.22. The average value of the compression ratio
Cr est Cr = 0.306. The compression index appears to
be higher for the tested waste than for soil even if
highly compressible soils like organic soils are considered. This result put forward that waste material
can be considered as a highly compressible material.
On the other hand the evolution of the compression
index with the initial gravimetric Liquid Content or
the initial dry density does not exhibit any remarkable
trend.
2.4 Material for gas permeability measurement

Gas permeability coefficient
For the gas permeability measurement using the same
oedopermeameter cell and the same waste material
than for the compression tests above, the procedure
with a constant differential gas pressure is satisfactory to measure gas permeability kG ranging between
2 1010 and 1 1014 m2 . However when a waste
sample is strongly compressed (very low porosity) and
exhibiting a high gravimetric Liquid Content, it is no
more possible to measure gas permeability (which in
this case is lower than 1014 m2 ) with the constant pressure head test and alternative falling pressure test is
used. More details are given in (Stoltz et al., 2010a).
Fig. 4 shows the underlying principle of these measurements. In permanent flow mode, pressure sensors
measure the pressure at entrance p1 and at exit p2 . A
mass flowmeter is used to measure flow rate qv at nitrogen entrance point. For gas, there is not conservation
of volumetric flow rate from one section to another
of the oedopermeameter since gas density varies with
pressure (Bouazza and Vangpaisal, 2003). However,
the flow of gas through porous medium is performed
at a low pressure gradient comparable to liquid flow
184
and follows Darcys law (Iversen et al. 2001, Stoltz

et al. 2010a):
with p1 and p2 as the relative pressures (compared with

atmospheric pressure Patm ) measured at the entry and
the exit of the sample respectively, H is the sample
height and N2 is the dynamic viscosity of nitrogen.
For this measurement, the pressures, p1 at entry and
p2 at exit, are registered with steady volume flow rate
qv . Several pairs of readings [qv ; p2 p1 ] are taken
keeping in view that the maximum volumetric flow
does not exceed 5 l/minute (limit of the mass flowmeter used). The pressure at entry is limited at 2 kPa. With
this low entry pressure, it is assumed that no movement
of liquid is resulting of gas flow, even in the sample
with high gravimetric Liquid Content. These readings
[qv ; p2 p1 ] showed a linear relationship between qv
and p2 p1 , which justifies the use of Darcys law.
In order to justify the laminar conditions and taking
into account that laminar conditions were checked for
hydraulic constant head test, the ratio of the hydraulic
Reynolds number ReL and the gas Reynolds number
ReG has been calculated. For this calculation, same
experimental values of 1 1010 m2 for the hydraulic
and for the gas permeability have been considered.
In this case, ReL /ReG = 0.5, which confirm laminar
conditions for hydraulic constant head test.
Procedure of the gas permeability tests
Samples with different initial gravimetric Liquid Content, presented before for the compressibility tests,
were subjected to gas flow in the oedopermeameter
at the end of every loading step. Initial physical state
of the porous medium (d0 , w0 and n0 ) was characterised according to the procedure described in section
2.2. Initial moisture content is given in Table 2. The
loading sequence under a selected vertical stress
varies depending of the test. Each compression step
was maintained for 24 hours in order to obtain stabilisation of liquid drainage out. At each compression
step:
Settlement measurement H is registered,
Leachate eventually drained out is weighed (MLout ),
Gas permeability measurement is taken.
Change in physical parameters for each compression stress is then calculated from the settlement
measurement H and possibly from mass of leachate
drained out MLout according to the Eqs. (9), (10)
and (11).
Determination of the gas permeability for the waste
material
Every test corresponds to a gravimetric Liquid Content
w0 constant all along the testing time (except in case
of liquid drainage out). A reduction in porosity causes
a reduction in gas permeability. For a given porosity, an increase in initial gravimetric Liquid Content
Figure 5. Vertical gas permeability kG against volumetric

Gas Content G .
induces a reduction in gas permeability (Stoltz et al.,

2010a). In the case of a sample heavily compressed and
presenting very high gravimetric Liquid Content, gas
permeability can be lower than 1015 m2 and Falling
head test is required.
The volumetric Gas Content is the most representative physical parameter since all the pores filled
with gas (except for possible unconnected pores) are
mobilised during gas flow in a not-saturated porous
medium. Volumetric Gas Content G can be calculated
from the values of d , w and n specific to every test
from Eqs (9), (10) and (11). Gas permeability is plotted
against volumetric Gas Content in Fig. 5.
Preliminary examination of the Fig. 5 reveals that
graphs for different gravimetric Liquid Content w0
tend to merge together with the exception of the one
plotted for the drier sample (T1, w0 = 0.26). The merging of graph lines, except for that of test T1, indicates
that a close value of gas permeability is achieved by
different ways (reduction in porosity by compression
or by increase in moisture), almost independently of
the value of gravimetric Liquid Content. Therefore
there is a poor sensitivity of the gas permeability to the
distribution of pores concerned by the gas flow on condition that the volumetric Gas Content is of the same
order. This is a basic result which has been already
confirmed on a second type of waste.
Gas permeability-volumetric gas content
relationship
This section focuses on the relationship between gas
permeability and volumetric Gas Content. For this
interpretation, gas permeability values of test T1 were
not considered, due to the specific preparation of this
very dry sample. In the case of an unsaturated medium
subjected to a gas flow, the gas uses the available pores.
The medium can be split into two parts: one part made
up of the solid and the liquid phase (assuming that
no movement of liquid is induced by gas flow), and
the other corresponding to the gas that is flowing.
The assumption that water is static like solid depends
185
Figure 6. Modeling gas permeability kG against volumetric

Gas Content G .
on several factors, for example, gas pressure gradient,

capillary pressure, etc. Nevertheless, this assumption
is considered within the framework of this study, and
was confirmed at the scale of the laboratory sample.
All the gas permeability values (T2 to T10) were
analysed with the Kozeny-Carman model for gases:
where u (m/s) is the filtration velocity, Kg (m2 ) is

the gas permeability tensor, g (2 105 kg/m/s) is
the gas dynamic viscosity, Pg (Pa) is the gas pressure, g (kg/m3 ) is the gas density, g (9.81 m/s2 ) is
the gravity acceleration, ez is the vertical unit vector,
g (m3 /m3 ) is the open gas porosity, Gp (kg/m3 /s) is
the volumetric gas production rate, rIG (276.85 J/kg/K)
is the gas specific constant and T (302 K) is the
temperature.
The equation system (eq. 202122) may be rewritten (eq. 23) considering the following assumptions are
considered:
The diphasic effects are neglected. In other words,
liquid flow is not considered and the presence of
water is taken into account using constant gravimetric moisture content. Besides, biogas overpressures
do not create gas porosity modification.
The landfill is considered isotherm.
The biogas/air mixture is considered an ideal gas,
the aerobic reaction conserves the gas molar quantity and methanogenesis inhibition due to oxygen
is neglected.
Regression applied to all the data obtained (Fig. 6)

gives CG = 3.3 1010 m2 . The coefficient of correlation obtained is R2 = 0.77. Poor agreement is observed
in the range of low G , i.e., high liquid saturation.
Consequently an empirical relationship was used to
express gas permeability kG as a function of volumetric
Gas Content G . It takes the form of:
where C3 [m2 ] and C4 are constants.

Regression of the data (Fig. 6) resulted in C3 = 3
108 m2 and C4 = 6.57. The coefficient of correlation
obtained is R2 = 0.88. The exponential relationship
happens to satisfactorily model the whole of gas
permeability experimental data.
3
MODELING GAS FLOW AT THE FIELD

SCALE
The experimental study put forward a strong effect of

the compaction state on the gas permeability (up to 5
decade of variation). In this section, a realistic configuration of gas flow is studied in order to evidence the
impact of this coupling between mechanics and gas
transfers on a landfill cell operation.
3.2 Landfill cell modeling

The numerical calculation uses an axisymmetric
domain centered on a biogas collection well. The cylindrical domain has a radius of 30 m and a depth of 30 m.
Nux flux boundary conditionsare used at the bottom
of the cell to simulate an engineered barrier and on
the vertical boundary away from the well for symmetry. The top boundary simulates a semi-permeable
cover of permeability similar to landfill covers. This
condition is modelled by a Robin (mixt) boundary condition that derives the biogas flux using an equivalent
conductance ke (eq. 24).
3.1 Mathematical model for gas flow

In this numerical model, biogas flow is simulated using
a generalized Darcys law (eq. 20) with a modified
mass balance equation including a biogas production
source term (eq. 21) and applying the ideal gas law
(eq. 22).
where Pa (Pa) is the atmospheric pressure at the cover

surface, ke = 5 108 m/s/Pa is the equivalent conductance for a composite cover, constituted of a bentonitic
geotextile and soil.
The collection well of radius r0 = 0.1 m is surrounded by a drainage layer of a determined thickness
(considered to be 0.25 m). The well radius also corresponds to the distance between the calculation domain
and its axis of symmetry. The well boundary condition
186
takes into account skin effect due to well perforation and possible head losses using a one-dimensional
model with steady-state mass and momentum balance
is developed (Tinet & Oxarango 2010).
MSW material properties evaluated from laboratory experiment (section 2) are used to estimate the
effect of settlement on the gas transfer properties.
The oedometric settlement model (Eq. 16) is applied
to obtain a vertical distribution of the material dry
density (Eq. 9) and porosity (Eq. 10). The MSW
initial dry density value corresponds to a low preconsolidation: d = 400 kg/m3 and MSW initial total
porosity n0 = 0.7. The gravimetric moisture content
is set arbitrarily constant to a value of 0.6 preventing full saturation at the bottom of the cell. The gas
permeability distribution is calculated from Eq. 19.
Gas production is described using a simple model:
the mass of biogas generated per time unit and per
dry MSW mass is considered constant. However, the
MSW density distribution creates a heterogeneity of
the volumetric biogas production rate. The amount
of biogas per volume unit thus increase significantly
with depth. Biogas specific production rate may vary
between 10 m3 /tDM /yr which corresponds to the production rate of a traditional landfill (Vigneault et al.
2004) and 70 m3 /tDM /yr that may occur in a bioreactor landfill. The gas generation rate is arbitrarily
set at 50 m3 /tDM /yr corresponding to a very efficient
bioreactor landfill cell.
In order to evaluate the effect of mechanical settlement, two models are compared using the same
landfill configuration and operating parameters. The
compressible heterogeneous model is compared to a
homogeneous equivalent model. This model considers constant porosity, gas generation rate and permeability such as the total production of biogas is the same
than the one obtained in the compressible model.
This corresponds to a dry density of 619 kg/m2 which
leads to a permeability of 2.2 1011 m2 and a gas
open porosity of 0.33.
3.3 Numerical calculation
The problem (Eq. 23) is solved in the configuration
presented in section 3.2 using a finite volume software developed at LTHE. In this study, a steady state
condition is considered.
3.4 Results and discussion
The pressure and velocity field distribution obtained
with each model is shown Fig. 7a and 7b.
The velocity field vectors were normalised but
their value is shown using a colour scale to consider the large variations of velocity. An increase
in gas flux towards the well can be observed. This
trend is enhanced by the axi-symmetry. In the compressible case the gas flux is much more important
at the top of the well unlike in the homogeneous
case where the influx in the well has little variation
with the depth. It can be observed that the important
Figure 7. Pressure and velocity fields model comparison.
compaction of the lowest part of the landfill generates a significant decrease in permeability. Moreover,
due to compaction, porosity decreases (eq. 10) and
gas generation increases, enhancing the decrease in
permeability. Consequently, gas overpressure at the
bottom of the landfill is much more important in the
compressible case as is the pressure gradient. It
enhances the development of a mainly vertical biogas
flow unlike in the homogeneous equivalent model.
The permeability ratio between the cap cover and
the MSW disposed at the top of the landfill tends
to promote horizontal gas flow under the cover. This
behaviour limits biogas leakage compared to the
homogeneous equivalent model. However, the air
entries could be increased and consequently lessen the
quality of the recovered gas.
The influence radius, as defined previously, is of
28 m for the homogeneous case compared to 61.5 m
for the compressible case. Fig. 8 shows the vertical
velocity at the top of the domain for both cases.
187
the fully coupled problem where an explicit model

of secondary settlement is used to describe the volume loss associated with the biodegradation of organic
matter as suggested by Conte et al., 2013.
REFERENCES
Figure 8. Vertical velocity at the top of the domain.
A negative value demonstrates air entry whereas

a positive value shows biogas leaching through the
cover. It can be seen that the air entry is more important close to the well in the homogeneous case but
rapidly leaching occurs. A zero velocity occurs at a
radius of 4 m and 18 m in the homogeneous case and
compressible case respectively. Because air entry
occur in a larger surface in the compressible case
the air entry in the compressible case with a radius
equal to its influence radius is over 3 times higher than
in the homogeneous case. It can also be noted that
the maximum leaking velocity in sensibly lower in the
compressible case as can be seen in Fig. 8.
4
CONCLUSION
This study put forward the high compressibility of

MSW and the resulting effect on gas permeability using an adapted oedopermeameter apparatus. At
high moisture content, the gas permeability could be
reduced up to 5 decade at depth about 30 m compared to the material under the cap cover. A steady
state numerical simulation is then used to evidence
the potential effect of this vertical heterogeneity of
permeability on the gas flow around a collection well.
The general structure of the flow distribution is drastically modified compared to a homogeneous material.
A drain effect occurs in the first meters under the
cap cover favoring ascending vertical flow at depth.
The next step of this study will consist in considering
Bouazza, A. and Vangpaisal, T. (2003) An apparatus to

measure gas permeability of geosynthetic clay liners.
Geotext. Geomemb. 21, 85101.
Conte, M., Gourc, J.P., Carrubba, P. and Oxarango, L. (2013)
Correlation between compressibility and degradation for
a PTMB waste. Proceedings of TC215 symposium CPEG,
Torino, Sept. 2013.
Gourc, J.P., Staub, M. and Conte, M. (2010) Decoupling
MSW settlement into mechanical and biomechanical
processes-Modelling and validation on large scale setups,Waste Management, 30, 89, pp. 15561568.
Iversen, B.V., Schjonning, P., Poulsen, T.G. and Moldrup, P.
(2001) In situ on-site and laboratory measurements of soil
air permeability: boundary conditions and measurementy
scale. Soil Science, 166, 97106.
Machado, S. L., Vilar, O. M. and Carvalho, M. F. (2008)
Constitutive model for long term municipal solid waste
mechanical behavior. Comput. Geotech. 35, 775790.
Olivier, F. and Gourc, J.-P. (2007) Hydro-mechanical behavior
of municipal solid waste subject to leachate recirculation in a large-scale compression reactor cell. Waste
Management. 27, 4458.
Powrie, W., Beaven, R.P. (1999) Hydraulic properties of
household waste andimplications for landfills. In: Proceeding of the Institution of Civil Engineering. Geotechnical Engineering Journal 137 (4), 235247.
Stoltz, G., Gourc, J.P. and Oxarango, L. (2010) Liquid and
gas permeability of unsaturated MSW under compression.
Journal of Contaminant Hydrology, 118, 12, pp. 2742.
Stoltz, G., Gourc, J.P. and Oxarango, L. (2010) Characterization of the physico-mechanical parameters of MSW, Waste
Management Journal, 30, 89, pp. 14391449.
Terzaghi, K., Peck, R.B., (1976). Soil Mechanics in Engineering Practice, second ed. John Wiley and Sons, New York
(p. 729).
Tinet, A-J. and Oxarango, L. (2010) Stationary gas flow to a
vertical extraction well in MSW considering the effect of
mechanical settlement on hydraulic properties. Chemical
Engineering Science, 65, (23), pp. 62296237.
Vigneault, H., Lefebvre, R. and Nastev, M. (2004) Numerical
simulation of the radius of influence for landfill gas wells.
Vadose Zone Journal, Vol. 3 No. 3, pp. 909916.
188

Mass loss and volume change: From sand-salt analogues to MSW

J.R. McDougall
Edinburgh Napier University, United Kingdom
I.R. Fleming
University of Saskatchewan, Saskatoon, Canada
R. Thiel
Thiel Engineering, Oregon House, USA
P. Dewaele
Golder Associates, Barrie, Canada
D. Parker
University of Saskatchewan, Saskatoon, Canada
D. Kelly
Edinburgh Napier University, United Kingdom
ABSTRACT: Results from two sets of laboratory test programmes on sand-salt mixtures in which the salt
fractions, of different amounts and size, are dissolved are brought together to gather insight into the mechanics
of mass loss and its controlling factors. The test results are then used to provide a framework in which to interpret
the long term behavior of two landfill reclaimed soils. There is some settlement and in all cases an increase in
void ratio. It is shown that settlement and void ratio increases are strongly influenced by the amount of salt but
almost insensitive to the particle size, at least over the ranges and size distributions tested. Both settlement and
void ratio increases are muted by wider particle size distributions.
INTRODUCTION
From initial design of a landfill site through to passive aftercare, coupled hydraulic, biodegradation and
mechanical phenomena give rise to environmental
hazards in the form of liquid and gas emissions. More
fundamental approaches to the analysis and design of
new landfills and the remediation of old landfills have
struggled with the interaction of these complex phenomena. One of the more challenging couplings, both
in its conceptualisation and in the practicality of its
measurement is the impact of biodegradation (or mass
loss) on the mechanical properties of the waste.A number of landfill models (e.g. HBM, LDAT, Moduelo,
see McDougall (2011) for a summary) account for the
impact of biodegradation on volume change. These
models either (i) use a secondary compression coefficient dimensioned according to the degradable content
of the waste or (ii) account directly for mass loss. In
the former, time is the controlling variable, sometimes
in the guise of a gas production model; in the latter,
a more fundamental account of mass loss is used but
some means of coupling mass loss to volume change
is then required.
The impact of mass loss on volume change is complex, although the amount and relative size of material
to be lost would seem to be significant controlling
factors. With these in mind, a programme of testing on
sand-salt mixtures both in the oedometer (McDougall
et al., 2013) and in the triaxial machine has been undertaken. A parallel investigation has been done at the
University of Saskatchewan.The replicability of MSW
by a sand-salt mixture may be difficult to justify in the
case of a typical MSW, i.e. highly compressible matrix
with wide particle size distribution, some of which
are elongated tensile elements, but it does provide a
framework within which to approach the problem.
Recent projects to reclaim soil from old landfills,
which offer impressive environmental benefits, process degraded MSW to produce a residual material
that is more closely represented by the sand-salt mixtures. The process uses a screening plant with a final
screen size of about 50 mm, depending on the waste
and the final goals (see Figure 1). If the old waste contains a substantial amount of concrete, especially for
construction and demolition debris dumps, a crushing
plant may be used in conjunction with the screening
plant to reduce the particle size of the concrete and
189
Figure 1. Photographs showing the landfill soil reclamation process: material post screening (left) and screening plant (right).
brick material so that it may also be re-used for the

controlled backfill. As with any type of recycling process, a small amount of contamination is intrinsic
to the soil reclamation effort. We define contamination, in this regard, as the carryover of undesirable
waste components, such as wood, paper, plastic, metals, glass, etc., into the reclaimed soil. The amount of
contamination depends on the original waste composition and the screening process used. However, the
small amount of contamination poses the question of
the acceptability of the reclaimed soil for purposes of
backfill to create a new geotechnical foundation for
structures.
While the inclusion of some glass, metal, and other
generally non-degradable materials is probably not of
concern from a geotechnical point of view (all of these
could legitimately be considered a soil material by
standard geotechnical definitions), it is the residual
organic fraction that may be of concern, principally
because of the potential for long-term settlement.
Considering that residential and commercial developments may have septic leach fields, water lines,
sewer lines, stormwater drainage, or other environmental features that may create a net liquid percolation
into the underlying soil, it would have to be assumed
that any degradable organic fraction in the soil mass
would eventually degrade. The question is what will be
the impact of the loss of an amount of organic particles on the stability and settlement of foundations built
over this soil?
Field studies show that the screened reclaimed soil
is very homogeneous, and the distribution of contaminants is likewise so. Laboratory testing, presented
in this paper, has shown for one case study that as
long as there is good compaction control, and the
reclaimed soil is compacted to a high level, potential
long-term settlement can be acceptable for foundations
of structures.
This paper presents the results of testing of sandsalt mixtures with a range of salt particle amounts
and sizes. A total of 158 samples were tested: 118 of
which were performed at University of Saskatchewan,
and the remaining 40 at Edinburgh Napier University. These results are brought together in order to
both reinforce the fundamental behaviours controlling
volume change due to particle loss and to provide an
a posteriori benchmark for the long term performance

of reclaimed landfill soils.
2
PREVIOUS STUDIES ON DISSOLVING

SAND-SALT MIXTURES
Studies by Fam et al. (2002), Shin and Santamarina

(2009) and Truong et al. (2010) show that dissolution
in coarse-grained soils leads to a change in soil fabric, i.e. structure. These studies focused primarily on
the amount of particle loss in mixes with a relatively
narrow range of sand to salt particle size diameter ratios (D50 sand /D50 salt between 1.44 and 2.34).
Consequently, there is little experimental data and
hence understanding of the influence of particle size on
fabric changes due to particle loss. This paper presents
the findings of experimental investigations undertaken
at Edinburgh Napier University and University of
Saskatchewan into the volumetric consequences of
both the amount and size of particles lost and host
particle size distributions and presents those results in
the context of the landfill reclaimed soils.
3
EXPERIMENTS EDINBURGH NAPIER

UNIVERSITY
Salt particle sizes ranging from 0.063 mm to 1 mm,

added to sand with a D50 of 0.85 mm, in amounts
from 2% to 15%, were tested. In such an investigation, the initial packing of the mixes assumes greater
importance.
The inert fraction is a coarse rounded to subrounded quartz sand from Leighton Buzzard in the
UK (D50 = 0.85 mm; CU = 1.4; GS = 2.65). The salt is
sodium chloride (GS = 2.165), chosen for its high solubility in water, which after crushing to the required
dimension, has a more angular form.
Particle size distributions are defined by a sieve
stack having a sequence of aperture dimensions that
double in size, i.e. 0.063, 0.125, 0.25, 0.5, 1.0 and
2.0 mm. Hence, in this paper, 0.063 mm particle fraction refers to particles between 0.0630.125 mm. Figure 2 shows the relative sizes of the sand and salt
particles for two of the gradings used: 1 mm salt
190
Figure 2. Photographs of sand-salt (Leighton Buzzard) mixes showing relative sizes and shapes: (left) 1.0 mm salt particles;
(right) 0.063 mm salt particles
Figure 3. Schematic diagram of modified oedometer allowing for circulation of pore fluid through sample and large (4 litre)
external reservoir (not shown to scale).
particles in Figure 2(a) and 0.063 mm particles in

Figure 2(b).
3.1 Equipment
The dissolution tests were performed in an oedometer that was modified to allow for (gravity-driven)
permeation of water, see Figure 3. The oedometer is
100 mm in diameter. It has an extended confining ring
to accommodate a sample height of up to 30 mm and
to retain a reservoir of water above the loading cap
and sample. The Perspex loading cap has a number
of small (1 mm) holes to improve the flow of water
from the overlying reservoir to the sample. Permeation
avoids the accumulation of ion-saturated solution in
the pores and encourages a more even distribution of
particle dissolution within the sample. Complete dissolution is usually obtained in less than an hour, which
is confirmed by electrical conductivity measurements
and final sample masses. Pore fluid exits through two
ports in the oedometer base into an external 4 litre
reservoir, which is continuously stirred. A peristaltic
pump transfers solution from the external reservoir
back to the oedometer. Both reservoirs are open to
atmosphere so pump flow rate is used to control the
level in the oedometer reservoir.
3.2 Test programme

The programme comprised 40 dissolution tests in 2
groups of 20. In each group of 20 tests, 5 different single salt particle size fractions (0.063, 0.125, 0.25, 0.50
and 1.00 mm) were added to Leighton Buzzard sand in
4 different proportions (2%, 5%, 10% and 15%). These
percentages are calculated as a fraction of the total dry
mass. In one group, dissolution occurred under a vertical stress of 62 kPa; in the other group, vertical stress
was 250 kPa.
EXPERIMENTS UNIVERSITY OF
SASKATCHEWAN
A sealed cylindrical load cell is used which allows

liquid flow from the bottom of the cell. Its diameter
is 159 mm and the sample is loaded to a height of
approximately 82 mm. Permeable filter paper on the
bottom and a porous stone at the top allow the flow of
water and brine in/out of the sample without the loss of
solid material. A known load is applied to the sample
by a calibrated pneumatic consolidation system (PCS)
that loads the sealed piston on top of the sample (Figure 4). Displacement is measured by a linear variable
191
dissolution process is allowed to run until settlement

has ceased and effluent conductivity measurements
show either (i) zero in the case of the Saskatchewan
tests or (ii) have stabilized at some non-zero value in
the case of the Edinburgh Napier tests.
Final, i.e. post-dissolution masses, of both sand and
salt, are checked by drying and weighing of the remaining sand samples. In the Saskatchewan tests, dissolved
salt masses are estimated from batch conductivity
measurements and corresponding salt solubility relations (NIST, 2007). In the Edinburgh Napier tests, the
salt solutions were evaporated to enable solute masses
to be checked.
CHARACTERISING THE TWO TEST

PROGRAMMES
Key characteristics of the two test programmes are

described below and summarised jointly in Table 1.
Figure 4. Sample cell seated in pneumatic consolidation
cell.
differential transformer (LVDT) interfacing with computer software. A pump circulates water to an inlet on
the piston or on the bottom of the cell.
4.1
Method
Each test begins by choosing, weighing, and mixing the appropriate sand-salt combination (e.g. 90%
SP-10% HM) and then loading the dry mixture into
the load cell in 34 layers, tamping each layer as it is
added. The cell is positioned in the PCS and subjected
to a 5 kPa seating load. Once settled, LVDTs are positioned and the height of the piston above or below the
top rim of the cell is measured manually. Data acquisition begins and, after a few minutes of sampling to
acquire a base line, the load is increased to 60 kPa.
The sample is then left for 300 minutes during which
time settlement is measured. Hoses are attached and
the sample is inundated with water entering from the
bottom of the cell. Filling is from the bottom to help
purge air from the sample.
Water is initially pumped intermittently at a rate
of 30 mL/min in intervals of 3 minutes pumping
with 3 minutes rest until the cell was saturated. Once
saturated, the hoses are reversed so that fresh water
enters from the top of the cell. Pumping continues in
3-minute intervals. Each pumping cycle produces a
mass of solution, which is collected and the mass of
solute determined based on the volume of effluent
collected and the known relationship between total
dissolved solids and electrical conductivity.
This control of the dissolution process used in
Saskatchewan differs from that in the Edinburgh
Napier tests, in which a single (much larger) volume
of water is continually circulated between the sample
cell and an external reservoir into which the solute
gradually accumulates. However, in both cases, the
5.1 Particle size: relative sand and salt particle

sizes diameter ratio
The relative size of the sand and salt particles is
shown to be one of the main factors controlling the
mechanical response to particle loss. It is here defined
by the diameter ratio, Dr, which is the ratio of D50
of sand particle to D50 salt particles. The two test
programmes cover diameter ratios ranging between
9 and 0.25. The larger diameter ratio corresponds
to a Leighton Buzzard sand (D50 = 0.85 mm) and
salt particles retained on a 0.063 mm sieve (median
D50 for the 0.0630.125 mm sieves = 0.094 mm). The
smaller diameter ratio corresponds to a poorly-graded
Ottawa sand (D50 = 2.36 mm) and coarse salt particles (D50 = 9.42 mm). Both Leighton Buzzard and the
Ottawa sand are poorly-graded with CU values of 1.4
and 1.18 respectively. Table 1 shows the combinations
of diameter ratio and amount of salt by percentage of
total mass.
There are other influences on the mechanical behavior that are not explicitly recognized in Table 1.
These are:
5.1.1 Vertical load
The Saskatchewan tests were performed under a vertical stress of 60 kPa. The Edinburgh Napier tests were
duplicates at two different vertical stresses: 62.5 kPa
and 250 kPa. In these tests no distinguishable difference between the test results, either in terms of initial
void ratio, settlement, or change in void ratio due to
dissolution, was found.
5.1.2 Particle size distribution
A second suite of tests was performed at the University of Saskatchewan using a well-graded Ottowa sand
fraction. Its D50 was also 2.36 mm but its Cu was 6.98.
Table 2 summarises the percentage and size characteristics. There is then a combined total of 158 tests, 95 of
192
Table 1. Summary of number of tests performed at Edinburgh Napier (ENU) and University of Saskatchewan (UoS) by
diameter ratio and amount of salt in each of tests poorly-graded sand fraction.
Diameter ratio = D50 sand /D50 salt
Amount
Percentage
salt
(by mass)
0
2
5
10
13
15
21
6.6
4.5
2.3
1.1
0.6
0.25
2 ENU
2 ENU
2 ENU
1 UoS
2 UoS
2 UoS
2 Uos
8 UoS
2 ENU
2 ENU
2 ENU
2 ENU
2 ENU
2 ENU
2 ENU
2 ENU
2 ENU
2 UoS
2 UoS
4 UoS
10 UoS
2 ENU
2 ENU
2 ENU
2 UoS
2 UoS
2 UoS
4 UoS
2 ENU
2 ENU
2 ENU
2 ENU
2 ENU
4 Uos
4 UoS
4 UoS
Table 2. Summary of number of tests performed at University of Saskatchewan (UoS) by diameter ratio and amount of salt
in each of tests well-graded sand fraction.
Diameter ratio = D50 sand /D50 salt
Amount
Percentage
salt
(by mass)
0
2
5
10
13
15
21
6.6
4.5
2.0
1 UoS
2 UoS
2 UoS
2 UoS
8 UoS
1 UoS
1 UoS
1.1
1
1 UoS
2 UoS
2 UoS
2 UoS
12 UoS
1 UoS
4 UoS
0.6
1 UoS
1 UoS
0.21
1 UoS
2 UoS
2 UoS
3 UoS
7 UoS
1 UoS
4 UoS
which were performed using poorly-graded sand and

63 tests using a well-graded sand.
6
RESULTS SETTLEMENT
Previous studies have reported relatively little settlement due to particle loss by dissolution. Fam et al.
(2002) conducted tests on salt-sand particle mixtures
with diameter ratios of 2.30 noted virtually no settlement whilst Shin and Santamarina (2009) measured
settlement between 1.2% and 8.2% for their range of
salt contents (5%15%) with diameter ratios of 2.33.
Truong et al. (2010) testing mixes with diameter ratios
of 1.44 showed settlements of 2% and less for mixes
containing up to 10% of salt particles.
6.1 Poorly-graded soils
Vertical settlements induced by dissolution for each of
the poorly-graded sand-salt mixes are shown in Figure 5. Also shown are (i) the settlement at constant void
ratio line, i.e. the settlement that would occur if solid
volume loss and corresponding void volume change
maintain a constant void ratio, and (ii) the Shin and
Santamarina data presented as mass fractions.
The settlement data show two principal features:
the first relates to the influence of the amount of
soluble particles, the second to the size of the soluble particles. Clearly settlement is directly related
to the amount of soluble material but not uniformly
Figure 5. Settlement of sand-salt (poorly-graded sand)

mixtures by particle size and percentage (by mass) of salt.
related. Compared to low salt contents, the rate of settlement is greater where the amount of soluble material
exceeds 10% or 15%. The second principal feature
is less well defined but discernable. Small particles,
i.e. particles with diameter ratios of 2 or greater, at
percentage amounts of 5% or greater, tend to occupy
the upper part of the settlement bandwidth. Larger particles, i.e. diameter ratios of 1 or less, occupy the lower
part of the bandwidth. Hence, it might be deduced that
small particles, which can nestle within the inter-sand
void spaces, are dissolved with little corresponding
settlement. For example, less than 2% vertical strain
was observed during dissolution in samples containing
10% of 0.063 mm salt particles.
193
Figure 6. Settlement of sand-salt (well-graded sand) mixtures by particle size and percentage (by mass) of salt.
The Shin and Santamarina (2009) data, obtained

from glass bead-salt mixtures with a diameter ratio
of 2.33, all settle more than the poorly-graded sandsalt mixtures shown here. This may be due to the
lower frictional resistance and shape of the glass beads
(Proctor & Barton, 1974) facilitating particle rearrangement. It is well recognised from experimental
and numerical DEM studies that both inter-particle
friction and particle shape have a significant effect
on the behaviour of granular materials (e.g. Cavaretta
et al., 2010).All settlements are, however, significantly
less than the settlement occurring under a constant
void ratio condition. Hence particle dissolution, for
the amounts and sizes of the soluble particles in the
poorly-graded sands, leads to an increase in void ratio.
Figure 7. Initial void ratio for both poorly-graded and

well-graded sand salt mixtures.
6.2 Well-graded soils

Consider now the settlement data for the well-graded
soils shown in Figure 6. All poorly-graded sand-salt
mix data shown in Figure 5 has been removed although
the bandwidth for these data is shown by the broken
lines. The bandwidth for the well-graded sand data is
shown by the solid lines. The muted response to small
amounts of salt particle loss is still evident. There is
also a change in the rate of settlement with amount
of particle additions but the change appears to occur
at a lower percentage of salt particles, i.e. about 5%
to 10%. Furthermore, the bandwidth in this part of
the settlement curve appears to be narrower and the
distribution of particle sizes lack the separation of
the poorly-graded samples. As with the poorly-graded
sand, the position of the experimental data in relation
to the constant void ratio line indicates, at least at percentage additions up to 20%, that particle loss leads to
an increase in void ratio.
7
7.1
RESULTS VOID RATIOS, PRE- AND

POST-DISSOLUTION
Figure 8. Change in void ratio due to dissolution in

(poorly-graded) sand-salt mixtures by particle size and percentage (by mass) of salt.
added salt particles is revealed. In the case of poorlygraded sands, increasing amounts of fine salt particles
lead to gradually reducing initial void ratios as the fine
particles fill up the voids surrounding the coarser sand
particles. The addition of coarser salt grains serves
only to displace sand particles with a range of initial
void ratios that is relatively insensitive to the amount of
added salt. In the case of the well-graded sand samples
there is little opportunity for nestling so the addition
of salt particles, either fine or coarse has little impact
on the initial void ratios, all of which are significantly
lower than the poorly-graded sand with coarse particle
mixes.
7.2 Poorly-graded salt samples
Initial void ratio
The initial or pre-dissolution sand-salt void ratios are

shown in Figure 7, from which the influence of the
Changes in void ratio due to dissolution in the poorlygraded samples with diameter ratio for all percentage
salt amounts are shown in Figure 8. The most striking
feature revealed is the difference in the void ratio
response to these two controlling factors. There is an
194
Figure 10. Average change in void ratio due to dissolution

in sand-salt mixtures by particle size and percentage (mass)
of salt.
Figure 9. Change in void ratio due to dissolution in

(well-graded) sand-salt mixtures by particle size and percentage (by mass) of salt.
increase in void ratio that is almost independent of

diameter ratio but clearly related to the amount of
added salt. Moreover, there appears to be no sensitivity to the wide range of initial void ratios evident in
Figure 7.
7.3 Well-graded salt samples

Compare now the changes in void ratio due to dissolution observed in the well-graded sands (Figure 9).
Where added salt amounts are in the range 2% to 5%,
regardless of diameter ratio, void ratio increases are
similar to those observed in the poorly-graded sand
mixes. However at 10% added salt and above, the
increase in void ratio is markedly less than is the case
in the poorly-graded sands. At 15% and 21% the insensitivity to diameter ratio appears to be lost, although in
the case of these tests the data are few and are obtained
for large salt particles only (Diameter ratio = 0.251.0
to 2.0).
7.4 Void ratio changes with amount of salt

Figure 10 shows increases in void ratio for the combined Edinburgh Napier/Saskatchewan data for both
the poorly- and well-graded sample tests, where the
void ratio change is read from the midpoint of the
fitted lines in Figures 8 & 9. In the poorly-graded samples, the relationship is almost linear over the range in
question, whereas in the well-graded samples, the rate
of increase is indeed seen to lessen beyond 5% salt
content.
8
RECLAIMED LANDFILL SOIL SAMPLES
Samples of fines produced from excavation and

screening at two different disposal sites, one in Canada
and one in the United States, were tested at the University of Saskatchewan. The samples were characterized
in terms of waste composition, potential for degradation and settlement, the main characteristics and
findings are presented in Table 3. In the remaining
part of this paper, the behaviour of these two samples
is considered within the context of the aforegoing sand
salt tests. However, it should be noted that at the time
of writing, the test on the sample from site 1 is still
in progress and so the insights drawn are necessarily
tentative.
From the data presented in Table 3, it can be seen
that the two soils represent two quite different materials: the soil from site 1 has the narrower particle
size range (CU = 10), high degradable content (nearly
20%) and a diameter ratio of degradable to inert material that is less than one. In contrast, Site 2 soils have a
much wider particle size range (CU = 200), less than
half the degradable content of Site 1 (based on LOI
data) and a diameter ratio that is similarly difficult to
estimate but probably much greater than one. In other
words, Site 1 soils have a degradable fraction of predominantly large particles, whereas at Site 2 they are
predominantly small particles.
Figure 11 shows the observed settlement of the
Site 1 and 2 soils superimposed on the settlement data
from the well-graded soil samples (Fig. 6).
While definitive conclusions cannot yet be reached
regarding the long term performance of the postscreening fines from the landfill mining project at Site
1 (at the time of writing, this test is still in progress),
measured gas production compared with gas potential indicates that the substantial organic content of
this material will be associated with greater long term
degradation-induced settlement. Site 1 settlements are
as shown in Figure 11 and predicted to move as shown
by the accompanying arrow. Settlement of the Site 2
test is located where it might be expected lying in
the upper part of the settlement bandwidth for a 4
5% settlement with large diameter ratio. This material
exhibited little measurable deformation.
195
Table 3. Comparison of key characteristics and measured behavior of two landfill

reclaimed soil samples.
Property
Site 1
Site 2
D50
Cu
Inert fraction (by dry mass)
Degradable fraction (by dry mass)
Loss on ignition
Size range of degradable fraction
Diameter ratio*
Biochemical methane potential
Initial compression
Compression (after 65 days)
Biogas production (L/kg dry mass)
Compression (after 244 days)
Long term settlement
2.0 mm
10
81.1%
18.9%
8.2%12.4%
90% >0.85 mm
0.20.8
4.79.3 ml/g
1.0 mm
2.8 mm
1.6
6.7 mm
1.1%
1.0 mm
200
95%
5%
4.2%
78% <1.2 mm
0.812
2.6 mm
3.5 mm
0.02
0.16%
*Degradable fraction size range spans several sieves complicating determination of

diameter ratio.
Figure 11. Comparison of current and predicted settlement

of landfill reclaimed soil samples with (well-graded) sand
salt mixtures.
However, it is interesting to speculate on the change

in void ratio accompanying mass loss, and the impact
on compaction characteristics and use of the materials as engineered fill. The two sites fit into the overall
scheme established by the Saskatchewan data and are
jointly presented in Figure 12. Of note is the implication that even though settlement is significantly
different in the two scenarios, the difference in the
change in void ratio is less marked.
While it seems unsurprising that the volume fraction of degradable material represents an important
control on the amount of settlement that may be
expected under load, the mechanisms are less easily
explained. The relative size of the degradable fraction plays a less significant role in determining the
amount of overall volume change, and the particle
size characteristics of the inert fraction (i.e. uniform
vs. well-graded) exerts some control on the degree
of particle redistribution into growing voids and thus
on the relationship between mass loss of degradable
solids and change in void ratio. The interaction of relative particle size and amount of degradable material
for the simpler mono-disperse mixtures is explored in
Figure 12. Comparison of current and predicted void

ratio increases of landfill reclaimed soil samples with
(well-graded) sand salt mixtures.
more detail in McDougall et al. (2013). In addition, the

effects of initial relative compaction of the soil matrix
with the degradable materials will likely influence the
ultimate settlement characteristics.
9
CONCLUSIONS
Given that rigorous, well-documented tests of this type

require many months for each test, rapid tests using
analogues have been undertaken and report here and
elsewhere. Such analogues include sand/salt mixtures
which can be degraded under load in a few hours
or days has the potential to relatively quickly shed
light on the underlying mechanisms and the relative
importance of the various parameters that may control
the amount of volume change. These are hypothesized
to include the following: the fraction of degradable
196
material; the relative size of the degradable and inert

particles; the applied load and the particle size characteristics of the inert soil. In the case of the landfill
reclaimed soils, the degree of initial compaction, relative to minimum and maximum potentials under
standard energy inputs, is also a factor deserving of
future evaluation.
The ultimate goal is to provide a rigorous, evidencebased standard for geotechnical engineers to follow
when using these sorts of non-ideal soil fills. As
green construction practices increasingly become
a societal goal and indeed an economic benefit, the
author suggests that geotechnical engineers may wish
to reconsider the age-old (safe and cautious) practice of
bulldozing away poor-quality soil to be replaced only
with high quality imported granular fill. The research
programme of which some first steps are described
in this paper represents an effort to provide industry
with such evidence and guidance.
Fam MA, Cascante G, Dusseault MB (2002) Large and

small strain properties of sands subjected to local void
increase. J. of Geotechnical and Geoenvironmental Eng.
128(12):10181025
McDougall, J.R. (2011) Settlement: the Long and the Short of
it. Geotechnical Special Publication 209, ASCE, Virginia,
ed. Zekkos, 76111
McDougall J, Barreto D, Kelly D (2013) Particle loss and volume change on dissolution: experimental results and analysis of particle size and amount effects. Acta Geotechnica
(Online First) doi: 10.1007/s11440-013-0212-0
NIST. (2007, 12 6). IUPAC-NIST Solubilty Database.
Retrieved 8 21, 2012, from http://srdata.nist.gov/solubility/
index.aspx
Procter DC, Barton RR (1974) Measurements of the angle of
interparticle friction. Gotechnique 24(4):581604
Shin H, Santamarina JC (2009) Mineral dissolution and the
evolution of k0. J. Geotechnical and Geoenvironmental
Eng. 135(9):11411147
Truong QH, Eom YH, Lee JS (2010) Stiffness characteristics
of soluble mixtures. Gotechnique 60(4):293298
REFERENCES
Cavarreta I, Coop M, OSullivan C (2010) The influence of
particle characteristics on the behaviour of coarse grained
soils. Gotechnique 60(6):413423
197

Modeling the coupled chemo-hydro-mechanical behavior

of compacted active clays
G. Musso
Department of Structural, Geotechnical and Building Engineering, Politecnico di Torino, Torino, Italy
G. Della Vecchia
Department of Civil and Environmental Engineering, Politecnico di Milano, Milano, Italy
E. Romero
Department of Geotechnical Engineering and Geosciences, Universitat Politcnica de Catalunya,
Barcelona, Spain
ABSTRACT: Engineering clay barriers are one of the main solutions considered for the containment of waste,
both within the context of landfills and, in perspective of longer operational times, of nuclear waste disposal.
Modeling the performance of these barriers in the long term can be difficult, since the behavior of active clays
highly depends on variables such as temperature, degree of saturation and suction, and chemistry of the wetting
fluid. This paper deals with the effects of chemical changes. Based on microstructural evidence, a double structure
model for coupled chemo-hydro mechanical processes is formulated. The model is then used to reproduce a salt
transport test run in oedometer conditions. Some implications of the proposed double porosity approach are
highlighted and discussed. The frame has the advantage of being based on direct microstructural evidence and
of allowing for joint interpretation of hydro and mechanical changes. The frame also shares characteristics
with models used to reproduce the behavior of clays in unsaturated conditions, so making it appealing for its
implementation within a wider context.
INTRODUCTION
Safe disposal of waste requires contaminants to be isolated from the surrounding geological layers and, in the
case of capping systems, also from the atmosphere.
Isolation is often provided by clay based engineered
barriers, that must guarantee that only a negligible
amount of contaminants, usually dissolved in water,
leaves the disposal. Some properties of active clays
make them very suitable as construction materials,
since allow satisfying the mentioned requirements.
These properties are the very low hydraulic conductivity and the high adsorption capacity, which contribute
in minimizing and retarding the fluxes of water and of
solutes under pressure and chemical gradients acting
in situ.
Compacted active clays have been extensively studied in the last years in the context of nuclear waste
disposal. Many research efforts have been devoted
to understanding and modeling their thermo-hydromechanical behavior in saturated and unsaturated conditions. From an engineering perspective, there are
two main reasons for this attention. On the one hand,
while initially cast in place in unsaturated conditions,
compacted active clays are exposed to local regimes

of water pressures. Therefore they are expected to
go from high values of initial suction to much lower
ones, and finally to become saturated. Upon wetting, swelling is expected to occur. In this context,
swelling has a positive effect since it allows sealing
joints or imperfections of the engineered construction.
On the other hand, significant temperature changes
are anticipated as a consequence of the decay of the
radioactive waste in the disposal. The activity of the
clay materials is reflected in a marked sensitivity
to thermal and hydraulic changes, claiming then for
careful modeling of these concomitant environmental
actions: a very comprehensive analysis is provided by
Gens (2010).
Further actions on the engineered clay barriers can
be exerted by chemical processes. Although they can
induce very important effects on the performance of
clay engineering barriers for nuclear waste, related literature has been often developing quite independently
from the one on thermal and hydraulic actions.
The present paper aims at providing an insight
on the coupled chemo-hydro-mechanical behavior of
compacted active clays. The main aspects connected
199
to chemical processes and chemically induced deformations in active soils are reported. Microstructural
effects induced by chemical changes in compacted
active clays are documented by means of Environmental Scanning Electrical Microscopy (ESEM) and
of Mercury Intrusion Porosimetry tests (MIP) carried
out at different saline concentrations of the saturating water. The resulting physical evidence is used to
formulate and calibrate a double porosity model, introduced to simulate the results of a salt transport test
run under controlled conditions. As an outcome, it is
found that the double porosity characteristics investigated with the microstructural analysis can contribute
explaining some of the peculiarities shown by the
behavior of the compacted material.
BACKGROUND
Chemo-mechanical interactions and pore-fluid chemistry effects on reconstituted clays are a well documented topic in the literature (to name only a few,
one can refer to Bolt 1956, Mesri & Olson 1971a,
Barbour & Fredlund 1989, Yang & Barbour 1992, Di
Maio 1996, Studds et al. 1998). Despite practical relevance for geo-environmental applications, relatively
few studies dealt with the effects of chemical changes
on the volumetric behaviour of compacted and natural soils (see, for instance, Di Maio & Fenelli 1997,
Musso et al. 2003, Castellanos et al. 2008, Rao et al.
2006, Siddiqua et al. 2011).
Chemically induced hydro-mechanical processes in
non reacting porous media are related to variations in
the short range interaction forces acting between clay
platelets or tactoids. Repulsion forces reduce when
salinity increases and this is macroscopically appreciated as an overall volume reduction of the soil.
High salinity also reduce the amount of swelling that
can be developed upon saturation under constant total
stress.
It is anyway evident that variations in the composition and saline concentration of the liquid phase do
not only induce volume or strength changes, but can
also have an impact on the fabric of the active clay.
This phenomenon can be readily appreciated in terms
of consequences on the permeability of the soil. At a
given void ratio, specimens saturated with saline solutions or with organic non polar liquids normally have
higher hydraulic conductivities than specimens saturated with distilled water (e.g Michaels & Lin 1954,
Mesri & Olson 1971a, 1971b, Fernandez & Quigley
1985, Calvello et al. 2005, Gajo & Maines 2007,
Musso et al. 2008).
Predicting volume strains simply on basis of the
equilibrium of molecular forces, e.g. by introducing
the diffuse double layer theory, is theoretically possible (see e.g. Bolt 1956). Nevertheless this approach is
not very feasible in practice, since it is very difficult to
correctly take into account the actual geometry (fabric,
or structure) of the soil and the number of phenomena
that actually govern the interaction between solid particles and the liquid phase (electrostatic and Van der
Waals forces, cation exchange, surface complexation,
to name a few).
In some works, cluster of particles have been
regarded as the microstructural unit at which chemical
phenomena could conveniently be taken into account
(Hueckel, 1997, Hueckel et al., 1997). Based on this
approach, some constitutive models have been proposed (e.g. Gajo et al., 2002 and Gajo & Lloret, 2007):
the solid phase has been considered as an assembly
of clusters of mineral sheets bathed in an electrolytic
solution (i.e. the fluid phase). Water and exchangeable
ions have been supposed to transfer between the solid
and the fluid phase. These models take into account
pH variations, cation exchange and changes in the
saline concentration of the electrolyte. The mechanical behavior is then modeled through an extension of
the Cam Clay model, into which chemical effects are
explicitly taken into account.
Alternatively, when discussing results of permeability measurements, Mesri and Olson (1971b) turned the
attention onto clay aggregates. As suggested by Yong
(1999) aggregates are made of many clusters of particles. Mesri and Olson (1971b) hypothesized then that
high water salinity and/or low dielectric constant of
the pore fluid cause the contraction of aggregates, consequently decreasing the volume of the pores within
the aggregates (intra-aggregate or micro-voids) and
increasing the volume of voids between the aggregates (inter-aggregate or macro-voids). The increase
of macro-void size could then be the dominant mechanism explaining the increase in permeability with
salinity.
The impact of aggregated fabrics on the hydromechanical behavior of compacted active clays is well
documented in unsaturated conditions. For instance
Lloret et al. (2003) performed an experimental study
on bentonite samples compacted to different dry densities: by investigating both the pore size distribution
and the water retention properties, they evidenced the
advantages of modeling the fabric of the soil through
a double-porosity frame.
Some microstructural experimental evidences of
the effects environmental actions, expressed in terms
of total suction, on the structure of compacted active
clay samples are shown in Figure 1. Pictures refer to
ESEM observations of FEBEX bentonite, an active
clay proposed as an engineered barrier for radioactive
waste disposal (Lloret et al., 2004). Samples were statically compacted at their hygroscopic water content at
an approximate initial dry density d = 1.55 Mg/m3 .
They were then inserted in the ESEM chamber and
different relative humidity values were imposed within
the chamber to check the effects of total suction
changes on the microstructure. In the pictures, aggregates are very evident and separated one from the
other when suction is high. Upon progressive wetting
aggregates swell adsorbing water, reducing the interaggregate space that separates them and eventually
starting to merge.
200
Figure 2. Effects of NaCl and CaCl2 concentration on the

liquid limit of FEBEX bentonite. Liquid limit is expressed as
the mass ratio distilled water/total mass of solids (salt + solid
particles). Data from Castellanos et al. (2006).
EFFECTS OF SALINITY CHANGES ON THE

MICROSTRUCTURE OF A COMPACTED
ACTIVE CLAY
3.1 Basic geotechnical characterisation
Figure 1. ESEM microphotographs of compacted FEBEX

samples exposed to approximate total suction values = 500 MPa (above), = 250 MPa (center) and
= 50 MPa (below).
In the following, specific reference will be made to

experimental characterization, and subsequent modeling, of a compacted active clay (FEBEX bentonite),
proposed as an engineered barrier for radioactive
waste disposal (Lloret et al., 2004). FEBEX is
a natural mixed bentonite (Cortijo de Archidona
deposit, Almera, South-eastern Spain). It has a very
high content of montmorillonite (around 90%), with
small quantities of quartz, plagioclase, cristobalite,
K-feldspars, trydimite and calcite. Its total Cation
Exchange Complex is equal to 102 meq/100 g: 38%
of which related to Calcium ions and 25 % to Sodium
ions. The total specific surface is Ss = 725 m2 /g, the
external specific surface is equal to 65 m2 /g. The average density of the solid phase is s = 2.70 Mg/m3 .
When saturated with distilled water, the material has a
liquid limit wL = 102% and a plastic limit wP = 53%.
A preliminary characterization of the sensitivity of
the material to salinity effects can be done by measuring the liquid limit obtained with different saturating
solutions. In fact empirical relationships exist relating
the strength and compressibility of the soil in reconstituted state to index properties (e.g. Wood, 1991) so as
stated by Gajo & Lloret (2007) the liquid limit can be
thought as a mechanical indicator. Figure 2 provides
the values of the liquid limit measured at increasing
NaCl and CaCl2 concentrations.
As also shown in previous works (e.g. Di Maio,
1996, Calvello et al., 2005) the liquid limit decreases
201
upon salinity increase. At the same molar concentration, CaCl2 appears to have a more pronounced effect
than NaCl.This is mainly a consequence of the osmotic
suction associated to salts with bivalent cations, which
is higher than the osmotic suction associated to monovalent salts; nonetheless also specific aspects related
to cationic exchange processes and size of the hydrated
radius of the cation are known to play a role (see e.g.
Mitchell & Soga, 2005).
3.2 Effects imparted on fabric by salinity changes
microscopic evidences
ESEM photomicrographs were performed to characterise the structure of samples saturated with different
saline solutions, so to provide information about the
effects of chemical changes on the structural arrangement. Samples were prepared statically compacting
the material at its hygroscopic water content (w = 12%
at a dry density d = 1.65 Mg/m3 ), at room conditions (temperature T = 20 C and relative humidity
R.H. = 50%). Samples were loaded in oedometer conditions at constant water content and then saturated
with different saline solutions at a constant vertical
stress of 200 kPa. All samples swelled upon saturation, although to a different extent depending on the
solution used. Details concerning the swelling process
and the influence of salt concentration on amount of
swelling can be found in Castellanos et al. (2008).
ESEM images documented that the original aggregated fabric is preserved by highly concentrated solutions. Through ESEM pictures it was also possible to
observe that, when the concentration of the pore fluid
decreases, samples develop a more uniform fabric,
similarly to what detected in Figure 1c.
Other microstructural observations consisted in
pore size curves data collected performing Mercury
Intrusion Porosimetry on freeze dried samples (Delage
et al. 1982) at different pore fluid concentrations. The
device used was a Micromeritics-Autopore IV which
makes possible the intrusion of pore throats from 6 nm
to 400 m. The cumulative intrusion curves, expressed
in terms of void ratio, are presented in Figure 3.
It can be appreciated that all samples swelled upon
saturation, since the total void radio is always greater
than the as compacted one. Nonetheless, not all the
pores are influenced in the same manner by the saturation process, suggesting that fabric and pore networks
associated to different water salt and concentrations
can be very different.
To make this information more useful for modeling
purposes, it can be of some interest to express the same
data in terms of difference between the cumulative
intruded volume obtained with a given solution and
the one obtained with distilled water, as presented in
Figure 4.
In Figure 4, eis represents the intruded void ratio
of samples saturated with a given saline solution,
while eiw is the intruded void ratio of the sample saturated with distilled water. It is clearly showed that
two regions of porosity exist which behave in a very
Figure 3. Selected mercury intrusion pore size curves

obtained with different saturating solutions.
Figure 4. Difference between the intruded void ratio of

samples saturated with saline solutions and the intruded void
ratio of the distilled water saturated sample.
different manner to salinity changes. The porosity fraction consisting of pores whose diameter is greater than
1000 nm appears to increase upon salinization, while
the opposite holds for the porosity fraction given by
pores having diameter smaller than 1000 nm. These
results help explaining the higher hydraulic conductivity of FEBEX samples saturated with saline solutions
in Castellanos et al., 2008 and in Musso et al., 2013.
3.3 Quantification of microstructural deformation
induced by salinity changes
Need for quantitative characterization of microstructural behavior proceeds from the necessity of populating double-structure models that can be used to
202
reproduce the behavior of structured active clays.

Double structure models have been proposed both in
order to model transport phenomena (e.g. Gerke &
Van Genuchten, 1993) and to model the mechanical
behavior (e.g. Gens & Alonso, 1992, Alonso et al.,
1999).
In double structure models, the overall void ratio
e is assumed to be the sum of a microstructural
void ratio em and a macrostructural void ratio eM ,
defined as:
where Vvm is the volume occupied by the micro-voids

(intra-aggregate voids), VvM is the volume occupied
by macro-voids (inter-aggregate voids) and Vs is the
volume of solids.
To quantify the evolution of the microstructural
void ratio upon chemical changes it is then important to identify a way to discriminate between microand macro-voids. This also allows evaluating the
volumetric deformation of the aggregates. Different approaches have been adopted in the literature,
depending on the type of phenomena investigated
and on the technique used. Starting from MIP data,
Delage & Lefebvre (1984) and Delage et al. (1996),
proposed a criterion based on data from mercury
intrusion/extrusion cycles. The first intrusion fills
all the accessible and interconnected pore space,
whereas on complete releasing of pressure, mercury is
extruded only from the non-constricted pores, which
are identified with the intra-aggregate pores. The difference between the intrusion and extrusion branches
describes the entrapped (constricted) porosity, which
is related to the inter-aggregate pore space.
Della Vecchia (2009) and Romero et al. (2011)
proposed a criterion based on the working assumption that intra-aggregate pores are always smaller than
inter-aggregate pores. Therefore a delimiting pore size
exists, being an upper bound for the intra-aggregate
porosity and a lower bound for the inter-aggregate
porosity at the same time. It can be determined from
MIP data referring to the condition that promotes the
most significant swelling of the aggregates (in presence of salinity changes, this corresponds to saturation
with distilled water). By adopting this criterion it is
found that this delimiting pore size for FEBEX samples compacted at the dry density d = 1.65 Mg/m3 is
xlim = 1000 nm (Musso et al., 2013). It is interesting to
note that this is the same size obtained as a threshold
between the two porosity regions discussed referring
to Figure 4.
Porosimetry data were thus used to evaluate Vvm
and then em through the expression:
where F(x) is the cumulative intruded volume corresponding to the diameter size x.
In order to formulate a simple model to describe
chemical effects on the microstructure, NaCl and
Figure 5. Evolution of the microstructural void ratio em with

osmotic suction.
CaCl2 concentrations have been simply expressed in

terms of osmotic suction. Assuming that concentration
remains relatively low, the vant Hoff expression can
be adopted:
where is osmotic suction, i is the number of constituents into which the salt molecule separates upon
dissolution (i = 2 for NaCl and i = 3 for CaCl2 ), c is
molar concentration, R is the universal gas constant
and T is absolute temperature.
The microstructural void ratio em of the investigated
FEBEX samples is found to depend on the osmotic
suction as in Figure 5.
Following the approach used inAlonso et al. 1999, it
can be assumed that microstructural strains are elastic
and that they can be reproduced by the expression:
being
where m is the osmotic suction of the solute in

the micro-voids, m
vol is the volumetric strain of the
microstructure and and are parameters describing the mechanical behaviour of the aggregates.
Parameter calibration of Eq. 4 performed on the
basis of the experimental data of Figure 3 provides
= 3.2102 MPa1 and = 2.0101 MPa1 .
Eq. (4) recalls the proposal of Alonso et al. (1994),
originally formulated in terms of total suction. It
can then be of interest comparing strains experienced
203
Figure 6. Microstructural volume strain induced on compacted FEBEX samples by varying total (right axis) and
osmotic suction (left axis). Total suction data taken from
Romero & Simms (2008).
by the microstructure when one or the other suction are imposed. In Figure 6, chemically induced
strains presented in this work are compared to the
results by Romero & Simms (2008). Those refer to
compacted FEBEX samples introduced in a ESEM
chamber where the temperature and the vapour pressure were varying. The pertinent total suction values
were deduced on basis of the psychrometric law:
where w is water density and Mw is water molecular

mass.
It is very interestingly found that strains induced
by total and osmotic suction changes follow a similar trend. Although a certain experimental dispersion
exists and further evidence is required, this seems
to suggest that mechanical phenomena caused by
hydraulic and chemical processes could be studied
within a single frame, as proposed by Guimares et al.
(2013).
4 A DOUBLE STRUCTURE MODEL FOR
COUPLED CHEMO-HYDRO-MECHANICAL
PROCESSES
4.1
Chemo-mechanical behavior
According to Eq. (1), the overall volumetric strain is

the sum of the volumetric strain associated with the
micro and to the macro-voids. From an experimental
point of view, the independent measurement of the two
contributions is not easy. In fact, phenomenological
observations at the laboratory scale, such as the displacements measured during an oedometer test, give
information only about the overall response of the sample and it is difficult to discriminate the contribution
Figure 7. Comparison of overall and expected microstructural void ratio obtained at different osmotic suctions. The
continuous line represents the expected behaviour of the
macrostructure at imposed constant aggregate volume.
of each domain. For example, during a water dilution test occurring at constant total stress, both the
aggregates and the soil sample swell: as a result, the
volume change of the macro-voids is the difference
between the volume change of the sample and the one
of the aggregates. Without a sound model for aggregate
size evolution, macro-void volume variation cannot be
inferred. Only in the limiting case of a constant volume
test, macropore volume reduction can be considered
equal in magnitude to aggregate volume expansion.
It can then be postulated that only a fraction * of
the microstructural volume strains dm
vol contributes
to the overall strains. The remaining fraction (1 *)
results in a reduction of the macro-porosity, or macrovoid invasion. The overall volumetric strain can be
defined then as follows:
where dM
vol is the volumetric strain that the macroporosity would experience at constant aggregate size.
Data presented in Figure 5 suggest that microstructural strains increments are negligible when
increases from 15 MPa to 40 MPa. Then the dependency of dM
vol on d could be obtained in this range:
under the assumption of oedometer conditions and
constant vertical total stress, this relationship can be
expressed by:
where is the compliance of the soil against osmotic

suction variations. The value of can thus be calibrated on the slope of the e log data in the high
range, as in Figure 7.
204
By choosing = 0.015 and imposing equation 7,

on basis of data of Figure 7 it is also found that void
ratio data at low suctions are fitted adopting = 0.45.
where s and Ss are the grain density and the specific

surface of the solid particles.
Total volumetric water flow v is obtained summing
Darcian and osmotic flows:
4.2 Chemo-hydraulic behavior

In double porosity models, the medium is considered
to be a superposition of the inter-aggregate and of the
intra-aggregate domains (Gerke and van Genuchten,
1993). Transport processes are modelled considering
these domains as homogeneous media with different
hydraulic and solute transport properties (Barrenblatt
1960, Warren & Root 1963).
Although in principle transport of the solute can
occur through both domains, in the present work it
is assumed that transport through the aggregates is
not very efficient, because of the very low permeability and effective diffusion properties associated
to very small pores. Therefore, the microstructure is
assumed to be able only to exchange mass with the
macrostructure and effective transport of mass to occur
only through the macropores.
Several mechanisms are known to govern the movement of solutes in active clays (e.g. Lu et al., 2004).
Fluxes are both direct (Darcian flow, for the mass of
water; diffusion fluxes, for the solute) and coupled
(among these, advection and osmosis are the dominant ones when no thermal or electrical gradients are
acting).
Osmotic flow vos , that is the flow of water under
chemical concentration gradients, can be written as:
The osmotic conductivity, K , depends on the capacity

of the soil to act as an impermeable barrier with respect
to the flow of the solutes. This is mostly controlled by
the size of the pores of the soil, by its specific surface and by the chemical concentration of the solute.
A convenient expression relates K to the hydraulic
conductivity Kh :
where is the osmotic efficiency coefficient. can

be evaluated for instance by means of the relationships
proposed by Bresler (1973) quoted in Mitchell & Soga
(2005). depends on the valence
of the cations in
solution and on the parameter b c, where b is the
average half distance between particles and c is the
concentration.
By assuming that flow occurs only through the
macrostructure, it follows that b has to be assumed as
the average half distance between aggregates (instead
of particles). Then b can be evaluated as:
where KM is hydraulic conductivity (associated to the

macroporosity), w is the density of water, pM is the
water pressure and z is the elevation head.
The salt flow in the macropores, j, is due to
both advection and diffusion. Neglecting the movement of the solid phase with respect to the reference
configuration, it can be written as
where DM is an effective diffusion coefficient,

accounting for the effects of porosity, tortuosity,
osmotic efficiency and retardation. In the following
the term DM will be express as DM = (1 )DM0 ,
where DM0 is the effective diffusion coefficient in
absence of osmotic effects.
4.3 Balance equations
Micro-porosity m and macro-porosity M are
defined as the volume of the corresponding pores
divided by the overall volume Vtot :
The mass balance of water is then given by:
where qEX
w is the water mass transfer term between
micro and macro pores.
Similarly, the mass balance equations for the
solute are:
where cm and cM are the saline concentrations in the

EX
micro and the macro pores. Both qEX
s and qw are positive when the mass exchanged is directed towards the
micro-porosity.
A key point in addressing the behavior of the double
porosity medium is an appropriate representation of
205
the mechanisms that allows exchange of mass between

the two phases. It should be noticed first that water and
solute mass transfer terms are related one to the other.
This relations exists because transfer of solute is due
to both advection and diffusion. From Eqs. (15), with
incompressible water, and (17) it follows that:
Mass exchange is usually assumed as proportional to

the difference between the driving potentials in the two
structural levels, thus q EX
s can be defined as:
where is a first-order transfer coefficient.

The first order transfer coefficient can be
expressed as a function of the effective hydraulic
conductivity K (Warren & Root, 1963), through a
relationship of the form:
where is a function of the geometry of the

aggregates.
Gerke & van Genuchten (1993) extended the to
unsaturated porous media by associating K to the permeability of the micro-porosity, which is a function
of the degree of saturation. Since relative permeability is related to the size and the distribution of the
interconnected pores (e.g. Romero et al., 1999), and
this distribution depends on salinity, PSD data allow
suggesting a dependency of the mass exchange transfer coefficient on salinity. Starting from Leong &
Rahardjo (1997), the role of evolving aggregates in
equation (20) can be taken into account as:
where xlim is the diameter representing the threshold

between micro-porosity and macro-porosity, that has
been set equal to 1000 nm, and f (c) is a function of salt
concentration. The ratio f (c)/f (c = 0) was evaluated
for all the available PDSs and its evolution is presented
in Figure 8.
A possible mathematical equation for the transfer
coefficient is then:
where and are two parameters to be calibrated on

the phenomenological response of the material.
5
SIMULATION OF SALT TRANSPORT AND

ASSOCIATED CHEMO MECHANICAL
PROCESSES
The double porosity model described in the previous

section has been used to simulate the results of a salt
Figure 8. Normalisation of the transfer term at a given concentration with respect to the null concentration condition.
It is hypothesized that the transfer term depends on the
permeability of the microstructrure, reconstructed upon MIP
measurements.
transport test run in oedometer. A 10.5 mm sample

of FEBEX bentonite (dry density d = 1.65 Mg/m3 )
was placed in a modified oedometer, loaded at the
hygroscopic water content up to v = 200 kPa and then
saturated by pouring a NaCl solution (c = 5.5 mol/l) in
two reservoirs connected to the opposite ends of the
sample. The sample swelled upon saturation. When
displacements ceased, the solution in the bottom reservoir was replaced by distilled water, allowing for the
salt within the sample to be transported towards the
reservoir, where the distilled water was renewed daily
to ensure a condition of constant null osmotic suction
at one sample end. The upper reservoir was kept filled
with the saline solution, whose concentration lowered
during the process as a consequence of salt transport.
This salinity evolution was monitored through electrical conductivity measurements, that were interpreted
in terms of NaCl concentration. To compensate for
evaporation occurring in the laboratory, the total volume of the liquid in this reservoir was kept constant
by daily pouring distilled water. Sample desalinization caused progressive swelling. Once that swelling
stopped and that salinity in the reservoir reached a
constant very low value, it was assumed that transport
and chemo mechanical phenomena had ceased, and
the lower reservoir was filled again with the 5.5 mol/l
NaCl solution. Evolution of salinity and displacement
were monitored again upon the reversal of the chemical process. Due to the very low permeability and
diffusion characteristics of the sample, the whole test
lasted about one year.
To allow processes taking place during the test to
be understood, the model exposed in the previous section was implemented into the Finite Element code
COMSOL and then analyses were run to simulate
the experiment. Parameters used in the simulation
206
Table 1. Mechanical parameters adopted in the double

porosity simulation.
Microstructure
Compressibility
MPa1
MPa1
Macrostructure
Compressibility
2 101
3.2 102
0.015
Interaction
Function
*
0.45
Table 2. Transport parameters used in the double porosity

simulation.
Hydraulic
Conductivity
KM
m s1
Effective
Diffusion
DM0
m2 s1
5 1012
9 1011
Transfer
Function
s1
l mol1
0.8
Figure 9. Comparison between experimental results and

model predictions: history of vertical displacement.
are reported in Table 1 and in Table 2. Microstructural compressibility values and were determined
through the procedure discussed in section 3. Data
in Castellanos et al. (2008) were used to estimate
the logarithm compliance against suction changes, ,
and concurred to the estimation of the hydraulic conductivity KM . The effective diffusion coefficient was
estimated on basis of literature data referring to analogous materials and porosity (Bourg et al., 2006).
The only parameters that were tuned during the
simulation were the ones of the micro-macro transfer
function of Equation 22, and .
The geometry of the model included both the soil

specimen and the reservoir where the salt concentration was left free to vary. The latter was considered
as a dummy rigid porous medium, characterised by a
very high hydraulic conductivity and with the diffusion coefficient of NaCl in water (Mitchell and Soga,
2005). This way, comparison between the experiment
and the model performance can be made in terms both
of displacements and concentration.
5.1
Figure 10. Comparison between experimental results and

model prediction: history of concentration in the upper
reservoir.
Simulation results reproduction of external

measurements
Swelling upon desalinization occurred in stages,

developing at different times (Figure 9). This seems to
depend on the way the sample was prepared, and then
on its fabric. An analogous behavior was documented
in other tests run on compacted specimens (e.g. Musso
et al. 2003; Rao et al. 2006), but not in works concerning reconstituted samples prepared starting from
slurries at high initial water contents (e.g. Di Maio
1996, Barbour & Fredlund 1989).
Swelling required more than 300 days to develop,
much more than the time required by shrinkage (less
than 10 days). As discussed in Gajo & Maines, 2007,
different time scales certainly depend on the nonlinearity of the relationship between strains and osmotic
suction, so that deformations are expected to be more
appreciable only when the salinity reaches very low
values. On the other hand, as shown in Figure 10,
salt concentration also evolves in stages, and although
to a far less remarkable extent desalinization and
salinization times were different as well.
Progress in time of both concentration and vertical displacement appears to be correctly captured by
the model, since both concentration and displacement
stages and time scales are well reproduced.
207
Figure 11. Profiles of microstructural osmotic suction m

(continuous lines) and macrostructural osmotic suction M
(dotted lines) during desalinisation.
5.2
Figure 12. Evolution of macrostructural void ratio eM as

predicted by the simulation based on the double porosity
model.
Simulation results an insight on the predicted

evolution of pressure and fabric
Displacements develop with time accordingly to the

stress history (stress being defined in a broad
sense, then accounting for osmotic suction and
pore pressure). Deformations arise as a consequence of variations in both salt concentration and
hydraulic pressure (e.g. Barbour & Fredlund, 1989;
Kaczmareck & Hueckel, 1998), the latter induced by
the continuity requirements of the water mass balance. Figure 11 provides data of predicted profiles of
microstructural and macrostructural osmotic suction.
After 2.5 and 5 days from the beginning of the
test, a strong difference between m and M values exists over a wide portion of the sample. This
difference reduces while the test progresses: at low
salinities the transfer coefficient is relatively high
(Figure 8), so that m quickly equalizes with M . At
times greater than 45 days, differences between micro
and macrostructural osmotic suction are negligible.
For the data set used in the analysis, the hydraulic
pressure changes were limited within 20 kPa, having
limited impact in terms of the void ratio changes.
As a consequence of the osmotic suction decrease
induced by desalinization, in the lower part of the specimen the macroscopic void ratio eM initially increases
(Figure 12, t = 2.5 days and t = 5 days). At intermediate times (t = 45 days), eM has been reduced in the
lower part of the sample, while it is still increasing in
the upper part. Towards the end of the test, eM reduces
everywhere and its final values are smaller than the
initial ones.
On the contrary, the microstructural void ratio em
increases monotonously with time all over the sample (Figure 13): in the very first days this increase
Figure 13. Evolution of microstructural void ratio em as

predicted by the simulation based on the double porosity
model.
is limited to the first millimiters close to the bottom of the specimen, later it is appreciable elsewhere.
Since the sensitivity of em to changes of osmotic suction is higher when suction is small, swelling of the
microstructure in the upper part of the specimen is
still evident in the time lapse between t = 240 days
and t = 315 days, a time lapse during which osmotic
suction changes are relatively limited (Figure 11).
208
Predicted profiles of em and eM would also suggest that, although swelling occurs monotonously,
microstructural and macrostructural effects are effective at different time scales. Expansion of the
macrostructure significantly contributes to swelling
during the first days. This depends on the high salinity
of the water, which implies a very low transfer coefficient and thus a difficult mass exchange between
micro- and macro-porosity. Within this concentration
range, microstructure is also quite insensitive to suction variations. Expansion of the aggregates is instead
dominant at higher times, when salinity has lowered:
mass exchange between the two structural levels is
also facilitated and the microstructure is very sensitive to suction changes. During this stage, expansion
of the microstructure does not completely reflect in
swelling, but is partially consumed by invasion of the
macro-pores.
6
CONCLUDING REMARKS
Upon compaction, active clays develop an aggregated

fabric, which is affected by salinity changes occurring
in the water phase. Microstructural data can be used to
characterize the evolution of fabric upon salinity variations: an information that can be conveniently used
to calibrate double porosity models, adopted to reproduce both transport and mechanical behavior of the
material. This opportunity is quite interesting from
a practical perspective, since it allows justifying the
adoption of the same modeling framework introduced
for thermal-hydro-mechanical processes.
Fabric changes have an impact on transport parameters, since the hydraulic conductivity of the soil, the
mass exchange coefficient between the macro porosity and the micro porosity and the osmotic efficiency
coefficient all depend on fabric. These phenomena,
together with the non linearity of the relationship
between salinity and volumetric strains at both the
macrostructural and the microstructural level, contribute explaining peculiar effects such as discontinuity in the evolution of both concentration and swelling
process.
REFERENCES
Alonso, E.E., Gens, A. & Gehling, W. 1994. Elastoplastic
model for unsaturated expansive soils. 3rd Conf. Num.
Meth. Geotech. Engng, 1118.
Alonso E.E., Vaunat J. & Gens A. 1999. Modelling the
mechanical behaviour of expansive clays. Eng Geol
54:173183.
Barbour, S.L. & Fredlund, D.G. 1989. Mechanisms of
osmotic flow and volume change in clay soils. Canadian
Geotechnical Journal 26, 551562.
Barrenblatt, G.I., Zheltov, I.P. & Kochina, I.N. 1960. Basic
concepts in the theory of seepage of homogeneous liquids
in fissured rocks. Journal of Applied Mathematics and
Mechanics 24(5), 12861303.
Bolt, G.H. 1956. Physico-chemical analysis of the compressibility of pure clays. Gotechnique, 6(2), 8693.
Bourg, I.C., Sposito, G. & Bourg A.C.M. 2006. Tracer diffusion in compacted, water-saturated bentonite. Clays and
Clay Minerals 54(3), 363374 DOI10.1346/CCMN.2006.
0540307.
Bresler, E. 1973. Anion exclusion and coupling effects in non
steady transport through unsaturated soils: I. Theory. Soil
Science Society of America Proceedings 37, 663669.
Calvello M., Lasco, M., Vassallo, R. & Di Maio C. 2005.
Compressibility and residual shear strength of smectitic
clays: influence of pore aqueous solutions and organic
solvents. Rivista Italiana di Geotecnica 1, 3446.
Castellanos E., Gens A., Lloret A. & Romero E. 2006. Influence of water chemistry on the swelling capacity of a
high_ density bentonite. Unsaturated soils. Geotechnical Special Publication N.147 ASCE Reston Virginia,
1, 962972.
Castellanos E., Villar M.V., Romero E., Lloret A. & Gens A.
2008. Chemical impact on the hydro-mechanical behavior
of high-density FEBEX bentonite. Physics and Chemistry
of the Earth 33, S516S526.
Delage P, Audiguier M., Cui YJ & Howatt M.D. 1996.
Microstructure of a compacted silt. Can. Geotech. J
33:150158.
Delage P. & Lefebvre G. 1984. Study of the structure of
a sensitive Champlain clay and of its evolution during
consolidation. Can Geotech J 21:2135.
Della Vecchia, G. 2009. Coupled hydro-mechanical
behaviour of compacted clayey soils. Phd Thesis, Politecnico di Milano.
Di Maio, C. 1996. Exposure of bentonite to salt solution:
osmotic and mechanical effects. Gotechnique, 46(4):
695707.
Di Maio, C. & Fenelli, G.B. 1997. Influenza delle interazioni chimico-fisiche sulla deformabilit di alcuni terreni
argillosi. Rivista Italiana di Geotecnica 1/97, 4457 (in
Italian).
Fernandez, F. & Quigley, R.M. 1985. Hydraulic conductivity
of natural clays permeated with simple liquid hydrocarbons. Canadian Geotechnical Journal, 22: 205214.
Gajo, A., Loret, B. & Hueckel, T. 2002. Electro-chemomechanical couplings in saturated porous media: elastoplastic behaviour of heteroionic expansive clays. Int. J.
Solids Struct. 39 (16), 43274362.
Gajo, A. & Loret, B. 2007. The mechanics of active clays
circulated by salts, acids and bases. J. Mech. Phys. Solids
55 (8), 17621801.
Gajo, A. & Maines, M. 2007. Mechanical effects of aqueous
solutions of inorganic acids and bases on a natural active
clay. Gotechnique 57 (8), 687699.
Gens, A. 2010. Soil-environment interactions in geotechnical engineering. 47th Rankine Lecture. Gotechnique 60:
374.
Gens, A. & Alonso E.E. 1992. A framework for the behaviour
of unsaturated expansive clays. Canadian Geotechnical
Journal 29, 10131032.
Gerke, H.H & van Genuchten, M.T. 1993. A dual-porosity
model for simulating the preferential movement of water
and solutes in structured porous media. Water Resources
Research 29 (2): 305319.
Guimares L. Do N., Gens, A., Sanchez M. & Olivella S.
2013.A chemo-mechanical constitutive model accounting
for cation exchange in expansive clays. Gotechnique 63
3, 221234.
Hueckel, T. 1997. Chemo-plasticity of clays subjected to
stress and flow of a single contaminant. Int. J. Numer.
Anal. Methods Geomech. 21, No. 1, 4372.
Hueckel,T., Kaczmarek, M. & Caramuscio, P. 1997.Theoretical assessment of fabric and permeability changes in clays
209
affected by organic contaminants, Canadian Geotechnical

Journal 34(4): 588603, 10.1139/t97-013.
Kaczmarek, M. & Hueckel, T. (1998). Chemo-Mechanical
Consolidation of Clays: Analytical Solutions for a Linearized One-Dimensional Problem. Transport in Porous
Media 32: 4974, 1998.
Lloret, A., Romero E. & Villar M. 2004. FEBEX II Project.
Final report on thermo-hydro-mechanical laboratory tests.
Publicacion tecnica 10/2004 ENRESA.
Lloret, A., Villar, M.V., Snchez, M., Gens, A., Pintado, X. &
Alonso, E. E. 2003. Mechanical behaviour of heavily compacted bentonite under high suction changes. Gotechnique 53(1), 2740, doi: 10.1680/geot.2003.53.1.27.
Leong, E.C. & Rahardjo, H. 1997. Permeability functions for
unsaturated soils. Journal of Geotechnical Engineering
123(12), 11181126.
Lu, N., Olsen, H.W. & Likos, W.J. 2004. Appropriate material properties for advective diffusive solute flux in
membrane soil. Journal of Geotechnical and Geoenvironmental Engineering, 13, 13411346.
Mesri, G & Olson, R.E. 1971a. Consolidation characteristics
of montmorillonite. Gotechnique 21(4), 341352.
Mesri, G & Olson, R.E. 1971b. Mechanisms controlling
the permeability of clays. Clays and Clay Minerals 19,
151158.
Michaels, A.S. & Lin, C.S. 1954. The permeability of
kaolinite. Industrial and Engineering Chemistry, 46:
12391246.
Mitchell, J.K. & Soga K. 2005. Fundamentals of Soil
Behavior. John Wiley & Sons. Inc., New York, 3rd ed.
Musso G., Chighini S. & Romero E. 2008 Mechanical sensitivity to hydro-chemical processes of Monastero Bormida
clay. Water Resources Research, 2008, Vol. 44, ISSN:
0043-1397, DOI: 10.1029/2007WR006533.
Musso G., Romero E. &, Della Vecchia G. 2013
Double-structure effects on the chemo-hydro-mechanical
behaviour of a compacted active clay. Gotechnique, 63,
206220, DOI: 10.1680/geot.SIP13.P.011.
Musso G., Romero E., Gens A. & Castellanos E. 2003. The

role of structure in the chemically induced deformations
of FEBEX bentonite. Applied Clay Science, 23: 229237.
Rao, S. M., Thyagaraj, T., & Thomas, H. R., 2006. Swelling
of compacted clay under osmotic gradients. Gotechnique
56 (10), 707713.
Romero, E., Della Vecchia, G. & Jommi, C. 2011. An
insight into the water retention properties of compacted clayey soils. Gotechnique 61(4), 313328. doi:
10.1680/geot.2011.61.4.313.
Romero, E., Gens, A. & Lloret, A. 1999. Water permeability,
water retention and microstructure of unsaturated compacted Boom clay. Engineering Geology 54, 117127.
Romero, E. & Simms, P. 2008. Microstructure investigation
in unsaturated soils: a review with special attention to
contribution of mercury intrusion porosimetry and environmental scanning electron microscopy. Geotech. Geol.
Engng 26, No. 6, 705722.
Siddiqua, S., Blatz, J. & Siemens, G. 2011. Evaluation of
the impact of pore fluid chemistry on the hydromechanical behaviour of clay-based sealing materials. Canadian
Geotechnical Journal, 48(2), 199213.
Studds, P.G., Stewart, D.I. & Cousens, T.W. 1998. The
effects of salt solutions on the properties of bentonite-sand
mixtures. Clay Minerals, 33: 651660.
Warren, J.E. & Root, P.J. 1963. The behavior of naturally fractured reservoirs. Society of Petroleum Engineers Journal
3, 245255.
Wood, D.M. 1991. Soil behaviour and critical state soil
mechanics. Cambridge: Cambridge University Press.
Yang, N. & Barbour, S.L. 1992. The impact of soil structure
and confining stress on the hydraulic conductivity of clays
in brine environments. Canadian Geotechnical Journal
29: 730739.
Yong, R.N. 1999. Overview of modeling of clay microstructure and interactions for prediction of waste isolation
barrier performance. Engineering Geology 54, 8391
210

Soil effective thermal conductivity from energy pile thermal tests

A. Bouazza
Department of Civil Engineering, Monash University, Melbourne, Australia
B. Wang
Golder Associates Pty. Ltd., Melbourne, Australia
R.M. Singh
ABSTRACT: The concept of transferring heat from a lower temperature circuit to a higher temperature circuit
utilising a heat transfer mechanism has been around for centuries. Recently, this concept has been extended to
pile foundations where the piles act as heat exchangers (energy piles) to extract from or store energy in the
ground. For these systems to operate efficiently, it is necessary to select suitable design methods and tools to
assess their thermal behaviour. An optimised system design requires the understanding of the built structure
energy (secondary circuit) requirement. However, an understanding of the relationship between temperature
transfer of the surrounding ground to and from the heat exchanger (primary circuit) is critical. As the knowledge
of the thermal process in the ground and its interaction with the heat exchanging energy pile are needed to
determine the amount of energy that can be extracted from or stored in to the ground. This paper reviews the
concept of utilising the primary circuit to assess the ground thermal properties and presents the effective thermal
conductivity results obtained from in-situ thermal response tests carried out on a field energy pile.
INTRODUCTION
The use of geothermal energy techniques integrated

with structural pile foundation has the capability of
being an efficient, cost effective, sustainable and environment friendly solution to space heating and cooling
of built structures. Generally known as geothermal
energy piles, these systems utilise plastic piping to circulate a heat built structures transferring medium from
the ground to the structure. The plastic tubing can be
installed in different types of piles, but generally the
replacement methods (bored piles) are most common.
Since the temperature within the ground remains relatively constant, the circulating medium maintains a
constant temperature throughout the year within the
structure. Effectively, geothermal energy piles will
reduce the need for traditional sources of heating
and cooling and will bring the idea of carbon-neutral
buildings closer to reality. In order to implement this
system, further studies of instrumented geothermal
energy piles are needed to provide insight into how
temperature variations affect the overall geotechnical
and structural performance of the system.
The performance of energy piles is determined
by the energy balance between the primary circuit
(i.e energy pile) and secondary circuit (i.e. building), the heat pump characteristics and the thermal
process between the heat exchanger and the surrounding ground. To design an efficient energy pile,
sufficient knowledge and understanding of the thermal behaviour within the primary circuit consisting
of the heat transfer medium, absorber pipes, heat
exchanger and the surrounding soil is required. In-situ
field tests conducted using borehole heat exchangers to assess the thermal properties of the ground
and the heat exchanger system are well established
(Kavanaugh et al., 2001). However, very limited literature is available on assessing these thermal properties
of the primary circuit utilising energy piles.
Vertical borehole of approximately 100 mm to
200 mm in diameter installed with absorber pipes to
depths of approximately 100 m to 200 m are commonly
utilised as vertical heat exchangers across the world
(Lund et al., 2004). The boreholes are generally backfilled following installation of the absorber pipes with
a relatively high conductive mixture consisting of bentonite, cement and/or quartz sand. Published literature
has shown that thermal properties of the ground can
be assessed by undertaking in-situ Thermal Response
Test (TRT) utilising borehole heat exchangers (Pahud
and Matthey, 2001; Witte et al., 2002; Lim et al.,
2007; Mattsson et al., 2008; Florides and Kalogirou,
2008; Saner et al., 2008; Sarbu and Bura, 2011). The
American Society of Heating, Refrigerating and AirConditioning Engineers (ASHRAE) has published a
set of guidelines (Kavanaugh et al., 2001), which
have been widely accepted as the standard requirements to undertake field investigation to determine
211
soil and rock thermal properties from short term field

tests utilising borehole heat exchangers. The principles used and reported byASHRAE are consistent with
work reported by the Swiss Federal Office of Energy
(Poppei et al., 2006) and other research work carried
out in the United States at Oklahoma State University (Austin, 1998) and in Sweden at Lund Institute
of Technology (Eskilson, 1987; Hellstrom, 1991) and
Lulea University of Technology (Gehlin, 2002).
However, there is limited literature on in-situ thermal response test utilising full-scale energy piles to
determine ground thermal properties. Brandl et al.
(2006), Gao et al. (2008a & 2008b), Lennon et al.
(2009), Ma and Grabe (2010) showed that thermal
properties of the ground are not consistent when
Thermal Response Tests (TRTs) were undertaken by
different heat exchange systems, such as energy piles
and heat borehole exchangers. This could be due to the
fact that the analytical concept utilised by ASHRAE
was developed through field trials of borehole heat
exchangers. Thus, key parameters such as diameter
and length of the energy pile, their structural elements (steel reinforcement and concrete compared to
cement/bentonite/sand mixture), the number of pipe
loops and distance between the pipes installed within
an energy pile could influence the variation of soil
thermal properties measured by an energy pile system.
This paper reviews the concept of utilising the
primary circuit to assess the ground thermal properties and presents the methodology and results of
the field in-situ thermal response tests carried out
on an energy pile installed at Monash University,
Melbourne, Australia.
2 THERMAL PROPERTIES
Thermal transport through soils is determined by three
thermal properties, these are described as follows:
(a) Thermal conductivity (W/m K): the rate of soil
to transfer thermal energy through a unit area at a
temperature gradient.
(b) Thermal capacity Cv (J/m3 K): the capacity of
soil to store thermal energy.
where Cv = Cm
Cv = volumetric heat capacity of soil (J/m3 K);
= density of soil (kg/m3 ); Cm = Specific heat
capacity of soil (J/kgK)
(c) Thermal diffusivity (m2 /s): the speed and
ability of soil to undergo temperature changes.
The relationship between the above three key
parameters can be expressed as:
The most important parameter required to optimise the design of energy piles or borehole ground
heat exchangers is the thermal conductivity of the
surrounding ground. For the preliminary design of
complex energy foundations or the detailed design
of standard geothermal systems, sufficient accuracy

of ground thermal properties can be obtained from
laboratory testing (Brandl et al., 2006; Brandl, 2006;
Sanner et al., 2008). However, for complex and
advanced designs in-situ measurements such as TRT
are recommended (Brandl et al., 2006) to determine
the specific soil thermal properties the primary circuit
would operate within.
A variety of analytical and numerical data analysis
methods have been developed, of which, the line source
model is the most commonly used method for evaluation of the TRT data because of its simplicity and speed
(Gehlin, 2002). TRT has so far been used primarily for
in-situ determination of design data for borehole heat
exchanger systems. Its theory and application have
been developed from Ingersoll and Plass (1948) where
the line source model was applied to design ground
loop heat exchangers. Mogensen (1983) was the first
to apply this concept on borehole heat exchangers to
estimate the ground thermal conductivity from experimental field tests by developing a method of measuring
three important parameters: 1) average thermal conductivity of the ground, 2) the thermal resistance of
the borehole heat exchanger and 3) the undisturbed
ground temperature. During the 1990s, independent
field tests of borehole heat exchangers were conducted
in Sweden at Lulea university of Technology, and in
USA at Oklahoma State University. These studies have
presented methods of evaluating site specific, effective soil thermal properties that are applicable to a soil
formation where the field TRT was carried out.
During a TRT, heat is transferred to or from a
vertical ground heat exchanger and causes a change
in temperature of the surrounding ground. The line
source method was utilised to analyse the temperature
field as a function of time and radius around a borehole
with a constant heat storage or extraction (Eskilson,
1987; Hellstrom 1991). Gehlin (2002) showed that the
temperature field around a borehole can be expressed
as follows:
Provided that:
where T is the change in temperature (K), q is the heat

storage/transfer rate per unit borehole length (W/m),
eff is the effective thermal conductivity (W/mK), is
the thermal diffusivity (m2 /s), t is the minimum time
criterion for test duration (s), rb is the borehole radius,
is the Eulers number (0.5772). The above formulation assumes the following as described by Brandl
et al. (2006) and Gehlin (2002):
212
A constant temperature along the borehole not

feasible in practice. However, the axial temperature gradient is relatively small in relation to radial
gradient. Therefore, is negligible.
Infinite length of borehole. In practice, the length

of borehole is considerably larger than the borehole
radius. Thus during short period of response tests
the borehole end effect can be ignored.
The effective thermal conductivity, the ground heat

capacity (proportional to thermal conductivity and
diffusivity Equation 1) and thermal resistance are
three important thermal design parameter required to
design an efficient energy pile. The following sections summarise the methods developed to assess the
ground thermal properties.
3
Gehlin (2002) showed that for t/r2b = 5 the error

for calculating the effective thermal conductivity is no
greater than 10% for individual TRT carried out in
the field, and decreases to 2.5% for t/r2b = 20. Equation 1 shows that thermal conductivity is a function
of volumetric heat capacity and thermal diffusivity,
this condition signifies that the accuracy increases as
the thermal front of the temperature wave (penetration
distance and velocity) moves further, beyond the heat
exchanger. The distance and velocity of the thermal
front is also dependent of the ratio between the thermal conductivity and volumetric heat capacity of the
ground, i.e. the ground thermal diffusivity.
EFFECTIVE THERMAL CONDUCTIVITY
Thermal conductivity of a soil is defined as the amount

of heat passing in unit time through a unit crosssectional area of soil under a unit temperature gradient
applied in the direction of the heat flow (Farouki,
1981). Field research of in-situ measurement of the soil
thermal conductivity was undertaken across Europe
and USA on borehole ground heat exchangers for a
number of years (Sanner et al., 2008; Mathsson et al.,
2008; Lund et al., 2004; Witte et al., 2002). Published literature (Kavanaugh et al., 2001; Gehlin, 2002;
Austin, 1998) showed that during a TRT, based on the
line source method, a defined energy is applied to the
heat exchanger whilst the power input and the inflow
and out flow temperature of the heat transfer medium is
recorded. This measures the entire length of the ground
heat exchanger, providing an effective thermal conductivity value whilst considering the borehole backfilling
(or pile properties), variable ground and groundwater
conditions. The effective thermal conductivity measured from a field TRT test can be calculated by
Equation 4:
where Q is the constant heat power (W), Lb is the

length of heat exchanging borehole or pile (m) and k is
the slope of semi log graph between the test duration
plotted on log scale and the mean temperature of the
heat transfer fluid.
The effective thermal conductivity calculated from
Equation 4 was based on the heat exchanger and its
immediate vicinity attaining steady-state conditions
(Eskilson, 1987). This requires a minimum time criterion, as shown in Equation 3, to be satisfied. The
test data prior to this period needs to be ignored to
reduce errors as during this initial heating stage, the
thermal front gradually moves further beyond the heat
exchanger wall. The average heating fluid temperature
rises rapidly during this initial period, then as the thermal front travels further into the surrounding ground,
the increase in average fluid temperature becomes
steady. Parameter k is found by plotting the regression line derived from the time temperature series of
a TRT, during the steady increase period of the fluid
temperature.
4 THERMAL DIFFUSIVITY
Thermal diffusivity is a measurement of how soils
undergo temperature changes. Soils with a high thermal diffusivity would have the capacity to undergo
fast and significant changes in temperature whilst
soils with low thermal diffusivity would have the
opposite capability. Thermal diffusivity can be easily
measured within a controlled laboratory environment.
However, in a field test environment, if the temperature of the ground was measured at a known distance
from the heat source during a thermal response test,
the thermal diffusivity of the ground can be calculated
using the amplitude decrease or the time lag methods (Brandl, 2006). Bristow et al. (1994a, 1994b and
1995) showed that by using dual-probe instrument, soil
thermal diffusivity could be measured by utilising a
short-duration heat pulse theory from an infinite line
source. They showed that at a distance r from the heat
source, the temperature change within a soil subjected
to a heat pulse with a duration time (to ), reaches to
a maximum value (Tm ) with respect to time (tm ).
Thermal diffusivity (), in m2 /s, of the soil was
expressed by the following:
This method can be applied to a field TRT if the heat

exchanger is the heat source. Temperature readings can
be collected at a known distance away from the heat
source by installing thermocouples or thermistors at
depth below where ground temperature is not affected
by daily or seasonal fluctuations.
5 THERMAL RESISTANCE
Another important and widely used factor for design
of heat exchanger systems is the thermal resistance
between the heat transfer fluid and the wall of the
213
heat exchanger. Developed utilising borehole heat

exchanger, this factor was referred to as thermal borehole resistance. Eskilson (1987), Hellstrom (1991)
and Gehlin (2002) indicated that heat transfer fluid
to borehole wall thermal resistance (Rb ), expressed in
K/(W/m), gives the temperature difference between
the fluid within the absorber pipes (Tf ) and the
borehole wall (Tb ) for a given heat flux (q). This
relationship is shown in Equation 6.
of the vertical heat exchanger in K/(W/m) can be

expressed as:
where Tf is the mean temperature of heat carrier fluid

and To is the undisturbed initial ground temperature.
6
Gehlin (2002) and Brandl et al. (2006) indicated that
thermal borehole resistance depends on the arrangement of the absorber pipes and the material utilised.
Gehlin (2002) reported that Rb values observed in
field tests range from 0.01 K/(W/m) for an open system to 0.20 K/(W/m) for single U-pipe in bentonite
grout borehole heat exchangers. The temperature difference between the heat carrier fluid and the borehole
wall is proportional to the heat transfer rate, which for
a typical heat transfer rate of 50 W/m the corresponding temperature difference becomes 0.5 C to 10 C.
Thus, the higher the thermal borehole resistance, the
higher is the temperature step between the heat transfer fluid and the surrounding ground, hence, higher
temperature losses. This is applicable to all vertical
heat exchangers such as boreholes, energy piles and
diaphragm walls (Brandl et al., 2006).
Gehlin (2002), Brandl et al. (2006), Sanner et al.
(2008) and Mattsson et al. (2008) reported that thermal borehole resistance caused temperature losses
that could affect heat transfer within ground heat
exchanger systems.
The equation used to assess the thermal borehole
resistance of the vertical heat exchanger was developed from Equation 2. The temperature field around
a borehole whilst considering the thermal borehole
resistance can be expressed by the following (Gehlin,
2002; Brandl et al., 2006; Sanner et al., 2008; Mattsson
et al., 2008):
HEAT EXCHANGE RATE
The thermal performance of a ground heat exchanger

can also be investigated by assessing the rate of heat
exchange between the heat exchanger system and its
surroundings. Koene and Geelen (2000) indicated that
the rate of heat exchange primarily depends on the
temperature of the heat transfer fluid and the thermal
conductivity of the surrounding soils. Thus, the higher
the thermal conductivity, the higher the rate of thermal
energy that could be extracted from or stored within
the surrounding soils. Brandl (2006) discussed further
the influence of a number of parameters on the heat
exchange rate. He indicated that the heat flux density
of the absorber pipe systems (where the heat transport within the system is forced by convection of the
fluid), the diameter and length of the absorber pipe;
roughness of the pipe wall; heat conductivity, specific
heat capacity, density and viscosity of the heat transfer fluid; flow velocity and flow conditions (laminar
or turbulent flow) within the absorber pipes, influence
the heat exchange rate.
Jalaluddin et al. (2011) and Jalaluddin and Miyara
(2012) examined the rate of heat exchange utilising different absorber pipe configurations, operation
modes and flow rates within a steel pipe with a diameter of about 140 mm acting as a 20 m pile founded
within clays, sandy clays and sands.The results showed
that the heat exchange rate was higher with higher
flow rate, ensuring turbulent flow of the heat transfer
fluid within the absorber pipes. Brandl (2006) reported
that for general feasibility studies and pre-design of
energy pile foundations the following assumptions can
be made regarding the amount of energy that can be
extracted from thermo-active foundations:
(a) Pile foundations with pile diameter = 0.3 m to
0.5 m: 40 to 60 watts per meter of pile.
(b) Pile foundations with pile diameter = 0.6 m: 35
watts per m2 of soil-pile contact area.
The transient process around a borehole can be

approximated into Equation 7 leading to the following
expression:
To calculate the heat exchange rate, Q, of an energy

pile, in W/m or W/m2 , the following equation can be
utilised:
To simplify to a linear relationship whilst utilising the effective thermal conductivity parameters
observed from a TRT (Equation 4), the thermal resistance between the heat carrier fluid and the wall
where m is the flow rate of the heat transfer fluid

(m3 /s), Cp is the volumetric heat capacity of the heat
214
transfer fluid (J/m3 K) and T is the temperature difference between Tinflow and Toutflow of the heat transfer
fluid (K).
7 THERMAL TESTING AT MONASH
UNIVERSITY
The field energy test is a 600 mm diameter bored
pile drilled to a depth of 16.1m in Brighton Group
material which consists primarily of very dense sands.
Two levels of O-cells were attached to the pile reinforcement cage at 10 m and 14 m depth as shown in
Figure 1. Three loops of heat transfer absorber pipes
were attached to the pile cage to the top of the lower
O-cell level (14 m depth). The thermal loading was
undertaken by a thermal response testing unit supplied
by GeoExchange Australia Pty Ltd.
Two boreholes equipped with thermocouples to
16 m depth were installed at 0.5 m and 2.0 m to the
edge of the test pile respectively (Figure 1).
Three field thermal response tests (TRT) were carried out by heating the energy pile. One TRT was carried out by circulating the heat transfer fluid through
one loop of absorber pipes closest to the boreholes
equipped with thermocouples. Two TRTs were carried
out by transporting the fluid through all three loops
of absorber pipes in a continuous series within the
energy pile. Inflow and outflow temperature of the heat
transfer fluid, ground temperature at every 2 m to 16 m
Figure 1. General layout of the field energy pile at Monash

University.
Table 1.
depth within two boreholes located at 0.5 m and 2.0 m

from the edge of the test pile as well as the pile concrete temperature were recorded continuously during
the heating periods. The test pile and the ground were
cooled naturally by letting the induced heat dissipate
into the surrounding environment following each TRT.
The subsequent TRT did not start until the temperature readings within the pile and the boreholes were
returned, as close as possible, to their initial undisturbed temperatures. The times and duration of each
TRT are summarised in Table 1.
All three TRTs were subjected to a thermal load
of approximately 2500 watts throughout the heating
periods. A constant flow rate of the heat transfer
fluid of approximately 10 L/min (0.000167 m3 /s) was
also maintained during the heating periods of the
three TRTs.
GROUND TEMPERATURE
Preene and Powrie (2009) indicated that at a certain

depth below ground surface the ground temperature
was not affected by seasonal air temperature fluctuations where the ground thermal behaviour was
relatively constant. This is generally achievable in
Europe (Brandl, 2009) and Australia (Bouazza et al.,
2011) within about 6 m to 8 m to the ground surface. The soil temperature below this level is cooler
than air temperature during the summer months and
warmer during the winter months. Thus, allowing for
cooler temperatures to be extracted from the ground
to cool the built structure in summer and utilising
warmer ground temperature in winter to heat the built
structure.
Ground temperature profile at the site of the field
energy pile at Monash University was monitored for
approximately eight months prior to the first application of the thermal load to the energy pile. The ground
temperature was measured by type K thermocouples
installed within two boreholes to 16 m below ground
surface. Figure 2 shows the initial ground temperature
profile from March 2011 to October 2011.
The ground conditions at the site of the energy pile
consists of very dense coarse grained material with low
moisture content and groundwater was not present.
Figure 2 indicates that the ground temperature consists of three distinctive zones. The upper surface zone,
from ground surface to about 2 m depth, which mostly
influenced by seasonal air temperature changes. The
middle shallow zone, from about 2 m to 6 m depth, was
affected by short term ambient temperature changes.
Duration of thermal response tests.
TRT Name
Start Time
End Time
Test Duration (Heating)
Rest After
Test (Cooling)
1 loop (3 days)
12:50 pm 10 October 2011
2:20 pm 13 October 2011 3 days, 1.5 hours 3 days
5 days
3 loop ST (9 days) 3:10 pm 18 October 2011
11:50 am 27 October 2011 8 days, 20.5 hours 9 days 47 days
3 loop LT (52 days) 12:10 pm 13 December 2011 3:40 pm 3 February 2012 52 days, 3.5 hours 52 days 78 days
215
Figure 4. Estimation of effective thermal conductivity

slope of average fluid temperature vs. ln(t).
Figure 2. General layout of the field energy pile at Monash
University.
spring). To approximately 2 m to 4 m below ground

level, the ground temperature was influenced by the
ambient temperature. Below 6 m to 8 m depth, the
ground temperature was relatively constant and was
not influenced by the seasonal air temperature fluctuations. This stable source of ground temperature of
approximately 17.5 C is a reliable heat source during
the cooler months and a heat sink during the warmer
periods in the Melbourne areas.
The energy test pile was subjected to thermal loading during the warmer months of the year, between
October 2011 and April 2012, simulating cooling of
the built structure whilst injecting heat into the ground
via the energy pile during the summer.
Figure 3. Daily air temperature near energy pile testing site

vs. ground temperature.
Below the middle shallow zone and entering the lower

deep zone, below 8 m depth, the ground temperature was relatively stable between 17 C to 18 C. The
constant temperature in the lower deep zone was not
affected by the seasonal changes, thus creating suitable conditions for geothermal energy pile systems to
operate within.
Figure 3 shows the observed ground temperature at
2 m depth intervals in comparison to the daily average minimum and maximum temperatures observed
by the Australian governments Bureau of Meteorology (BOM) Moorabbin airport weather station located
approximately 8 km from the energy test pile site.
Figure 3 shows the daily minimum and maximum
ambient temperatures from March 2011 to April 2012
and the undisturbed ground temperature from March
2011 to the start of thermal response testing in October
2011. The daily average ambient temperature reached
its lowest around June and July, during the winter
months, and peaked around January and February, during the summer months. Figure 3 also shows that the
undisturbed ground temperatures were recorded from
March 2011 (early autumn) to early October 2011 (mid
EFFECTIVE THERMAL CONDUCTIVITY

RESULTS
In-situ field estimation of the ground systems effective thermal conductivity consists of incorporating the
energy pile ground heat exchanger and the surrounding soils as a whole system. This study presents an
estimate of the effective thermal conductivity utilising
the three TRTs. The average heat transfer fluid temperatures were plotted against time for each of the tests
and the regression lines are shown in Figure 4.
To satisfy the minimum-time criterion set out in
Equation 3 and utilising the thermal diffusivity measured during 3 loop ST TRT, the first 100 hours (4.2
days) of the three TRT were ignored in the effective thermal conductivity calculations. However, the
1 loop TRT was terminated within the minimumtime, therefore, the data for the first 48 hours was
not included in the effective thermal conductivity
calculations.
Effective thermal conductivity calculated following
Equation 4 for the three TRTs are shown in Table 2.
The effective thermal conductivity values were not
consistent; the 3 loop ST TRT achieved the highest
value whilst the 3 loop LT TRT achieved the lowest effective thermal conductivity of the energy pile
system.
Austin (1998) showed that the line source model
utilised to estimate thermal conductivity was very sensitive to temperature fluctuations. Figure 4 shows that
216
Table 2.
Duration of thermal response tests.
TRT Name
Average
eff
Q (W) Lb (m) k (slope) (W/mK)
1 loop (3 days)
2404
3 loop ST (9 days) 2423
3 loop LT (52 days) 2422
14.2
14.2
14.2
3.2168
2.7215
3.6205
4.19
4.99
3.75
there were fluctuations of the heat transfer fluid temperature during the heating periods of each TRT. The
HDPE absorber pipes were insulated between the top
of the energy pile to the testing unit with a combination
of insulation foam, aluminium foil and soil. However,
the energy pile was exposed to the summer environment and direct solar radiation. The fluctuation of average fluid temperature shown in Figure 4 was likely to
be caused by solar radiation. The direct sunlight would
heat up the concrete of the energy piles upper surface
section whilst increasing the average fluid temperature within the absorber pipes. Subsequently, during
cooler nights where solar energy was not present, the
pile concrete cooled down significantly and decreased
the average fluid temperature.
The estimated effective thermal conductivity
showed in this study is comparable to other published
literature utilising energy piles as the ground heat
exchanger. Published data (Brandl et al., 2006; Gao
et al., 2008b; Lennon et al., 2009; Brettmann andAmis,
2011) show that utilising energy piles of at least 0.6 m
in diameter for the TRTs, gave an effective thermal
conductivity between 4 W/mK to nearly 7 W/mK in
sandy and clayey soils. However, with smaller diameter piles the effective thermal conductivity was found
to be 2 W/mK and 3W/mK. The long term TRT (3 loop
LT) carried out over 52 days shown in this study is not
a practical test to carry out due to the length of the
testing period.
The estimated field in-situ thermal conductivity of
4.19 W/mK and 4.99 W/mK was compared to results
of laboratory thermal conductivity testing carried out
on the same soil as shown in Figure 5. Barry-Macaulay
(2013) carried out laboratory thermal conductivity
testing on the clayey sand and sand material excavated
during energy pile installation. He used a commercially manufactured thermal needle probe which is
also based on infinite line source theory to measure
the thermal conductivity.
Figure 5 showed that thermal conductivity for both
the Brighton Group clayey sand and sand increased
with increase in soil density and moisture content. The
in-situ standard penetration testing undertaken during
borehole drilling encountered N values greater than 50
throughout the installed depth of the pile, indicating a
very dense material surrounding the energy pile. The
in-situ soil moisture content of the Brighton Group
materials was also investigated and the results showed
the moisture content range from 10% to 20% at the
upper surface (clayey sand) and uniformly decreased
Figure 5. Thermal conductivity vs. dry density at different moisture contents: (a) Brighton Group clayey sand;
(b) Brighton Group sand (after Barry-Macaulay, 2013).
to about 2% to 3% at the base of the energy pile

(sand).
The results of in-situ soil density and moisture
content and Figure 5 indicates a thermal conductivity of between 2 W/mK and 3W/mK was achievable in Brighton Group materials. However, the
estimated effective thermal conductivity utilising an
energy pile were greater than 4 W/mK, considerably
higher than laboratory results. This may be influenced by laboratory testing unable to represent the
field in-situ conditions accurately where the laboratory soil samples were not able to be compacted
to the same degree of density as in the field. Thus,
in-situ field thermal conductivity testing achieved a
higher result. Other uncertainties that may also influence estimating thermal conductivity of the ground
systems include the arrangement of the absorber
pipes and the material immediately surrounding the
absorber pipes.
Brandl (2006) indicated that the arrangement of the
absorber pipes within an energy pile play a significant role in transferring energy between the ground
heat exchanger and the surrounding ground as well as
determining the efficiency of the GSHP system. For
an energy pile of at least 0.6 m in diameter equipped
with 3 U-loops of absorber pipes, the spacing between
217
the absorber pipes would be up to 200 mm. However,

for a borehole heat exchanger with a diameter of about
150 mm, on which estimation of the effective thermal
conductivity model was based on, the spacing between
the absorber pipes would be less than 100 mm. The
line source model currently utilised during TRTs is
best represented by the size borehole heat exchanger
compared to an energy pile, where the heat source
generated from a borehole performs more closely to
an infinite line source than an energy pile. Thus, this
method of estimating thermal properties utilising an
energy pile may introduce some uncertainties.
Further research is recommended on investigating
the suitability of the currently popular line source
model (Eskilson, 1987; Gehlin 2002; Austin, 1998 and
Kavanaugh et al., 2001), and if appropriate, develop a
more adequate method to estimate the effective thermal conductivity of the ground system utilising energy
pile ground heat exchangers.
10
CONCLUSIONS
Three thermal response tests were undertaken on an

energy pile heat exchanger to determine the in-situ
effective thermal conductivity. The thermal response
testing consisted of utilising 1 loop of absorber pipe
within the energy pile for a heating period of 3 days,
then 3 loops (Short Term) for a heating period of 9
days and finally 3 loops (Long Term) for a heating
duration of 52 days. Sufficient rest periods were also
undertaken between the heating periods to allow the
ground to cool and return as closely as possible to its
undisturbed initial temperatures. The effective thermal
conductivity was found to range between 3.75 W/mK
(3 loop LT) and 4.99 W/mK (3 loop ST). The results
are high compared to published thermal conductivity results undertaken for borehole heat exchangers
and laboratory thermal conductivity tests carried out
on the same soil the energy pile was installed in.
However, the results shown in this study were consistent with the available published thermal conductivity
data undertaken for energy piles of at least 0.6 m in
diameter equipped with more than 1 loop of absorber
pipe, where it ranged between 3.9 W/mK to 6.9 W/mK.
The variation of the effective thermal conductivity
observed between borehole heat exchanger data, as
reported in literature, and energy pile heat exchangers (including the present tests) is primarily due to the
placement of the absorber pipes (number of loops and
spacing between the pipework), the dimensions of the
heat exchanger (diameter and depth), and the backfill materials (thermal grout/sand for boreholes and
concrete for energy piles).
To accurately measure the insitu thermal conductivity of the ground, the authors recommend further
research on assessing the suitability of the current
method of investigating thermal conductivity whilst
utilising energy piles as the ground heat exchanger,
and possibly develop a more suitable method to
assess the thermal properties whilst utilising energy
piles.
ACKNOWLEDGMENTS
This project was funded by the Victorian Government Sustainability Fund, Golder Associates Pty.
Ltd., Vibropile Pty. Ltd., Geoexchange Australia Pty.
Ltd. and Genesis Now. Their support is gratefully
acknowledged.
REFERENCES
Austin, W, A. 1998. Development of an in situ system for
measuring ground thermal properties. M.Sc. thesis,
Oklahoma State University, USA.
Brandl, H. 2006. Energy foundations and other thermo-active
ground structures. Geotechnique, 56(2): 79122.
Brandl, H., Adam, D., and Markiewicz, R. 2006. Groundsourced energy wells for heating and cooling of buildings.
ACTA Geotechnica Slovenia, 1: 527.
Brettmann, T., and Amis, T. 2011. Thermal conductivity evaluation of a pile group using geothermal energy piles. Proc.
ASCE Geo-Frontiers 2011 conference, Dallas, TX, USA.
Barry-Macaulay, D. 2013. An investigation on the thermal
and thermo-mechanical behaviour of Melbourne Soils.
M.Sc. thesis, Monash Univesity, Australia.
Eskilson, P., 1987. Thermal Analysis of Heat Extraction
Boreholes. Ph.D. thesis, University of Lund, Sweden.
Farouki, O, T. 1981 Thermal propserties of soils. U.S. Army
corps of engineers cold region research and engineering
laboratory monograph, Hanover, NH, USA.
Florides. G., and Kalogirou, S. 2008. First in situ determination of the thermal performance of an U-pipe borehole
heat exchanger, in Cyprus. Applied Thermal Engineering,
28: 157163.
Gao, J., Zhang, X., Liu, J., Li, K. S., and Yang, J. 2008a.
Numerical and experimental assessment of thermal performance of vertical energy piles: An application. Applied
Energy, 85: 901910.
Gao, J., Zhang, X., Liu, J., Li, K. S., and Yang, J. 2008b.
Thermal performance and ground temperature of vertical
pile-foundation heat exchangers: A case study. Applied
Thermal Engineering, 28: 22952304.
Gehlin, S. 2002. Thermal response test: method development and evaluation. Ph.D. thesis, Lulea University of
Technology, Sweden.
Hellstrom, G. 1991. Ground Heat Storage. Ph.D. thesis,
University of Lund, Sweden.
Ingersoll, L, R., and Plass, H, J. 1948. Theory of the ground
pile heat source for the heat pump. ASHVE transactions,
54: 339348.
Jalaluddin., and Miyara, A. 2012. Thermal performance
investigation of several types of vertical ground heat
exchangers with different operation mode. Applied thermal engineering, 3334: 167174.
Jalaluddin., Miyara, A., Tsubaki, K., Inoue, S., and Yoshida,
K. 2011. Experimental study of several types of ground
heat exchanger using a steel pile foundation. Renewable
energy, 36: 764771.
Kavanaugh, S. P., Xie, L., and Martin, C. 2001. Investigation of methods for determining soil and rock formation
thermal properties from short-term field tests. ASHRAE
1118-TRP.
Koene, F., and Geelen C. 2000. Energy piles as an efficient
way to store heat. CADDET Energy Efficiency, 46.
Lennon, D. J., Watt, E., and Suckling, T. P. 2009. Energy piles
in Scotland. Deep foundations on bored and auger piles,
Taylor and Francis Group, London, UK.
218
Lim, K., Lee, S., and Lee, C. 2007. An experimental study on the thermal performance of ground heat
exchanger. Experimental Thermal and Fluid Science, 31:
985990.
Lund, J., Sanner, B., Rybach, L., Curtis, R., and Hellstrom, G.
2004. Geothermal ground source heat pumps: A world
overview. Geo-Heat Center (GHC) Bulletin (Klamath
Falls, Oregon, USA), November 2004, 110.
Ma, X., and Grabe, J. 2010. Field test of a geothermal system in HafenCity Hamburg. ASCE geotechnical special
publication No. 204, GeoShanghai, 2010, 159166.
Mattsson, N., Steinmann, G., and Laloui, L. 2008. Advanced
compact device for the in situ determination of geothermal
characteristics of soils. Energy and Buildings, 40: 1344
1352.
Mogensen, P. 1983. Fluid to duct wall heat transfer in duct
system heat storages. Proc. Int. Conf. Subsurface heat
storage in theory and practice, Stockholm, Sweden.
652657.
Pahud, D., and Matthey, B. 2001. Comparison of the thermal

performance of double U-pipe borehole heat exchangers
measured in situ. Energy and Buildings, 33, 503507.
Poppei, J., Schwarz, R., Mattsson, N., Laloui, L., Wagner, R.,
and Rohner, E. 2006. Innovative improvements of thermal
response test. Rep. No. 101690, Swiss Federal Office of
Energy (BFE), Switzerland.
Sanner, B., Mands, E., Sauer, M, K., and Grundmann, E.
2008. Thermal response test, a routine method to determine thermal ground properties for GSHP design. Proc.,
9th International IEA Heat Pump Conference, Zrich,
Switzerland.
Sarbu, I., and Bura, H. 2011. Thermal tests on borehole heat
exchangers for ground-coupled heat pump systems. Int. J.
Energy and Environment, 3(5): 385393.
Witte, H, J, L., van Gelder, G, J., and Spitler, J, D. 2001. In situ
measurement of ground thermal conductivity: A Dutch
perspective. ASHRAE Transactions, 108(1): 263272.
219

Modification of clay consolidation properties using microbial

gas production
A.M. Puzrin
ETH Zurich, Switzerland
Two main sources of protracted primary consolidation of clayey soils are the low permeability of such
soils and the incompressibility of water in the voids.
This presentation describes a pilot research carried
at the ETH Zurich to examine the use of microbial
gas production to engineer the consolidation properties of clayey soils by changing the settlement time
profile (Puzrin et al., 2011). The research focused
on increasing the compressibility of the pore-fluid,
by introducing gas bubbles via in situ microbial gas
production. The major effect of these occluded gas
bubbles is that they allow the material to have a higher
portion of settlement as immediate settlement.
The presented feasibility study provided simple
analytical tools allowing for the above effects to be
quantified and establishes that by properly adjusting
the rate of gas generation to permeability of soil, the
proposed technique of in situ microbial gas production takes place without excessive swelling of soil and
can shift about 50% of total settlement from primary
consolidation settlement to immediate settlement.
For experimental validation of the suggested technique, Clostridia acetobutylicum, an anaerobic fermentative microorganism was mixed into clayey-sand
samples, fed with a nutrient-rich media and allowed
to produce gas in situ. A series of standard oedometer
tests were performed on gas-containing and gas-free
soil specimens. Results showed that in agreement with
the model predictions, in situ gas production caused

negligible swelling of the samples, while shifting a
significant portion (up to 50%) of total displacement to immediate settlement without changing total
displacements and the coefficient of consolidation.
This work introduced well-characterized fermentative microorganisms into the experimental system,
but it is projected that future applications can harness
naturally occurring organisms and readily available
substrates to achieve the same end point.
In soils of low permeability the stage of gas production and partial displacement of pore water by
the gas bubbles may require significant periods of
time. The main potential benefit of the proposed technique is, therefore, likely to be in reducing the time of
preloading or eliminating the necessity of preloading
completely, rather than in reducing of the total time
period necessary for preparing the site for the future
loading. More work, both theoretical and experimental, is required to justify applicability of the proposed
method in the field.
REFERENCE
Puzrin, A.M., Tront, J., Schmid, A., Hughes, J.B. (2011),
Engineered Use of Microbial Gas Production to Decrease
Primary Consolidation Settlement in Clayey Soils,
Geotechnique, Vol. 61, No. 9, pp. 785794.
221

Injection of zero-valent iron micro- and nano-particles for groundwater

remediation: Laboratory tests and transport modelling
T. Tosco, F. Gastone & R. Sethi
Politecnico di Torino DIATI (Dipartimento di Ingegneria dellAmbiente, del Territorio e delle Infrastrutture),
corso Duca degli Abruzzi, Torino, Italy
ABSTRACT: Microscale iron particles (MZVI) have been recently studied for application in groundwater
remediation. Thanks to their small size, they can be directly injected into the subsurface for a targeted treatment
of contamination. The present work reports laboratory tests and numerical modelling concerning the mobility of
MZVI suspensions in porous media. The efficacy of shear thinning green fluids in improving colloidal stability
and mobility of MZVI was proved with column transport tests, injecting highly concentrated iron suspensions
(20 g/L) dispersed in xanthan gum (3 g/L) and guar gum (1.54 g/l). A numerical model for MZVI transport
was then developed (E-MNM1D, Enhanced Micro- and Nanoparticle transport Model in porous media in 1D
geometry). It is designed as a tool for inverse modelling of laboratory transport tests, and as a support in the
design of field-scale applications of MZVI and NZVI-based remediation.
INTRODUCTION
The efficacy of zero-valent iron (ZVI) for the remediation of polluted groundwater is well known, and has
been applied in the last twenty years for the realization
of permeable reactive barriers (PRBs). The concept of
PRB is relatively simple: granular reactive material
(usually ZVI with particle size in the order of 0.2
5 mm) is placed in the subsurface into a trench, to intercept a contaminated plume that moves through it under
natural gradient (Fig. 1a). As the groundwater passes
through the PRB, the contaminants are degraded, precipitated or adsorbed by the reactive media in the PRB.
Physical and chemical processes that occur inside the
barrier transform the contaminants to less harmful or
immobile species (Orth & Gillham 1996, Gillham &
Ohannesin 1994, Di Molfetta & Sethi 2006). Nowadays the application of zero-valent iron in permeable
reactive barriers (PRBs) has become a standard practice for the remediation of contaminated aquifer by
a wide variety of contaminants (Blowes et al. 1995,
OHannesin & Gillham 1992).
Although the PRB is a successful technology, the
granular sizes of iron (0.25 mm) can often restrict its
applications. Injectable Fe-based materials (nano- and
micro-sized zero-valent iron particles, named NZVI
and MZVI) can be suspended in a slurry and injected
directly into the source of contamination, overcoming
most of the limitations of zerovalent iron permeable
reactive barriers (PRBs) (Di Molfetta & Sethi 2006).
Furthermore, thanks to the reduced particle size, and
thus to the extremely high specific surface area, MZVI
and NZVI exhibit an improved reactivity compared to
granular iron (Fig. 1b).
Critical points for successfully full-scale applications are stability against aggregation and sedimentation, mobility in subsurface environments, and
longevity in the subsurface. Iron particles should
remain in suspension for a time sufficient for slurry
preparation, handling and injection in the subsurface.
Also, they should have a sufficient mobility in the subsurface to be transported for some extent around the
injection point. However, several studies have shown
MZVI and NZVI to be scarcely mobile and stable in
both laboratory studies and field-scale tests, due to fast
sedimentation (MZVI) or aggregation and subsequent
sedimentation (NZVI).
Enhanced transport of engineered colloids for
groundwater remediation is a major topic extensively
addressed in recent years (Zhang 2003, Tiraferri &
Sethi 2009, Tiraferri et al. 2008). Coating of the iron
nanoparticles with hydrophilic polymers and increasing the viscosity of the MZVI and NZVI slurry were
found successful approaches for improving both colloidal stability and mobility in lab-scale experiments
(Tiraferri et al. 2008, Tiraferri & Sethi 2009, Dalla
Vecchia et al. 2009). Surface modification of zerovalent iron particles with partial coating of other
metals was also studied as an alternative approach
(Zhang 2003, Hosseini & Tosco 2013). Nevertheless,
if this approach can be successful when applied to
nanosized particles, providing a surface stabilization
cannot be sufficient when working with larger colloids, like microsized iron, that are in the size range of
most favourable transport in groundwater, but exhibit a
strongly limited mobility due to gravitational sedimentation (Dalla Vecchia et al. 2009). In this case, xanthan
gels proved to be able to prevent sedimentation of both
223
iron particles, in particular during their injection in

the subsurface, and in the early stages of migration through the porous matrix, namely E-MNM1D
(Tosco & Sethi 2010). Numerical modelling of transport processes of iron particles transport in porous
media, is necessary for the interpretation of laboratory tests, and for the simulation of injection at
the field scale. E-MNM1D couples a modified dualphase advection/dispersion differential equation with
clogging phenomena (i.e. reduction of porosity and
permeability due to particles deposition), and the rheological properties of the carrier fluid (in this project,
guar gum solution) for a correct estimate of pressure
drops. This model extends the MNM1D code previously developed by the authors for the description
of colloid transport under transients of ionic strength
(Tosco et al. 2009, Tosco & Sethi 2009). Although
here implemented for a one-dimensional geometry,
the model is intended to be used as a tool in the
development of an efficient injection technology for
field-scale applications of nano- and microscale iron
suspensions.
2
Figure 1. Schematic representation of contaminated plume
treatment using a permeable reactive barrier (a) and the
injection of NZVI and MZVI slurries at the sources of
contamination (b).
micro- and nanoscale iron particles for several hours

(Comba & Sethi 2009, Dalla Vecchia et al. 2009).
Polymer chains of xanthan adsorb on particle surface,
providing steric stabilization as the additives discussed
so far. Moreover, xanthan solutions exhibit a nonNewtonian shear-thinning behaviour, that facilitates
injection and delivery of the colloidal suspension in
the subsurface. Shear thinning fluids are characterized
by a high viscosity at zero shear rate, that decreases
with increasing shear rate till to the value of pure
water. Consequently, xanthan solutions can enhance
stability of the iron dispersion when the product is
stored (as high viscosity prevent gravitational settling
and aggregation of the particles) without hindering the
injection.
In this work the mechanisms controlling the transport of colloidal suspensions in saturated porous
media, the improvement of stability of highly concentrated micro- and nanosized iron suspensions, and
the assessment of their transportability via addiction
of green polymers are studied. Laboratory transport tests in one-dimensional columns are reported,
exploring the effects of both flow rate and polymer
concentration (in this study, guar gum) on particles
mobility. The tests were performed injecting MZVI
suspensions at a concentration of 20 g/L, amended
with xanthan gum (3 g/l) or guar gum (1.5 to 4 g/l),
in 0.4 m long columns.
A numerical model is also presented that describes
the transport of highly concentrated suspensions of
Iron particles provided as dry powder (HQ from

BASF, d10 , d50 , d90 equal to 0.7, 1.2, 1.8 m, respectively) were dispersed in xanthan gum (Jungbunzlauer,
Switzerland) and in guar gum (RANTEC HV7000
from Ranchester, United States) solutions and used for
column transport tests. For sedimentation tests, also
larger particles were used, namely BASF MS200 (d50
equal to 4.8 m).
The protocol for the preparation of the iron suspensions included the following steps: the iron particles
were dispersed in water, sonicated for 5 minutes and
mixed with the guar gum solution using a high speed
rotor-stator (UltraTurrax). The guar gum solution was
prepared 24 h prior to use, to ensure complete hydration of the polymer powder. The dispersion of the iron
particles into the polymer solution was carried out
immediately prior to each experiment, to avoid settling
of the particles (Dalla Vecchia et al. 2009).
Sedimentation tests were performed to evaluate the
colloidal stability of iron particles dispersed in polymeric solutions. For each test, the iron slurry was put
in an elongated cuvette and the evolution over time of
the iron concentration was continuously monitored at
a fixed height measuring the magnetic susceptibility
of the suspension, and thus following the sedimentation in time. The sedimentation rate was quantified in
terms of sedimentation half time (t50 ), corresponding
to the time required to reduce the initial concentration
to 50%.
Column transport tests of HQ particles dispersed
in xanthan gum followed the experimental procedure
described in details by Dalla Vecchia et al (2009).
Iron suspensions were injected in horizontal sandpacked saturated columns (0.46 m long, inner diameter
of 2.5 cm), with an average porosity of 0.49, average
224
Figure 2. Column setup for iron transport experiments, and measured parameters (iron concentration and pressure, solid line).
hydraulic conductivity of 5.7 103 m/s, and average

darcyan flow rate of 1.49 103 m/s. Particles were
injected at a concentration of 20 g/l dispersed in 3 g/l
xanthan gum solutions. Dispersivity determined using
the approximation x d50 , proposed for laboratoryscale dispersion phenomena (Bear 1988), resulted in
an average dispersion coefficient of 2.1 106 m2 /s.
Column tests for HQ particles dispersed in guar
gum were performed in 42 cm long sand-packed (Dorsilit 8) columns with 2.4 cm inner diameter, average
porosity of 0.37, at constant flow rate (2.8 104 m/s,
1.2 103 m/s, 5.3 103 m/s). Particles were injected
at a concentration of 20 g/l dispersed in 4, 3 or
1.5 g/l guar gum solutions. Three values of flow
rate (2.8 104 m/s, 1.2 103 m/s, 5.3 103 m/s) and
guar gum concentration (1.5 g/l, 3 g/l and 4 g/l) were
applied. Inline continuous measurements of inlet and
outlet iron concentration, as well as profiles of total
iron concentration along the column were performed
using susceptibility sensors (Bartington Instruments,
UK), as reported by Dalla Vecchia et al. (2009) (Fig. 2).
Also, pressure drop at column ends was monitored
during the whole test.
For all tests, the injection protocol included:
pre-conditioning with DI water;

injection of iron suspensions (7 or 26 PVs for
xanthan gum tests, 10 PVs for guar gum tests);
flushing with DI water.
3
RESULTS OF SEDIMENTATION AND

TRANSPORT TESTS
Stability against sedimentation was first tested for

the particles HQ and MS200, using a guar gum
Figure 3. Sedimentation tests results for 2 particle samples

(HQ and MS200) in a guar gum solution at concentration of
3 g/l (a), and for HQ particles (20 g/l) in guar gum solutions
at different polymer concentrations (4, 3 and 1.5 g/l).
concentration of 3 g/l. Results are reported in Figure 3a.The sedimentation half time is larger for smaller
particles, shorter for larger particles, being equal to
5.92 h for HQ and 4.77 h for MS200. Sedimentation
225
Figure 4. Viscosity as a function of shear rate for guar gum

solutions with polymer concentration equal to 1.5, 3 and 4 g/l.
curves for HQ particles in guar gum solutions at different polymer concentrations are reported in Figure
3b, showing that increasing guar gum concentration
the colloidal stability of the suspension increases. This
is consistent with the increased viscosity of the guar
gum dispersions with increasing guar gum concentration, which exhibited clear non-Newtonian rheological
properties, well modeled by the shear thinning Cross
model (Fig. 4).
Figure 5 reports an example of results of the
transport tests for HQ particles dispersed in guar
gum performed at the maximum explored flow rate
(v = 5.3 103 m/s) for three guar gum concentrations
(1.5, 3 and 4 g/l). Results include breakthrough curves
and pressure drop at column ends. It is evident that
increasing guar gum concentration the breakthrough of
HQ particles is improved, and the pressure increases.
The direct dependence of the mass of iron retained
inside the column as a function of guar gum concentration, expressed as the ratio of the retained mass to the
injected mass of particles is reported in Figure 6a. Iron
particles are more retained in the column for low guar
gum concentrations and lower flow rates. The pressure
increase at the beginning of the iron slurry injection,
due to the higher viscosity of guar gum if compared
to water, (almost) linearly increases when increasing guar gum concentration (Fig. 6b), coherently with
rheological curves (Fig. 4) and Darcy law.
IRON TRANSPORT MODELLING
The transport of iron micro- and nanoparticles in

saturate porous media, particularly if a viscositymodifier polymer is employed, cannot be described
only with advection-dispersion-deposition equations.
These equations are usually applied for many colloidal
systems, like latex particles, natural colloids, and bacteria, when the low particle concentration do not have
any effect on flow nor on medium properties (Tosco
et al. 2009, Tufenkji & Elimelech 2005, Bradford
et al. 2002). Conversely, for concentrated iron colloids,
Figure 5. (a) Breakthrough curves and (b) evolution of

pressure drop at column ends over time for HQ particles
dispersed in guar gum at three different concentrations
(1.5, 3 and 4 g/l) injected in sand packed columns at a flow
rate v = 5.3 103 m/s.
changes in pore velocity, viscosity, density and porosity must be accounted for. Flow and transport are then
coupled problems.
In the literature, clogging phenomena are usually
simulated with a filter-averaged approach, mostly for
waste water treatment applications (Mays & Hunt
2005), or with lattice Boltzmann approaches for 3D
simulations at the pore scale (Chen et al. 2009).
However, none of these is suitable for laboratory or
field-scale simulations. Studies are also available that
account for changes in pore space geometry due to
mineral precipitation (Emmanuel & Berkowitz 2005)
or microbial growth (Kildsgaard & Engesgaard 2001),
but none of these models can simulate the nonNewtonian nature of the carrier fluid, nor the influence
of the concentration of suspended colloids on the fluid
properties.
A novel approach for modelling transport of highly
concentrated shear-thinning slurries of NZVI and
MZVI was then developed by the authors (Tosco &
Sethi 2010). Colloid transport is modelled with a
dual-site (physico-chemical interactions plus straining) advection-dispersion-deposition equation. A general formulation for attachment/detachment dynamics
is adopted. The influence of colloid transport on
226
The transport equations for the iron particles were

modelled with a physico-chemical (interactions plus
straining) advection-dispersion-deposition equation:
Figure 6. Dependence of (a) fraction of particles retained

inside the column after injection and (b) pressure increase
at the beginning of iron + guar gum slurry injection both
reported as functions of the guar gum concentration, for
different flow rates.
porosity, permeability, and fluid viscosity is explicitly embedded into the model through correlations
from the literature, or derived on purpose. The shearthinning behaviour of the iron slurries is described with
a modified Darcys law. Coupling of flow and transport is solved under the hypothesis that clogging of the
porous medium is not a repent process, thus handling
the problem as a quasi-stationary phenomenon. The
proposed formulation relies on the hypotheses of 1D
horizontal quasi-stationary flow, constant discharge,
negligible compressibility of particles, porous matrix,
pore fluid and particles deposits.
4.1 Advection-dispersion-deposition equation

When transport of highly concentrated colloid suspensions is considered, the void space of the clean
medium (characterized by a porosity n) is occupied
by (i) the immobile deposed colloids s, with a volume
fraction s , (ii) the mobile suspended colloids c, with
a volume fraction c , and (iii) the pore fluid f, with a
volume fraction f = n c s . When a distinction
between suspended colloids and carrier fluid is not
required, the sum of the two can be identified as the
mobile phase, with m = f c .
where m is the porosity, D is the hydrodynamic

dispersion, qm is the darcyan velocity, c is the concentration of suspended particles, s is the concentration
of deposited particles (expressed as mass of deposited
particles per unit mass of the porous medium), ka,i ,
kd,i , i (i = 1, 2), and A1 , are the kinetic parameters of
colloid deposition and release kinetics. The first equation represents the mass balance for the liquid phase,
the second and third ones the mass balance for the
solid phase, following, respectively, a blocking and a
straining dynamics.
The form for attachment/detachment dynamics was
formulated on purpose, that can be adapted to all
commonly used interaction kinetics, namely linear
attachment, blocking, ripening dynamics (Tosco &
Sethi 2010).
4.2 Darcys law for Newtonian and non-Newtonian
fluids
When modelling flow of non-Newtonian fluids in
porous media, the usual form of Darcys law can be
applied, and all non-Newtonian effects can be modelled using a proper viscosity parameter (m ), function
of both porous medium characteristics and fluid properties, called porous medium viscosity (Pearson &
Tardy 2002, Lopez et al. 2003). Consequently, under
the hypothesis of constant darcyan velocity, the pressure drop p at the ends of a column can be directly
obtained from the Darcys law.
The apparent viscosity is a function of both porous

medium characteristics (permeability K and porosity
m ) and fluid properties (via a modified Cross model,
see below).
4.3
State equations and hydrodynamic parameters
The following parameters were considered functions

of the concentration of deposed or suspended particles,
and/or of polymer (guar gum) concentration.
Porosity available for fluid flow. The pore space
available for the fluid flow m can be related to the
227
concentration of deposed particles s: increasing concentration of deposed particles, the medium porosity
decreases:
Being n the initial porosity (i.e. before the injection

of the particles), b the bulk density of the porous
medium, and s the bulk density of deposited particles.
Viscosity of the pore fluid. Dynamic viscosity of
polymeric solutions is known to be a function of shear
rate, polymer concentration and particles concentration. The dependence on shear rate is described by a
modified Cross model (Tosco & Sethi 2010):
where
M(c) is a function that can be derived from experimental data, and depends only on the concentration of
suspended particles, m (c) and m (c) are two parameters of the Cross model, both functions of c, and m
is the shear rate of the fluid in the porous medium.
Permeability coefficient. Following Kozeny, the
permeability of a porous medium depends on the third
power of porosity, and on the square of the specific
surface area of the matrix, through a constant. When
colloids depose on the soil grains, the pore space available for the fluid flow decreases, and the specific
surface area increases, thanks to a further contribution
of deposed particles. Therefore, both contributions
were considered and the permeability coefficient was
obtained by their combined effects:
where K0 is the clean bed permeability (corresponding

to porosity n and specific surface area a0 ), is the
fraction of the surface area of the deposited particles
ap that contributes to the increase of the surface area
of the matrix, and p is the density of the particles.
4.4
Numerical solution
Coupling of flow and transport was solved under the

hypothesis that clogging of the porous medium is
not a repent process, thus handling the problem as
Figure 7. Breakthrough curves (a) and pressure drop at column ends (b) for six transport tests with microiron dispersed
in xanthan solution. Experimental data (grey) and model fitted curves (black) are reported for each test. The tests include
a short-lasting injection (labeled Short inj) of 7 pore volumes of slurry and a long-lasting injection (labeled Long
inj) of 26 pore volumes.
a quasi-stationary phenomenon. The proposed formulation relies on the hypotheses of 1D horizontal

quasi-stationary flow, constant discharge, negligible
compressibility of particles, porous matrix, pore fluid
and particles deposits.
The set of model equations is implemented in a
finite-differences code, that represents the extension of
the MNM1D model, originally designed for the simulation of colloid transport under transients in ionic
strength (Tosco et al. 2009, Tosco & Sethi 2009). The
system is solved iteratively using a Picards iteration
scheme.
An example of the model results is provided in
Figure 7, which shows a breakthrough curve of iron
concentration for an injection of iron HQ particles dispersed in a xanthan gum (3 g/l), followed by a flushing
with particle- and polymer-free water. The experimental data are well reproduced by the inverse-fitting
provided by E-MNM1D.
5
CONCLUSIONS AND PERSPECTIVES
The exploitation of the magnetic properties of

microparticles injected in the column to measure
the breakthrough curve and the amount of material blocked along the length of the column proved
extremely effective in our column experiments. This
method allowed to overcome the limitations (in terms
228
of concentration of the slurry and of continuity of

the measurement) that affect traditional breakthrough
curve acquisition methods, i.e. spectrometric analysis of the elute (Kanel et al. 2007, Saleh et al. 2007,
Sirk et al. 2009). To our knowledge, this is the first
column transport study using iron dispersions with
concentrations as high as 20 g/L and able to provide
continuous and reliable eluted concentration measurements, concentrations used in previous studies ranging
from 0.1 g/L to 6 g/L (Cantrell et al. 1997, Kanel
et al. 2008, Saleh et al. 2007, Tiraferri & Sethi 2008,
Kanel & Choi 2007, Phenrat et al. 2009). The efficacy of the magnetic susceptimeter in our bench scale
experiments and its being originally designed for geological and geophysical investigations on the field,
suggest that measuring susceptivity to keep track of
ZVI particles could be feasible also at a larger scale. In
this sense, it could also be considered as a monitoring
technique for in field applications during groundwater remediation. Magnetic sensors could be installed
in monitoring wells at increasing distances from the
injection points to test the spatial distribution of the
iron slurry.
Before this study, no model was available in the
literature for the simulation of iron transport under
the conditions described in the experimental section:
although solutions exist for the simulation of clogging
in deep bed filtration, or changes in pore space geometry due to mineral precipitation, none of these models
can simulate the non-Newtonian nature of the carrier
fluid, nor the influence of the concentration of suspended colloids on the fluid properties. The numerical
model developed in this work, although implemented
for the simulation of 1D laboratory column tests,
could be extended to more complex geometries, thus
becoming a useful tool for the design of the injection
and early stages of migration of iron slurries in field
applications.
As well as suggesting an innovative and valuable
method for measuring indirectly ZVI concentration in
suspension and present within sand packed columns,
in this study we show evidence that 20 g/L ZVI suspensions can be eluted through a porous medium with
outstanding yields, if amended with xanthan gum or
guar gum.
The effectiveness of the polymers relies on their
capability of enhancing suspension stability and hindering particle aggregation. The mechanism of stabilization is kinetic rather than thermodynamic in
nature, meaning that not only it reduces interaction potentials or avoids their expression, as in the
case of electrosteric stabilization, but it also reduces
motion of the particles by trapping them within the
network of polymer chains, thus preventing aggregation. Moreover the high viscosity of the polymer
solution opposes higher drag forces to the particles
motion, which would otherwise lead to aggregation
and sedimentation (Comba & Sethi 2009). The shear
thinning behavior further improves the capability of
the considered biopolymers to act as a carrier of colloidal particles in porous media: particles can remain
suspended as a consequence of the high viscosity in

the centre of the pore space (where shear rates are
low). Furthermore, the dispersion can easily flow in
the porous media since viscosity is low near the pore
walls where high shear rates are experienced (Comba
& Sethi 2009, Cantrell et al. 1997, Oostrom et al.
2007).
In terms of field application, obtaining such a high
yield is extremely promising, since successful transport of highly concentrated iron dispersions is an
essential prerequisite for real scale remediation interventions on site. In order to minimize the volumes
of material injected in the subsurface, the concentration of required reactive material to address the
contamination can rise very easily and is likely to
be of the order of several tens of g/L (Varadhi et al.
2005). Highly concentrated slurries are preferable
mainly due to the risk of contaminant displacement
and iron passivation which may occur if large volumes
of dilute suspensions are employed (Gavaskar et al.
2005). Additionally, higher solid loadings are likely to
reduce transport costs from the synthesis facility to the
contaminated site.
ACKNOWLEDGEMENTS
The work was partly funded by the European Union
project AQUAREHAB (FP7 Grant Agreement
Nr. 226565).
REFERENCES
Bear, J. 1988. Dynamics of fluids in porous media. New York:
Dover.
Blowes, D. W., Ptacek, C. J., Cherry, J. A., Gillham, R. W. &
Robertson, W. D. (1995) Passive remediation of groundwater using in situ treatment curtains. 46 /2 ed.
Bradford, S. A.,Yates, S. R., Bettahar, M. & Simunek, J. 2002.
Physical factors affecting the transport and fate of colloids
in saturated porous media. Water Resources Research 38:
Doi 10.1029/2002wr001340.
Cantrell, K. J., Kaplan, D. I. & Gilmore, T. J. 1997. Injection of colloidal Fe-0 particles in sand with shear-thinning
fluids. Journal of Environmental Engineering-Asce 123:
786791.
Chen, C., Lau, B. L. T., Gaillard, J. F. & Packman, A. I.
2009.Temporal evolution of pore geometry, fluid flow, and
solute transport resulting from colloid deposition. Water
Resources Research 45: Doi 10.1029/2008wr007252.
Comba, S. & Sethi, R. 2009. Stabilization of highly
concentrated suspensions of iron nanoparticles using
shear-thinning gels of xanthan gum. Water Research 43:
37173726.
Dalla Vecchia, E., Luna, M. & Sethi, R. 2009. Transport
in Porous Media of Highly Concentrated Iron Microand Nanoparticles in the Presence of Xanthan Gum.
Environmental Science & Technology 43: 89428947.
Di Molfetta, A. & Sethi, R. 2006. Clamshell excavation of a
permeable reactive barrier. Environmental Geology.
Emmanuel, S. & Berkowitz, B. 2005. Mixing-induced precipitation and porosity evolution in porous media. Advances
in Water Resources 28: 337344.
229
Gavaskar, A., Tatar, L. & Condit, W. (2005) Cost and performance report nanoscale zero-valent iron technologies
for source remediation. Port Hueneme, California, Naval
Facilities Engineering Command.
Gillham, R. W. & Ohannesin, S. F. 1994. Enhanced degradation of halogenated aliphatics by zero valent iron. Ground
Water 32: 958967.
Hosseini, S. M. & Tosco, T. 2013. Transport and retention of high concentrated nano-Fe/Cu particles through
highly flow-rated packed sand column.Water Research 47:
326338.
Kanel, S. R. & Choi, H. 2007. Transport characteristics of
surface-modified nanoscale zero-valent iron in porous
media. Water Science & Technology 55: 157162.
Kanel, S. R., Goswami, R. R., Clement, T. P., Barnett, M. O. &
Zhao, D. 2008. Two dimensional transport characteristics of surface stabilized zero-valent iron nanoparticles in
porous media. Environmental Science & Technology 42:
896900.
Kanel, S. R., Nepal, D., Manning, B. & Choi, H. 2007.
Transport of surface-modified iron nanoparticle in porous
media and application to arsenic(III) remediation. Journal
of Nanoparticle Research 9: 725735.
Kildsgaard, J. & Engesgaard, P. 2001. Numerical analysis of
biological clogging in two-dimensional sand box experiments. Journal of Contaminant Hydrology 50: 261285.
Lopez, X., Valvatne, P. H. & Blunt, M. J. 2003. Predictive
network modeling of single-phase non-Newtonian flow in
porous media. Journal of Colloid and Interface Science
264: 256265.
Mays, D. C. & Hunt, J. R. 2005. Hydrodynamic aspects of particle clogging in porous media. Environmental Science &
Technology 39: 577584.
Ohannesin, S. F. & Gillham, R. W. 1992. Permeable reaction
wall for in situ degradation of halogenated organic compounds. Canadian Geotechnical Conference: 69/169/5.
Oostrom, M., Wietsma, T. W., Covert, M. A. & Vermeul, V. R.
2007. Zero-valent iron emplacement in permeable porous
media using polymer additions. Ground Water Monitoring
and Remediation 27: 122130.
Orth, W. S. & Gillham, R. W. 1996. Dechlorination of
trichloroethene in aqueous solution using Fe0. Environmental Science and Technology 30: 6671.
Pearson, J. R. A. & Tardy, P. M. J. 2002. Models for flow of
non-Newtonian and complex fluids through porous media.
Journal of Non-Newtonian Fluid Mechanics 102: 447
473.
Phenrat, T., Kim, H. J., Fagerlund, F., Illangasekare, T., Tilton,
R. D. & Lowry, G. V. 2009. Particle size distribution,
concentration, and magnetic attraction affect transport

of polymer-modified Fe0 nanoparticles in sand columns.
Environmental Science & Technology In Press.
Saleh, N., Sirk, K., Liu, Y. Q., Phenrat, T., Dufour, B.,
Matyjaszewski, K., Tilton, R. D. & Lowry, G. V. 2007.
Surface modifications enhance nanoiron transport and
NAPL targeting in saturated porous media. Environmental
Engineering Science 24: 4557.
Sirk, K., Saleh, N., Phenrat, T., Kim, H. J., Dufour, B.,
Ok, J., Golas, P. L., Matijaszewski, K., Lowry, G. V. &
Tilton, R. D. 2009. Effect of Adsorbed Polyelectrolytes on
Nanoscale Zero Velent Iron Particle Attachment to Soil
Surface Models. Environmental Science & Technology. In
Press.
Tiraferri, A., Chen, K. L., Sethi, R. & Elimelech, M. 2008.
Reduced aggregation and sedimentation of zero-valent
iron nanoparticles in the presence of guar gum. Journal
of Colloid and Interface Science 324: 7179.
Tiraferri, A. & Sethi, R. 2008. Enhanced transport of zerovalent iron nanoparticles in saturated porous media by guar
gum. Journal of Nanoparticle Research: Submitted.
Tiraferri, A. & Sethi, R. 2009. Enhanced transport of zerovalent iron nanoparticles in saturated porous media by guar
gum. Journal of Nanoparticle Research 11: 635645.
Tosco, T. & Sethi, R. 2009. MNM1D: a numerical code for
colloid transport in porous media: implementation and
validation. American Journal of Environmental Sciences
5: 517525.
Tosco, T. & Sethi, R. 2010. Transport of non-newtonian suspensions of highly concentrated micro- and nanoscale
iron particles in porous media: A modeling approach.
Environmental Science and Technology 44: 90629068.
Tosco, T., Tiraferri, A. & Sethi, R. 2009. Ionic Strength
Dependent Transport of Microparticles in Saturated
Porous Media: Modeling Mobilization and Immobilization Phenomena under Transient Chemical Conditions.
Environmental Science & Technology 43: 44254431.
Tufenkji, N. & Elimelech, M. 2005. Breakdown of colloid
filtration theory: Role of the secondary energy minimum
and surface charge heterogeneities. Langmuir 21: 841
852.
Varadhi, S. N., Gill, H., Apoldo, L. J., Liao, K.,
Blackman, R. A. & Wittman, W. K. (2005) Full-scale
Nanoiron Injection for Treatment of Groundwater contaminated with Chlorinated Hydrocarbons. Natural Gas
Technologies 2005 Conference. Orlando, Florida.
Zhang, W. X. 2003. Nanoscale iron particles for environmental remediation:An overview. Journal of Nanoparticle
Research 5: 323332.
230

Modelling gas and water flow in environmental problems

Sebasti Olivella
Department of Geotechnical Engineering and Geosciences, Escola Tecnica Superior dEnginyers de Camins,
Canals i Ports Universidad Politcnica de Catalunya, Barcelona, Spain
ABSTRACT: The general formulation for water and gas flow in porous media is described. This includes
mass, momentum and energy balance. The mechanical models for unsaturated soils are defined in terms of
stress, suction and temperature and permit to obtain deformations caused by variations in these variables. Problems involving multiphase flow under nonisothermal conditions are first described. Afterwards gas flow in
large permeability variation materials is presented. A section is devoted to present the modeling of injection
of CO2 in aquifers. Through these examples, the large versatility of the formulation and computed program
CODE_BRIGHT is demonstrated.
INTRODUCTION
At present, a general approach for nonisothermal multiphase flow and deformation in geological porous
media is applied to a variety of environmental and
geotechnical problems. The formulation is solved
usually by numerical methods as it includes several nonlinear functions. Existing programs have been
developed by combination of existing ones or newly
developed. Numerical solutions using finite elements
are quite standard.
Finite element method was introduced in unsaturated soil geotechnics mainly because the mechanical
part of the problem (Lloret and Alonso, 1980), while in
unsaturated flow and multiphase flow finite volumes
(Pruess, 1987) was most commonly used.
The modeling of water and vapour motion in geoenvironmental problems is described, with special
consideration of the formation of salt crusts.
A field of research which is receiving significant
attention corresponds to the development of formulations including double structure which are required
in some cases for unsaturated soils and especially
for expansive materials (Alonso et al 1999). In this
context, Sanchez et al (2005, 2008) developed a formulation that permits to analyze problems involving soils
which can be idealized as having two pore volumes
undergoing different processes.
The problem of discontinuities and interfaces is
receiving attention for unsaturated soils and for gas
movement in geological materials. In case of environmental problems such us storage of nuclear waste
interfaces between clay barriers elements and host rock
require special treatment. Zandarin et al (2009) have
developed a special element which incorporates an
elastoplastic mechanical response under stress, temperature and suction effects. In addition, transport
properties (permeability, diffusivity, thermal conductivity) specialize depending on the normal and longitudinal directions using aperture variations and step
functions.
In the case of environmental problems it is also
necessary to incorporate geochemical aspects as processes of clay deformations are influenced greatly by
the presence of solutes (Barbour, S.L. and Yang, N.,
1993; Di Maio, C., 1996; Hueckel T., 1997), but little
investigations have been done considering unsaturated
conditions. Saaltink et al (2004) coupled a geochemical formulation with a multiphase flow problem and
solved problems related to evaporation in mine tailings (Acero et al 2008). The work by Guimaraes et al
(2007) coupled reactive transport and multiphase flow
in deformable materials and was applied to swelling
clay barriers subjected to heating and hydration. Actually, swelling and collapse processes show significant
influence of the presence of the solutes, and therefore
the mechanical model requires the incorporation of
the effect of solutes. Liu et al (2005) have developed
a constitutive model for the chemo-hydro-mechanical
response of unsaturated clays. As concrete is also
viewed as porous unsaturated medium some common features with unsaturated geological media are of
interest. Chemo-thermo-hygro-mechanical response
has been investigated and of particular interest is
the work by Gawin et al (2003) which incorporates the thermo-chemical and mechanical degradation
of the porous material into a coupled THMC model.
An emerging related field of research is the sequestration of carbon dioxide in geological formations.
Vilarrasa et al (2010) have investigated the mechanical
response of the rocks when injection at high pressures of carbon dioxide is carried out. The conditions
of high confinement, two phase flow and chemical
interactions require also specific characterization of
231
properties of the cap rock which is the element that is

foreseen to confine the injected gas.
2
GENERAL FORMULATION FOR WATER

AND GAS FLOW IN DEFORMABLE
POROUS MEDIA
A general formulation for multiphase flow in porous

media is described briefly in this section. It will be considered that the state variables (unknowns) are: solid
displacements, u (three spatial directions) and liquid
pressure, P l . Balance of momentum for the medium
as a whole is reduced to the equation of stress equilibrium together with a mechanical constitutive model
which relates stresses with strains. Strains are defined
in terms of displacements. Small strains and small
strain rates for solid deformation are assumed. Advective terms due to solid displacement are neglected once
the formulation is written in terms of material derivatives (in fact, material derivatives are approximated
as eulerian time derivatives). In this way, volumetric
strain is properly considered.
The governing equations for non-isothermal multiphase flow of water and gas through porous
deformable saline media have been presented by
Olivella et al. (1994). A detailed derivation is given
there, and a brief description of the formulation for
thermo-hydro-mechanical problems is presented here.
Mass balance of solid present in the medium is
written in the eulerian form as:
where s is the mass of solid per unit volume of solid

and js is the flux of solid. From this equation, an expression for porosity variation can be obtained using the
flux of solid that has been written as the velocity of the
solid multiplied by the volumetric fraction occupied
by the solid phase and the density:
phase densities, jil , jig are the mass fluxes of the component i in each phase and f i is an external supply of mass
of component i. In this formulation the components in
most cases considered in this paper are the water and
air, but air can be substituted by generated hydrogen
in case of corrosion (or when modeling gas injection
tests). The mass flux of components is a combination
of a non-advective flux (diffusion + dispersion) written as iil , iig the advective Darcy flux written as ql , qg
and another advective term caused by the solidmotion,
which is proportional to the solid velocity du dt:
Expansion of time derivatives and spatial derivatives

and the use of the material derivative with respect to
the solid velocities, leads to the following equation
which is actually a lagrangian:
The mass balance of solid (equation, 2) is introduced in the mass balance of components to obtain,
after some algebra:
Equation (2) expresses the variation of porosity

caused by volumetric deformation and solid density
variations. Two fluid phases are considered: gas and
liquid. The total mass balance of a component i present
in each phase (for instance water that may evaporate)
is expressed as:

The volumetric deformation rate term [dv dt =

du dt ] couples the mass balance equations with
the deformation of the porous medium. This requires
the coupled solution of the mechanical equations. The
momentum balance for the porous medium reduces
to the equilibrium of stresses if the inertial terms are
neglected:
where Sl , Sg are the phase degree of saturations; li , gi

are the mass fractions of the component i in each phase
(mole fraction can be considered as well); l , g are the
where is the stress tensor and b is the vector of body

forces.
In addition, the balance of energy is introduced
in case of non-isothermal conditions. The equation
232
for internal energy balance for the porous medium is

established taking into account the internal energy of
each phase (E s , E l , E g ):
where ic is energy flux due to conduction through

the porous medium, the other fluxes (jEs , jEl , jEg ) are
advective fluxes of energy caused by mass motions and
f Q is an internal/external energy supply. Energy generation by deformations can be considered included in
this source term and can be incorporated without loss
of generality (for instance to analyze failure problems
with high velocities). The use of the material derivative allows obtaining an equation formally similar to
the mass balance of components.
The simultaneous solution of the coupled equations
given above produces the spatial and temporal evolution of displacements, liquid pressure, gas pressure
and temperature; which are considered as state
variables or unknowns in this THM approach.
If the mass of i is put into a single variable
mi = li l Sl + gi g Sg the balance equation can be
transformed into:
This equation contains three storage terms,

related to:
I. Storativity due to changes in pore pressure which
i
can be obtained from m1i dm
and the coupled terms
dpl
obtained as:
1 dmi 1 dmi
;
.
mi dpg mi dT
s
II. Solid deformability since 1s d
= K1x is the
dx
deformability of the soil particles with respect
to the variable x, which can be pressure or temperature. This term is not significant (sometimes
neglected) for soils but it is essential for rocks as
it combines with the bulk deformation to produce
the changes in porosity which is very small and
suffers relatively large variations.
III. Soil skeleton deformability since the divergence of solid
velocity can be transformed

into du
= dtd ( u) = ddtv , and by means
dt
of an appropriate constitutive model that
contains suction and temperature dependen
cies the volumetric strain rate is: dv (dt,s,T ) =

dv d
v ds
v dT
+ d
+ d
. The mechanical model
d dt
ds dt
ds dt
may include effective or net stress terms or
suction terms and temperature terms.
Coupling the above described set of equations to

transport of solutes including chemical interactions
can be found in Saaltink et al (2004) and Guimaraes

et al (2007).
A model for unsaturated soils such as the elastoplastic model BBM (Alonso et al, 1990) including the
effect of effective or net stresses and the suction is
written as:
where suction can be defined as s = max (pg pl , 0)

and net stress as = + max (pg , pl )m. This is
valid for saturated and unsaturated conditions assuming stresses in compression as negatives. The model
parameters are included in D which is the stiffness
tensor (6x6) or constitutive matrix for changes in net
or effective stress and h which is the constitutive vector
for changes in suction. Both are nonlinear functions.
The preceding set of constitutive relations was cast
in the form proposed by Perzyna (1966) to develop
a family of viscoplastic constitutive models for soils
which can be used either as viscoplastic or, for very
low viscosities, as elastoplastic. Incorporation of time
effects in deformation of geological materials can be
done in several ways: viscoelasticity, viscoplasticity
(the type of model described here) and creep (the type
of model that is used for saline materials, for secondary
consolidation of soils or for long term response of
rockfills).
In the case of viscoplasticity, total strain rates ()
are calculated as a sum of elastic (e ) and viscoplastic
components (vp ):
With the the viscoplastic component computed

through:
where is a fluidity parameter and <(F)> is a flow

function defined as:
A usual expression for the flow function is (Perzyna,

1966):
The shape adopted for F and G and the functions for

irreversible deformations are summarized in Table 1
233
Table 1.
Basic relationships for BBM.

Barcelona Basic Model (BBM)
Irreversible
volumetric
deformation
Compressibility
Hardening law
Loading-collapse
curve (LC)
Shear strength
critical state
parameter
Tensile strength
yield surface
Plastic potential
dv =
(s) dp
1+e p
(s) = (0)[(1 r)t + r]

t = exp (s)
(1 + e)po p
dv
dpo =
(0)
(0)
(s)
po
po (s) = pc
pc
M (s) = M
ps (s) = ps0 + ks s
F(p, q, s)
= q2 M 2 (p + ps (s))(po (s) p)
G(p, q, s)
= q2 M 2 (p + ps (s))(po (s) p)
Figure 3.1. Diagram of the design of the evaporation column experiments and their conceptual model from day one
(saturation) to the end of the experiment (oven dry conditions). The water fluxes are on the left, the salt fluxes are on
the center and the energy ones on the right. The dashed and
point lines show the location of the evaporation front and the
minimum in salt concentration respectively.
for the particular case of unsaturated models BBM

(Barcelona Basic Model).
The use of viscoplasticity for engineering applications has the advantage that the solution is regularized
and, to some extent, problems involving failure can be
solved. It is also robust to account for tensions which
are not significant for the solution of the problems but
may appear in practical solution of field problems (for
instance near the boundary the slope of an earth dam).
The elastic part can be written in the following way
(for isothermal conditions):
Depending on the values of the parameters, very high

compressibility can be obtained and in such case
the nonlinear volumetric stiffness should be limited
with a K min . When the irreversible part is formulated
as described above (i.e. addition of elastic plus viscoplastic) the elastic deformations can be modified
independently. For instance, a linear elastic equation
can be combined with a viscoplastic part without loss
of generality. The non-isothermal case including a linear thermal expansion of the porous material (and solid
phase) is a first approximation for incorporation of
thermo-mechanical processes.
3 WATER MOVEMENT INDUCED BY
HEATING IN POROUS MEDIA
Thermal, suction and osmotic gradients interact during evaporation from a salty soil. Vapor fluxes become
the main water flow mechanism under very dry conditions. A coupled nonisothermal multiphase flow and
reactive transport model was developed to study mass

and energy transfer mechanisms during an evaporation experiment from a sand column (Gran et al 2011).
Figure 3.1 shows an idealization of the experiment
performed to investigate evaporation in a salty soil.
Very dry and hot conditions, including the formation of a salt crust, necessitate the modification
of the retention curve to represent oven dry conditions. Experimental observations (volumetric water
content, temperature and concentration profiles) were
satisfactorily reproduced using mostly independently
measured parameters (Figure 3.2), which suggests that
the model can be used to assess the underlying processes. Results show that evaporation concentrates at
a very narrow front and is controlled by heat flow,
and limited by salinity and liquid and vapor fluxes.
The front divides the soil into a dry and saline portion
above and a moist and diluted portion below. Vapor
diffusses not only upwards but also downwards from
the evaporation front, as dictated by temperature gradients. Condensation of this downward flux causes
dilution, so that salt concentration is minimum and
lower than the initial one, just beneath the evaporation front. More details, interpretations and sensitivity
calculations can be found in Gran et al (2011).
The investigations described above are essential
for the study of tailing mines in arid and semiarid
zones. Blanco (PhD Thesis, UPC, 2009) has investigated the long term behavior of tailing deposits at
Monte Romero in the Piritic Belt (Huelva).
Modelling the evolution of the tailings subjected to
atmospheric conditions has been done in order to see
what is the effect of the crust, for instance whether it is
convenient or not to remove it or not. Figure 3.4 shows
the results of the thermo-hydrological calculations for
the base case model.
234
Figure 3.2. Profiles of saturation, temperature and salinity

measured at the end of the experiment (symbols) and computed (lines). The time evolution is shown for four different
times (after 1.1 days, 3.3 days, 6.6 days and, at the end of the
experiment, 12 days). Note that salinity is expressed as concentration in water to facilitate the analysis of mass transport
processes.
Figure 3.4. Soil profile at MonteRomero including hard

later SC-1 and HP-1.
Figure 3.3. Profile and evolution of volumetric water content at different times and flux. Experimental results and
model calculations.
Figure 3.5. Heat flux and piezometric head as a function of

time. Experimental results and model calculations.
In this work, various assumptions of modeling have

been considered, some of the including the presence of
the crust (low permeability layer) while some other did
not include the low permeability feature. More details
can be found in Blanco (PhD Thesis, UPC, 2009).
According to these investigations, along the first
60 cm of the tailing materials, the vertical flux explains
most of water movement. This flux is clearly influenced by the existence of intermediate layers and
cemented materials (HP), and during dry seasons by
the existence of surface crusts. Only in the case of
intense rains, the lateral flow would be significant due
to the relatively low permeability of some intermediate layers. On the other hand, the intense evaporation
in summer leads to the dominance of vapour flow

which stabilizes the contaminants as they remain in
the liquid phase which has limited movement due to
its low relative permeability. Saline efflorescences are
an indication of evaporation near the surface.
In another context, Olivella and Gens (2005) have
carried out thermo-hydro-mechanical (THM) analyses simulating the Drift Scale Test (DST) performed
atYucca Mountain. A double structure approach based
on two superimposed domains was adopted. Intrinsic permeability changes with deformations imply full
THM coupling. Temperatures and gas permeabilities
were measured during four years and are used to
validate the model.
235
Figure 3.6. Location of instrumentation for temperature

(Borehole 158 and 160) and gas permeability measurement
(Borehole 74 and Borehole 76). DST in situ test at Yucca
Mountain.
Figure 3.7. Calculated (thick lines) and measured (thin

lines) temperatures for selected points in borehole 158. DST
in situ test at Yucca Mountain.
Figure 3.8. Measured and calculated gas permeability variations in Borehole 76 and Borehole 74. DST in situ test at
Yucca Mountain.
4 WATER MOVEMENT IN ISOLATION

BARRIERS
Measured gas permeability variations show patterns that are successfully explained by the model
calculations. These gas permeability variations may
be attributed to thermo-hydraulic effects but also to
mechanical effects. Different cases of intrinsic permeability variations have been considered in the model
and their influence on the calculated temperatures,
degree of saturations and gas permeabilities are presented. Volumetric deformation, in contraction or dilatancy, implies changes in the aperture of rock fractures
that in turn lead to changes in intrinsic permeability.
Dilatancy, caused by shear stresses, increases intrinsic
permeability. Consideration of this factor contributes
significantly to improve the agreement of calculated
gas permeability with the measured values obtained
during the DST experiment.
Figure 3.7 shows the evolution of measured and
calculated temperature at different distances from the
drift. Figure 3.8 shows the measured and calculated
intrinsic permeability at different times during the
heating test.
Swelling and expansivity of the soils takes place for a

more dense materials (overconsolidated clays) and it is
a process that is desired in some cases. Construction of
isolation barriers requires expansive materials and this
is expected to happen during humidification or hydration. In the last 20 years, a number of experimental
investigations both in the laboratory and in situ have
been performed in order to understand the behavior
of such waste isolation schemes. The FEBEX experiment (Gens et al, 2009) has been a successful exercise
including a comprehensive number of investigations
to learn about the response of the bentonite barriers.
Another investigation in this field of isolation of
waste is the EB experiment which consists in the
hydration of a barrier made of pellets. The experiment
consists in the artificial hydration of an isolation barrier made of bentonite blocks and pellets (Figure 4.1a).
The artificial hydration was achieved using a system of
pipelines which permitted to inject water at different
locations within the blocks and pellets (Figure 4.1b).
236
the micropores and macropores. The model used for

the modeling presented in this paper is a relatively
simple extension of BBM model which uses a suction
micro (sm ) to calculate the swelling deformations.
The suction micro is calculated from the suction that

prevails in the macropores (sM ) according to a transfer
equation:
Figure 4.1. EB in situ test: a) Scheme in vertical section,

b) Calculated water pressure during hydration (artificial),
c) Calculated evolution of porosity (micro and macro) in the
pellets and in the blocks, d) Measured and calculated volume
of injected water.
Further details can be found in Alonso and Hoffmann

(2001) including modeling of the EDZ.
One of the key issues in this experiment was to
discriminate the ratio the injected water was stored in
From deformations, the porosity variations are considered with the two components, the micro and the
macroporosity. These can be decomposed using the
following equation for which micro porosity may
be defined in two ways: = M + (1 M )m or
= M + m with the corresponding equivalence.
This is represented in Figure 4.1c. Although the
components of porosity undergo significant variations, the overall porosity shows little variations. Actually, the total porosity may change only if some zones
compress and other expand because the total volume
is practically constant in this problem, as it is limited by the drift walls. Both in the pellets and in the
blocks the assumed microporosity increases while the
macroporosity reduces. The impact on permeability is
expected to be significant (it was not considered in this
modeling).
The possibility to calculate these two components
of porosity from the model permits to obtain a more
accurate estimate of the water inflow. This is because
a higher density of the water in the micropores can
be considered. Using a density of 1400 kg/m3 for the
water in the micropores the calculated water inflow
was much more consistent with the measured values
(Figure 4.1d). This is considered a confirmation of
the assumed and calculated distribution of porosity
between micro and macro.
This is an evidence of the necessity of the double
structure models because it not only influences the rate
at which the water hydrates the clay components but
also controls the amount of water that inflows. And
this is important for the swelling process.
As mentioned above, a general formulation that captures in a very appropriate way the phenomena taking
place in these double structured materials has been
developed by Sanchez et al (2005, 2008). The formulation includes an elastoplastic model for expansive
materials which is formulated by means the generalized plasticity approach and an extended form of the
balance equations for double porosity materials. The
formulation has been applied to model in situ and mock
up large scale tests involving THM processes.
The application of this model to the Mock-up test
has lead to a large improvement of swelling stress
development induced by swelling under nonisothermal conditions. Figure 4.2 shows an sketch of the 1:1
237
Swelling and collapse deformation appear in unsaturated soils in different magnitude depending on the
pore size distribution, average and variance. In general
the collapse processes are related to capillary forces
while the swelling are more explained by water migration towards the micropores (double layer) driven by
osmotic suction.
Various problems involving unsaturated soils
require considering variable intrinsic permeability in
modeling. Porosity variations induced by volumetric
deformations of collapse or swelling cannot explain
the large intrinsic permeability variations taking place
in soils. The large variations are caused either by
changes in soil structure for instance in Olivella and
Gens (2001) it is shown that intrinsic permeability of
low permeability clays changes dramatically as water
content changes); by formation of highly permeable
preferential paths induced by dilatancy and damage
(Rummel and Weber, 2000) and by gas migration
(Harrington and Horseman, 2003; Arnedo et al, 2008)
or gasfrac processes. An appropriate constitutive
model is essential for calculating intrinsic permeability as a function of variables that depend on
deformations or changes in structure. First, a model
of an embedded discontinuity is applied to model and
slope in an unsaturated soil and afterwards the model
will be applied to problems of gas migration coupled to
deformations.
Figure 4.2. Mock up test. Simulation and measurements of

relative humidity and radial stress. Modelling is performed
using a double structure model.
scale mock-up experiment together with the calculated and measured relative humidities and stresses
(Sanchez et al, 2008). The heating up to 100 C of
the simulated heaters produces evaporation of water
(decrease of relative humidity) but this is compensated
by artificial hydration on the outer boundary. Overall,
the bentonitic clay tends to full saturation and therefore swells progressively. The comparison of the model
predictions of stresses with measurements was greatly
improved by the incorporation of the double structure
model.
GAS FLOW IN POROUS MATERIALS WITH

LARGE INTRINSIC PERMEABILITY
VARIATIONS
In the case of dilatance deformations in zones of

localization, it is known that permeability increments
may be significant (see for instance Rummel and
Webber, 2000) and this in turn may lead to important
consequences on the deformation process itself.
The incorporation of the model simulating a discontinuity in the form of embedded element was actually
motivated by the necessity to simulate gas migration
in barriers. An interesting example has been developed and is described in Olivella and Alonso (2008).
The simulation of the GMT in situ test (Figure 5.1)
was very challenging as several phenomena were combined including high rates of water and gas flow. The
presence of the discontinuities was a key factor for
both, the water flow taking place induced by an artificial water pressurization period and the gas flow taking
place during gas injection.
The test consists in various events of water and gas
injection with different objectives. The most important
ones are: the pressurization and gas injection tests. The
first was carried out with the objective of achieving full
saturation and a level of pressure in shorter time than
under natural conditions. The gas injection tests simulate the possibility of gas generation by corrosion and
have the objective to understand the process regarding preferential path development and coupling with
mechanical interactions.
238
Figure 5.1. Picture of the GMT in situ test together with a

schematic representation. A concrete silo simulating a waste
package is surrounded by a buffer material (sand + bentonite)
in contact with the host granitic rock.
Figure 5.2a shows the development of water fluxes

during water injection which took place in a ring above
the layers in the basis of the upper cavity. Figure 5.2b
shows the development of gas fluxes which begin at
the upper part of the concrete silo. Both processes are
strongly influenced by the presence of discontinuities
between the buffer and the host rock and by the presence of a permeable zone that hits the upper cavity.
The assumption of anisotropy was a limiting factor of
this model.
Several measurements were used for comparison
with the model results. Figure 5.3 shows some of the
comparison plots, indicating that an overall picture
of the problem was achieved. The pressure evolution at injection points, the water content and the
normal stresses are compared with in situ measurements. In addition, the calculated permeability is
shown for the zones that undergo larger variations.
Water pressurization phase and gas injection phases
can be clearly identified in the pressure, stress and
permeability plots.
Figure 5.2. Contour fields of water fluxes (m/s) during

pressurization (a) and gas fluxes (m/s) during gas injection
tests (b).
6 CO2 SEQUESTRATION
Geologic carbon storage has been proposed to reduce
carbon dioxide (CO2 ) emissions to the atmosphere.
To achieve an efficient storage, pressure (P) and temperature (T ) conditions should guarantee that CO2 will
remain in supercritical (SC) state (Bachu, 2003). However, CO2 will rarely reach the reservoir in supercritical
state because CO2 injection conditions at the well-head
239
Figure 6.1. (a) Temperature and (b) CO2 density after 1 year
of liquid CO2 injection at an injection rate of 1 Mt/yr.
Figure 5.3. Evolution of pressures, water content, stresses

and permeabilities for GMT in situ test.
can be diverse and CO2 may not reach thermal equilibrium with the geothermal gradient (Vilarrasa et al
2013). Therefore, temperature differences will arise in
the vicinity of the injection well, which will induce
thermal strains.
The situation may become compromising if CO2
is injected colder than the reservoir because the rock
will contract, bringing the stress state closer to failure conditions (Goodarzi et al, 2012, Gor et al, 2013).
Nevertheless, injecting cold CO2 is very advantageous
because injection costs are significantly reduced.
Thus, understanding to what extend the rock mechanical stability (of both the reservoir and the caprock)
may be compromised due to cold CO2 injection is crucial. On the one hand, shear slip within the reservoir
opens up fractures, enhancing injectivity. On the other
hand, CO2 leakage may occur if fracture instability
propagates into the caprock, which is not desirable.
The objective of the study by Vilarrasa et al (2013)
is to analyze the thermal stress changes caused by cold
CO2 injection and how they affect the rock mechanical
stability. Determining whether the caprock integrity is
maintained or not is a central point for decision-makers
to consider cold CO2 injection as a feasible injection
strategy. To do this, we perform non-isothermal twophase flow simulations in a deformable porous media.
Figure 6.1 shows the shape of the plume of CO2
injected in an aquifer and the corresponding distribution of temperatures.
Figure 6.2 shows the variations of vertical and horizontal stresses caused by the injection of CO2 , while
for the SC injection the stresses increase (confinement increases), in the case of cold CO2 stresses
decrease (confinement decreases). Therefore, for this
cold injection case, failure conditions are approached
in the aquifer.
In the work presented by Vilarrasa et al (2013), the
following conclusions are drawn. Cold CO2 injection
is a feasible injection strategy. Though it may induce
microseismicity in the reservoir, which is not necessarily negative because it enhances injectivity, it does not
jeopardize the caprock mechanical stability. In fact, it
even improves it, at least when the maximum principal
stress is the vertical, because the increase of horizontal stresses reduces the size of the Mohr circle. This
result is very interesting because injecting cold CO2
(in liquid state) rather than in supercritical conditions
reduces significantly the energy consumption that is
necessary for the injection.
The applications presented in this paper demonstrated that coupled THM problems of a wide variety
240
Figure 6.2. Changes in total stress as a function of depth 3 m

away from the injection well for liquid and SC CO2 injection
after 8 months of injection.
are solved at present. The modeling is of great help to

understand the processes, to make predictions in short
and long term, to modify engineering designs, to study
the safety of waste disposal. Advances are required in
different directions for instance to investigate in detail
the couplings between chemical and thermo-hidromechanical problems at different levels (mechanical
constitutive model, hydraulic model, transport properties, osmotic effects, etc). This may help to understand
real problems and to help to improve designs.
7
CONCLUSIONS
Another field of research is the gas migration and

transport in soils and rocks. This is motivated by
various problems; one of them is the gas generation
induced by corrosion in problems of waste storage.
This problem is numerically highly demanding and
non-continuous. Probably different numerical techniques are required which is actually related to strain
localization and therefore using alternative numerical methods would help to solve different difficulties
within the THMC coupled problem. Finally the THMC
problems with highly nonlinear calculations in three
dimensions will be performed and normally time steps
are small.
ACKNOWLEDGMENTS
This paper shows a number of research works
performed at UPC, all the contributors including
professors, researchers and doctorate students are
acknowledged.
REFERENCES
Acero P., C. Ayora, M.W. Saaltink, S. Olivella,
J. Carrera, 2008. Modeling the interaction between geochemical and hydraulic processes during evaporation
in tailings. 8th Annual V M Goldschmidt Conference,
Vancouver, Canada. Geochimica et Cosmochimica Acta,

72, A3A3.
Alonso, E.E., Gens, A. and Josa, A. (1990) A constitutive
model for partially saturated soils, Geotechnique, Vol.40,
No. 3, pp. 405430.
Alonso E, Vaunat J, Gens A. (1999) Modelling the mechanical behaviour of expansive clays. Engineering Geology
54:173183.
Alonso, E., S. Olivella, N-M. Pinyol (2005) A review of
Beliche dam. Gotechnique, 55(4): 267285.
Alonso E.E. and C. Hoffmann (2007) Modelling the field
behaviour of a granular expansive barrier. Physics and
Chemistry of the Earth 32, 850865.
Arnedo D., E.E. Alonso, S. Olivella, E. Romero (2008) Gas
injection tests on sand/bentonite mixtures in the laboratory. Experimental results and numerical modelling,
Physics and Chemistry of the Earth 33, S237S247.
Bachu, S. (2003) Screening and ranking of sedimentary
basins for sequestration of CO2 in geological media in
response to climate change. Environmental Geology 44:
277289.
Barbour, S.L. and Yang, N. (1993). A review of the influence
of clay-brine interactions on the geotechnical properties
of ca-montmorillonitic clay soils from Western Canada.
Canadian Geotechnical Journal, 30, 920934.
Blanco, Alejandro, Ph D Thesis, Technical University of
Catalunya, 2009.
Di Maio, C. (1996). Exposure of bentonite to salt solution:
osmotic and mechanical effects. Geotechnique 46 (4),
695707.
Gran, M., Carrera, J., Olivella, S., and Saaltink, M. W. Modeling evaporation processes in a saline soil from saturation
to oven dry conditions, Hydrol. Earth Syst. Sci. Discuss.,
8, 529554, doi:10.5194/hessd-8-529-2011, 2011.
Gawin, D, F. Pesavento, B.A. Schrefler, (2003) Modelling
of hygro-thermal behaviour of concrete at high temperature with thermo-chemical and mechanical material
degradation, Computer Methods in Applied Mechanics
and Engineering, Volume 192, Issues 1314, 28 Pages
17311771
Gens, A; M. Snchez; L. Do N. Guimares; E. E. Alonso;
A. Lloret; S. Olivella; M. V. Villar; F. Huertas, (2009)
A full-scale in situ heating test for high-level nuclear
waste disposal: observations, analysis and interpretation,
Gotechnique, Volume 59, Issue 4, pages 377 399.
Guimaraes, L.M.; A. Gens and S. Olivella, (2007) Coupled Thermo-Hydro-Mechanical and Chemical Analysis
of Expansive Clay Subjected to Heating and Hydration
Transport in Porous Media, (2007) 66:341372.
Goodarzi, S., A. Settari, and D. Keith. 2012. Geomechanical
modeling for CO2 storage in Nisku aquifer in Wabamun
lake area in Canada. International Journal of Greenhouse
Gas Control 10: 113122.
Gor, Y.G., T.R. Elliot, and J.H. Prevost. 2013. Effects of thermal stresses on caprock integrity during CO2 storage.
International Journal of Greenhouse Gas Control 12:
300309.
Harrington, J. F. & Horseman, S. T. (2003). Gas migration in
KBS-3 buffer bentonite: Sensitivity of test parameters to
experimental boundary conditions, SKB Technical Report
TR 03-02. Stockholm:SKB.
Hueckel T., (1997). Chemo-plasticity of clays subjected to
stress and flow of a single contaminant. International journal for numerical and analytical methods in geomechanics,
21: 4372.
Lloret A. and E.E. Alonso (1980). Consolidation of unsaturated soils including swelling and collapse, Geotechnique,
30, 4, 449477.
241
Liu, Z; N. Boukpeti, X. Li, F. Collin, J.-P. Radu,T. Hueckel and

R. Charlier, (2005) Modelling chemo-hydro-mechanical
behaviour of unsaturated clays: a feasibility study, Int. J.
Numer. Anal. Meth. Geomech.; 29:919940.
Oldecop, L. A. & Alonso, E. E. (2001). A model for rockfill
compressibility. Gotechnique 51, No. 2, 127139.
Oldecop, L. and Alonso, E. E. (2003) Suction effects on
rockfill compressibility. Gotechnique. 53(2): 289292.
Olivella, S., Carrera, J., Gens,A. &Alonso, E. E. (1994). Nonisothermal multiphase flow of brine and gas through saline
media. Transport in Porous Media, 15, No. 3, 271293.
Olivella S. and Alonso, E.E. (2008) Gas flow through clay
barriers, Geotechnique, Geotechnique 58, No. 3, 157176.
Perzyna, P. (1966) Fundamental problems in viscoplasticity. Advances in Applied Mechanics, Academic Press,
New York, 9, 244368.
Pruess, K. (1987) TOUGH Users Guide, Lawrence Berkeley
Laboratory.
Rummel, F. & Weber, U. (2000). GS experiment: Rockmechanical testing on gasfrac self-healing effect, Mont
Terri Project, TN99-36.
Saaltink M.W.; F. Batlle; C. Ayora; J. Carrera and S. Olivella.

(2004) RETRASO, a code for modeling reactive transport
in saturated and unsaturated porous media, GeologicaActa
Vol. 2, N3, 2004, 235251.
Sanchez, M., A. Gens, L. D. Guimaraes and S. Olivella,
(2005), A double structure generalized plasticity model
for expansive materials, Int. Journal for Numerical and
Analytical Methods in Geomechanics, 29, 751787.
Sanchez, M. A Gens, L Guimaraes, S Olivella, (2008), Implementation algorithm of a generalized plasticity model for
swelling clays, Computers and Geotechnics, 35, 860871
Vilarrasa, V; D. Bolster; M. Dentz; S. Olivella; J. Carrera,
(2010), Effects of CO2 Compressibility on CO2 Storage
in Deep SalineAquifers,Transp Porous Med, 85: 619639.
Vilarrasa, V., O. Silva, J. Carrera, and S. Olivella. (2013).
Liquid CO2 injection for geological storage in deep
saline aquifers. International Journal of Greenhouse Gas
Control, doi:10.1016/j.ijggc.2013.01.015.
Zandarin, M., T E Alonso, S Olivella (2010) Influence
of suction on the mechanical behaviour of rock joints,
UNSAT.
242

Migration of aggressive solutions through cement-bentonite slurry walls

E. Fratalocchi, V. Brianzoni, M. Di Sante, F. Mazzieri & E. Pasqualini
Universit Politecnica delle Marche, Ancona, Italy
ABSTRACT: Although cement-bentonite cut-off walls are frequently used as passive barriers against migration of pollutants, very few data are available in the literature on their long-term performance and migration
parameters, usually assumed equal to those of cement pastes for most chemical species. A research has been
set up aimed at investigating interaction, diffusion and sorption of sulphate solutions through cement-bentonite
mixtures. Different curing times and sulphate solutions were investigated taking into account the reduction of
hydraulic conductivity with time typically occurring in cement-bentonite mixtures. The results show that both
curing time and sulphate concentration significantly influence the hydraulic conductivity and sorption capacity.
Values of the diffusion coefficient different than those of cement paste were found from a preliminary analysis
of sulphate migration.
INTRODUCTION
Low hydraulic conductivity and diffusion coefficients

as well as high retention capacity are the necessary
requirements for passive barriers since they have to
prevent migration of pollutants. These requirements
must be guaranteed in the long term, (decades or
longer times), depending on the site characteristics and
use. Moreover, risk assessment and site-specific risk
analyses require the knowledge of the parameters
governing the migration of pollutants.
Although cement-bentonite (CB) cut-off walls are
very frequently used as passive barriers against migration of pollutants, few data are available in the literature on their long-term performance (Tedd et al.
1993; Garvin & Hayles 1999; Jefferis 2003; Soga &
Joshi 2010) and migration parameters (e.g. Manassero
et al. 1995; Visudmedanukul et al. 2007; Fratalocchi
et al. 2006; 2010). In particular, referring to the effective diffusion coefficients, values similar to those
of compacted clays or cement pastes are generally
assumed for most chemical species. Sorption, equally
important, is frequently neglected.
A research program is currently ongoing at the Universit Politecnica delle Marche (Ancona, Italy) aimed
at investigating the long term performance and migration mechanisms of acidic or salt solutions through
cement-bentonite mixtures. In this paper results will
be discussed on the performance of a CB mixture
exposed at different curing times to a wide range of
sulphate concentrations, taking into account the variation in the microstructure and hydraulic conductivity
with time typically occurring in this kind of mixtures.
Sulphate solutions are indeed known to be aggressive
for cementitious materials.
2.1
MATERIALS AND TESTING METHODS

CB mixture
As a passive barrier, a CB cut-off wall must be

as impervious as possible to reduce advective flow.
Therefore, the investigated CB mixture was selected
in order to get a low hydraulic conductivity when
permeated with water. To this purpose, a blast furnace slag cement was chosen (III-B, 70% of slag,
produced by Cementir S.p.A., Italy) as well as an activated Na-bentonite (Bentosund AUSuper, by Laviosa
Chimica Mineraria S.p.A., Italy). The mix design was:
cement/water = 0.22, bentonite/water = 0.045. Blast
furnace slag cement was selected also for its resistance
to sulphate attack.
Following the procedure by ETC8 (1993), bentonite
was added to tap water, mixed at 1500 round/minute for
5 minutes and left at least one night for hydration; then,
the required quantity of cement was added and mixed
for further 5 minutes and the main characteristics at
the fluid state were measured: typical values of density (11.4 kN/m3 ), Marsh viscosity (50 s) and bleeding
(1.0%) were obtained, in accordance with the practice requirements. The mixture was poured in plastic
moulds and submerged in tap water a few hours after
casing, until the testing time.
2.2 Batch sorption tests
Since there are no standard procedures for batch sorption tests for CB mixtures, tests were performed taking
into consideration the guidelines of batch tests for
fine grained soils (US EPA 1992; ASTM D4319-93),
and the typical sorption experiments on cement pastes
(e.g. Hong & Glasser 1999).
243
Table 1. Chemical composition of the solutions used for

batch tests (B) and column tests (C).
Solution
Test
SO2
4
(mg/l)
K+
(mg/l)
Ca2+
(mg/l)
pH
Distilled water
Tap water
K2 SO4 0.2 g/l
K2 SO4 0.9 g/l
K2 SO4 5.0 g/l
K2 SO4 27.5 g/l
K2 SO4 50 g/l
K2 SO4 95 g/l
B
C
B
B
B, C
B, C
B, C
B, C
0.0
71
110
496
2756
15159
27562
52368
0.0
8
90
404
2244
12341
22438
42632
0.4
86.1
3.5
0.1
1.3
1.6
10.4
5.0
6.75
7.35
6.94
7.30
6.80
7.02
7.34
8.03
Figure 2. Sorption degree (S) of K+ as a function of time

of contact. Batch tests with solutions of potassium sulphate
(concentration in the legend) on CB samples at 45 days of
curing.
for equilibrium), all the batch tests were performed

with a time of contact of 72 hours.
Liquid and solid phases were always separated by
centrifuge and concentrations of the liquid phase were
analyzed by ionic chromatography.
Batch sorption tests were also performed with
aqueous solutions of KOH with K+ concentration
ranging from 39 mg/l (pH = 10.41) to 60584 mg/l
(pH = 13.18).
2.3 Column tests
Sorption degree (S) of SO2
4
Figure 1.
as a function of time
of contact. Batch tests with solutions of potassium sulphate
(concentrations in the legend) on CB samples at 45 days of
curing.
CB samples were crumbled by forced passing

through the N.10 ASTM sieve (2 mm opening).
Mixture-solution ratio of 1:2 (50 g of mixture:100 ml
of solution) was always tested instead of 1:4, considering the high water content of the mixture (about
250%). The mixture was tested after different curing
times ranging from 13 days up to 207 days of curing in
water at room temperature. The CB samples werent
dried to avoid increase in the cement hydration ratio.
They were added to the solution, tightly sealed and
kept at about 20C with continuous agitation. Details
of the K2 SO4 solutions used in batch tests are given in
Table 1.
Different times of contact (agitation) were investigated in order to define the minimum lasting time to
reach chemical equilibrium (US EPA 1992). The time
of contact necessary to reach equilibrium was investigated on CB samples cured in water for 7 days and for
45 days. As shown in Figure 1 and 2, concentrations
of both K+ and SO2
4 were constant after 2448 hours
of contact. Similar results were obtained on samples
cured for 14 days and for about 8 months. On the basis
of this evidence and considering results in the literature
of cement pastes (showing that few days are sufficient
After a curing time of 1216 days, column tests were

performed with solutions of K2 SO4 at different concentrations. Details of the permeant solutions are given
in Table 1. The thickness of all the samples varied in
the range 4.05.7 cm. Flexible wall permeameters with
bladder accumulators were used in order to control the
effective confining pressure (3045 kPa), the sample
volumetric strains (reduction in volume < 0.1% were
always measured after consolidation before starting
the permeation phase) and to collect the effluent liquid.
Constant head tests with continuous source concentration were performed at hydraulic gradient, i, ranging
from 20 to 120. Samples of the effluent liquid were
periodically taken for concentration measurements in
order to get the breakthrough curves. Permeation was
kept for long times (months up to almost 2 years) to
evaluate the long-term performance.
3
3.1 Sorption
Sorption is intended as the overall mechanisms which
remove contaminant from solution. These processes
include cation exchange, precipitation, adsorption and
in general any binding between solutes and the solid
matrix.
Referring to CB mixtures and K2 SO4 solutions,
sorption of K+ can be due to both hydrated cement
products and bentonite whereas sorption of SO2
4 can
244
Table 2. Data of the solutions at equilibrium from batch

sorption tests with K2 SO4 on the CB mixture at 207 days of
curing.
Initial K2 SO4
concentration
SO2
4
(mg/l)
K+
(mg/l)
Ca2+
(mg/l)
pH
Distilled water
0.2 g/l
0.9 g/l
5.0 g/l
27.5 g/l
50 g/l
95 g/l
4.6
21.8
23.6
30.0
4795
14185
34074
7
69
311
1346
8126
14894
29842
428
375
396
93
19
16
25
12.75
12.21
12.35
12.54
12.95
13.04
13.13
be considered related only to cement. In this last

case, sorption of SO2
4 is essentially due to gypsum
formation:
Figure 3. K+ isotherms from batch tests and from mass
balance in column tests with K2 SO4 solutions on CB samples
at different curing times (in the legend).
and to a number of different reactions mainly

involving gypsum, tricalcium aluminate (C3A, where
C=CaO, A=Al2 O3 ), tricalcium aluminate monosulfate (C3A CaSO4 12H2 O) or the hydrated phases of
tricalcium aluminate, e.g. C3AH6 (where H=H2 O),
that give rise to ettringite; for example:
Table 2 shows the concentrations at equilibrium of

sulphate, potassium and calcium ions of the batch test
with K2 SO4 solutions on the CB mixture after 207
days of curing. The high alkalinity of the solutions is
mainly due to dissolution of calcium hydroxide and
to the reaction of sulphates with calcium hydroxide to
form gypsum, according to eq. (1).
These reactions explain why the higher the initial
concentration of sulphate in the solution, the lower the
calcium concentration and the higher the pH at equilibrium (Tab. 2). Results analogous to those in Table 2
were obtained from all the batch tests with the K2 SO4
solutions performed on the CB samples at different
times of curing.
X-ray diffractometry are being performed on some
of the solid samples to evaluate reaction products and
in particular to identify ettringite and K+ compounds
(e.g., syngenite: K2 Ca(SO4 )2 2H2 O). With reference
to K+ , as shown in Figure 3, a linear isotherm:
was found to well fit the data from batch tests

(S = sorbed mass of solute per unit mass of solid
[], C e = equilibrium concentration of solute [ML3 ],
k d = distribution coefficient [L3 M1 ]). The curing
time of the CB mixture does not seem to appreciably
influence the distribution coefficient, which resulted in
Figure 4. K+ isotherms from batch tests with KOH

solutions on CB samples at different curing times.
the range of 3.33.9 ml/g. Very similar isotherms were

obtained from the batch tests performed with the KOH
solutions, as shown in Figure 4.The distribution coefficients measured on differently cured samples (14 days
and 4.5 months) are similar (k d = 3.9 and 3.2 ml/g,
respectively) and lie in the range of the k d values
obtained from the batch tests with the K2 SO4 solutions. On the basis of this result, formation of syngenite
should not be the main reason for K+ sorption.
It is important to point out that the same kind of
isotherm is documented in the literature for sorption
of K+ on cement pastes. In particular, batch tests
performed on CSH (calcium hydro-silicate, where
C=CaO, S=SiO2 and H=H2 O) with KOH solutions
(Hong & Glasser 1999) gave k d values ranging from
0.5 to 5 ml/g depending on Ca/Si ratio (Fig. 5). The k d
values of the tested CB mixture belong to this range
and correspond to a low Ca/Si ratio, typical of blended
cements, such as the slag cement of the investigated
mixture. Therefore, sorption of K+ by CB mixtures
245
can be explained with the same theories presented in

literature for cement pastes (Taylor 1987; Stade 1989;
Hong & Glasser 1999). In general, it is known that
the binding of alkali in hardened cement pastes containing blastfurnace slag cement or Portland cement
is restricted to sorption at the CSH or CASH phases
(VDZ Activity Report 20032005).
As far as the role of the bentonite is concerned, its
sorption of K+ is negligible since its cation exchange
capacity (77 meq/100 g) and its mass (about 15% of the
total solid mass in the mixture) give rise to a negligible
mass of sorbed K+ in comparison with the overall K+
mass sorbed by the mixture. Indeed, k d values of the
CB mixtures are found to be in the range measured by
Hong & Glasser (1999) on C-S-H pastes.
Together with the sorption isotherms from batch
tests, Figure 3 shows the linear isotherm from the performed column tests based on the mass balance of each
test (equilibrium was assumed to be reached in the pore
liquid on the basis of the measured effluent concentrations, as shown later). The distribution coefficient,
k d = 3.14 ml/g or 3.86 ml/g, obtained with or without
the highest value of concentration, respectively, lies in
the range of values from batch tests.
With reference to sorption of SO2
4 , the Langmuir
isotherm was found to give the best fitting (Fig. 6):
The reduced sorption capacity with curing time can

be ascribed to the reduction in mixture porosity with
curing time; as documented in Fratalocchi (1996), significant decrease of porosity occurs in CB mixtures
with curing as well as a decrease of the mean void size.
Therefore, sorption sites tend to be less accessible.
The results in Figure 6 are consistent with results by
Soga & Joshi (2010) who found that CB mixtures can
better resist sulphate attack after hydration over time
as a result of reduced migration through the mixture.
The Langmuir constants S m and b obtained from
all the batch tests are plotted as a function of curing
time of the CB mixture in Figures 8 and 9, respectively. Looking at the S m coefficient, its value tends to
decrease with time during the first months of curing;
this trend, which needs to be confirmed by further data
(given the small range of values), can be explained in
terms of a lower availability of sites with curing time,
due to the reduced porosity and pore size of the mixture. For the same reason the b coefficient tends to
decrease with curing time, particularly after the first
months of curing.
where S m is the solid-phase concentration corresponding to all available sites being occupied [] and b is a
parameter representing the rate of sorption [L3 M1 ].
The results of fitting at different curing times of testing are shown in Figure 6 together with the isotherm
from the mass balance of the column tests. It is clear
that sorption of SO2
4 is appreciably influenced by the
curing time and tends to be lower the higher the curing
time. This result does not seem to depend on the time
of contact between CB sample and solution of batch
tests; in fact, as shown in Figure 7, the isotherm related
to the time of contact equal to 72 hours is the same of
that obtained after 10 days of contact, curing time of
the CB mixture being equal.
Figure 6. SO2
isotherms from batch tests with K2 SO4
4
solutions on CB samples at different curing times.
Figure 5. Distribution coefficient (distribution ratio) for

potassium as a function of the Ca/Si ratio of C-S-H pastes
(Hong & Glasser 1999).
Figure 7. SO2
isotherms from batch tests with K2 SO4
4
solutions performed with different times of contact (in the
legend) on CB samples at 97 days of curing.
246
With reference to sorption assessed from the mass

balance of column tests, the experimental data were
also well fitted by the Langmuir isotherm (Fig. 6).
The isotherm is similar to those from batch tests on
samples with some months of curing; indeed, sorption
in the column tests occurred starting from about 14
days of curing (beginning of the test) and ended after a
period of some months, depending on the inlet sulphate
concentration.
3.2
Hydraulic performance and durability
Figure 10 shows the hydraulic conductivity coefficient, k, versus curing time of the CB mixture
Figure 8. Sm versus curing time of the CB mixture.
Figure 9. Values of the reaction rate, b, versus curing time

of the CB mixture.
permeated with the salt solutions of K2 SO4 at different concentrations together with the k trend of the
same mixture permeated with tap water. With reference to the performance in water, the mixture shows
the decrease in k with time typical of good performance CB mixtures: k values of the order of 108 cm/s
can be reached after two months of curing. The reduction in k remains significant during the first year of
curing.
Referring to hydraulic conductivity trend over time
of the samples permeated with the salt solutions, it
is evident from Figure 10 that the samples show the
same qualitative trend consisting of three phases: (1)
initial decrease in the k value with a correspondent
2
2+
sorption of SO2
concentration
4 (no SO4 and Ca
were detected in the effluent during this period, e.g.
Fig. 11 and 12); (2) increasing hydraulic conductivity (with increasing effluent concentration of SO2
4 ),
and (3) an almost constant trend of k with curing time,
with effluent SO2
4 concentration equal to the inlet one
and Ca2+ concentration of 50100 mg/l. From Figure
10 it is also evident that both the rate of reduction
in k and its subsequent increase depend on the inlet
SO2
4 concentration: the higher the concentration, the
faster the reduction and the subsequent increase in
the k value with time, as well as the shorter the curing time at which the inversion of k trend occurs.
The trend of reduction of k exhibited by the mixture during the first month of curing is equal or even
more rapid than that of the mixture permeated with
water.
The results of Figure 10 can be explained considering the reactions occurring between sulphate and
hydrated cement products. Sulphate attack initially
develops from the reaction between SO2
4 and dissociated calcium hydroxide (and partially CSH), with the
consequent precipitation of gypsum (CaSO4 2H2 O)
and release of OH . The subsequent (damaging)
reactions take place between the sulphate and the
hydrated calcium aluminate or calcium aluminate
monosulphate hydrate (Bensted 1995; Gollop &Taylor
Figure 10. Hydraulic conductivity trend with curing time of the CB sample permeated with tap water (R) and with solutions
of K2 SO4 at different concentrations (in the legend).
247
Figure 11. Column test with K2 SO4 influent

concentration = 5 g/l: hydraulic conductivity and effluent
K+ and SO2
4 concentrations over curing time.
Figure 12. Column test with K2 SO4 influent

concentration = 27.5 g/l: hydraulic conductivity and
effluent K+ and SO2
4 concentrations over curing time.
1992; 1995) to produce ettringite with a consequent

expansion.
All the aforesaid interaction mechanisms are progressive through the sample with curing time: the
initial dissociation of Ca(OH)2 and gypsum precipitation into the pores (reaction confirmed by the high pH
values measured in the effluent, pH = 1212.4) causes
clogging of pores and contributes to the reduction of
the hydraulic conductivity with time. This contribution
tends to reduce and become negligible as expansive
reactions start and ettringite is produced; such expansive reactions are able to progressively invert the
hydraulic conductivity trend with time up to a rapid
increase. The expansive reactions producing ettringite
were confirmed by the increase in volume of the samples equal to 710% (Fratalocchi & Pasqualini 2007)
and by a net of diffuse small fissures observed on the
samples at the end of the tests. The fissures give rise
to preferential paths that are responsible for the overall almost constant value of k (23 106 cm/s) at
the end of the interaction mechanisms. X-ray analyses were performed on the sample permeated with the
Figure 13. XRD diffractogram of the CB sample after column testing with the K2 SO4 solution at
concentration = 95 g/l.
most concentrated salt solution at the end of column

testing. They confirmed gypsum and ettringite formation (Fig. 13), together with the presence of syngenite,
K2 Ca(SO4 )2 2H2 O.
Considering the chemical reactions occurring
between the salt solutions and the CB mixture, it is
necessary to point out that the overall hydraulic conductivity values measured with time (Fig. 10) depend
on the sample mass and on the sulphate influx, i.e., on
the flow rate. Therefore, in order to define how long
the mixture is able to keep a good performance (k not
increasing with time), the curing time cannot be considered as a reference parameter. To this purpose, the
number of pore volumes of flow, PV, can be appropriate. Different criteria can be adopted to establish
a satisfactory hydraulic performance for the CB mixture: for example, hydraulic conductivity lower than a
maximum allowable value, or k lower than the k value
measured with water, etc.; among them, the number of
pore volumes of flow until k is decreasing, PV*, can
be appropriate for the following reasons:
at brief curing, regardless of the concentration of
K2 SO4 , the reduction of k with time is equal or
lower than that with water;
in the long term, if there is no inversion of the k
trend with time, low k values can be reached;
it is not necessary to establish a target k value (that
would be related to a particular curing time);
an increasing k trend with time does not imply a
bad performance at least immediately; therefore,
the criterion is on the safe side.
Therefore, the number of pore volumes of flow at
which the k value stops decreasing (named critical
pore volumes, PV*) can be assumed as the reference
value for the CB mixture good performance.This value
(calculated assuming a porosity of the mixture equal to
0.6) was related to the concentration of SO2
4 for the
different salt solutions. Data from permeability tests
performed with acidic solutions of H2 SO4 , were also
considered (Brianzoni et al. 2013). Figure 14 shows
that PV* decreases as the sulphate ion concentration
248
Figure 14. Critical number of pore volumes, PV*, as a function of the SO2
4 concentration. Column tests with solutions
of sulphuric acid and potassium sulphate (Brianzoni et al.
2013).
increases; the relation is well represented by a linear

correlation on a bi-logarithmic plot. In particular, for
concentrations of the order of g/l or more, that is, in
very aggressive conditions, the critical number of pore
volumes assumes the same trend for both the solutions
(saline and acidic). For concentrations of SO2
4 < 1 g/l,
data are currently available only for the H2 SO4 solution; the critical PV* should be lower for the H2 SO4
solution than that of the salt solution considering its
combined deleterious effect due to expansive reactions
and dissolution of cement hydration products. Data are
necessary to confirm this hypothesis.
Even assuming the criterion of good performance
as the PV value until the k trend over time be not
increasing (instead of decreasing), the data in Figure
14 would be practically the same except for the data
with the lowest concentration of SO2
4 which would
have a PV* slightly higher (equal to about 57). It is
important to point out that a little increase in PV*
means a significant longer durability for the mixture
when the k value is low.
The results in Figure 14 can be useful from the
practical point of view to estimate the durability of a
cement-bentonite cut-off wall on the basis of the flow
rate through it (i.e. on the basis of the hydraulic boundary conditions and assuming, safely, the hydraulic
conductivity value measured at brief curing). Since
a high value of PV* results in case of low concentration of SO2
4 , a good long term performance can
be expected because the hydraulic conductivity values
after few months of curing are sufficiently low so that
very few pore volumes of flow can pass through a cutoff wall in case of typical hydraulic gradients of the
order of units.
3.3 Migration parameters
Breakthrough curves of potassium and sulphate in
Figure 11 and 12 show a higher retardation of SO2
4
relative to K+ , due to the higher sorption (i.e. reactions
to form gypsum and ettringite). The delay between the
+
breakthrough curves of SO2
4 relative to K tends to
be more remarkable the lower the influent concentration (see also Fratalocchi et al. 2010), that is, when the
sulphate has not yet significantly changed the mixture
macrostructure.
Due to cement hydration and chemical interactions,
the seepage velocity in the column tests significantly
changed with time. Therefore, the analytical models frequently used to analyse column tests (e.g. van
Genuchten & Alves 1982) could not be used to analyse
migration of K+ and SO2
4 through the CB mixture.
The experimental breakthrough curves of K+ and
SO2
4 were fitted by the POLLUTEv7 (2004) program able to solve the one dimensional contaminant
migration equation subjected to different boundary
conditions, with time varying properties and linear
or non-linear sorption isotherm. Breakthrough data of
+
SO2
4 and K were fitted assuming constant concentration at the top boundary (C o ) and infinite thickness
of the sample. The assumption that the effluent (measured) concentration is equal to the resident concentration was considered not to have a major impact on
the results as the Peclet number resulted always greater
than 5 (Shackelford 1994).
The overall testing time was subdivided in intervals
with constant seepage velocity equal to the average
in that interval. The coefficient of longitudinal dispersivity, L [L], was assigned for each period. The
values of L were chosen on the basis of laboratory
measured values on fine grained soils, which commonly range between 0.1 and 10 mm (e.g. Freeze &
Cherry 1979). Fittings were tried with L = 0.01L or
0.1L (L = samples thickness).
Figure 15 shows the breakthrough curve of potassium of the column test with the K2 SO4 solution
at concentration of 5 g/l (Fig. 11). No significant
microstructure modifications were supposed to occur
since the hydraulic conductivity was close to that of
the test with water.
Several fitting curves were attempted assuming
combinations of different values of L and diffusion
coefficients (D*) in each time interval, and assuming k d values in the range of those of the isotherms in
Figure 3. On the basis of the decreasing hydraulic conductivity trend with time (Fig. 11), the hydrodynamic
dispersion coefficient, D:
was considered not to increase with curing time

(ve = seepage velocity = v/n, where v = Darcy velocity, n = effective porosity, assumed equal to 0.6).
The fitting curve with k d = 4, L = 0.1 L and
D* = 2E-6 cm2 /s gives the best fitting of the experimental breakthrough curve (Fig. 15). The distribution
coefficient is very close to the values from batch
sorption tests (equal to that from column test mass balance). Since the Peclet number ranges between 9 and
5, both advective and dispersive transport processes
are significant. The diffusion coefficient is found to
be higher than that of cement pastes, where D* is of
the order of 107 cm2 /s for K+ (Alonso et al. 2001).
249
Figure 15. Potassium breakthrough and fitting curves with

different values of dispersivity coefficient and diffusion
coefficient (column test with potassium sulphate = 5 g/l).
As long as the mean seepage velocity is of the order

of 105 cm/s (first 45 days of curing) mechanical dispersion is predominant, whereas both diffusion and
mechanical dispersion are important when the seepage
velocity is of the order of 106 cm/s.
Starting from the same value of k d and with the
same procedure adopted for the test in Figure 11, the
breakthrough curves of K+ from the other column tests
were fitted. In these cases, due to the high sulphate
concentration, expansive reactions and fissuring start
prevailing during the K+ transient phase and the k
values of each of these samples are higher than the
k values in water, k w , curing time being equal. The
micro and macro-structure of these samples starts to
be significantly modified.
In the presence of macroscopic fissures, contaminant migration should be modelled by a fractured
medium that requires assumptions on the fissures
distribution and geometry. This type of analysis is currently being developed. An initial estimate of contaminant migration was done maintaining the assumption
of an intact medium. The increase in seepage velocity
during the transient phase, due to progressive occurrence of preferential paths through each sample, was
simulated by a decrease in the effective porosity with
time. It was assumed n = 0.6 when k is close to k w ,
down to n = 0.05 when k > 10 k w , i.e. when a diffuse
net of fissures has formed through the sample.
Figure 16 shows the experimental potassium breakthrough with the best fitting curve of the column
test with the influent K2 SO4 concentration of 27 g/l.
Also in this case and in all the other tests performed
with higher influent concentrations, a good fitting was
obtained with k d values in the range of those from the
batch tests.
The best fitting curve in all the tests was found
with D* = 2 106 cm2 /s and L = 0.1 L until the
hydraulic conductivity trend with time was close to
the trend with water (up to 60 days of curing in the test
of Figure 16). Both diffusive and dispersive transport
processes are significant when ve is of the order of
106 cm/s.
Figure 16. Potassium breakthrough and best fitting curve.

Column test with concentration of potassium sulphate =
27.5 g/l.
When k starts to increase with time, D* values

up to 6 106 cm2 /s were used for fitting, in order
to take into account the reduction in tortuosity due
to the preferential paths. However, diffusion in this
phase is negligible in comparison to mechanical dispersion since hydrodynamic dispersion coefficients of
the order of 104 cm2 /s were obtained in this second
phase of the tests.
As far as migration of sulphate is concerned, in all
the column tests the breakthrough occurred simultaneously with the increase in hydraulic conductivity. Only
the SO2
4 breakthrough of the test with K2 SO4 = 5 g/l
was analysed because only in this test an extended
part of the breakthrough can be assumed to have
occurred before extensive fissuring, as shown by the
increase in hydraulic conductivity.
Consistently with the potassium migration, L was
assumed equal to 0.1L and the same assumption was
done regarding the increase in the seepage velocity due
to fissuring.
Experimental data were fitted assuming the
Langmuir isotherm from the mass balance of column tests (Fig. 6) as a first attempt, and modifying
the parameters b and S m , together with diffusion
coefficient, until a good fitting was obtained.
Figure 17 shows the best fitting curves until k is
not increasing (540 days of curing). Until 415 days
of curing, a good fitting can be obtained assuming n = 0.6, S m = 0.21, b = 270 ml/g (exactly those
from mass balance of column tests) and D* = 1
106 cm2 /s (fitting curve a); a good fitting was also
obtained with n = 0.6, S m = 0.20, b = 250 ml/g and
D* = 1 106 cm2 /s (fitting curve b). In the subsequent time interval (415540 days of curing), n was
decreased to 0.2 and then to 0.05 and D* was increased
to 3 106 cm2 /s. It was impossible to get reasonably
good fittings after 540 days of curing, that is, when the
hydraulic conductivity rapidly increases.
On the basis of this preliminary analysis, the
Langmuir isotherm with the parameters equal or very
close to those from mass balance of column tests gives
the best fitting. The diffusion coefficient of SO2
4
through CB mixture is estimated to be of 106 cm2 /s
250
Figure 17. Sulphate breakthrough ad best fitting curves.

Column test with concentration of potassium sulphate = 5 g/l.
when the mixture microstructure is not yet significantly modified. Similar values were also found by
previous analyses performed on some of the column
tests presented in this paper, assuming a linear sorption
with k d different in each of the time intervals (Fratalocchi et al. 2010). Such values of D* are two orders
of magnitude higher than those of cement pastes (of
the order of 108 cm2 /s; e.g. Gospodinov et al. 1996).
This result deserves further investigations and analyses
since it implies a worse performance of CB mixtures
against migration of sulphates (ad of potassium, too)
relative to cement paste. Values of D* of cement pastes
should not be used in case of CB mixtures because they
may lead to unsafe prediction of sulphate migration.
As far as sorption is concerned, a linear isotherm

was found to well fit the data from batch tests with
reference to K+ , with k d = 3.13.9 ml/g. This value is
consistent with data in the literature on sorption of K+
by CSH. No significant variation in k d were found with
curing time. Mass balance from column tests gave the
same values of k d obtained from the batch tests.
Sorption of sulphates can be modelled by the
Langmuir isotherm. Sorption appreciably reduces
with curing time: both S m and b tends to decrease with
curing time of the mixture. From a practical point of
view this means that the curing time is a fundamental
parameter to consider also in terms of sorption capacity for the design of CB barriers; in particular, curing
time is beneficial for durability since CB mixtures
become less reactive to sulphate.
The sorption isotherm of sulphate evaluated by
mass balance from column tests resulted similar to
those of batch tests on the same mixture at curing times
of the order of months.
As long as the hydraulic conductivity of the CB
mixture permeated with the sulphate solutions is equal
or of the same order of magnitude of the k measured in water, D of both K+ and SO2
4 were found to
be 2 106 cm2 /s and 1 106 cm2 /s, respectively.
Higher values of D were found when expansive reactions start to prevail. Such values, to be confirmed
by further investigation and analyses, are greater than
the values documented in the literature for cement
pastes and therefore the D* of cement pastes should
not be used in case of CB mixtures because they may
lead to unsafe prediction of sulphate and potassium
migration.
CONCLUSIONS
REFERENCES
On the basis of the available results it is possible to

state that permeation with aqueous solutions of K2 SO4
adversely affects the hydraulic performance of CB
mixture when SO2
4 concentration is of the order of
g/l or more. Sulphate produces an initial decrease in
the k value, followed by an increase and finally an
almost constant trend of hydraulic conductivity (of the
order of 106 cm/s) with curing time. Swelling and a
dense net of fissures were observed on the samples
after permeation mainly due to ettringite formation.
The number of pore volumes of flow at which the
change in the k trend occurs (PV*) is greater the lower
the SO2
4 concentration. A correlation between PV*
and SO2
4 concentration was found that can be useful
from the practical point of view to estimate a cut-off
wall durability on the basis of the expected flow rate
through it. This criterion is on the safe side because it
is based on the number of PV for a decreasing k with
time (a constant or increasing k with time does not
necessarily imply a bad performance, at least immediately). Moreover, the chemical conditions adopted in
the tests are not expected to occur continuously in the
field if a pumping system is provided so that there is
an advective counter-flow of groundwater and not of
pollutant across the barrier.
Alonso C., Andrade C., Chadbourn G.A., Page C.L. 2001.

Oxygen and chloride diffusion in cement pastes as a
validation of chloride diffusion coefficients obtained
by steady-state migration tests. Cement and Concrete
Research, 31 (4): 621625.
Bensted J. 1995. The standardisation of sulphate-resisting
cements. Part I + Part II. World Cement, Vol. 7 and 8.
Brianzoni V., Fratalocchi E., Pasqualini E. 2013. Long
term performance of cement-bentonite cut-offs in saline
and acidic solutions. Proc. 18 ICSMGE, Paris, in press.
ETC8, European Technical Committee N.8 (1993) Technical recommendation GLC. Geotechnics of landfills and
contaminated land. German Geotechnical Society Ed. for
the ISSMFE.
Fratalocchi 1996. Hydraulic behaviour of cement-bentonite
mixtures. PhD. Thesis, Politecnico di Torino (in Italian).
Fratalocchi E. & Pasqualini E. 2007. Vertical barriers for side
containment. Proc. XXI Turin Geotechnical Conference,
CGT, Torino (Italy).
Fratalocchi E., Giorgini S. & Pasqualini E. 2010. Migration of
sulphate solutions through cement-bentonite diaphragms.
Proc. VI Int. Congress on Envir. Geotech., New Delhi,
India.
Fratalocchi E., Pasqualini E., Balboni P. 2006. Performance
of a cement-bentonite cut-off wall in an acidic sulphate
environment. Proc. V Int. Congress on Environmental
Geotechnics, Cardiff (UK), Vol. 1, 133139.
251
Freeze R.A. & Cherry J.A. 1979. Groundwater. Prentice-Hall

Inc.
Garvin S.L. & Hayles C.S. 1999. The chemical compatibility
of cement-bentonite cut-off wall material. Construction
and Building materials 13: 329341.
Gollop R.S. & Taylor H.F.W. 1992. Micro-structural and
microanalytical studies of sulphate attack. I. Ordinary
Portland cement paste. Cem. & Concr. Res., 22(6):
10271038.
Gollop R.S. & Taylor H.F.W. 1995. Micro-structural and
microanalytical studies of sulphate attack. III. Sulphateresisting Portland cement: reactions with sodium and magnesium sulphate solutions. Cem. & Concr. Res., 25(7):
15811590.
Gospodinov P., Kazandjiev R. & Mironova M. 1996. The
effect of sulphate ion diffusion on the structure of cement
stone. Cement and Concrete Composits, 18: 401407.
Hong S.Y. & Glasser F.P. 1999. Alkali binding in cement
pastes. Part I. The C-S-H phase. Cem. & Concr. Res., 29:
18931903.
Jefferis S.A. 2003. Long term performance of grouts and
the effects of grout by-products. Proc. 3rd International Conference on Grouting & Ground Treatments,
New Orleans, ASCE Geotechnical Special Publication,
No. 120.
Manassero M., Fratalocchi E., Pasqualini E., Spanna C.,
Verga F. 1995. Containment with vertical cut-off walls.
ASCE GSP 46: 11421172.
Shackelford C.D. 1994. Critical concepts for column testing.

Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 120 (10): 18041828.
Soga K. & Joshi K. 2010. Long-term engineering performance of cement-bentonite cut-off walls: a case study.
Proc. VI Int. Congress on Environmental Geotechnics,
New Delhi, Vol. 1, pp. 151164.
Stade H. 1989. On the reaction of C-S-H with alkali hydroxides. Cem. & Concr.Res., 19: 802810.
Taylor H.F.W. 1987. A method for predicting alkali ion
concentrations in cement pore solutions. Advanced in
Cement Res., 1 (1): 516.
Tedd P., Paul V., Lomax C. 1993. Investigation of an eight
year old slurry trench wall. Proc. Green 93, Int. Symp.
On Waste Disposal by Landfill, 581590.
US EPA, United States Environmental Protection Agency
(1992) Batch-type procedures for estimating soil sorption
of chemicals. EPA/530/SW-87/006-F.
Van Genuchten M.J. & Alves W.J. 1982. Analytical solutions
of the one dimensional convective-dispersive solute transport equation. US Department of Agriculture, Technical
Bulletin n. 1551.
VDZ Activity Report 2003-2005. III. Performance of
Cement. pp. 6284.
Visudmedanukul P., Kamon M., Katsumi T. 2007. Chloride
transport through cement-bentonite barriers. Journal
133(2): 175185.
252
General papers
Landfill bottom and sides lining systems

Comparison of natural, polymer treated, and sodium activated Ca-bentonites

G. Di Emidio, R.D.F. Verastegui & A. Bezuijen
Laboratory of Geotechnics, Ghent University, Ghent, Belgium
ABSTRACT: The hydraulic conductivity of geosynthetic clay liners (GCLs) for the confinement of waste disposal facilities depends on the hydraulic conductivity of the bentonite clay encased between the two geotextiles.
Sodium bentonites are widely used for GCLs due to their low hydraulic conductivity to water. Although, calcium
bentonites (Ca-bentonites) are low cost and readily available, the hydraulic conductivity of calcium bentonites
is higher compared to that of sodium bentonites (Na-bentonites). This study compares the impact of the treatment of Ca-bentonite with an anionic polymer, sodium carboxymethyl cellulose (Na-CMC) using HYPER clay
technology, relative to standard sodium activation treatment. The swelling ability and hydraulic conductivity to
CaCl2 solutions of a natural (untreated) Ca-bentonite was compared to those of sodium activated bentonite and
to polymer treated bentonite. Test results showed the superior performance of the polymer treatment compared
to sodium activation.
INTRODUCTION
The spread of landfill leachates is prevented by the

use of hydraulic barriers such as compacted clay liners
(CCLs), geosynthetic clay liners (GCLs) and geomembranes. High quality bentonites, like sodium bentonite,
are used in GCLs and mixed to soil in CCLs. Sodium
bentonite clays are widely used because of their low
hydraulic conductivity to water and favorable swelling
properties (Mitchell, 1993; Shackelford et al., 2000;
Egloffstein, 2001; Bouazza, 2001). Ca-bentonite is a
lower quality bentonite but is more readily available
and may be enhanced by sodium activation (i.e., saturating the exchange complex of the Ca-bentonite with
sodium ions). However, exposure to high concentrations of organic or inorganic compounds present in
landfill leachates can change the clay fabric of natural sodium or sodium-activated bentonites, increasing
their hydraulic conductivity (Norrish, 1954; Mitchell,
1993). Another critical disadvantage of sodium activation is the consequent formation of calcium carbonate
precipitates into the soil. This calcium carbonate may
re-dissolve during contact with a dilute permeant,
releasing calcium ions that exchange with the sodium
in the clay (Guyonnet et al., 2005). This exchange leads
to the development of a more permeable clay.
To overcome these problems, modified clays have
been recently introduced in barrier applications to
improve their chemical resistance to aggressive permeants (Kondo, 1996; Onikata et al., 1996, 1999;
Flynn and Carter, 1998; Schroeder et al., 2001;
Ashmawy et al., 2002; Simon and Mller, 2005;
Lorenzetti et al., 2005; Filippi et al., 2007; Betega de
Paiva et al., 2008; Mazzieri et al., 2010; Katsumi et al.,
2008; Di Emidio, 2010a&b; Bohnhoff & Shackelford,
2010; Palomino, 2010; Scalia et al., 2011; Malusis
et al., 2011). In terms of pollutant containment applications, it is of great interest to evaluate the potential
benefits of these treated clays by analyzing their
hydraulic behaviour, investigating the mechanisms
through which the amendments can improve the sealing capacity of a clay, highlighting more suitable
treatment techniques and further investigating the
weaknesses of the available treatments with the aim
of future advancements and improvements.
Recent studies on the effect of polymer treatment
on the properties of a sodium bentonite (Di Emidio,
2010a), showed a reduction of hydraulic conductivity
of the amended bentonite (called HYPER clay) compared to the non-amended clay, in presence of various
permeant solutions. The anionic polymer penetrates
the interlayer regions between clay platelets enhancing the osmotic swell of the clay. Intercalation of the
anionic polymer in the interlayer space between clay
platelets tends to protect the clay from chemical attack,
by preventing the contraction of the interlayer and the
consequent increase of hydraulic conductivity in the
presence of aggressive permeants.
The objective of this paper is to compare treatment
of a Ca-bentonite using HYPER clay technology to
standard sodium activation. The swelling ability and
hydraulic conductivity to CaCl2 solutions of a natural (untreated) Ca-bentonite is compared here to those
of sodium activated bentonite and to polymer treated
bentonite.
2
MATERIALS
In this study three materials are compared: a natural calcium bentonite (CaB), the same CaB sodium
activated (SA CaB) and the same CaB treated with
255
Table 1.
Characteristics of tested soils.
Characteristics
CaB1
SA CaB1
Table 2.
HCaB1
Solution
Swell index (ml/2g)

Specific gravity
Liquid limit ()
Plasticity index ()
Smectites Mica (%)
Quartz Opal (%)
Feldspars (%)
12
2.5
310
246
820
20
2
19
2.5
414
349
820
20
2
42
2.2
832
632
820
20
2
Deionized
water
KCl
CaCl2
CaB = calcium bentonite; SA CaB = sodium-activated
CaB; HCaB = HYPER CaB.
the HYPER clay technology (HCaB). The latter treatment method consists of mixing the base clay with a
polymeric solution containing 8% (by dry weight of
clay) of sodium carboxymethil cellulose (Na-CMC)
with a mechanical stirrer for 30 minutes. This slurry
is then oven dried at 105 C for 16 hours. After drying, the HYPER (CaB) clay is ground first manually
using a mortar and pestle and then mechanically using
a Retsch Mortar Grinder RM 200. This treatment
method can also be applied to other types of clay
with different polymer dosages (Di Emidio, 2010b).
Some properties of the materials tested here are listed
in Table 1. As shown in the table, the specific gravity
of the HYPER clay is lower compared to the untreated
Ca-bentonite. This result may depend not only on the
lower specific gravity of the polymer (about 1.59),
but also on the dispersed configuration of the treated
clay. Moreover, the table shows that swell index and
liquid limit of HYPER clay are higher compared
to sodium activated and untreated clays, indicating
that the polymer addition may improve (more than
sodium activation) the hydraulic performance of the
clay. There is in fact an inverse relationship between
hydraulic conductivity and swell index (Jo et al., 2001),
and between hydraulic conductivity and liquid limit of
clays (Lee et al., 2005).
The electrolyte solutions used in this investigation
are deionized water, KCl and CaCl2 solutions with different concentrations. The deionised water was used as
base solution for the HYPER clay and electrolyte solutions preparation. The deionized water was produced
using a water purification system PURELAB OptionR 7/15. The electrical conductivity of the deionized
water was EC = 0.0039 mS/cm, the pH = 7.57, and the
redox potential was Eh = 293 mV. The electrolyte solutions were used to test the swelling and the hydraulic
conductivity of the materials. The solutions were prepared by dissolving different concentrations of KCl
and CaCl2 2H2 O (>99.7% pure) in deionised water.
Some properties of the solutions are listed in Table 2.
3
METHODS
The tests performed on the three samples were: (1)

swell index tests using KCl and CaCl2 solutions with
concentrations between 0.0001 M and 0.1 M; (2) swell
Chemical properties of the electrolyte solutions.

Concentration
M
0.0001
0.001
0.01
0.1
0.0001
0.001
0.005
0.01
0.1
EC
mS/cm
Salinity
pH
Eh
mV
0.004
0.0
7.57
293
0.014
0.143
1.392
12.76
0.025
0.251
1.211
2.220
18.54
0.0
0.0
0.5
7.3
0.0
0.0
0.4
0.9
11
6.45
6.22
6.26
6.59
6.35
7.28
6.56
6.69
6.77
258
243
304
324
297
330
262
238
260
pressure tests and (3) hydraulic conductivity tests

using 5 mM and 100 mM CaCl2 solutions. The swell
index tests were performed to have a qualitative evaluation of the impact of various electrolyte solutions
with different concentrations (0.0001 M up to 0.5 M)
and valences (K+1 and Ca+2 ). The swell pressure tests
were executed to prepare the samples for the hydraulic
conductivity test and also for the qualitative evaluation
to overcome some limitations of the swell index test
(such as false measurements caused by turbidity of the
solutions and macropores formation).
3.1 Swell tests
Swell index tests were performed following the ASTM
D5890. The bentonites were oven dried at 105 C and
were ground using a mortar and pestle until 100%
passed the 0.075 mm (#200 mesh U.S. standard) sieve.
Ninety ml of the testing solutions were poured into a
100 ml graduated cylinder. Two grams of sieved bentonite were poured in the aqueous solutions in 0.1 g
increments. After the 2 g were added, the remaining 10 ml solution was poured to fill the cylinder
to 100 ml. This additional solution was also used to
rinse any particles of bentonite adhered to the internal
sides of the cylinder. After 16 hours of hydration, the
final temperature and volume of the swollen bentonite
were measured.
The swell pressure test apparatus used consisted
of a stainless steel ring (7.1 cm diameter) accommodated in a one-dimensional cell (similar to a standard
oedometer cell) located in a frame provided with a
load cell connected to a computer. The specimen was
prepared by spreading a thin layer of dry clay (0.45 g
dry bentonite/cm2 ) into the ring and placing the ring
into the oedometer cell. The aim of this procedure was
to obtain samples that represent the bentonite core of
a standard GCL with typical dry bentonite per unit
area of 0.45 g/cm2 . After assembling, the sample was
inundated with the testing solution (5 mM CaCl2 ). For
these powder samples a fixed height (h = 0,665 cm)
was chosen to obtain a similar initial porosity (with a
dry unit weight of 6.65 kN/m3 ) for all specimens. The
swelling pressure was measured by the load cell keeping the height of the sample constant. The achievement
256
Figure 1. Swell indices in (a) KCl and (b) CaCl2 solutions

for the CaB, SA CaB and HCaB.
Figure 3. Swell pressure vs. time of CaB, SA CaB and

HYPER CaB in a 0.005 M CaCl2 solution.
Figure 2. (a) Macropore formation in the swell index test

of the sodium activated clay in a 5 mM CaCl2 solution. (b)
Swell presure of CaB, of SA CaB and of HCaB.
of a steady maximum swelling pressure was chosen as

termination criterion.
3.2 Hydraulic conductivity test
Figure 1 shows the swell index of the materials versus
concentration of the KCl and CaCl2 solutions used.
The swell indices of the treated clays were higher compared to that of the untreated bentonite. The HYPER
clay showed a higher swell index compared to both the
untreated clay and the sodium activated clay for KCl
solutions 0.0001 M up to 0.5 M. For CaCl2 solutions
from 0.0001 M to 0.01 M (which is similar to a typical
maximum pore water content that may be expected,
Lin & Benson, 2005) the HYPER clay showed a
higher swell index compared to the untreated clay.
For concentrations of CaCl2 higher than 0.05 M the
swell indices of the three clays were comparable. This
result indicates that the HYPER clay treatment could
potentially perform better compared to the sodium
activation.
For concentrations of CaCl2 between 0.005 M and
0.01 M it was observed an unusual behavior for the
sodium activated clay: a card house structure with
marked macropores was detected (Figure 2a); this
behavior was highlighted with a letter M in Figures 1b and 2b on the columns that showed this
phenomenon. The card house structure with marked
macropores evidently produced false measurements
of the swelling ability of the sodium activated clay
within that range of concentrations. To overcome this
limitation and to represent the actual swelling performance of the clays, a one-dimentional swell pressure
test was performed using a 0.005 M CaCl2 solution.
The swell pressure test, in fact, is not expected to be
affected by this type of macropores formation, because
of the mechanisms that govern the swell pressure phenomenon: when hydrated, the bentonite clay adsorbs a
large amount of water molecules and ions; as a consequence a swell stress is raised by this water entering the
pores among montmorillonite particles and the interstitial layers in individual montmorillonite crystals,
leading to an increase in volume and manifested as
a stress on the surrounding materials. On the other
hand, the macropores formation, in the swell index
test, leads to an increase in volume (due to the structure and the presence of macropores), but do not raise
the interstitial stress between clay layers and on the
surrounding materials.
The results of the swell pressure are given in Figure 2 and 3. The swell pressure of the HYPER clay
is higher compared to that of the sodium activated
clay in the 0.005 M CaCl2 solution, in contradiction
with misleading swell index test results that showed
an apparent opposite behavior due to the macropore formation (Figure 2b). Figure 3 clearly shows
the variation of the swell pressure for the three clays
with time. As expected, both treated clays showed a
higher swell pressure compared to the untreated calcium clay.As shown in the Figure 2b, the swell pressure
of the HYPER clay was higher compared to that of
the sodium activated clay, overcoming the technical
limitations of the swell index test observed and suggesting the potential higher hydraulic performance of
the HYPER clay.
Figure 4 shows the hydraulic conductivity of the
three clays as a function of time (Fig. 4a) and of
pore volumes of flow, PVF (Fig. 4b). As expected, the
hydraulic conductivity of both treated clays were lower
than that of the untreated clay due to the typical thinner
diffuse double layer between clay platelets in a calcium
bentonite. Figure 4 also shows that the hydraulic conductivity of the HYPER clay was lower than that of
the sodium activated clay, indicating a higher performance of the HYPER clay not only in terms of swelling
ability (as seen above) but also in terms of hydraulic
conductivity. These results were in good agreement
with swelling tests results. As expected, the higher the
swell pressure, the lower the hydraulic conductivity of
the materials tested.
The reference hydraulic conductivity to deionized
water of a natural sodium bentonite is of the order
257
The lower hydraulic conductivity of the HYPER

clay is due to its thicker diffuse double layer (DDL) that
is maintained in the long term due to the irreversible
adsorption of the polymer into the clay (Di Emido,
2010b). The polymer, in fact, intercalates between
the clay particles maintaining the interlayer open and
adsorbing lots of immobile water molecules and ions.
This phenomenon restricts the pore space available for
the flow and causes tortuous flow pathways, which is
typical of dispersed structured low permeable clays.
Given that, the thickness of the adsorbed layer is
inversely related to the ions concentration and valence,
bentonites are particularly sensitive to changes in the
composition of the pore fluid. In particular, electrolyte
solutions with high valence (such as the CaCl2 solution
used in this work) cause the thickness of the diffuse
double layer to collapse and therefore the hydraulic
conductivity to increase.
Figure 4. Hydraulic conductivity to 5 mM and 100 mM

CaCl2 solution of the clays analyzed as a function of (a) Time
and (b) Pore volumes of flow.
of 6.4 1012 m/s (for further details see Di Emidio,

2010a). The hydraulic conductivity to 5 mM and
100 mM CaCl2 of the HYPER clay shown in Figure
4 is of the same order (8.76 1012 m/s after 2.79
PVF of 5 mM CaCl2 , and 8.71 1012 m/s after 0.41
PVF of 100 mM CaCl2 ) whereas the hydraulic conductivity of the sodium activated clay was half order of
magnitude higher (2.96 1011 m/s after 8.38 PVF of
5 mM CaCl2 , and 4.44 1011 m/s after 1.42 PVF
of 100 mM CaCl2 ). The hydraulic conductivity of the
calcium clay was 1.38 1010 m/s after 45.15 PVF
of 5 mM CaCl2 solution, and it was 1.43 1010 m/s
after 4.34 PVF of 100 mM CaCl2 solution.
The hydraulic conductivity tests to 5 mM CaCl2
reached both short-term termination criteria (as
defined by ASTM D5084) and long-term termination
criteria for chemical equilibrium between the influent and effluent electrical conductivity (the ratio of
effluent to influent electrical conductivity, EC, was
within the range 1.0 0.05). The EC ratio criterion is a
practical criterion for ensuring that final hydraulic conductivities are reasonably representative of long-term
conditions in which chemical equilibrium has been
achieved in the test specimens, that is, chemical reactions between the permeant liquid and the bentonite
are complete. Conversely, the hydraulic conductivity
tests to 100 mM CaCl2 solution have not reached yet
the termination criteria for chemical equilibrium, for
this reason these tests are still ongoing.
CONCLUSIONS
Swell index and swell pressure tests results showed that

the treatment with the anionic polymer studied here
improved the swelling ability of the untreated calcium
clay compared to the sodium activation method. The
swell index test of the sodium activated clay showed
some limitations in its accuracy due to the macropore
structure of the clay in contact with CaCl2 solutions.
The macropores formed lead to a card house structure with an apparent high swell index. On the other
hand, the swell pressure test was not affected by this
phenomenon. Therefore, the swell pressure provided
more reliable information on the swelling ability of
the clays.
Hydraulic conductivity tests were conducted on
untreated calcium bentonite clay, sodium activated
clay and polymer treated clay. The hydraulic conductivity of the untreated clay was the highest due to a
collapsed double layer thickness. The hydraulic conductivity of the HYPER clay was the lowest. The low
hydraulic conductivity of the HYPER clay is due to
its thick diffuse double layer (DDL), which is maintained in the long term. Further research is in progress
to confirm these results for higher concentrations of
the permeant solution (such as 100 mM and 500 mM
CaCl2 ).
ACKNOWLEDGEMENT
The authors acknowledge J. Seurynck and J. Van Der
Perre for the help provided with the experiments.
REFERENCES
Ashmawy, A. K., El-Hajji, D., Sotelo, N. and Muhammad, N.
2002. Hydraulic performance of untreated and polymertreated bentonite in inorganic landfill leachates. Clays and
Clay Minerals, 50(5): 546552.
258
Betega de Paiva, L., Morales, A.R., Valenzuela Daz, F.R.

2008. Organoclays: Properties, preparation and applications. Applied Clay Science, Elsevier, 42: 824.
Bohnhoff, G.L. & Shackelford, C.D. 2010. Global Geoenvironmental Engineering Challenges. The First USIndiaWorkshop on Global Geoenvironmental Engineering
Challenges. November 7, New Delhi, India.
Bouazza, A. 2001. Geosynthetic clay liners, Geotextiles and
Geomembranes, Elsevier 20: 317.
Di Emidio, G. 2010a. Hydraulic and Chemico-Osmotic Performance of Polymer Treated Clays, PhD thesis, Ghent
University.
Di Emidio, G. 2010b. Clayey barriers. Patent Pending:
PCT/EP2011/064542, WO2012/025564 A1.
Egloffstein, T.A. 2001. Natural bentonites-influence of the
ion exchange and partial desiccation on permeability
and self-healing capacity of bentonites used in GCLs,
Geomembranes and Geotextiles, Elsevier 19: 427444.
Filippi, S., Mameli, E., Marazzato, C., Magagnini, P.
(2007). Comparison of solution-bleding and meltintercalation for the preparation of poly(ethylenecoacrylic acid)/organoclay nanocomposites., European
Polymer Journal. Elsevier, 43(5): 16451659.
Flynn, B. N. & Carter, G. C. 1998. Waterproofing material
and method of fabrication thereof. US 6537676B1
Guyonnet, D., Gaucher, E., Gaboriau, H., Pons, C.-H., Clinard, C., Norotte, V., Didier, G. 2005. Geosynthetic Clay
Liner Interaction with Leachate: Correlation between Permeability, Microstructure, and Surface Chemistry, Journal of Geotechnical and Geoenvironmental Engineering,
ASCE, 131(6): 740749.
Jo, H., Katsumi, T., Benson, C. H., Edil, T. B. 2001. Hydraulic
conductivity and swelling of non-prehydrated gcls premeated with single species salt solutions, Journal of
Geotechnical and Geoenvironmental Engineering, ASCE
127(7): 557567.
Katsumi, T. 2010. Geosynthetic Clay Liners for Waste Containment Facilities, Bouazza and Bowders eds., Chapter 4.
Katsumi, T., Ishimori, H., Onikata, M., Fukagawa, R. 2008.
Long-term barrier performance of modified bentonite
materials against sodium and calcium permeant solutions,
Geotextiles and Geomembranes, Elsevier 26: 1430.
Kondo, M. 1996. Method of activation of clay and activated
clay. 5573583.
Lee, J.M., Shackelford, C. D., Benson, C. H., Jo, H. Y.
and Edil, T. B. 2005. Correlating index propertis and
hydraulic conductivity of geosynthetic clay liners, Journal of Geotechnical and Geoenvironmental Engineering,
131(11): 13191329.
Lin, L.-C. & Benson, C.-H. 2005. Effect of wet-dry cycling on
swelling and hydraulic conductivity of GCLs. Journal of
Geotechnical and Geoenvironmental Engineering, ASCE,

126(1): 4049.
Lorenzetti, R. J., Bartelt-Hunt, S. L., Burns, S. E. and Smith, J.
A. 2005. Hydraulic conductivities and effective diffusion
coefficients of geosynthetic clay liners with organobentonite amendments. Geotextiles and Geomembranes, 23:
385400.
Malusis, M., McKeehan, M.D., LaFredo, R.A. 2010. Multiswellable bentonite for soil-bentonite vertical barriers.
Proceedings of the 6th ICEG, November 812, New Delhi,
India.
Mazzieri, F., Di Emidio, G. and Van Impe, P. O., 2010. Diffusion of CaCl2 in a modified bentonite: impact on osmotic
efficiency and hydraulic conductivity, Clays and Clay
Minerals, 58(2): 351363.
Mitchell, J. 1993. Fundamentals of Soil Behaviour, John
Wiley & Sons.
Norrish, K. 1954. The swelling of montmorillonite, Transactions Faraday Society, 18: 120134.
Onikata, M., Kondo, M., Hayashi, N. and Yamanaka, S.
1999. Complex formation of cation-exchanged montmorillonites with propylene carbonate: Osmotic swelling in
aqueous electrolyte solutions., Clays and clay minerals.
47(5): 672677.
Onikata, M., Kondo, M. and Kamon, M. 1996. Development and characterization of a multiswellable bentonite,
in M. K. ed. Balkema Rotterdam (ed.), Environmental
Geotechnics, pp. 587590.
Palomino, A.M. 2010. Polymer-enhanced geomaterials for
use in geoenvironmental applications, The First USIndiaWorkshop on Global Geoenvironmental Engineering
Challenges. November 7, New Delhi, India.
Petrov, R. J. and Rowe, R. K. 1997. Geosynthetic clay liner
(gcl) chemical compatibility by hydraulic conductivity
testing and factors impacting its performance, Canadian
Geotechnical Journal, 34: 863885.
Scalia, J., Benson, C.H., Edil, T.B., Bohnhoff, G.L.,
Shackelford, C.D. 2011. Geosynthetic Clay Liners containing bentonite polymer nanocomposite. Proceedings of
the Geo-Frontiers 2011 Congress, Dallas, Texas.
Schroeder, C., Monjoie, A., Illing, P., Dosquet, D. and
Thorez, J. 2001. Testing a factory-prehydrated gcl under
several conditions, in CISA (ed.), Proc. Sardinia 2001,
Cagliari, Italy, Vol. 1.
Shackelford, C. D., Benson, C. H., Katsumi, T., Edil, T. B.
and Lin, L. 2000. Evaluating the hydraulic conductivity
of gcls permeated with non-standard liquids, Geotextiles
and Geomembranes, Elsevier 18: 133161.
Simon, F. G. and Mller, W. W. 2005. Standard and alternative landfill capping design in Germany, Environmental
Science and Policy, Elsevier 7(4): 277290.
259

Dynamic shear behavior of geosynthetic-soil interface and numerical

implementation considering chemical effect
C.W. Kwak
Seoul National University, Seoul, S. Korea
I.J. Park
Hanseo University, Seosan-si, Chungcheongnam-do, S. Korea
J.B. Park
Seoul National University, Seoul, S. Korea
J.K. Kim
Gyeonggi Railway Co. Ltd., Yongin-si, Gyeonggi-do, S. Korea
ABSTRACT: Geosynthetics are normally used in waste landfill sites for various purposes. Geosynthetics in
soil compose geosynthetic-soil interface by contacting soil particles. The remarkable differences of the intrinsic
material characteristics at the interface induce complex stress-strain behaviors. In particular, the chemical elements in the leachate affect the dynamic shear behavior of geosynthetic-soil interfaces. In this study, the dynamic
shear tests of geosynthetic-soil interfaces are performed using multi-purpose interface apparatus (M-PIA), which
was recently developed in South Korea. The specimens are submerged into acid, neutral and basic solutions for
a period and the effect of chemical element is analyzed. Test results show different responses according to the
pH value of the solutions. The Disturbed State Concept (DSC) and disturbance function have been employed to
suggest better approaches to analyze the dynamic shear behavior of the interface considering chemical effect.
Consequently, the numerical implementation of the modified DSC equations has been performed.
1
1.1
INTRODUCTION
General
Geosynthetics are some of the most widely used

materials in waste landfills as they exhibit great
performance in filtering, separation, drainage, reinforcement, and lining. Various types of soils are also
also used in landfills in conjunction with geosynthetics. Therefore, geosynthetics in soil compose
geosynthetic-soil interface by contacting soil particles and leachate in the case of landfills. Since these
interfaces typically involve side slopes, stability is
a critical consideration for the waste landfill site
design (Triplett et al., 2001). The cover system of the
landfill is illustrated in Figure 1. It can contain various geosynthetics such as geogrids, geomembranes,
geocomposites, geonets, geotextiles, etc.
As shown in Figure 1, geosynthetics have significant functions in waste landfills, therefore, the
behavior of geosynthetics has been intensively studied.
The intrinsic characteristics of the geosynthetic-soil
system with respect to the interface was studied by
1988 failure case in Unit B-19 at the Kettleman Hills
Class hazardous facility (Byrne et al., 1992).A number
of conventional researches were performed to study
Figure 1. Geosynthetics in landfill design (redrawn after

Zornberg & Christopher, 1999).
static shear behavior of geosynthetic-soil interface by

direct shear test (Seo et al., 2004, P. J. Fox, 1998,
Gilbert et al., 1996).
In the past years the dynamic shear behavior
of geosynthetic-soil interface was studied by various researchers. Yegian and Lahlaf (1992) performed
shaking table tests to measure the dynamic interface shear strength properties between geotextiles and
geomembranes, then an innovative use of geosynthetics as base isolation for earthquake hazard mitigation
261
was proposed and its applicability was demonstrated.

Al-Douri and Poulos (1992) examined the cyclic
behavior of different types of carbonate sediments and
one type of silica sand under direct shear test conditions. They found that the cyclic behavior depends
significantly on the number of cycles and the cyclic
displacement amplitude and the amount of crushing
caused by cyclic loading is depend strongly on particle shape and grading characteristics. P.J. Fox (2011)
studied the damage of geomembrane due to static and
cyclic sharing.
A waste containment facility liner system must
not only provide a sound hydraulic barrier but must
also be structurally stable during all phases of a
project (i.e. during construction, operation, and closure) (Bouazza et al., 2002). Geosynthetics must
secure static and dynamic stability and also chemical
resistance. Geosynthetic-soil interface is a critical part
which behavior is difficult to analyse. Especially, the
chemical elements in the leachate affect the dynamic
properties of the geosynthetic-soil interface (B. P.
Shiels, 2005, S. A. Jefferis, 2001).
1.2
2.1
According to the DSC, the disturbance (D), can define

the degree of material damage quantitatively. Internal micro structural damage is assumed to be mainly
affected by the deviatoric plastic strain trajectory, D ,
hence,
where t is time or the number of loading cycle and

i is environmental factors (Kwak et al., in press).
In the functional form based on the modification of
Weibull (1951) function is,
where A and Z are intrinsic material parameters and

Du asymptotically approaches 1.0 but can never reach
1.0 (Desai, 2001). In this study, cyclic shear test
results are used to estimate the disturbance, therefore,
the disturbance can be calculated by Equation (3) as
following,
Objectives
In this study, the dynamic shear test, to simulate an

earthquake condition, of geosynthetic-soil interface
using a multi-purpose interface apparatus (M-PIA)
which was recently developed in South Korea, were
utilized to investigate the effect of chemical element on the interface. Not a real leachate nor a
synthetic leachate but basic, neutral and acid conditions are considered to represent chemical element.
The Disturbed State Concept (DSC) and the disturbance function have been employed and modified to
express the chemical effect on the dynamic behavior of the interface. Furthermore, simple numerical
implementation of the modified DSC equations was
conducted to verify the compatibility of the suggested
equations.
2
2.2 Disturbance function
DISTURBED STATE CONCEPT
where i, a, and c denote Relative Intact (RI), observed,

and Fully Adjusted (FA) states, respectively and the
equation evaluates the normalized shear stress degradation, D from each hysteretic loop (Kwak et al., 2013,
in press). Refer to Figure 2 and related Equations (4),
(5) to calculate D and D values directly from the test
results.
Deviatoric plastic strain trajectory, D represents

the accumulated plastic strain at each stress-strain
loop, therefore, it can be calculated by summation of
projected length on the strain axis in each loop.
Outline
The main concern of this study is on the geosyntheticsoil interface. Even though there are a number of
advanced constitutive models such as Mohr-Coulomb,
Modified Cam-Clay, etc., these conventional models
are not enough to predict or simulate the realistic
damage progress at the geosynthetic-soil interface
(Park & Desai, 2000).
The disturbed state concept (DSC) is based on
the basic physical consideration that the observed
response of a material can be expressed in terms of
the responses of its constituents, connected by the disturbance function (Desai, 2001). In other words, the
current damage of a material can be estimated by the
disturbed state with respect to its initial and final states.
The detailed contents of the theory are explained in the
reference by Desai.
EXPERIMENTAL RESEARCH
3.1 New apparatus

A new apparatus, M-PIA has been developed and
modified to examine the geosynthetic-soil interface
considering chemical effect under dynamic loading
condition. Figure 3 shows the whole M-PIA unit.
A geosynthetic-soil specimen is placed on top of
the circular shear box. 5 shear rings which are 2 mm
in thickness each are placed on the top of shear box to
simulate simple shear conditions. Wire draw transducers are linked with 1st, 3rd and 5th shear rings from
the top to record shear displacement. Figure 4 displays
shear box and wire draw transducers.
Table 1 below demonstrates the mechanical specification of M-PIA.
262
Figure 4. Shear box and wire draw transducers.

Table 1. Mechanical specification of M-PIA (Kwak et al.,
2013, in press).
Device
Specifications
Loading unit
Horizontal load: Max. 50 kN

Vertical load: Max. 50 kN
Electric-servo control type servo
motor
Precise helical structure
Linear motion slides
Max. 1 Hz (horizontal)
Embedded strain control and
logging program
Min. 0.03 second of data logging
interval
Effective wire length: 0250 mm
Max. resolution: 0.01 mm
Outer diameter: 125 mm
Inner diameter: 100 mm
Thickness: 2.0 mm
Loader
Figure 2. Calculation of D and D (Kwak et al, 2013, in

press.)
Range of vibration
Data logger
Wire draw transducer

Shear rings
Figure 3. M-PIA outline.
3.2 Test conditions

A composite type of geosynthetic which is one of the
most commonly applied to the waste landfill site in
South Korea was utilized in this study. For this type
of geosynthetic, a high density polyethylene (HDPE)

geogrid is covered with a nonwoven fabric on both
sides. 15 kN of normal loads to simulate the in-situ
vertical stress was applied. Jumunjin sand was also
used as soil specimen. The dry unit weight of Jumunjin
sand is 15.4 kN/m3 , and the coefficient of uniformity is 1.9. Jumunjin sand is spread in the shear
box by air-pluviation method with 60% of relative
density. Figure 5 shows the cross section of the geosynthetic and Table 2 displays the specification of the
geosynthetics.
Since the chemical effects on the geosynthetic-soil
interface is the main issue in this study, the chemical
conditions of the specimen are important. In this study,
not a real leachate nor a synthetic leachate but basic,
neutral and acid conditions are considered to represent
chemical element and both geosynthetic and soil were
submerged in those solutions. Chemical solutions were
prepared by using standard pH solutions and an automatic pH meter was used for pH measurement. Table 3
shows the chemical conditions.
263
Table 4.
Summary of test conditions.
Solutions
Remarks (15 kN of normal load)
Neutral
Acid
Basic
Sum
1 10 = 10 sets of tests
Table 5. Estimated A and Z parameters (Kwak et al., 2013,

in press).
Average of 10 tests
Figure 5. Cross section of the composite type of

geosynthetic.
Chemical Condition
Acid
Neutral
Basic
0.626
0.256
0.454
0.9756
1.2524
1.1837
Table 2. Specification of geosynthetics (Kwak et al., 2013,

in press).
Specifications
Remark
Manufacturer: GOLDENPOW
(Seoul, Korea, www.goldenpow.com)
Product name: Geocomposite
Type: Composite type of geotextile
and nonwoven fabric
Thickness: 7.0 mm
Mass per unit area: 9.4 kg/m2
Carbon black content: 2.2%
Min. yield tensile strength: 14.7 MPa
Min. breaking strength: 26.5 MPa
Yield tensile stain: 12%
Oxidative induction time: 100 min.
(standard condition)
Oxidative induction time: 400 min.
(pressured condition)
Table 3.
ASTM D5261
ASTM D4595
ASTM D4632
ASTM D4632
ASTM D3895
Figure 6. Shear stress-strain loops under basic condition

(Kwak et al., 2013, in press).
ASTM D3895
thanks to equations (3), (4), and (5). Based on the calculation results, the intrinsic material parameters, A
and Z could be estimated using Equation (6) which is
the linear regression form of Equation (2).
Chemical conditions (Kwak et al., 2013, in press).
Solutions
Basic
Neutral
Acid
pH
Submerging period
12.5
200 days
6.9
200 days
3.1
200 days
1 Hz of cyclic load which reproduces an earthquake

condition is applied to the upper module of shear box.
100 mm of the maximum shear movement is exerted
after the normal load reached 15 kN, as a reference
value.
Those material, chemical, and dynamic conditions
are summarized in Table 4.
3.3 Test results
30 sets of cyclic shear tests were conducted in total.
Disturbance (D) and deviatoric plastic strain trajectory (D ) were calculated as explained on Figure 2 and
Table 5 displays the calculated A and Z values based

on the test results and Equation (6).
The estimated A and Z values represent the degree
of damage of geosynthetic-soil interface under the test
conditions. A representative plot of shear stress-strain
loops of the interface under chemical condition are
shown in Figure 6.
As shown in Figure 6, the geosynthetic-soil interface is considered to fail after 50 cycles of cyclic
loading since shear stress degradation converges.
Therefore, the interface reaches the Fully Adjusted
state after the 50th cycle. The shear stress degradation which means the accumulation of damage as the
cyclic loads appears distinctly.
264
Figure 9. Algorism flow of FISH programming.
Figure 7. Comparison of disturbance function (basic

condition) (Kwak et al., 2013, in press).
Figure 8. Reproduced disturbance function (Kwak et al.,

2013, in press).
Figure 7 displays the estimated disturbance by

the linear regression method in comparison with test
results, under the basic condition which marks 12.5 of
pH. The reproduced disturbance functions under the
all chemical conditions are shown in Figure 8.
Based on Figure 8, the acid condition shows the
most rapid damage at the same deviatoric plastic
strain level and the neutral condition displays the
least damage. It is induced that different A and Z
values according to chemical conditions result in different disturbance function shapes because the damage
of geosynthetic-soil interface depends on the characteristics of chemical aggressor. Furthermore, it is
known that the degree of damage of the geosyntheticsoil interface can be estimated quantitatively by the
disturbance function parameters.
those two conditions. If the behaviors of geosyntheticsoil interface under the chemical condition are reproducible, it can be applied to solve various practical
problems such as design and analysis of complicated
waste landfill site. The reproducibility of the interface
behavior mainly depends on the simulation of the test
by numerical implementation of disturbance function.
The disturbance (D) can be directly obtained by
the test results and equation (3). The intrinsic material
parameters, A and Z are also known, hence, the shear
stress-strain curves ( a ) are reproducible by combining
Equations (2) and (3).
In this study, the commercial code, Fast Lagrangian
Analysis of Continua (FLAC) and its embedded language FISH are employed to build up the user-defined
shear stress-strain relationship. FISH enables the user
to define new variables and functions. Figure 9 represents the essential algorism flow of the numerical
implementation of the disturbance function.
The initial disturbance, D is calculated by using the
equation (2) and the observed (present) shear stress
states, a are estimated, since the initial shear stresses,
i , and shear stresses at failed, c , are known. a
is updated by calculated a as the calculation step
marching. Then the all procedures are repeated.
In the numerical implementation, Mohr-Coulomb
model is utilized to represent the RI state, and sinusoidal wave function is applied on the surface of mesh.
Figure 10 displays the reproduced shear stress-strain
curves as result and the estimated disturbance function
curve is shown in Figure 11.
As shown in Figure 11, shear stress degradation is
obviously observed. Increase of shear strain is also
noticeable. Disturbance (D) shows dramatic increase
after the initial strain level.
Note that this try is only to confirm that the possibility of the implementation of DSC function into
the commercial program, therefore, the specific application and modification of the FISH program, and
comparison between test results will be conducted for
further study.
NUMERICAL IMPLEMENTATION
5
Cyclic shear tests were performed to obtain the disturbance function parameters. DSC has been employed
to define quantitatively the initial condition (RI state),
failed condition (FA state) and the process between
SUMMARY AND CONCLUSIONS
The chemical effects on the cyclic shear behavior

of a geosynthetic-soil interface were reviewed and
a numerical implementation of disturbance function
265
ACKNOWLEDGEMENTS
This work (No.2012045040) was supported by
Mid-career Researcher Program through NRF grant
funded by the MEST.
REFERENCES
Figure 10. Reproduced shear stress-strain curves (basic

condition).
Figure 11. Estimated disturbance function curve D.
using the commercial code was also tried in this study.

Based on the test results and Disturbed State Concept
equations, the intrinsic material parameters, A and Z
which express the degree of damage were obtained
(Kwak et al., 2013 in press). A newly developed apparatus, M-PIA was briefly introduced and utilized in the
tests. It is induced that the damage of geosynthetic-soil
interface depends on the characteristics of different
chemical aggressor. Furthermore, it is confirmed that
the developed algorism based on FISH to simulate
the tests has been successfully simulate shear stress
degradation and disturbance function.
For further study, practical test results shall be compared with the numerical implementation results.After
that, it is quite sure that the numerical analysis will
provide convenient and strong solution to predict the
cyclic shear behavior of geosynthetic-soil interface
considering chemical conditions.
Al-Douri, R H. & Poulos, H. G. 1992. Static and cyclic

direct shear tests on carbonate sands. Geotechnical Testing
Journal, 15(2): 138157.
Bouazza,A., Zornberg, J. G. &Adam, D. 2002. Geosynthetics
in waste containment facilities: recent advances. Delmas,
Gourc & Girard (ed.), Geosynthetics 7 ICG, Swets &
Zeitlinger, Lisse.
Byrne, R. J., Kendall, J. & Brown, S. 1992. Cause and
mechanism of failure, Kettleman Hills Landfill B-19, unit
IA. Proc., ASCE Spec. Conf. on Perf. And Stability of
Slopes and Embankments II(2) ASCE, New York, NY :
11881215
Desai, C.S. 2001. Mechanics of Materials and Interfaces
The disturbed state concept CRC Press LLC, Boca Raton,
Florida, USA: 33209.
Fox, P. J., Rowland, M. G. & Scheithe, J.R. 1998. Internal
Shear Strength of Three Geosynthetic Clay Liners. Journal of Geotechnical and Geoenvironmental Engineering,
124(10): 933944.
Fox, P. J., Ross, J. D., Sura, J. M., & Thiel, R. S. 2011.
Geomembrane damage due to static and cyclic shearing
over compacted gravelly sand. Geosynthetics International, 18(5): 272279.
Jefferis, S. A. 2001. Permeability a dynamic property of
barrier materials. Proceedings of the International Containment and remediation technology Conference and
Exhibition, Orlando, Florida: 427429.
Kwak, C.W., Park, I.J., & Park, J.B. 2013. Modified Cyclic
Shear Test for Evaluating Disturbance Function and
Numerical Formulation of Geosynthetic-Soil Interface
Considering Chemical Effect. ASTM Geotechnical Testing
Journal, In Press at May.
Park, I.J. & Desai, C.S. 2000. Cyclic Behavior and Liquefaction of Sand using Disturbed State Concept. Journal of
Geotechnical and Geoenvironmental Engineering,ASCE,
126(9): 834846.
Seo, M. W., Park, I.J. & Park, J.B. 2004. The Influence
of Moisture on the Interface Shear Strength Between
Geosynthetics. Journal of Korean Geotechnical Society,
20(2): 7585.
Shiels, B. P. 2005. Performance Evaluation of Chemical
Protective Clothing Materials under Dynamic Mechanical Deformation. A M.S thesis at North Carolina State
University: 5573.
Triplett, E. & Fox, P. 2001. Shear Strength of HDPE Geomembrane/Geosynthetic Clay Liner Interfaces. Journal of
543552.
Weibull, W. 1951. A Statistical Distribution Function of Wide
Applicability Journal of Applied Mechanics. September,
293297.
Yegian, M. & Lahlaf, A. 1992. Dynamic Interface Shear
Strength Properties of Geomembranes and Geotextiles.
Journal of Geotechnical. Engineering, 118(5): 760779.
Zornberg, J. G. & Christopher, B. R. 1999. Geosynthethics.
Chapter 27, The Handbook of Groundwater Engineering,
CRC Press, Inc., Boca Raton: Florida.
266

Influence of membrane behavior on solute diffusion through GCLs

M. Malusis
Bucknell University, Lewisburg, Pennsylvania, USA
J. Kang
Engineering Analytics, Inc., Fort Collins, Colorado, USA
C. Shackelford
Colorado State University, Fort Collins, Colorado, USA
ABSTRACT: The relationship between membrane efficiency and diffusion for a geosynthetic clay liner (GCL)
was investigated by conducting multi-stage membrane/diffusion tests on four GCL specimens at different effective stresses ( = 34.5242 kPa) in flexible-wall cells. Each test was conducted by circulating five different
source KCl solutions, sequentially from lowest to highest concentration (Co = 3.9, 6.0, 8.7, 20, 47 mM), across
the top specimen boundary, while circulating de-ionized water across the bottom boundary. Membrane efficiency coefficients () were determined from differential pressure measurements, and effective salt-diffusion
coefficients (Ds ) were inferred from electrical conductivity measurements. Increases in Ds with increasing Co
were observed for all specimens and were correlated to decreases in . In each case, Ds approached a maximum
as 0 and decreased linearly toward Ds = 0 as 1, regardless of the applied . The results support the
hypothesis from prior theoretical studies that Ds for clay membranes may be expressed in terms of a restrictive
tortuosity factor, r (or effective porosity ratio, ), represented by the simple linear function r (=) = 1 .
INTRODUCTION
The ability of clays to act as semipermeable membranes that inhibit the passage of solutes while allowing relatively unimpeded migration of water is well
recognized. While much of the historic literature on
membrane behavior pertains to natural clays and shales
(e.g. Young & Low 1965, Greenberg et al. 1973,
Kharaka & Berry 1973, Marine & Fritz, 1981, Neuzil
1986), many of the more recent studies have investigated such behavior in engineered clay barriers,
including geosynthetic clay liners (GCLs), compacted
clay liners, and soil-bentonite vertical cutoff walls
(e.g. Keijzer et al. 1999, Malusis & Shackelford
2002a,b, Yeo et al. 2005, Henning et al. 2006, Kang &
Shackelford 2010, 2011, Mazzieri et al. 2010). These
studies show that GCLs are likely to exhibit significant membrane behavior due to the high content
of sodium bentonite (Na-bentonite) in these barriers.
Such behavior can reduce the flux of aqueous miscible
contaminants (solutes) through a GCL due to hyperfiltration, chemico-osmotic flow, and reduced diffusion
(Malusis et al. 2003).
The influence of membrane behavior on solute
diffusion in GCLs has been investigated in a limited number of experimental studies (e.g. Malusis &
Shackelford 2002b, Di Emidio 2010). These studies
indicate that the effective salt-diffusion coefficient,
Ds , decreases with increasing membrane efficiency
as represented by the membrane efficiency coefficient, . Because this effect of decreasing Ds with
increasing is not explicitly captured in theoretical
formulations for coupled solute flux based on nonequilibrium thermodynamics (e.g. see Malusis et al. 2012),
the effect has been referred to as implicit coupling
(e.g. Malusis & Shackelford 2004a,b).
Manassero & Dominijanni (2003) proposed that Ds
for clay membranes may be expressed as a simple
linear function of . Although this linear function is
supported theoretically by pore-scale physical modeling (Dominijanni 2005, Dominijanni & Manassero
2012), experimental support is limited (Malusis et al.
2012). In this study, the influence of on Ds for
GCLs is investigated based on multi-stage membrane
tests conducted on GCL specimens by Kang (2008).
Although the values for these specimens were
reported by Kang & Shackelford (2011), the diffusion results inferred from the work of Kang (2008)
heretofore have not been published. Thus, these results
will be used herein to evaluate the proposed linear
relationship between Ds and .
2
BACKGROUND
Consider a salt-diffusion experiment in which a clay

membrane is placed between two sealed reservoirs
(Fig. 1). The source reservoir contains binary salt
267
Figure 1. Salt diffusion of KCl through a clay in a closed

system (DIW = de-ionized water).
(KCl) solution, whereas the collection reservoir contains de-ionized water (DIW). The diffusive molar
fluxes, J d , of the salt cation (c) and the salt anion (a)
may be expressed as follows based on Ficks law:
where n is total porosity and Cc and Ca are the molar

concentrations of the cation and anion, respectively.
The effective salt-diffusion coefficient, Ds , is the
product of the salt-diffusion coefficient in free solution, Dso , and the apparent tortuosity factor, a (i.e.
Ds = a Dso ). Also, a can be defined as the product of a matrix tortuosity factor, m , and a restrictive
tortuosity factor, r , as follows:
where m accounts for tortuosity associated with the

geometry of the interconnected pores and r accounts
for any number (N ) of other mechanisms (represented
by i ) that restrict diffusion, such as solute exclusion
and solute drag near particle surfaces (e.g. Kemper
et al. 1964, Shackelford & Daniel 1991, Shackelford &
Moore 2013). Based on Eq. 2, Ds may be written as
follows:
where Dse (=m Dso ) is the effective salt-diffusion coefficient that accounts only for the matrix tortuosity.
Since m is associated solely with the geometric interconnectivity of the pores, m and Dse generally are
considered constant for a given arrangement of soil
particles and, therefore, independent of solute concentration. In contrast, r for clay membranes decreases
with increasing . Theoretically, r = 0 for ideal membranes ( = 1) that completely exclude solutes. However, higher solute concentrations cause shrinkage of
the diffuse double layers (DDLs) surrounding the clay
particles and a decrease in , such that r 1 as
0, assuming that all other potentially restrictive effects are insignificant. Under this assumption,
Ds ( = 0) = Dse based on Eq. 3.
Alternatively, restricted diffusion in clay membranes may be represented as a porosity restriction
where the effective porosity, ne , or the portion of the
porosity that allows migration of both the solutes and
Figure 2. Restrictive tortuosity factors for bentonite specimens in KCl or CaCl2 solutions (replotted after Malusis
et al. 2012). Data for GCL from Malusis & Shackelford
(2002b); data for Na-bentonite and HYPER clay from Di
Emidio (2010).
the solvent, is less than n (e.g. see Shackelford &

Moore 2013). On this basis, Manassero & Dominijanni (2003) expressed r in Eq. 3 as an effective
porosity ratio, (=ne /n, where ne = n and = 1
when = 0, and ne = = 0 when = 1). Furthermore, Manassero & Dominijanni (2003) proposed
that (=r ) may be approximated as a simple, linear
function of , as follows:
Substitution of Eqs. 34 into Eq. 1 yields the following alternative expressions for diffusive flux of the
cation and anion in Figure 1:
Although other expressions relating r to may be

possible (Manassero & Dominijanni 2003), Eq. 4 has
been shown to be theoretically valid based on porescale physical modeling under the assumption that
pore-scale variations in pressure, ion concentration,
and water velocity within the membrane are negligible (Dominijanni 2005, Dominijanni & Manassero
2012). Also, r values from Malusis & Shackelford
(2002b) and Di Emidio (2010) for specimens of
Na-bentonite or polymer-modified Na-bentonite
(HYPER clay) generally follow the linear trend defined
by Eq. 4, as illustrated in Figure 2. However, given the
paucity of data and the scatter in Figure 2, additional
data are needed to bolster the validity of Eq. 4.
3
3.1 GCL and KCl solutions

The GCL in this study is the same as that used
by Malusis & Shackelford (2002b) and is sold as
Bentomat DN by Colloid Environmental Technologies Co. (CETCO, USA). As described by Kang &
Shackelford (2011), the bentonite component of the
GCL contained 71% smectite (montmorillonite) and
268
Figure 3. Relationship between measured electrical conductivity and salt concentration for KCl and NaCl solutions
at 25 C.
53% sodium on the exchange sites. The measured

cation exchange capacity was 47.7 cmolc /kg, and the
liquid limit and plasticity index were 478 and 439,
respectively.
The KCl solutions used in this study ranged in concentration from 3.9 mM to 47 mM and were prepared
by dissolving KCl crystals in DIW. The relationship
between electrical conductivity, EC, and salt concentration for these solutions is illustrated in Figure 3
along with the same relationship for similarly prepared
NaCl solutions, for comparison.
3.2 Testing apparatus
The testing apparatus, described by Kang &
Shackelford (2011), consisted of a flexible-wall cell
and a hydraulic control system (syringe pump and
stainless steel tubing) to circulate different solutions across the boundaries of a GCL specimen (see
Fig. 4). During membrane/diffusion testing, a source
KCl solution (Cot > 0) and DIW (Cob = 0) are circulated across the top and bottom of the specimen,
respectively, under closed-system conditions such that
volume change within the system is prevented. When
the specimen acts as a semipermeable membrane, the
applied difference in KCl concentration across the
specimen induces a pressure difference (due to prevention of chemico-osmotic liquid flux through the
specimen) that is measured with a differential pressure transducer. Furthermore, salt diffusion through
the specimen results in collection of solutes in the DIW
circulated across the bottom boundary, such that the
solute concentration exiting the bottom boundary, Cb ,
is greater than that in the DIW (i.e. Cb > Cob ). This
difference between Cb and Cob provides the basis for
determining Ds for the specimen. Further details of
this apparatus are provided by Kang & Shackelford
(2009, 2011).
3.3
Specimen preparation
Four circular specimens of the GCL, with nominal

diameters of 102 mm and thicknesses of 10 mm, were
Figure 4. Schematic of flexible-wall membrane/diffusion

test apparatus (redrawn after Kang & Shackelford 2009).
cut from a larger GCL sheet and placed on the base

pedestal of a flexible-wall permeameter. Each specimen was subjected to an effective stress, , of 34.5 kPa
(5 psi) under 172 kPa (25 psi) back pressure and permeated with DIW to saturate the specimen, measure
the baseline hydraulic conductivity, k, and remove
most of the soluble salts initially contained within
the specimen. After permeation, the specimens were
transferred to the flexible-wall membrane cells (Fig. 4)
and again subjected to = 34.5 kPa. Once consolidation reached completion under = 34.5 kPa, three of
the four specimens were further consolidated under
final values of = 103 kPa (15 psi), 172 kPa (25 psi),
or 241 kPa (35 psi) by increasing the cell pressure in
a single loading step (the back pressure of 172 kPa
[25 psi] was maintained constant in all tests). Changes
in specimen height were estimated based on changes
in porosity computed using measured changes in cellwater volume. The drainage (back-pressure) lines were
closed after the consolidation stage and prior to the
start of the membrane/diffusion tests. See Kang &
Shackelford (2011) for more details.
3.4 Membrane/diffusion testing
The membrane/diffusion tests were initiated by circulating DIW through the top and bottom boundaries of
each specimen at a circulation rate of 4.2 1010 m3 /s
for 7 d to establish a steady baseline pressure difference. This circulation rate has been proven to
be sufficiently fast to maintain reasonably constant
concentration boundaries (Malusis et al. 2001). The
membrane efficiency measurements then were initiated by circulating the 3.9 mM KCl solution across the
top specimen boundary while continuing circulation
of DIW across the bottom boundary. The differential pressure induced across each specimen, P, and
269
the EC of the solutions exiting the top and bottom

boundaries (EC t and EC b , respectively) were measured until P, EC t , and EC b became steady. Once
the tests were completed for the 3.9 mM KCl solution,
four additional stages were performed in which the
source KCl concentration was increased sequentially
from 3.9 mM to 6.0, 8.7, 20, and 47 mM. Each stage
lasted 7 d, which was sufficient in all cases to achieve
steady P, EC t , and EC b .
In each stage, the Cl concentrations in samples of
the circulation outflows from the top and bottom specimen boundaries (Ct and Cb , respectively, in Fig. 4)
were estimated using the calibration curve for KCl
shown in Figure 3. These estimated concentrations are
based implicitly on the assumption that the only solutes
contributing to the EC in the circulation outflows
from the specimen boundaries were chloride (Cl )
and potassium (K+ ). These estimated concentrations
are considered reasonably accurate, in part because
the GCL specimens were permeated with DIW to
remove excess soluble salts from the pore water of
the specimens before testing. As a result of this permeation step, the contribution of soluble salts to EC t
and EC b was small. Also, although cation exchange
of K+ for Na+ (the predominant exchangeable cation
species) likely was occurring during the tests, the EC
calibration curves for KCl and NaCl in Figure 3 are
reasonably similar. Based on these calibration curves,
Cl concentrations estimated from the NaCl curve
would be 14 % higher than those estimated from the
KCl curve.
Determination of Ds for each testing stage was
based on the steady-state approach in which the
estimated Cl concentrations were converted to cumulative moles per unit area, Qt , and plotted as a
function of elapsed time, t (Shackelford 1991).
Values of Ds were then computed using the following
expression:
where Qt /t is the steady-state diffusive molar flux

(i.e. the slope of the Qt vs. t data at steady state), L
is the specimen thickness, and Cb,ave and Ct,ave are the
average molar Cl concentrations at the bottom and
top specimen boundaries, as follows:
Membrane efficiency coefficients at steady-state

also were computed from average boundary concentrations, as follows:
Figure 5. Representative results from a multi-stage membrane/diffusion test on GCL specimen ( = 34.5 kPa): (a)
induced pressure difference versus time; (b) boundary EC
versus time (replotted after Kang & Shackelford 2011).
4.1 Pressure difference and boundary EC

Differential pressures induced across the GCL specimens, P (>0), and EC values measured in the
outflows from the top and bottom specimen boundaries (EC t and EC b , respectively) were plotted as a
function of cumulative elapsed time for each test (see
Kang & Shackelford 2011). Two of these plots are
illustrated in Figure 5 for the specimen consolidated
at = 34.5 kPa. In each test, an initial P was
induced across the specimen while circulating DIW
across both boundaries (see Fig. 5a). This initial P
was nearly zero in all cases and was subtracted from
the P measured after introducing the KCl solutions when computing using Eq. 8. Likewise, initial
EC t and EC b values greater than that of the DIW were
measured during DIW circulation across both boundaries (see Fig. 5b), due to release of residual salts that
were not removed during permeation. These initial EC
values remained steady during DIW circulation and
were subtracted from the EC t and EC b values used to
estimate the boundary KCl concentrations based on
Figure 3.
4.2
where R is the universal gas constant (8.314 J/mol K)

and T is absolute temperature (K).
Effective diffusion coefficient
The results of the multi-stage tests are summarized

in Table 1. The steady-state diffusive fluxes, Qt /t,
were obtained from the slopes of the steady (linear) portions of the Qt versus t plots for each stage,
270
Table 1.
Summary of multi-stage membrane/diffusion test results for four GCL specimens (Cob = 0 in all stages).
Values at steady state
(kPa)
Cot
(mM)
L
(mm)
n
()
EC t
EC b
Ct
(mS/m) (mS/m) (mM)
Cb
(mM)
Cave
(mM)
P
(kPa)
()
Qt /t
(mol/m2 d)
Ds
(1010 m2 /s)
34.5
0
3.9
6.0
8.7
20.0
47.0
103
0
3.9
6.0
8.7
20.0
47.0
0
3.9
6.0
8.7
20.0
47.0
0
3.9
6.0
8.7
20.0
47.0
9.5
9.3
9.1
9.0
8.7
8.7
9.5
9.4
9.3
9.3
9.2
9.2
8.2
8.2
8.1
8.0
7.9
7.9
6.4
6.3
6.1
5.8
5.7
5.6
0.81
0.80
0.80
0.79
0.79
0.79
0.80
0.80
0.80
0.80
0.80
0.79
0.77
0.77
0.77
0.76
0.76
0.76
0.70
0.69
0.68
0.67
0.66
0.66
2.95
39.2
58.5
81.5
177
415
2.08
46.8
67.6
95.9
203
460
2.62
47.6
67.5
94.9
202
452
2.95
46.3
71.4
101
212
464
0.63
1.25
2.19
6.31
15.2
0.41
1.12
1.93
5.25
13.6
0.441
0.791
1.43
4.47
12.7
2.98
4.43
6.16
13.3
31.1
3.35
4.86
6.83
14.7
33.5
3.36
5.01
7.07
15.1
33.7
8.18
9.01
8.57
4.90
2.29
9.48
11.0
11.0
11.0
11.2
12.1
15.2
16.6
16.6
16.1
0.561
0.418
0.286
0.076
0.015
0.584
0.461
0.328
0.152
0.068
0.719
0.628
0.484
0.226
0.098
2.76
5.56
9.87
28.1
68.2
2.13
4.99
8.70
23.2
60.2
1.96
3.50
6.53
19.9
57.0
1.26
1.66
2.08
2.69
2.80
0.86
1.39
1.71
2.12
2.41
0.72
0.86
1.13
1.59
2.02
0.301
0.711
1.40
4.23
10.7
3.39
5.17
7.26
15.6
35.0
12.9
16.0
16.2
17.6
18.2
0.784
0.635
0.459
0.230
0.106
1.37
3.18
6.24
18.5
48.6
0.43
0.64
0.87
1.18
1.38
172
241
3.10
11.6
20.0
32.7
88.3
211
1.65
7.17
16.8
27.8
72.6
185
3.14
9.09
13.8
22.5
63.5
175
3.63
7.70
13.2
22.5
60.8
148
2.68
4.12
5.81
12.9
30.3
3.22
4.84
6.91
14.7
33.5
3.25
4.81
6.87
14.7
33.2
3.18
5.05
7.22
15.5
33.7
= effective stress; Cot = source KCl concentration; L = specimen thickness; n = total porosity; EC t and EC b = EC of
outflows from top and bottom boundaries, respectively; Ct and Cb = molar Cl concentrations in outflows from top and
bottom boundaries, respectively; Cave = average boundary concentration difference; P = induced pressure difference;
= membrane efficiency coefficient computed based on Cave (see Kang & Shackelford 2011); Qt /t = diffusive molar
Cl flux; Ds = effective salt-diffusion coefficient.
as illustrated in Figure 6 for the test conducted at

= 34.5 kPa. The resulting values of Ds (computed
using Eqs. 6-7) are shown in Table 1 along with the
steady-state values of reported previously by Kang &
Shackelford (2011).
The results in Table 1 and Figure 7a show that
the values of for a given specimen decreased with
increasing source KCl concentration, Cot . This trend of
decreasing with increasing Cot is consistent with previous research and is attributed to progressively greater
collapse of the DDLs surrounding the clay particles
as the salt diffuses into the GCL pores at progressively higher concentrations (see Kang & Shackelford
2011). The progressively greater collapse of the DDLs
with increasing Cot also increases the sizes of the
pore spaces available for solute diffusion and, therefore, causes an increase in Ds (Malusis & Shackelford
2002b). This trend of increasing Ds with increasing
Cot was evident in each of the tests in this study, as
illustrated in Figure 7b.
The results in Figure 7 also reveal trends of increasing and decreasing Ds with increasing for a given
Cot , indicating an overall reduction in pore size with
increasing consolidation. Although a higher did not
yield a lower total porosity, n, in all cases, a general
trend of decreasing n with increasing is evident in

Table 1. As noted by Kang & Shackelford (2011), a
decrease in void space would be expected to restrict
solute passage and increase membrane efficiency.
The relationships between Ds and for the specimens in this study are illustrated in Figure 8. Each of
the specimens exhibited a decrease in Ds with increasing , consistent with the prior results of Malusis &
Shackelford (2002a) and Di Emidio (2010). Moreover,
the decreases in Ds are approximately linear, such that
the best-fit linear regressions shown in Figure 8 intersect the Ds = 0 axis within the range = 1.0 0.1
in all cases. Theoretically, Ds = 0 should correspond
to = 1, since an ideal membrane, by definition,
completely restricts solute passage. This theoretical
consideration is well supported by the data in Figure 8.
The theoretical maximum value of Ds for each
specimen corresponds to the minimum membrane
efficiency (i.e. = 0) and may be estimated from the
linear regressions in Figure 8. As discussed previously,
values of Ds at = 0 are interpreted as values of Dse ,
or effective diffusion coefficients that account solely
for the tortuosity associated with the geometry of the
interconnected pores (i.e. the matrix tortuosity). The
linear regressions yielded Dse values that range from
271
Figure 6. Cumulative moles of Cl per unit area, Qt , diffused through a GCL specimen ( = 34.5 kPa) as a function of
elapsed time.
Figure 8. Effective salt-diffusion coefficients, Ds , as a

function of the membrane efficiency coefficient, , for
GCL specimens consolidated at different effective stresses
( = 34.5, 103, 172, and 241 kPa).
Figure 7. (a) Membrane efficiency coefficients (replotted after Kang & Shackelford 2011) and (b) effective
salt-diffusion coefficients as a function of source KCl concentration for GCL specimens consolidated at different
effective stresses ( = 34.5, 103, 172, and 241 kPa).
1.5 1010 m2 /s to 2.9 1010 m2 /s and decrease

with increasing (see Fig. 9). This trend of decreasing
Dse with increasing is expected, given the aforementioned trends of decreasing Ds and increasing (for a
given Cot ) with increasing .
4.3
Restrictive tortuosity factor
Values of the restrictive tortuosity factor, r , for the

GCL specimens were computed using Eq. 3 (i.e. r =
Ds /Dse ) for each testing stage based on the Ds values in Table 1 and the Dse values in Figure 9. These
values of r are plotted as a function of in Figure
10 along with the best-fit regression of the linear relationship given by Eq. 4 (i.e., r = 1 ). The results
illustrate that r closely follows the linear trend given
Figure 9. Effective salt-diffusion coefficients at zero membrane efficiency, Dse , as a function of effective confining
stress, .
by Eq. 4 (R2 = 0.976), regardless of the employed

in the test. Thus, the results in this study provide more
compelling evidence that Eq. 4 is a valid expression
for relating r to for GCLs relative to the prior experimental results of Malusis & Shackelford (2002b)
272
decrease linearly with increasing , such that the relationship between r and is well represented by the
expression r = 1 proposed in previous theoretical studies. Thus, the results presented in this study
provide compelling evidence in support of the hypothesis that r = 1 is a valid expression for clay
membranes. However, since this study was limited to
KCl solutions and GCLs containing 100% conventional Na-bentonite, additional testing is warranted to
assess the applicability of this expression for other
chemical solutions and other barrier materials that
exhibit membrane behavior.
ACKNOWLEDGEMENTS
Figure 10. Computed values of the restrictive tortuosity factor, r , as a function the membrane efficiency coefficient, ,
for GCL specimens consolidated at different effective stresses
( = 34.5, 103, 172, and 241 kPa).
and Di Emidio (2010) shown in Figure 2. However,

these results are limited to KCl solutions and GCLs
containing 100 % conventional Na-bentonite. Similar
analyses as described herein must be performed for
other chemical solutions and for other types of barrier
materials exhibiting membrane behavior before a more
robust conclusion can be made regarding the general
applicability of Eq. 4 for clay membranes.
CONCLUSIONS
The results of this study demonstrate that effective

salt-diffusion coefficients, Ds , for a geosynthetic clay
liner (GCL) are dependent upon both the source concentration of the salt, Co , and the effective stress, ,
applied to the GCL. In addition, the concentration
dependence of Ds was related directly to the concentration dependence of membrane efficiency, which has
been reported in several previously published studies on clay membranes. In this study, Ds for each
GCL specimen approached a maximum at zero membrane efficiency ( = 0) and decreased linearly toward
Ds = 0 as the membrane behavior approached the ideal
condition ( = 1). Thus, the results also were consistent with the theoretical consideration that Ds = 0
for an ideal membrane that completely restricts solute
passage.
The maximum values of Ds corresponding to = 0,
designated herein as Dse , are considered to be governed solely by the matrix tortuosity, or the tortuosity
associated with the geometry of the interconnected
pores. These Dse values were shown to decrease with
increasing due to an overall reduction in pore size
with increasing . In contrast, the values of Ds corresponding to > 0 are considered to be a function of
both the matrix tortuosity factor, m , and a restrictive
tortuosity factor, r , that accounts for solute exclusion due to membrane behavior. Values of r computed
for the GCL specimens in this study were shown to
Financial support for portions of this work was provided by the US National Science Foundation (NSF),
Arlington, Virginia, under Grants CMS-0099430 entitled, Membrane Behavior of Clay Soil Barrier Materials and CMS-0624104 entitled, Enhanced Clay
Membrane Barriers for Sustainable Waste Containment. The opinions expressed in this paper are solely
those of the writers and are not necessarily consistent
with the policies or opinions of the NSF.
REFERENCES
Di Emidio, G. 2010. Hydraulic and Chemico-Osmotic
Performance of Polymer Treated Clays. PhD Dissertation,
Ghent University, Ghent, Belgium.
Dominijanni, A. 2005. Osmotic Properties of Clay Soils. PhD
Dissertation. Politecnico di Torino, Torino, Italy.
Dominijanni, A. & Manassero, M. 2012. Modelling the
Science 51: 5173.
Greenberg, J., Mitchell, J., & Witherspoon, P. 1973. Coupled
salt and water flows in a groundwater basin. Journal of
Geophysical Research 78(27): 63416353.
Henning, J., Evans, J., & Shackelford, C. 2006. Membrane behavior of soil-bentonite slurry trench cutoff wall
backfill. Journal of Geotechnical and Geoenvironmental
Engineering 132(10): 12431249.
Kang, J. 2008. Membrane Behavior of Clay Liner Materials.
PhD Dissertation, Colorado State University, Fort Collins,
Colorado, USA.
Kang, J. & Shackelford, C. 2009. Clay membrane testing
using a flexible-wall cell under closed-system boundary
conditions. Applied Clay Science 136(10): 13681382.
Kang, J. & Shackelford, C. 2010. Membrane behavior of
compacted clay liners. Journal of Geotechnical and
Kang, J. & Shackelford, C. 2011. Consolidation enhanced
membrane behavior of a geosynethtic clay liner. Geotextiles and Geomembranes 29(6): 544556.
Keijzer, T., Kleingeld, P., & Loch, J. 1999. Chemical osmosis
in compacted clayey material and the prediction of water
transport. Engineering Geology 53(2): 151159.
Kemper, W., Maasland, D., & Porter, L. 1964. Mobility of
water adjacent to mineral surfaces. Soil Science Society of
America Proceedings 28: 164167.
Kharaka,Y. & Berry, F. 1973. Simultaneous flow of water and
solutes through geological membranes I. Experimental
273
investigation. Geochimica et Cosmochimica Acta 37(12):

25772603.
Malusis, M. & Shackelford, C. 2002a. Chemico-osmotic
efficiency of a geosynthetic clay liner. Journal of Geotechnical and Geoenvironmental Engineering 128(2): 97106.
Malusis, M. & Shackelford, C. 2002b. Coupling effects during steady-state solute diffusion through a semipermeable
clay membrane. Environmental Science and Technology
36(6): 13121319.
Malusis, M., Shackelford, C., & Olsen, H. 2003. Flow and
transport through clay membrane barriers. Engineering
Geology 70(23): 235248.
Malusis, M., Shackelford, C., & Maneval, J. 2012. Critical
review of coupled flux formulations for clay membranes
based on nonequilibrium thermodynamics. Journal of
Contaminant Hydrology 138139: 4059.
Malusis, M. & Shackelford, C. 2004a. Explicit and implicit
coupling during solute transport through clay membrane
barriers. Journal of Contaminant Hydrology, 72(14):
259285.
Malusis, M. & Shackelford, C. 2004b. Predicting solute flux
through a clay membrane barrier. Journal of Geotechnical
and Geoenvironmental Engineering, 130(5): 477487.
Manassero, M. & Dominijanni,A. 2003. Modelling the osmosis effect on solute migration through porous media.
Marine, I. & Fritz, S. 1981. Osmotic model to explain

anomalous hydraulic heads. Water Resources Research
17(1): 7382.
Mazzieri, F., Di Emidio, G., & Van Impe, P. 2010. Diffusion
of calcium chloride in a modified bentonite: Impact of
osmotic efficiency and hydraulic conductivity. Clays and
Clay Minerals 58: 351363.
Neuzil, C. 1986. Groundwater flow in low-permeability environments. Water Resources Research 22(8): 11631195.
Shackelford, C. 1991. Laboratory diffusion testing for waste
disposal A review. Journal of Contaminant Hydrology
7(3): 177217.
Shackelford, C. & Daniel, D. 1991. Diffusion in saturated
soil: I. Background. Journal of Geotechnical Engineering
117(3): 467484.
Shackelford, C. & Moore, M. 2013. Fickian diffusion of
radionuclides for engineered containment barriers: Diffusion coefficients, porosities, and complicating issues.
Yeo, S., Shackelford, C., & Evans, J. 2005. Membrane behavior of model soil-bentonite backfill mixtures. Journal of
418429.
Young, A. & Low, P. 1965. Osmosis in argillaceous rocks.
Bulletin of the American Association of Petroleum Geologists 49(7): 10041008.
274

Hydraulic behavior of activated calcium bentonite mixed with polyionic

charged polymers for landfill and earthwork applications
A. Razakamanantsoa
Now at: Department of Earthworks and Earth structure French Institute of Sciences Technology of Transport
and Network, Bouguenais, Nantes Facility, France
Previously: University of Lyon, INSA of Lyon, LGCIE, Villeurbanne, Cedex France
I. Djeran-Maigre & G. Barast

University of Lyon, INSA of Lyon, LGCIE, Villeurbanne, Cedex France
ABSTRACT: This paper aims to study the hydromechanical behaviour and the clay polymer interaction of
amended Ca-bentonite. Specimens were formed by mixing of Ca-bentonite with two soluble polyelectrolyte
polymer powders.
Some important parameters are studied: swelling, water adsorption and hydraulic performance for landfill and
earthwork applications. Tests are performed with tap water and synthetized leachate (SL) in order to reproduce
the hydrochemical phenomena. Hydraulic performance tests were performed with an oedopermeameter. Tests
results show that polymers tend to reduce the permeability, in contact with the SL. Water adsorption and free
swell index tests confirmed that adsorption, swelling and permeability parameters depend on the clay polymer
mixtures and that polymers adding improves the clay properties. Each polymers charges have a different effect:
the anionic polymer gives a low permeability to the mixture; the cationic polymer enhances the bentonite swell
ability and water retention which can be used also as a performance index.
INTRODUCTION
Bentonite corresponds to high reactive clay with high

specific surface. These properties are suitable for
hydraulic containment use or as a soil amendment
(Alther 1982; Chapuis et al., 1992; Daniel 1993).
Direct exposure of bentonite to aggressive solution may deteriorate the barrier performance because
the bentonite has insufficient resistance against electrolytic chemical solutions (Katsumi et al., 2007;
Barast et al., 2008). The hydraulic conductivity value
increases as the concentration and valence of the electrolyte solution increases (Petrov and Rowe 1997; Ruhl
and Daniel 1997; Shackelford et al., 2000; Jo et al.,
2001; Shan and Lai 2002; Kolstad et al., 2004; Katsumi
et al., 2007).
The use of aqueous polymers with bentonite was
discussed during the last decades as an issue to stabilize clayey soil (Liao 1989; Green et al., 2000; Elhajji
et al., 2001; Yu et al., 2003; Inyang et al., 2005; Liu
2007; Razakamanantsoa et al., 2008). Theng (1970;
1979; 1982), explains in detail the mechanism of interaction between the clay polymer mixture and shows
how important is the surface charge of the polymer.
Various charged polymers have been shown to
improve flocculation or dispersion of soil particle
suspension. Negatively charged polymers tend to be
repelled by repulsion force from the net negative
surface of the clay particle but held to the positive
charge of the clay edge particles and cause delamination (Theng 1982; Lu et al., 2002). In a light gel
system, weak hydrogen bonding constitutes the important bridging mechanism between clay particles and
anionic charged polymer. According to Theng (1982),
anionic polymer adsorption is limited to the outer
surface of the clay particles because of the charge
repulsion between polymer and clay surface which
makes it difficult for the polymer to enter in the interlayer of clay minerals. Nevertheless the presence of
soluble salts increases the amount of adsorbed polymer on the clay surface (Lu et al., 2002). Uncharged
polymer adsorption on clay surfaces is led by the gain
of entropy during adsorption phases. Cationic polymers bond with negatively charged sites on mineral
surfaces by Coulombic attraction. Clay particles stack
to form aggregates. These mechanisms change the
physical properties of treated bentonite: soil wettability increases, liquid is retained, and the infiltration rate
of soil is affected (Ben-Hur et al., 1992). According
to Theng, (Theng 1970) an uncharged and polyanionic
polymer is attached to the edge of the crystal and forms
peripherical complexes.
The problematic of formulation of bentonitepolymer mixtures becomes of interest. The applications extend from the industry to the geotechnical
field. For geotechnical application, commercialized
polymer component characteristics are frequently not
communicated by manufacturer.
275
Actually, the hydraulic performance of bentonite

based on geomaterial is qualified with the permeability
and the swelling ability of the bentonite. According the
studies of Razakamanantsoa et al. (Razakamanantsoa
et al., 2012), permeability testing is still accurate but
the swelling index testing needs to be upgraded or reinforced by using other performance index testing in case
of polymer adding. The test should take into account
the hydraulic properties of the polymer adding.
This paper is focused on the investigation of the
polymer treatment influence, on the hydraulic properties of bentonite.
The study aims first to analyse the hydromechanical
behaviour of an activated Ca-bentonite amended with
polyionic charged polymers and permeated by multispecies fluid. And additionally, the target is to investigate the accuracy of the adsorption index for qualifying
the bentonite properties. Hydraulic conductivity tests
are performed using oedopermeameter.
The second step supports the characterization of
the swelling and adsorption ability of amended bentonite. Two testing methods are used in this case: the
Ensling and the Free Swell Index (FSI) tests. The aim
is to investigate the accuracy of the adsorption index
for qualification of the bentonite polymer mixture
properties.
2
2.1
MATERIALS
Bentonite
The present bentonite (B) is a commercial bentonite

that currently sells in France as a substitute to natural sodium bentonite. This bentonite consists of a
smectite with minor quartz calcite and pyrite. B is
a powdered natural calcium-rich bentonite (<75 m)
activated with Na2 CO3 by thermic treatment. The
main major soluble cations which composed the bentonite B are: Na+ (17.35 mg/L), Mg2+ (0.83 mg/L),
K+ (0.67 mg/L), Ca2+ (5.49 mg/L).
2.2
Table 1.
Charge distribution of the polymer.
Charges
P1
%
P2
%
Non charged
Anion
Cation
97.05
0.25
2.70
94.53
4.06
1.41
Figure 1. Oedopermeameter cell.
leachate (SL): with pH ranging from 5.5 0.5 and

measured electrical conductivity (EC) at 20 C around
13.5 mS/cm (Razakamanantsoa et al., 2011). SL component is similar to the one used by (Guyonnet et al.,
2009) but with lower concentration. The mixture is
composed of a cation, anion and organic matter.
High concentration of Ca2+ (952 mg/L) and Mg2+
(503 mg/L), which are known to be aggressive to the
bentonite, against lower concentration of Na+ (431
mg/L) and K+ (488 mg/L). The major anions are
Cl (3031 mg/L) and SO2
4 (877 mg/L).
3
METHODS AND TESTING PROGRAM
3.1 Specimen preparation
Polymers
Two manufactured powder polymers (P1, P2, <80 m)

were used. The main soluble element on the polymer
surfaces were quantified with Inductively Coupled
Plasma- by Atomic Emission Spectroscopy (ICP AES)
(Razakamanantsoa et al., 2012). Both Polymers contain major cation Na+ and K+ . None of them contains
major cations, such as Ca2+ and Mg2+ . P1 contains
SO2
4 as major anion and P2 presents two varieties of
main soluble anions: Cl , NO
3.
97.05% of P1 and 94.91% of P2 of the polymers
dry mass is uncharged, which is important, compared
to the anionic and cationic surface charges (Table 1).
The bentonite-polymer mixture was prepared by mixing dry bentonite with 2% of polymer P1 or P2, by dry
weight of the bentonite.
Three mixtures were prepared and used for the
study: bentonite alone (B), Bentonite and polymer P1
(BP1) and Bentonite and polymer P2 (BP2).
For this study, intact material tested with standard
Free Swell Index (FSI) or Adsorption test is called
reference specimen. The specimen collected after permeability tests performed with SL and tested with
standard FSI and Adsorption test is called leached
specimen.
3.2
2.3 Testing liquids

The permeant liquids consisted of low mineralized
water (NaCl 103 M), and a multispecies synthetic
Hydraulic conductivity testing
Hydraulic conductivity tests are conducted with an

oedopermeameter (Figure 1) and performed according
to NF P84-705.
276
Three hydraulic performance tests were performed

on a layer of: B, BP1 and BP2 with 5 kg per square
meter. 5 kg/m2 , correspond to the standard Geosynthetic Clay Liner (GCL) commercialized in France.
The bentonite mixtures are set between two geotextiles
filters inside the oedopermeameter cell. A geotextile
filter was used to avoid the bentonite particles migration during the flow and the inner wall of the oedopermeameter is greased in order to avoid preferential flow
path. When the upper part of the oedopermeameter cell
is placed, the specimen is encapsulated between two
porous stones.
A confining stress equal to 160 kPa was applied
with pressurised air on the top of the upper part of the
oedopermeameter cell (NF P84-705). The confining
stress used in this study corresponds to the standard
confining stress observed in the bottom layer of the
landfill liner in France. The hydraulic performance
testing is performed in two consecutive stages. The
first stage consists of controlling the swelling ability
of treated bentonite permeated by SL.The second stage
controls the hydraulic behaviour of untreated and polymer treated bentonite under hydro-chemo-mechanical
stress. The moisture content of each mixture was controlled at the end of the hydraulic performance test and
the specimen was dried and crushed for the adsorption
and free swell tests.
The swelling and adsorption behaviour of the bentonite specimen were monitored during the saturation
phase and the performance phase with a displacement
sensor in the upper part of the oedopermeameter cell.
Swelling ability is intimately linked to the water
adsorption, and then the end of swelling phase (or settlement) gives information concerning the volume of
adsorbed water.
Saturation phase is considered complete when
95% of the infinite swell and adsorption ability are
obtained. It is known that clay material may adsorb
water until the end of the swelling process. When those
conditions are satisfied, then, the upper part of the
oedopermeameter cell is blocked at the last height of
the specimen in order to maintain constant thickness
during the hydraulic conductivity measurement. The
saturation is then maintained by the influent applied
constant hydraulic head. The hydraulic conductivity is
usually calculated with the measured thickness.
The hydraulic performance test can be performed
when the saturation phases are achieved and the influent hydraulic flows are at steady state. To have an
effective steady state and to validate the applicability of Darcys law, three hydraulic heads (0.3, 0.6 and
0.9 m) were applied with a Marriotte bottle linked to
the oedopermeameter cell. The hydraulic conductivity
was determined according to the equation (1):
Figure 2. Ensling test device (NF EN ISO 10769).
been used for measuring hydraulic conductivity of

Compacted Clay Liners (CCL) (Didier and Comeaga,
1997; Petrov et al., 1997). The average hydraulic gradient for B, BP1 and BP2 are respectively: 43, 19.98,
and 82.70.
Hydraulic conductivity testing is considered to be
achieved when two criteria were satisfied: stabilization of the influent and effluent flow and chemical
equilibrium. This last criterion is practically evaluated
and based with pH and EC measurements. The number of pore volume of flow as described by Paumier
and Touze-Foltz 2012 are not considered since the last
results comparison are based only on the tests duration.
3.3 Performance indexes
Free swell index tests (FSI) are performed according to
ASTM 5890 standard to assess the free swell ability of
bentonite.Test is performed in isolated aqueous system
in two steps. At first, a standard FSI test is performed
with Demineralized Water DIW and SL. At second
time, the FSI test is performed with lixiviated bentonite
material collected after the hydraulic conductivity tests
permeated with SL.
BP1 bentonite has much important swelling ability
than that of BP2 or B. These results show that polymer
adding increases the swelling ability of the bentonite.
Ensling test are performed according to the NF EN
ISO 10769. The device is presented in the Figure 2.
This test is used to quantify the water adsorption and
retention ability of clayey material until saturation.The
adsorption cell is composed of a horizontal graduated
cylinder, mounted with a porous stone plate placed
inside a cylinder cell. Test is considered to be achieved
when 95% of the infinite adsorbed volume is reached.
The water content is then tested in order to evaluate the
water retention ability of specimen. Ensling tests were
performed with the same specimen as with FSI test.
4
where F corresponds to the average fluid flow (m/s),

i is the corresponding hydraulic gradient.
High hydraulic gradients might underestimate the
hydraulic conductivity due to consolidation by seepage
forces. Hydraulic gradient ranging from 50 to 550 have
METHODS AND TESTING PROGRAM
4.1 Hydraulic conductivity testing

Hydraulic conductivity tests were performed during 90
days. The device used for the study is presented in the
Figure 3. This test duration is relatively short comparing to those in the literature (Shackelford et al., 2000;
277
Table 3.
Figure 3. Hydraulic conductivity testing of the 3 bentonites

mixtures: B, BP1, BP2.
Table 2.
Hydraulic conductivity testing results.
Mixture
k
m/s
Moisture content
%
PVF
B
BP1
BP2
8 1010
5 1011
2 1011
95
92
86
10.56
1.73
6.64
Ashmawy et al., 2002; Katsumi et al., 2008), but the

termination criteria were satisfied. Table 2 summarizes
the test results performed with the oedopermeameter
cell.
During the saturation phase, all of the permeation
fluid is adsorbed by the bentonite materials and the
effect of the fluid on the permeability takes place
immediately.
Tests results confirm that hydraulic conductivity
of polymer treated bentonite is much lower than that
of untreated bentonite B (Table 2). French regulation recommend an average hydraulic conductivity
value less than 1 109 m/s for landfill or earthwork
applications; all of the bentonite materials used for
this study presented acceptable values: 8 1010 m/s
for B, 5 1011 m/s for BP1 and 2 1011 m/s for
BP2. According to these results, both polymer adding
enhance hydraulic conductivity of the bentonite.
For the same test duration, the difference of Pore
Volume of Flow (PVF) calculated for each bentonite
material was very important. Bentonite B treated with
polymer P1 is very resistant to the fluid flow (1.73
PVF). 6.64 pore volumes of flow have been percolated from BP2. The higher value corresponds to
the untreated bentonite with PVF = 10.56. Moisture
content was controlled at the end of the hydraulic conductivity test. BP1 (w = 92%) retains more water than
that BP2 (w = 86%).
4.2
Free swell index tests.
Fluids
Mixture
Reference
mL/2g
Leached specimen
mL/2g
DIW
DIW
DIW
SL
SL
SL
B
BP1
BP2
B
BP1
BP2
30
28
27
8
12
13
10
23
17
6
12
11
testing. Table 3 presents the testing results obtained on

reference and leached specimen with DIW and SL.
For the FSI test on polymer-bentonite mixture with
DIW, the polymer led to flocculation of big clay
aggregates. Under this latter form, the resulting clay
particles do not possess high specific surface that is
available for water hydratation.
Thus, in that state, the clay could presumably not
swell as much as expected with polymer treatment.
This would explain why the FSI values of BP1 and BP2
are lower than B with DIW and express also the limit of
the FSI for qualification of bentonite polymer mixture.
FSI values of BP1 and BP2 are respectively equal to
28 mL/2g and 27 mL/2g with DIW. FSI values for BP1
decrease from 28 mL/2g with DIW to 23 mL/2g and
show that FSI corresponding to BP1 is less affected by
the percolated solution comparing to B. FSI values of
BP2 decrease from 27 mL/2g with DIW to 17 mL/2g
when specimen is leached.
In contrast with DIW, little difference occurs when
the FSI is performed directly with SL. Results of FSI
performed with SL show that B and BP2 FSI values
decrease respectively from 8 mL/2g to 6 mL/2g and
from 13 mL/2g to 11 mL/2g. It is not the case for BP1
since its FSI value is still equal to 12 mL/2g. The overall comparison between initial and final steps shows
that FSI test performed with SL give relatively close
results but loss of FSI occurs in case of DIW.
Bentonite material reacts immediately with the
permeant solution and shows a very important loss
of swell ability. FSI of polymer treated bentonite
percolated with SL is less affected.
An order of classification between the three bentonite materials can be easily proposed. Different
reaction occurs when FSI tests were performed directly
with the SL. BP2 has the higher FSI and B has the lower
one. Swelling ability of BP1 are less affected with
DIW and maintained with SL.As discussed previously,
the specimen B has the higher FSI before hydraulic
conductivity testing, presumably due to the flocculation of clay aggregates by the polymer. However,
it was observed as expected that the SL percolation
reduces considerably its swelling ability, proving its
high sensitivity to chemical fluids.
Free swell index
The FSI value of the pure bentonite performed with

DIW decreases from 30 mL/2g to 10 mL/2g, meaning
that the swelling ability of the material is reduced to
a third of its initial value after hydraulic conductivity
4.3 Adsorption test

Adsorption tests are performed with reference specimen. B bentonite has little higher adsorption ability
than the polymer treated bentonite (Table 4).
278
Table 4. Adsorption index test with DIW.
Mixture
Infinite value*
mL
Measured**
mL
Moisture
%
B
BP1
BP2
33.67
59.88
38.31
33.7
49.8
35.5
457
694
700
*Forecasted infinite adsorbed volume of water.

**Measured adsorbed volume.
Table 5. Adsorption index test with leached specimen.
Mixture
Infinite value*
mL
Measured**
mL
Moisture
%
B
BP1
BP2
18.42
30.86
28.17
18.70
31.30
27.30
218
403
352
*Forecasted infinite adsorbed volume of leachate.

**Measured adsorbed volume.
At the end of hydraulic conductivity tests, the

leached bentonite mixture specimens are collected and
dried, crushed and poured through Ensling device. The
water adsorption of BP1 increases up to 31.3 mL. So
the adsorbed volume is equal to 27.3 mL for BP2, and
18.7 mL for B (Table 5).
The hierarchy of adsorption ability is maintained
for standard test and also for leached specimens. BP1
adsorbed water value is still higher than for B and
BP2 and the difference is more important in case
of test runned with leached bentonite mixture. Test
results show that polymer adding increases the adsorption ability of bentonite but the mechanism takes
place in different manner: immediately for BP1 and
progressively for BP2.
Table 4 and Table 5 summarize respectively the
specimen characteristics, controlled at the end of the
standard adsorption test and with leached specimens.
The measured adsorbed volume, for B, BP1 and BP2,
goes respectively up to 33.7 mL, 49.8 mL and 35.5 mL.
Compared to the infinite adsorbed volume, BP1 and
BP2 adsorption is not yet achieved.
Measured adsorbed volume for bentonite B
decreased from 33.7 to 18.7 mL and the moisture
content decreased from 457% to 218%. The water
adsorption and moisture content are reduced by half
for bentonite B due to the SL effects. Similar result
appears with BP2. BP2 water adsorption decrease
from 35.5 to 27.3 mL and the corresponding water
content goes from 700 to 352%. 20% of adsorption
ability is lost for BP2.
For BP1, water adsorption decreased from 49.8
to 31.3 mL and the corresponding moisture content
decreased from 694 to 403 mL. These results confirm
that P1 improves considerably the water adsorption
ability of bentonite B.
DISCUSSION
It is known that SL may lead to dissolution of the major

element of the bentonite and in that case the increase
of the hydraulic gradient lead to a new microstructure
arrangement where the pore water can move easily
in a shorter time laps. Theng (1970) stated that the
presence of an anionic polymer lead to an extralamellar sorption of polymer. Filling of extralamellar voids
may block the flow path and reduces the flow. This
statement may explain the reaction of the bentonite
polymer mixture. These results show that the polymer
adding may delay the damage of the SL on the bentonite mixture; this mechanism is materialized by the
hydraulic conductivity variation with the increase of
the hydraulic gradient.
The adsorption and swelling abilities of the
bentonite materials are affected by the nature of
the permeant solution. Clay collapse and loss of
water adsorption are the immediate parameters
that could occur during hydraulic conductivity testing and that could express the sensitivity of the
materials.
The swelling index and the water adsorption of the
pure bentonite specimen with DIW decrease greatly
after hydraulic conductivity testing with SL. This
implies the high chemical sensitivity of B with electrolytes fluids. It is observed that P1 and P2 increase
the water adsorption and swell ability of B. The
increase of water adsorption can be associated to the
high cationicity of P1 comparing to P2. The difference in adsorption velocity may be suggested by the
fact that hydroxyl groups from water molecules are
more attracted by the cation charged surface of the
polymer, so the free hydrogen of the water may bond
easily with clay particles. Hence, the water adsorption
is relatively faster with BP1 and relatively slowly with
BP2. This information is important and it cannot be
obtained with swelling index testing.
Several studies have shown the correlation between
the swelling index and the hydraulic conductivity of
GCL (Shackelford et al., 2000, Lee et al., 2005). Jo
et al. (2001) and Lee et al. (2005) suggested that a
FSI value of 20 mL/2g is the minimal threshold to
determine whether or not GCL would be susceptible to
increase its hydraulic conductivity value with leachate.
This suggestion can be ambitious to prove in case of
bentonite polymer mixtures tested with DIW. All of the
bentonite materials have a free swell index superior
to 20 mL/2g with DIW, meaning that these materials should be acceptable for landfill applications, as
suggested by Jo et al., (2001).
The higher swelling index with DIW corresponds
to the lower hydraulic performance with SL. FSI
test performed with DIW gives important indications
about the quality of bentonite as a hydraulic barrier. The swelling index tests performed directly with
SL matches well with the corresponding hydraulic
conductivity order.
It can be concluded that BP1 bentonite has a better
swelling ability with SL than BP2. B bentonite material
279
is immediately affected by the aggressiveness of the

synthetized leachate.
This is not the case for BP2. BP2 reaction is different
because P2 can bond to the edge of the clay particle
and increases also the net anionic surface exchange of
the clay particles, leading to dispersion.
CONCLUSION
This paper focused on the influence of the polymer

adding on the properties of an activated Ca-bentonite
B permeated with synthetic leachate SL for hydraulic
barrier purposes. Two soluble polyelectrolyte powder
polymers P1 and P2 were used, with the same proportion of uncharged surfaces, but different amount of
cations and anions.
Hydraulic performances tests performed with an
oedopermeameter method give the following classification order of hydraulic conductivity: BP2 < BP1
< B, where B has the higher hydraulic conductivity
value.
The polymer P1 greatly affects the swelling and
water retention of the bentonite, whereas the polymer
P2, better improves the hydraulic flow behaviour of the
specimen. Both polymers enhanced the characteristics
of the clay, comparing to the untreated one.
Free swell index test is relatively difficult to perform for bentonite polymer mixture with DIW.Anionic
polymer spreads and render difficult the swelling
ability measurement with DIW, cationic polymers
flocculate and might underestimate the measurement
of FSI.
The water adsorption seems to be a good performance index because this test highlights the effect of
polymer adding by the water adsorption and moisture
content measurement. Adsorption tests give specimen
information concerning its wettability and specimen
reactivity after leaching.
Good correlation can be found with Free Swell
Index and adsorption parameters of leached bentonite
mixture. Theses observations match with the hydraulic
conductivity results.
ACKNOWLEDGMENT
This study was conducted entirely at INSA of Lyon.
The authors thank OSEO ANVAR for supporting
the project and Julie POLAK and Fouad BOURICH
INSAs student for their assistance on the research
project.
REFERENCES
Alther, G.R. (1982). The role of bentonite in soil sealing
applications. Bulletin of the Association of Engineering
Geologists, 19, 401409.
Ashmawy, A.K., El-Hajjy D., Sotelo N., Muhammad N.,
(2002), Hydraulic performance of untreated and polymer
treated bentonite in inorganic landfill leachate. Clays and

Clay Minerals, 50(5), 546552.
Barast, G., Razakamanantsoa, A.R., Djeran-Maigre, I.,
Didier, G. and Couradin, A. (2008), Filtration and rheology study of characterization of Ca-activated bentonites.,
Proceedings of Journes Nationales de Gotechnique et de
Gologie de lIngnieur, Nantes, 483490.
Ben-Hur, M., Malik, M., Letey, J., Mingelgrin, U., (1992).
Adsorption of polymers on clays as affected by clay
charge and structure, polymer properties, and water
quality. Soil Sciences. 153, 349356.
Chapuis, R.P., Lavoie, J. and Girard, D. (1992). Design, construction performance, and repair of the soil bentonite
liners of two lagoons. Canadian Geotechnical Journal.
Ottawa, Canada, 29(6), 638649.
Daniel, D. E. (1993). Clay liners. Geotechnical practice
for waste disposal, D.E. Daniel, ed., Chapman & Hall,
London, England, 137163.
Didier, G., and Comeaga, L. (1997). Influence of initial
hydration conditions on GCLs leachate permeability.
Testing and acceptance criteria for geosynthetic clay
liners. ASTM STP 1308, Philadelphia, 181195.
Elhajji, D., Ashmawy, A.K., Darlington, J. and Sotelo
N. (2001). Effect of inorganic leachate on polymer
treated GCL material. Proceedings of the Geosynthetics,
Portland, Oregon, pp. 663670.
Green, V.S., Stott, D.E., Norton, L. D., Graveel, J.G., (2000).
Polyacrylamide molecular weight and charge effect
on infiltration under simulated rainfall. Soil Sciences
Society of America Journal. 64, 17861791.
Guyonnet, D., Touze-Foltz, N., Norotte, V., Pothier, C.,
Didier, G., Gailhanou, H., Blanc, P., Warmont, F., (2009)
Performance based indicators for controlling geosynthetic clay liners in landfill applications. Geotextiles and
Geomembranes, 27, (5), 321331.
Inyang, H.I., Bae, S. (2005). Polyacrylamide sorption on
interlayer and external pore surfaces of contaminant
barrier clays. Chemosphere, 58, 1931.
Jo, H.Y., Katsumi, T., Benson, G.H., Edil. (2001). Hydraulic
conductivity and swelling of non prehydrated GCLs permeated with single species salt solutions. Journal of
Geotechnical and Geoenvironmental Engineering, 127
(7), 557566.
Kolstad, D. C., Benson, C.H. Edil, T.B. (2004). Hydraulic
conductivity and swell of nonprehydrated geosynthetic
clay liner permeated with multispecies inorganic solutions. Journal of Geotechnical and Geoenvironmental
Engineering, 130 (12), 12361249.
Katsumi, T., Ishimori, H., Ogawa, A. Yoshikawa, K.,
Hanamoto, K., Fukagawa. R., (2007). Hydraulic conductivity of nonprehydrated geosynthetic clay liners permeated with inorganic solutions and waste leachates. Soils
and Foundations, 47 (1), 7996.
Katsumi, T., Ishimori, H., Onikata, M., K., Fukagawa. R.,
(2008). Long-term barrier performance of modified
bentonite materials against sodium and calcium permeant solutions. Geotextiles and Geomembranes. 26,
1430.
Lee, L.M., Shackelford, C.D., Benson, C.H., Jo, H.Y. and
Edil T.B., (2005). Correlating index properties and
hydraulic conductivity of geosynthetic clay liners. Journal of Geotechnical and Geoenvironmental Engineering,
131(11), 13191329.
Liu, P. (2007) Polymer modified clay minerals: A review.
Applied Clay Science, 38(12), 6476.
Liao, W.A. (1989). Polymer bentonite soil admixtures
as hydraulic barriers. SPE Drilling Engineering. 4(2),
153161.
280
Lu, J.H., Wu, L., Letey, J. (2002). Effect of soil and

water properties on anionic polyacrylamide sorption.
Soil Sciences Society of America Journal. 66, 578584.
NF EN ISO 10769 (2011). Geosynthetic Clay Liners, Determination of water adsorption of bentonite. Association
Franaise de Normalisation.
NF P 84 705 (2008). Determination of the swelling,
flow and permeability characteristics of geosynthetic clay
liner (GCL) using an oedopermeameter characterization test and performance test. Association Franaise de
Normalisation.
Petrov, R. J., Rowe, R.K., and Quigley, R.M. (1997). Comparison of laboratory measured GCL hydraulic conductivity based on three permeameter types. Geotechnical
Testing Journal, 20(1), 4962.
Paumier, R. S., Touze-Foltz, N., (2012). Hydraulic and
chemical evolution of GCLs during filter press and
oedopermeametric test performed with real leachate
Geotextiles and Geomembranes, 33, 1524.
Razakamanantsoa, A.R., Barast, G., Djeran-Maigre, I.,
Didier, G. and Couradin, A. (2008), Hydraulic performance of bentonite soil mixture reinforced by polymer
in contact of different fluids. Proceedings of Journes
Nationales de Gotechnique et de Gologie de lIngnieur,
Nantes, 133140
Razakamanantsoa, A.R., Barast, G., Djeran-Maigre I. (2011).
Hydraulic conductivity determination of compacted
sand-bentonite mixture using filter press. Geotechnical
Testing Journal, 31(1), 917.
Razakamanantsoa, A.R., Barast, G., Djeran-Maigre I. (2012).

Hydraulic performance of activated calcium bentonite
treated by polyionic charged polymer. Applied Clay
Science, 5960, 103114
Ruhl, J.L. and Daniel, D.E., (1997). Geosynthetic clay liners
permeated with chemical solutions and leachates. Journal of Geotechnical and Geoenvironmental Engineering.
123, 369381.
Shan, H.-Y. and Lai, Y.-J. (2002). Effect of hydrating liquid
on the hydraulic properties of geosynthetic clay liners.
Geotextiles and Geomembranes. 20(1), 1938.
Shackelford, C.D., Benson, C.H., Katsumi, T., Edil, T.B., Lin,
L. (2000). Evaluating the hydraulic conductivity of GCLs
permeated with non standard liquids. Geotextiles and
Geomembranes. 18(2-4), 133162.
Theng, B.K.G. (1970). Interactions of clay minerals with
organic polymers. Some practical applications. Clays and
Clay Minerals. 18, 357362.
Theng, B.K.G. (1979). Formation and clay properties of
clay polymer complexes. Elsevier scientific Publishing
Company. Netherland.
Theng, B.K.G. (1982). Clay polymer interactions: summary
and perspectives. Clays and Clay Minerals. 30, 110.
Yu, J., Lei, T., Shainberg, I. Mamedov, A.I. Levy, G.J. (2003).
Infiltration and erosion in soils treated with dry PAM and
gypsum. Soil Sciences Society of America Journal. 67,
630636.
281

Interaction between clogging of a leachate drainage layer and leakage

through a composite liner
R.K. Rowe, Y. Yu & M.S. Hosney
GeoEngineering Centre at Queens-RMC, Department of Civil Engineering, Queens University, Kingston, Canada
ABSTRACT: This paper examines the clogging of leachate collection systems (LCSs) in municipal solid
waste (MSW) landfills and evaluates leakage through a composite liner arising from an increase in leachate
head due to clogging. Accumulation of clog mass within the drainage layer is modeled and used to
calculate the reduction in hydraulic conductivity and the subsequent leakage through the underlying composite liner at different times. For the cases and leachate history examined, the hydraulic conductivity of
the granular drainage layer is reduced by one to five orders of magnitude (depending on the grain size
of the drainage material) when the leachate mound reaches the LCS thickness. Leakage analyses show
that while the reduction in the hydraulic conductivity of the drainage layer is sufficient to increase the
head on the liner and hence leakage, the reduction is not sufficient for the clogged LCS to provide
significant resistance to leakage through a hole in the geomembrane.
INTRODUCTION
The disposal of municipal and industrial solid waste

in landfills is still one of primary waste management methods in many countries and is likely to
remain so for the foreseeable future. To prevent
the contaminants in the leachate generated within
landfills from escaping into the surrounding environment, regulations generally require installation of
a barrier system (e.g., Ontario Regulations 232/98;
Subtitle D regulations from US EPA 1991) below
the waste (Rowe et al. 2004). Two components are
included in a barrier system (Rowe 2005): (i) a
leachate collection system (LCS) with a highly permeable granular drainage material to collect and remove
leachate from landfills for treatment, and (ii) a low
permeability composite liner system (e.g., geomembrane (GMB) + geosynthetic clay liner (GCL) or
GMB + compacted clay liner (CCL)) to provide resistance to leachate leakage.
Based on Subtitle D regulations for municipal solid
waste and nonhazardous industrial waste landfills (US
EPA 1991), a LCS with a hydraulic conductivity higher
than 1 104 m/s is likely to satisfy the requirement
that the leachate head within the LCS is less than 0.3
m for many common situations, at least in the short
term (i.e., before clogging). However, field observations (e.g., Young et al. 1982; Bass 1986; Brune et al.
1991; Koerner et al. 1993, 1994; McBean et al. 1993;
Rowe 1998a; Fleming et al. 1999; Craven et al. 1999;
Maliva et al. 2000; Bouchez et al. 2003; Levine et al.
2005) and laboratory studies (e.g., Paksy et al. 1995,
1998; Peeling et al. 1999; Rowe et al. 2000a, b, 2002;
Fleming & Rowe 2004; VanGulck & Rowe 2004a, b;
McIsaac & Rowe 2006, 2007) show that permeation
by landfill leachate causes clog mass to accumulate

within the pore space of the drainage material. The
clogging reduces the hydraulic conductivity of the
drainage layer and causes leachate mounding within
the LCS. This buildup of leachate head provides the
driving force for increased leakage through the underlying composite liner at locations where holes are
present in the geomembrane.
Several studies have quantified the leakage through
defects in a GMB overlying a CCL (e.g., Touze-Foltz
2002; Cartaud et al. 2005) or GCL (e.g., Harpur et al.
1993; Koerner & Koerner 2002; Rowe & Abdelatty
2012a). Based on both field and laboratory measurements for leakage through composite liners, empirical
(e.g., Giroud & Bonaparte 1989; Giroud 1997), analytical (e.g., Jayawickrama et al. 1988; Rowe 1998b), and
numerical (e.g., Walton et al. 1997; Foose et al. 2001)
solutions have been developed. However, no consideration has been given to the effect of a clogging-induced
reduction in drainage layer hydraulic conductivity
on the flow rate (leakage) through an underlying
composite liner.
The objective of this paper is to examine the
interaction between the clogging of a LCS and the
leakage through a composite liner. Drainage material
of three different grain sizes (dg = 1, 2, and 6 mm;
dg = D60 ) and two different composite liner systems
(GMB + GCL and GMB + CCL) are considered.
2
MODELING SUMMARY
The modeling effort for this study was divided into

two stages. First, the clogging of granular drainage
material was modeled using the recently enhanced
283
BioClog model (Yu & Rowe 2012a). The hydraulic

conductivity distribution within the drainage layer and
leachate mound was evaluated assuming that the leakage through the liner is sufficiently small compared to
the leachate generation rate that the leakage does not
affect the leachate mound. Second, the leakage through
the composite liner was calculated by the computer
program SEEP/W (GEO-SLOPE International Ltd.
2012) based on the leachate head and LCS hydraulic
conductivity distribution obtained from the BioClog
model. These two modeling stages are described in
further detail below.
2.1
governing differential equation for the 2D steady-state

seepage is:
Clogging of LCSs by BioClog
BioClog (Cooke et al. 2005a; Cooke & Rowe 2008;

Yu 2012; Yu & Rowe 2012a) models the fate and
transport of nine species (i.e., acetate, butyrate,
propionate, suspended acetate degraders, suspended
butyrate degraders, suspended propionate degraders,
suspended inert biomass, and suspended inorganic
solids) within a saturated drainage medium using the
advection-diffusion-reaction equation and simulates
the accumulation of five different films (i.e., acetate
degraders, butyrate degraders, propionate degraders,
inert biofilm, and inorganic solids) on the particle
surfaces. The increase in total film thickness reduces
the void space within the porous medium and the
consequent change in porosity and surface area are
computed using a geometric model (Yu & Rowe
2012b). The hydraulic conductivity, k, of the drainage
medium then is computed based on the porosity, n,
using the following empirical relationship (e.g., Rowe
et al. 2002; VanGulck & Rowe 2004b; Cooke et al.
2005b; Cooke & Rowe 2008):
where Ak and bk = empirically determined coefficients for the granular material. Based on Cooke &
Rowe (2008) the following values were adopted
herein: medium sand (dg = 1 mm)Ak = 1.6
1010 m/s and bk = 38.2; coarse sand (dg = 2 mm)
Ak = 7.6 1010 m/s and bk = 38.2; pea gravel
(dg = 6 mm) Ak = 1.7 108 m/s and bk = 38.2.
2.2
Figure 1. Schematic showing a modern leachate collection

system and an underlying composite liner system (not to
scale).
Leakage through composite liners by SEEP/W
The 2D software SEEP/W was used to simulate steadystate flow through the composite liners (GMB + GCL
and GMB + CCL) for the head and hydraulic conductivity distributions calculated at different times by
BioClog. Consideration was given to leakage through
a hole coinciding with a wrinkle in the GMB (see
Rowe 2012 for a discussion of factors affecting holes
in wrinkles). The distribution, width, and length of
wrinkles in GMBs have been documented by Chappel
et al. (2012a, b) and Rowe et al. (2012). The general
where h = total hydraulic head; kx and kz = hydraulic

conductivity in x and z direction, respectively; and
Q = applied boundary flux.
3
PROBLEM DEFINITION
The barrier system examined in this paper comprises

a LCS and an underlying composite liner system
(Fig. 1). The drainage path length (L) is 30 m and
the LCS thickness (B) is 0.3 m. The drainage layer
is graded to the drainage pipes at a slope () of
1%. As stated previously, consideration was given
to drainage material with three different nominal
grain sizes: dg = 1, 2, and 6 mm. The medium sand
(dg = 1 mm) had an initial porosity of 0.35 and initial hydraulic conductivity of 1 104 m/s. The initial
porosity of the coarse sand (dg = 2 mm) was 0.37
and initial hydraulic conductivity was 1 103 m/s.
The pea gravel (dg = 6 mm) had an initial porosity
and hydraulic conductivity of 0.39 and 5 102 m/s,
respectively. The leachate strength and infiltration rate
into the LCS were assumed to vary with time based on
inferred values for the Keele Valley landfill (Yu 2012;
Rowe & Yu 2012). The leachate infiltration rate into
the drainage layer was held constant at 0.3 m/year for
the first 18 years, was reduced to 0.2 m/year at 19
years, and was kept stable thereafter.
For the leakage analysis, two specific composite
liners were examined. Case 1 examined a composite GMB/GCL liner where the GCL was 0.0075 m
thick. Two saturated GCL hydraulic conductivities, k,
were considered: (i) k = 5 1011 m/s which is typical for a virgin GCL; and (ii) k = 2 1010 m/s which
corresponds to a GCL affected by cation exchange
from the underlying soil and/or clay-leachate interaction (Rowe & Abdelatty 2012a, b). The water retention
curve for the GCL was modeled as per Beddoe et al.
(2011). The interface between the GMB and the GCL
was modeled as a 0.0025 m thick layer with a hydraulic
284
conductivity corresponding to an interface transmissivity of 3 1011 m2 /s (Rowe 2012). The GCL rested
on an attenuation layer with a hydraulic conductivity of
1 107 m/s. The distance between the top of the GCL
and the top of an underlying aquifer was 3.75 m (i.e.,
minimum allowed for a single composite liner under
regulations in Ontario Canada). The piezometric level
in the aquifer was assumed to be at 3.5 m above the
top of the aquifer.
Case 2 examined a composite GMB/CCL liner
where a 0.9 m thick CCL rested on an attenuation
layer with a hydraulic conductivity of 1 107 m/s
and thickness such that the distance between the top
of the CCL and the top of an underlying aquifer was
3.75 m (as in Case 1). The piezometric level in the
aquifer also was the same as in Case 1. The saturated
hydraulic conductivity of the CCL was 1 109 m/s.
The interface between the GMB and the CCL was
modeled as a 0.0025 m thick layer with a hydraulic conductivity corresponding to an interface transmissivity
of 2 108 m2 /s or 1 107 m2 /s. This represents the
cases of good and poor contact conditions between
the GMB and the CCL, respectively (Rowe 1998b).
In both cases, the GMB was modeled as a 0.0015
m thick non-permeable layer with a hole having a
radius of 0.00564 m coincident with one interconnected wrinkle of width = 0.2 m and height = 0.06 m
(Rowe 2012). The wrinkle was located at distance
X = 15 m (i.e., at the mid-point between the drainage
pipe and the drainage divide in Fig. 1).
4
4.1
RESULTS AND DISCUSSIONS

Leachate mounding in LCSs due to clogging of
granular drainage material
Figure 2 shows the profiles of the leachate mound

at 0.5, 1, 1.5, and 2 years as well as the hydraulic
conductivity distribution in the leachate mound at 2
years for a medium sand drainage layer (grain size
dg = 1 mm). The leachate mound at X = 15 m (i.e.,
at the wrinkle location) increased to 0.06 m (i.e., the
wrinkle height) within the first 0.1 years. The leachate
mound continued increasing after 0.1 years and the
maximum leachate mounds at 0.5, 1, and 1.5 years
were 0.19, 0.23, and 0.26 m, respectively (Fig. 2a). At
about 2 years the maximum leachate mound reached
the design thickness 0.3 m; thus the service life of
this system is about 2 years. At 2 years (Fig. 2b),
the maximum hydraulic conductivity at the upstream
end was about 8 105 m/s (reduced from the initial
hydraulic conductivity of 1 104 m/s) and the minimum hydraulic conductivity at the downstream end
was about 1 105 m/s (Fig. 2c). Thus the hydraulic
conductivity of medium sand within the leachate
mound had decreased by up to one order of magnitude.
For the LCS with coarse sand (dg = 2 mm), the
leachate mound at X = 15 m reached 0.06 m at 0.7
years. The maximum leachate mounds at 1.5, 3, and
4.5 years were 0.09, 0.16, and 0.23 m, respectively
Figure 2. (a) Profiles of leachate mound with time, (b)

hydraulic conductivity within the LCS at 2 years due to the
clogging of medium sand with grain size dg = 1 mm, and
(c) detail of hydraulic conductivity distribution in last 1 m
approaching pipe.
(Fig. 3a). After approximately 6 years, the maximum leachate mound reached 0.3 m (i.e., the service
life of this LCS is 6 years). Figure 3b shows the
hydraulic conductivity distribution within the leachate
mound at 6 years. The maximum hydraulic conductivity was 8 104 m/s (reduced from the initial value of
1 103 m/s) at the upstream end, and the minimum
hydraulic conductivity was 5 107 m/s at the downstream end (Fig. 3c). Thus the hydraulic conductivity
within the leachate mound had reduced by up to four
orders of magnitude at the downstream end.
The profiles of leachate mound at 4, 8, 12, and 15
years for the LCS with pea gravel (dg = 6 mm) are
shown in Figure 4a. The leachate mound at X = 15 m
exceeded 0.06 m at about 4 years. The maximum
leachate mounds at 8 and 12 years were 0.16 and
0.25 m, respectively. The maximum leachate mound
reached 0.3 m at about 15 years (i.e., the service life
of this system is about 15 years). The hydraulic conductivity of pea gravel within the leachate mound
at 15 years is shown in Figure 4b. The maximum
hydraulic conductivity at the upstream end was about
2 103 m/s (reduced by an order of magnitude from
the initial value of 5 102 m/s) and the minimum
hydraulic conductivity at the downstream end was
about 8 107 m/s (Fig. 4c). Thus the hydraulic conductivity of pea gravel within the leachate mound was
reduced by up to five orders of magnitude.
The results show that, other conditions being equal,
increasing the grain size of granular drainage material
decreases the rate of clogging and hence the rate at
285

hydraulic conductivity within the LCS at 6 years due to
the clogging of coarse sand with grain size dg = 2 mm, and
approaching pipe.
which the leachate mounding grows within the granular drainage layer. The increase in service life of the
LCS with increasing particle size (assuming a uniform material with a uniformity coefficient < 2) arises
from: (i) a higher initial hydraulic conductivity for a
coarser particle size and hence a greater decrease in
hydraulic conductivity needed to cause the leachate
mound to reach the top of the drainage layer, (ii) the
larger pore size to be clogged associated with the
larger grain size, such that a greater mass of clog
material is needed to cause a drop in hydraulic conductivity, and (iii) the smaller surface area for film
attachment associated with the larger grain size, which
results in slower growth of the clog film. The maximum hydraulic conductivity occurs at the upstream
end because the relatively low flow rate in this region
gives rise to less clogging in this zone. The minimum
hydraulic conductivity is at the downstream end near
the pipe since this is where the maximum mass loading
occurs.
4.2
Leachate leaking through composite liners due

to wrinkle with a hole in geomembrane
Leakage through the composite liner comprised of a

GMB and a GCL due to a hole in the GMB coincident
with a wrinkle of length of 125 m is illustrated in Figure
5 for two cases: (a) when there is no significant change
in k of the GCL within the service life of the LCS compared to the k of virgin GCL (i.e., k = 5 1011 m/s,
Fig. 5a), and (b) when k of the GCL has increased

hydraulic conductivity within the LCS at 15 years due to
the clogging of pea gravel with grain size dg = 6 mm and
approaching pipe.
to 2 1010 m/s (Fig. 5b). In both cases, the interface transmissivity () is assumed to be 3 1011 m2 /s
(Rowe 2012).
In case (a), when a medium sand layer (dg = 1 mm,
ksand = 1 104 m/s) was placed above the composite
liner as a drainage layer, the leakage was zero between
0 to 0.1 years because the leachate head at the wrinkle
location (at X = 15 m) was lower than the level of the
hole (<0.06 m). Once the leachate reached the level of
the hole, the leakage increased to 6.0 lphd (litres per
hectare per day) (Fig. 5a). The combination of both
the increase in leachate head and the reduction in the
hydraulic conductivity of the drainage layer with time
(see Fig. 2) lead to an increase in the leakage through
the composite liner. For example, after 1, 1.5, and 2
years, the leakage increased to 8.6, 9.1, and 9.8 lphd,
respectively.
Using drainage layers with dg > 1 mm (e.g., coarse
sand with dg = 2 mm or pea gravel with dg = 6 mm,
Fig. 5a) leads to (i) an increase in time before the
leachate mound is high enough to reach the hole and
for leakage to begin (e.g., 0.7 years for coarse sand
and 4 years for pea gravel), and (ii) an increase in the
time required to reach a given leakage rate once the
leachate did enter the hole in the GMB (e.g., it took
1.9, 5.3, and 11 years for the leakage to increase from
6.9 lphd to 9.8 lphd for medium sand, coarse sand, and
pea gravel, respectively).
The leakage through the composite liner was mainly
dependent on the leachate head regardless of the distribution of the hydraulic conductivity in the drainage
layer for the three grain sizes examined. This is
286
Figure 5. Leakage through a hole in a 125 m long wrinkle

for a composite liner (GMB and GCL; = 3 1011 m2 /s)
with a GCL having: (a) k = 5 1011 m/s, and (b) k =
2 1010 m/s.
because the minimum k at the wrinkle location for

the cases examined only ranged between 1 105 and
2.5 105 m/s (see Figs 2b, 3b, and 4b) and therefore was still very high compared to the k of the
GCL (5 1011 m/s). Thus, when the drainage layer
reached the end of its service life, the leakage was
essentially independent of the nature of the granular
material; however, the time to reach the service life
was highly dependent on the nature of the granular
material.
A four-fold increase in k of the GCL from
5 1011 m/s to 2 1010 m/s leads to an increase in
leakage by about 2.7 times (Fig. 5b). For example, at
the service life of the LCS containing medium sand
(i.e., 2 years) the leakage increased from 9.8 lphd to
26.4 lphd.
Leakage through composite liners involving a CCL
(k = 1 109 m/s; = 2 108 m2 /s for good contact and 1 107 m2 /s for poor contact) is presented
in Figure 6 for a 125-m long wrinkle. For a good
contact condition and a medium sand drainage layer
(k = 1 104 m/s), the calculated leakage was 48 lphd
at 0.1 years (Fig. 6a). As the leachate head increased
with progressive clogging of the drainage layer, the
leakage through the composite liner increased to 65,
68, and 72 lphd at 1, 1.5, and 2 years, respectively,
despite the reduction in the hydraulic conductivity of
the drainage layer. Again, because the hydraulic conductivity of the drainage layer at the wrinkle location
Figure 6. Leakage through a hole in a 125 m long

GMB wrinkle for a composite liner with a CCL
with k = 1 109 m/s having: (a) good contact, = 2
108 m2 /s, and (b) poor contact, = 1 107 m2 /s.
at the end of the service life of LCS (1 105

2.5 105 m/s) was still highly permeable relative to
the k of the CCL (1 109 m/s), the k of the drainage
layer had negligible impact on leakage in terms of providing hydraulic resistance. In case of poor contact
between GMB and CCL, the leakage essentially doubled compared to the case of good contact (Fig. 6b).
For example, when pea gravel was used as drainage
layer above GMB/CCL composite liner, the leakage at
15 years was 72 lphd for good contact and 146 lphd
for poor contact between the GMB/CCL.
For all cases examined, the reduction in the
hydraulic conductivity of the drainage layer at the
wrinkle location had negligible effect on the leakage through a hole coincident with a wrinkle in the
GMB forming part of a composite liner since the
total head loss predominantly occurred at the GMB
clay liner interface and through the clay liner, not
the drainage layer. To quantify the reduction in k
of the drainage layer that would be required to significantly affect the leakage through the composite
liners, four cases were simulated. In each case, the
leachate head was assumed to be 0.3 m above the hole,
and the k of the drainage layer was varied between
1 103 m/s and 1 109 m/s. In the first two cases,
the GMB/GCL interface had = 3 1011 m2 /s and
the effect of the hydraulic conductivity of the GCL
of (i) k = 5 1011 m/s or (ii) k = 2 1010 m/s
was examined. The second two cases considered
287
Figure 7. Leakage through a hole in a GMB with a

125 m long wrinkle with a hole for a composite liner
for a (i) GCL k = 5 1011 m/s, = 3 1011 m2 /s, (ii)
GCL k = 2 1010 m/s, = 3 1011 m2 /s, (iii) CCL with
k = 1 109 m/s, = 2 108 m2 /s, (iv) k = 1 109 m/s,
= 1 107 m2 /s.
a CCL with k = 1 109 m/s and examined (iii)

= 2 108 m2 /s, and (iv) = 1 107 m2 /s.
For the GCL, a reduction in the hydraulic conductivity of the drainage layer from 1 103 m/s to
1 107 m/s led to only a very slight (1.5 to 4.2%)
reduction in the leakage (Fig. 7). At a hydraulic conductivity of 1 108 m/s, the leakage was reduced by
14% and 31% for k of 5 1011 and 2 1010 m/s,
respectively.
With a CCL, there was negligible change in leakage
for cases in which the k of the drainage layer ranged
from 1 103 m/s to 1 106 m/s. When the k of the
drainage layer was reduced to 1 107 m/s, the leakage was reduced by 9% and 17% for the cases of good
and poor contact, respectively. A reduction in the leakage by 49% and 66% was calculated for the cases of
good and poor contact, respectively, when the k of the
drainage layer was reduced to 1 108 m/s.
5
CONCLUSIONS
Based on the results presented in this paper, the following conclusions were reached for the conditions
examined:
When the clogging of granular drainage material is

taken into account, the leachate mound in a LCS
that satisfies the typical design requirements can be
expected to eventually exceed the thickness of the
drainage layer even though initially the mound is
well within this layer thickness.
Increasing the grain size of granular drainage material (e.g., from medium sand to pea gravel) decreases
the rate of clogging and extends the time it takes
for the leachate mound to reach the drainage layer
thickness (i.e., the service life of the LCS).
The clog mass accumulated within the pore space
of porous media reduced the hydraulic conductivity
of the granular material in the drainage layer by one
and five orders of magnitude for medium sand and

pea gravel, respectively, before the drainage layer
was considered to have reached its service life.
For the three drainage layer grain sizes examined,
the flow through composite liners depends only on
the leachate head regardless of the hydraulic conductivity distribution in the drainage layer up to the
point where the LCS service life is reached.
A four-fold increase in the k of GCL from
5 1011 m/s to 2 1010 m/s resulted in an
increase in leakage through a GMB/GCL composite
liner by 2.7 times.
Leakage through GMB/CCL liner was much higher
than the leakage through GMB/GCL liner.
The leakage through GMB/GCL system would not
be affected by a reduction in the hydraulic conductivity of the drainage layer until it dropped to below
about 107 m/s at the location of the hole in the
GMB.
Leakage through GMB/CCL composite liner would
not be affected by a reduction in the hydraulic conductivity of the drainage layer until it dropped to
below about 106 107 m/s at the location of the
hole in the GMB.
ACKNOWLEDGEMENTS
This research was funded by a grant from the Natural
Sciences and Engineering Research Council of Canada
(NSERC).
REFERENCES
Bass, J.M. 1986. Avoiding failure of leachate collection and
cap drainage systems. EPA-600/2-86-058, USEPA, Land
Pollution Control Division, Hazardous Waste Engineering
Research Laboratory, Cincinnati, OH.
Beddoe, R.A., Take, W.A., & Rowe, R.K. 2011. Water
retention behaviour of geosynthetic clay liners, Journal of Geotechnical and Geoenvironmental Engineering 137(11): 10281038
Bouchez, T., Munoz, M.L., Vessigaud, S., Bordier, C., Aran,
C., & Duquennoi, C. 2003. Clogging of MSW landfill leachate collection systems: prediction methods and
in situ diagnosis. In Proceedings of Sardinia 2003, 9th
International Waste Management and Landfill Symposium, S. Margherita di Pula, Cagliari, Italy, 610 October
2003. CISA, Environmental Sanitary Engineering Centre,
Cagliari, Italy (CD-ROM).
Brune, M., Ramke, H.G., Collins, H.J., & Hanert, H.H. 1991.
Incrustation processes in drainage systems of sanitary
landfills. In Proceedings of Sardinia 91, 3rd International
Landfill Symposium, S. Margherita di Pula, Cagliari,
Engineering Centre, Cagliari, Italy: 9991035.
Cartaud, F., Touze-Foltz, N., & Duval, Y. 2005. Experimental
investigation of the influence of a geotextile beneath the
geomembrane in a composite liner on leakage through a
hole in the geomembrane. Geotextiles and Geomembranes
23(2): 117143.
Chappel, M.J., Brachman, R.W.I., Take, W.A., & Rowe, R.K.
2012a. Large-scale quantification of wrinkles in a smooth,
288
black, HDPE geomembrane. Journal of Geotechnical and

Chappel, M.J., Rowe, R.K., Brachman, R.W.I., & Take, W.A.
2012b. A comparison of geomembrane wrinkles for nine
field cases. Geosynthetics International 19(6): 453469.
Cooke, A.J., Rowe, R.K., & Rittmann, B.E. 2005a. Modelling
species fate and porous media effects for landfill leachate
flow. Canadian Geotechnical Journal 42(4): 11161132.
Cooke, A.J., Rowe, R.K., VanGulck, J.F., & Rittmann, B.E.
2005b. Application of the BioClog model for landfill
leachate clogging of gravel-packed columns. Canadian
Geotechnical Journal 42(6): 16001614.
Cooke, A.J. & Rowe, R.K. 2008. 2D modelling of clogging in landfill leachate collection systems. Canadian
Craven, W., Townsend, T.G., Vogel, K., & Laux, S. 1999.
Field investigation of landfill leachate collection system
clogging. Practice Periodical of Hazardous, Toxic, and
Radioactive Waste Management 3(1): 19.
Fleming, I.R. & Rowe, R.K. 2004. Laboratory studies of clogging of landfill leachate collection & drainage systems.
Fleming, I.R., Rowe, R.K., & Cullimore, D.R. 1999. Field
observations of clogging in a landfill leachate collection
system. Canadian Geotechnical Journal 36(4): 685707.
Foose, G.J., Benson, C.H., & Edil, T.B. 2001. Predicting
leakage through composite landfill liners. Journal of
510520.
GEO-SLOPE International Ltd, 2012. SEEP/W 2012
Groundwater Seepage Analysis. GEO-SLOPE International Ltd., Calgary, Alberta, Canada.
Giroud, J.P. 1997. Equations for calculating the rate of liquid
migration through composite liners due to geomembrane
defects. Geosynthetics International 4(34): 335348
Giroud, J.P. & Bonaparte, R. 1989. Leakage through liners constructed with geomembranes Part II. Composite
liners. Geotextiles and Geomembranes 8(2): 71111.
Harpur, W.A., Wilson-Fahmy, R.F., & Koerner, R.M. 1993.
Evaluation of the contact between geosynthetic clay
liners and geomembranes in terms of transmissivity. GRI
Seminar on Geosynthetic Liner Systems, Philadelphia, PA,
IFAI: 143154.
Jayawickrama, P., Brown, K., Thomas, J., & Lytton, R. 1988.
Leakage rates through flaws in geomembrane liners. Journal of Environmental Engineering 114(6): 14011420
Koerner, G.R. & Koerner, R.M. 2002. Geomembrane leakage
arising from broken needles with GCLs. In Proceedings
of the International Symposium on Geosynthetic Clay
Barriers, Nuremberg, Germany: 209217.
Koerner, G.R., Koerner, R.M., & Martin, P.M. 1993. Field
performance of leachate collection systems and design
implications. In Proceedings of 31st Annual International
Solid Waste Exposition, San Jose, Calif., August 1993.
Solid Waste Association of North America, Silver Spring,
Md.: 365380.
Koerner, G.R., Koerner, R.M., & Martin, J.P. 1994. Design
of landfill leachate collection filters. Journal of Geotechnical Engineering 120(10): 17921801.
Levine, A.D., Cardoso, A.J., Nayak, B., Rhea, L.R., Dodge,
B.M., & Harwood, V.J. 2005. Assessment of biogeochemical deposits in landfill leachate drainage systems. Project
Final Report.Center for Solid and Hazardous Waste Management, Gainesville, Florida, USA.
Maliva, R.G., Missimer, T.M., Leo, K.C., Statom, R.A.,
Dupraz, C., Lynn, M., & Dickson, J.A.D. 2000. Unusual
calcite stromatolites and pisoids from a landfill leachate
collection system. Geology 28(10): 931934.
McBean, E.A., Mosher, F.R., & Rovers, F.A. 1993.

Reliability-based design for leachate collection systems.
In Proceedings of Sardinia 93, 3rd International Landfill Symposium, Santa Margherita di Pula, Cagliari,
Engineering Centre, Cagliari, Italy, 3: 433441.
McIsaac, R. & Rowe, R.K. 2006. Effect of filter-separators
on the clogging of leachate collection systems. Canadian
McIsaac, R. & Rowe, R.K. 2007. Clogging of gravel
drainage layers permeated with landfill leachate. Journal of Geotechnical and Geoenvironmental Engineering
133(8): 10261039.
Ontario Regulation. 1998. Ontario Regulation 232/98,
Ontario Ministry of the Environment, Environmental
Protection Act Ontario, PIBS 3651E.
Paksy, A., Peeling, L., Robinson, J.P., Powrie, W., & White,
J.K. 1995. Incrustation processes in drainage systems
of sanitary landfills. In Proceedings of Sardinia 95, 5th
International Landfill Symposium, S. Margherita di Pula,
Cagliari, Italy, 26 October 1995. CISA, Environmental
Sanitary Engineering Centre, Cagliari, Italy: 681690.
Paksy, A., Powrie, W., Robinson, J.P., & Peeling, L. 1998. A
laboratory investigation of microbial clogging in granular
landfill drainage media. Geotechnique 48(3): 389401.
Peeling, L., Paksy, A., Robinson, J.P., & Powrie, W. 1999.
Removal of volatile acids from synthetic landfill leachate
by anaerobic biofilms on drainage aggregates: a laboratory study. Waste Management and Research 17:
141149.
Rowe, R.K. 1998a. From the past to the future of landfill engineering through case histories. In Proceedings
of 4th International Conference on Case Histories in
Geotechnical Engineering, St.Louis, Mo., 912 March
1998. University of Missouri-Rolla, Rolla, Mo.: 145166.
Rowe, R.K. 1998b. Geosynthetics and the minimization of
contaminant migration through barrier systems beneath
solid waste. In proceedings of 6th International Conference on Geosynthetics, Atlanta, IFAI, St .Paul. MN. 1:
27103.
Rowe, R.K. 2005. Long-term performance of contaminant
barrier systems. Geotechnique 55(9): 631678.
Rowe, R.K. 2012. Short and long-term leakage through
composite liners. Canadian Geotechnical Journal 49(2):
141169.
Rowe, R.K. & Abdelatty, K. 2012a. Modeling contaminant transport through composite liner with a hole in
the geomembrane. Canadian Geotechnical Journal 49(7):
773781.
Rowe, R.K. & Abdelatty, K. 2012b. Effect of a calcium-rich
soil on the performance of an overlying GCL. Journal of
423431.
Rowe, R.K. & Yu, Y. 2012. A practical technique for estimating service life of MSW leachate collection systems.
Canadian Geotechnical Journal, (in press).
Rowe, R.K., Armstrong, M.D., & Cullimore, D.R. 2000a. Particle size and clogging of granular media permeated with
leachate. Journal of Geotechnical and Geoenvironmental
Rowe, R.K., Armstrong, M.D., & Cullimore, D.R. 2000b.
Mass loading and the rate of clogging due to municipal solid waste leachate. Canadian Geotechnical Journal
37(2): 355370.
Rowe, R.K., Chappel, M.J., Brachman, R.W.I., & Take, W.A.
2012. Field monitoring of geomembrane wrinkles at a
composite liner test site. Canadian Geotechnical Journal
49(10): 11961211.
289
Rowe, R.K., Quigley, R.M., Brachman, R.W.I., & Booker,

J.R. 2004. Barrier systems for waste disposal facilities.
Taylor & Francis Books Ltd (E & FN Spon) London, 587 p.
Rowe, R.K., VanGulck, J.F., & Millward, S. 2002. Biologically induced clogging of a granular media permeated with
synthetic leachate. Canadian Journal of Environmental
Engineering and Science 1(2): 135156.
Touze-Foltz, N. 2002. Evaluation of the hydraulic transmissivity in soil liner-geomembrane interfaces. In Proceedings of the Seventh International Conference on
Geosynthetics, vol. 2, Nice, France: 799802.
US EPA. 1991. Solid Waste Disposal Facility Criteria: Final
Rule, Part II. Federal Register, 40 CFR Parts 257 and 258.
US Environmental Protection Agency, Washington, D.C.
VanGulck, J.F. & Rowe, R.K. 2004a. Evolution of clog formation with time in columns permeated with synthetic
landfill leachate. Journal of Contaminant Hydrology 75:
115135.
VanGulck, J.F. & Rowe, R.K. 2004b. Influence of landfill
leachate suspended solids on clog (biorock) formation.
Waste Management 24(7): 723738.
Walton, J., Rahman, M., Casey, D., Picornell, M., & Johnson,
F. 1997. Leakage through ?aws in geomembrane liners.
Young, C.W., Nummo, T.J., Jasniksi, M.P., Cogley, D.R., &
Capone, S.V. 1982. Clogging of leachate collection systems used in hazardous waste disposal facilities. Draft
White Paper, prepared for USEPA, Research Triangle
Park, North Corolina.
Yu, Y. 2012. Modelling MSW leachate characteristics and
clogging. Ph.D. thesis. Department of Civil Engineering,
Queens University, Kingston ON, Canada.
Yu,Y. & Rowe, R.K. 2012a. Modelling leachate-induced clogging of porous media. Canadian Geotechnical Journal
49(8): 877890.
Yu, Y. & Rowe, R.K. 2012b. Improved solutions for porosity and specific surface of a uniform porous medium
with attached film. Journal of Environmental Engineering
138(4): 436445.
290

Some geotextiles properties useful for HDPE geomembrane

puncture protection
G. Stoltz, D. Croissant & N. Touze-Foltz
Irstea, Hydrosystems and bioprocesses Research Unit, Antony, France
ABSTRACT: Geotextiles are used in the constitution of landfill lining systems in order to protect the geomembrane against puncturing effects during installation of the granular drainage layer, then during the life of the
landfill site, in relation with the weight of waste. The objective of this paper is to evaluate the performance
of nonwoven needlepunched geotextiles having an average mass per unit area equal to 1000 g/m2 in protecting
geomembrane, based on static puncture tests performed in 1 m diameter cells. Recommendations are given in
the following as regards the length of fibers, the elasticity modulus, the mass per unit area and the thickness of
the geotextile depending on the nature of the granular material used.
INTRODUCTION
The liner systems of landfills are subjected to puncturing stresses directly linked to the granular layer. The
role of geomembrane liners under most loading conditions is strongly influenced upon the interaction and
compatibility of the various components within the
liner system. To avoid any geomembrane deformation
that could decrease its watertightness performance
and, in the worst case, leads to a failure by puncturing,
protective geotextiles are often used. These protective
geotextiles permit to reduce the occurrence of local
stresses and thus to control potential puncture of the
geomembrane by the granular material installed immediately above. The design of such geotextiles requires
a thorough understanding of the interaction between
the liner system components (i.e. granular material,
geotextile, geomembrane, foundation).
Some studies tried to depict the puncturing mechanisms by a theoretical approach (Giroud et al., 1995;
Wilson-Fahmy et al., 1996). Nevertheless, the selection of adequate protective geotextiles is often made
after experimental tests. Two types of tests can be
made: short term tests and long term tests. The general
principle of short term test is to measure the puncture
pressure (e.g. the ASTM standard method D 5514).
In case of long term tests, the loading pressure, corresponding to the one of the site possibly major by safety
factor, is applied during a given time and the deformations of the geomembrane are observed at the end of
the test. This is for example the case of the standard
cylinder test (NF EN 13719, AFNOR, 2003). Some
design methods of protective geotextiles are derived
from these tests (Narejo et al., 1996; Environmental
Agency, 2006).
To optimise the design of a protective geotextile, its physical and mechanical properties that affect
puncture protection have to be precisely characterised. When non-woven needle-punched geotextiles

are considered as protective geotextiles, it has been
demonstrated that some physical properties like the
mass per unit area are essential (Koerner et al.,
1996). Moreover, following Shercliffe (1996), the
CBR puncture resistance obtained following the Standard NF EN ISO 12236 (AFNOR, 2006), seems to
be a mechanical property that significantly influences
the HDPE geomembrane puncture protection. Other
extern parameters control the puncture behavior of
geomembranes like the normal stress level, the particle
size distribution and protrusion shape of the granular
drainage layer and the nature of the substratum directly
below the considered liner.
The aim of this study is to reveal some other geotextiles properties that can be taken into account to
design the adequate protective geotextile.
In a first step, the method consisted in applying
a static stress during four days on a simulated liner
system using a 1m diameter specific device. The
liner system from the bottom to the top was composed by a dense rubber pad simulating the compacted
clay layer; a 2 mm thick High Density Polyethylene
(HDPE) geomembrane, a protective geotextile and
0.3 m thick 20/40 mm granular layer composed by
either angular or rounded particles. Eleven non-woven
needle-punched geotextiles from various manufacturers were tested. Their mass per unit area was close to
1000 g/m2 . A previous study showed that, in the range
of this mass per unit area, the protective level of the
geotextile varied depending of the nature of the products that thus could be adequate or not. This aspect
is consistent with the results obtained by Jones et al.
(2000).
In a second step, after unloading, the geomembrane was visually inspected and a biaxial tensile test
291
Table 1. Some physical and mechanical properties of the

tested geotextiles (PP: polypropylene and PE: polyester).
Figure 1. Schematic of the geotextile during puncturing test

following NF P 84507.
was carried out. Following a criteria depending on

the observations and on the results of the biaxial tensile test, the protective geotextiles were classified as
efficient or not. Then, the physical and mechanical
properties of the geotextiles were analysed regarding
their protective efficiency.
2
2.1
Thickness
(mm)
Mass
EN ISO
per unit
Ea
GTX 9863-1
area (g/m2 ) (MPa) Polymer
Fiber
length
1
2
3
4
7.5
6.5
7.0
7.0
1004(18)
987(89)
997(21)
985(27)
26.1
21.0
23.1
10.0
Short
Long
Short
Short
5
6
7
8
9
10
7.2
6.1
7.5
6.3
6.6
7.0
1014(57)
973(38)
918(34)
1076(23)
1029(52)
971(32)
14.9
28.3
15.3
27.9
24.2
9.6
11
7.1
1003(28) 19.4
Virgin PP
Virgin PP
Virgin PP
PP &
recycle PE
Virgin PP
Virgin PP
Virgin PP
Virgin PP
Virgin PP
Multi-color
PP
Virgin PP
Long
Short
Short
Short
Short
Short
Long

Geotextiles
Table 1 presents a part of useful properties of geotextiles used in this study. All geotextiles have a mass per
unit area close to 1000 g/m2 . This property was measured following the NF EN ISO 9864 (AFNOR, 2005).
The measurement is given with an uncertainty corresponding to two standard deviations. The apparent
elasticity modulus Ea of the products was measured.
To do so, the standard test NF P 84507 (AFNOR,
1996) that permits to measure the puncture resistance
of a geomembrane specimen was adapted to geotextile. The apparent elasticity modulus Ea is determined
by the Hooke relationship (1) applied on the rim of the
puncturing rod (Figure 1):
where F = force applied by the puncturing rob;

rrob = puncturing rob radius; e = geotextile thickness;
L0 = initial length of the geotextile; L = length of the
geotextile during the test.
It is useful to note that the geotextile is supposed
to be elastic, homogeneous, isotropic and to have a
constant thickness during the test.
The heterogeneity of the geotextiles depended on
the various products as displayed by Figure 2 showing the variability of the mass per unit area from 10
specimens for three types of geotextiles. For each geotextile, specimens were sampled across de roll width
following NF EN ISO 9862. This heterogeneity could
affect the puncturing protection efficiency.
Figure 2. Variability of the mass per unit area from 10

samples for three types of geotextiles.
292
instead of 1000 h, duration required to be representative of long term conditions). Without considering any
safety factor on the mechanical stress, a vertical stress
of 535 kPa has to been applied on the liner system.
Gaillard et al. (2011) showed that the experimental device (Figure 3) does not permit to uniformly
spread the vertical stress that is applied onto the gravel
material, principally due to not controlled boundary
conditions. Indeed, there is a stresses concentration
directly spread on the liner system. Calibration curves
established using total stress sensors showed that a vertical stress of 300 kPa applied onto the gravel layer
permitted to apply a vertical stress of 535 kPa onto the
liner system.
2.5 Method for geotextile classification
After the static puncture test in the large-scale device,
the analysis of the protection efficiency of the geotextiles were both based on a visual inspection and
on a biaxial tensile test performed on the geomembrane. In any case, the major criterion that determined
whether the protection was efficient or not was the
biaxial tensile test.
2.5.1 Visual inspection
The visual inspection consisted in detecting any damage on the geomembrane as suggested by the French
Chapter of IGS (CFG) (CFG, 2001). Five type of
damage are noticed:
Figure 3. Schematic of the experimental device for the static

puncture tests.
2.2 HDPE geomembrane
Scrath with no reduction of the geomembrane

thickness,
Notch with low reduction of the geomembrane
thickness,
indentation: significant reduction of the geomembrane thickness but without any hole,
deformation with no reduction of the geomembrane thickness, and
hole or opening
There was only one tested HDPE geomembrane that

was 2 mm thick. All geomembrane specimens tested
disks were sampled in a single roll.
2.3 Granular material
Two types of 20/40 mm granular material that are
expected to be used in landfills were tested: one with
angular particles and the other with rounded particles.
2.4 Experimental device
The tests were performed in a metric-scale experimental device (Figure 3) that permits to reproduce the
liner system with a good representativeness.
The regulation on waste landfilling in Europe states
that the substratum of a liner system in a landfill has
to be a compacted clay layer. To reproduce this layer,
a dense 0.95 m diameter rubber pad 25 mm thick was
used. This material has a hardness of 50 Shore A and
was chosen as proposed in the standard NF EN 13719
(AFNOR, 2003).
In French landfills, the typical height of the solid
waste heap is about 30 m. By taking into account the
0.5 m thick granular layer and the 1 m thick cover
soil layer, the vertical stress onto the liner system is
about 320 kPa. Consistently to what is described in
NF EN 13719 (AFNOR, 2003), a safety factor of 1.67
was applied on the loading stress to take into account
the temperature effect (test performed at 23 C instead
of 40 C and the duration of the test that was 100 h
The visual inspection was performed following

two steps. First, a detailed analysis was made on the
geomembrane to notice any damage described above.
Then, the damages were classified following their
weight and their number by two operators. Finally, if
an unacceptable damage was identified on a geomembrane specimen, the geotextile that was used to protect
it was considered as non-efficient. In any case, this
designation was confirmed by a biaxial test on the
geomembrane specimen.
2.5.2 Biaxial tensile test
This test was performed on geomembrane specimens
following NF EN 14151 (AFNOR, 2010). It consisted
in applying a pressure under the specimen with a circular shape, which deforms the geomembrane and,
in the specimen center, induces an isotopic stress in
the tangent plane of the geomembrane (Lambert and
Duquennoi, 2001). During this test, the bending strain
at the center of the specimen was measured. Then the
vessel pressure and the bending strain permitted to
determine the stress and strain of the geomembrane
293
basing on some hypothesis that consisted in considering the strain as spherical and homogeneous during the
whole test, incompressibility of the geomembrane and
stress field constant in the geomembrane thickness.
For each geomembrane specimen subjected to a

static puncture test, a spherical piece with a 0.3 m
diameter was sampled.As far as possible, the geomembrane specimen for the biaxial test was centered on the
most noticeable damage.
The property considered for the biaxial tensile test
analysis is the strain at failure since it permitted a better
distinction of the damage effects. In this study it was
considered that the protective geotextile was efficient
if the loss of geomembrane mechanical characteristics
at failure was less than 25% of its virgin performances.
It is worth noting that this criterion is used by in NF EN
13493 (AFNOR, 2006). Figure 4 shows the difference
between a virgin geomembrane specimen subjected
to biaxial tensile test and a specimen from the static
puncture tests.
3
Figure 4. Schematic of the geomembrane sample after a

biaxial tensile test: virgin sample on the top and sample
subjected to static puncture test on the bottom.
RESULTS
The results obtained for the rolled granular materials are presented on Figure 5. The big rectangles
correspond to the geotextile mass per unit area while
the small rectangles correspond to the apparent elastic
modulus. The rectangles are green and noticed with an
S when the protective performances of geotextile are
assumed as satisfactory. They are red and noticed
with a U when the protective performances of geotextile are assumed as unsatisfactory. It is observed
that the protection was satisfactory when the mean
mass per unit area of the geotextile was larger than
1000 g/m2 , or very close, and the apparent elastic modulus was larger than 10 MPa. For the geotextile 6,
although the mass per unit area was locally superior to
1000 g/m2 , this was not sufficient to ensure a sufficient
protection of the geomembrane.
The results obtained for rounded granular materials
are presented on Figure 6. In this case, it is observed
that the protection function brought by the geotextile
Figure 5. Mass per unit area (big rectangles) and apparent elastic modulus (small rectangles) of protective geotextiles
tested with the rounded granular material; the rectangles are green and noticed with a S when the protective performance of
geotextile are assumed as satisfactory unlike the red rectangle noticed with a U that indicates an unsatisfactory protective
performance.
294
Figure 6. Mass per unit area (big rectangles) and thickness (small rectangles) of protective geotextiles tested with the angular
granular material; the rectangles are green and noticed with a S when the protective performance of geotextile are assumed as
satisfactory unlike the red rectangle noticed with a U that indicates an unsatisfactory protective performance.
was not sufficient when the thickness was smaller

than 6.5 mm. For higher thicknesses, the geotextiles
consisted of short fibers assembly did not bring satisfactory protection. Based on these results and taking
into account the experimental conditions, it can be recommended to use geotextiles with a mass per unit area
larger than or equal to 1000 g/m2 , with a thickness
larger than 6.5 mm and made of short fibers.
CONCLUSION
Eleven geotextiles from various companies and with

a mass per unit area larger than or equal to 1000 g/m2
were tested in a large scale device to evaluate
their protection efficiency of the geomembrane liner.
The major aim of this study was to correlate the protection efficiency of the geotextiles with some physical
and mechanical properties.
To qualify the protection efficiency of the geotextiles as satisfactory or unsatisfactory, a visual
inspection and a biaxial tensile test were performed
on the tested geomembrane samples. The results were
obtained for specific conditions as a normal stress corresponding to 30 m of a household solid waste column
and for a 2 mm thick HDPE geomembrane.
The results obtained for rounded granular material
suggested a minimum mass per unit area of 1000 g/m2
for the geotextile and an apparent elastic modulus
superior to 10 MPa. For rounded granular material, the
mass per unit area of the geotextile has to be superior to
1000 g/m2 , the thickness has to be superior to 6.5 mm
and the short fibers assembly was more efficient.
This study suggested that some specific physical
and mechanical properties of the geotextile should be
taken into account for the design of geotextile for the
puncture protection of geomembrane liners for landfill applications. Moreover, it can be noticed that the
technical sheet of the various products may not be sufficient as they generally do not include any indication
about the heterogeneity of the mass per unit area. This
study is a first step to define a design method based on
relevant geotextile properties but these first results are
not currently sufficient to serve for designing a project.
REFERENCES
AFNOR, 2010. NF EN 14151, Geosynthetics
Determination of burst strength
AFNOR, 2006 NF EN 13493, Geosynthetic barriers
Characteristics required for use in the construction of solid
waste storage and disposal sites.
AFNOR, 2005. NF EN ISO 9864, Geosynthetics Test
method for the determination of mass per unit area of
geotextiles and geotextile-related products.
AFNOR, 2005. NF EN ISO 9862, Geosynthetics Sampling
and preparation of test specimens.
AFNOR, 2003. NF EN 13719, Geotextiles and geotextilerelated products Determination of the long term protection efficiency of geotextiles in contact with geosynthetic
barriers.
AFNOR, 1996. NF EN ISO 12236, Geotextiles and
geotextile-related products Static puncture test (CBR
test)
AFNOR, 1996. NF P 84507, Testing of geomembranes.
Determination of the static puncture resistance of
geomembranes and geomembrane lining systems. Case
of cylindrical punch without support.
ASTM D5514-94, 2001 Standard Test Method for Large
Scale Hydrostatic Puncture Testing of Geosynthetics.
Environmental Agency, 2006. Methodology for cylinder
testing of protectors for geomembranes on landfill sites,
Environmental Agency U.K.
Gaillard, G., Croissant, D., Touze-Foltz, N., 2011. Evaluation
de la protection contre lendommagement des gomembranes en PEHD. 8mes Rencontres Gosynthtiques,
Tours, France.
Giroud, J.P., Badu-Tweneboah, K. & Soderman, K.L. 1995.
Theoretical analysis of geomembrane puncture. Geosynthetics International 2 (6), 10191048.
295
Jones et al., 2000. Difficulties associated with the specification of protection geotextiles using only unit weight,
Proceedings Eurogeo2, Bologna, Italy, 551555.
Koerner, Wilson-Fahmy, Narejo, 1996: Puncture Protection
of Geomembranes: Part III: Examples, Geosynthetics
International, Vol. 3, no. 5, pp. 655675
Lambert S., Duquennoi C. (2001) From burst test to bi-axial
tensile test, Geosynthetics Conference 2001, Portland,
USA, 701714.
Narejo, Koerner, Wilson-Fahmy, 1996: Puncture Protection

of Geomembranes: Part II: Experimental, Geosynthetics
International, Vol. 3, no. 5, pp. 629653
Shercliffe D.A., 1996: Optimisation and Testing of Liner
Protection Geotextile Used in Landfills, EuroGeo 1, First
European Geosynthetics Conference, Maastricht
Wilson-Fahmy, Narejo, Koerner, 1996: Puncture Protection of Geomembranes: Part I: Theory, Geosynthetics
International, Vol. 3, no. 5, pp. 605628.
296
Landfill waste characterization

Geotechnical characterization of waste dredged sediments

for Algerian dams
M.A. Bourabah, B. Serbah & N. Abou-Bekr
Dpartement de Gnie-Civil, Facult de Technologie, Universit Abou-Bekr Belkaid,
Laboratoire Eau et Ouvrages dans leur Environnement, Tlemcen, Algrie
S. Taibi
Laboratoire Ondes et Milieux Complexes LOMC, Universit du Havre, Le Havre, France
ABSTRACT: Each year, more than 32 million m3 of sediments settle to the bottom deductions ofAlgerian dams.
The management of these waste sediments after dredging has become increasingly complex. Many researchers
have shown the possibility to use these dredged sediments in road construction, which allow a better use of this
natural resource. This study is part of research a methodology of waste management of dredged sediments in the
western Algeria for using in road engineering. The first objective is to determine the physical and mechanical
characteristics of fine dredged sediments from two deductions of the west of Algeria (Cheurfas and Bakhadda
dams). In the second stage, the characteristics of the fine sediments are enhanced for use as a road material
because, in their original condition, the waste dredged sediments do not have the mechanical characteristics
required by the technical guide to be used in foundation layer. To do this, the sediment can be treated with
cement, lime, or substituting other materials like quarry sand and dredged sand. We show in this study, the
influence of adding granular material and binders as both particle size and mechanical corrector for improving
these geotechnical state settings, the optimal characteristics of compaction and the bearing capacity of the
material. Various formulations were elaborated in this direction. The paper presents a preliminary synthesis of
the influence of these materials components on the mechanical behaviour for studied formulations.
INTRODUCTION
Internationally, road engineering is concerned on the

use of waste materials of various origins. Development
must comply with the following objectives:
Contribute to the protection of the environment by

providing an outlet for the waste, thus avoiding
stockpiling;
Contribute to the economies of traditional materials;
Provide road construction materials of sufficient
quality and having a certain intrinsic stability.
Among the waste, in terms of an overall assessment
of their properties for road use, the sediment dredging
of dams offer opportunities.
These materials are indeed likely to have better characteristics, either as such or treated or in combination
with others.
In North Africa, nearly 122 million m3 of sediments
are deposited annually at the bottom of 230 dams with
a total capacity of 23 billion m3 , with, in order of
decreasing importance: Morocco with 65 106 m3 , the
Algeria with 32 106 m3 and the Tunisia with 25 106 m3
(Remini et al., 2007).
In 2002, the volume of sediments deposited in 57
large dams was estimated at 0.9 km3 . The total amount
is 1.1 km3 in 2010 and 1.35 km3 in 2020 for the same
number of dams. The annual rate of silting exceeds
60 hm3 widely from year 2010 for 100 large dams
planned (Remini et al., 2003).
Algeria lost a volume of water storage of 32 million
m3 on a volume of 5.2 billion m3 of large deductions
in operation (Mekerta et al., 2008). Thus, more than
30 million m3 of sediments are deposited annually
at the 57 Algerian large dams; fifteen are seriously
threatened by siltation (Figure 1). The desilting by
dredging has become an inevitable option to extend
their lifespan (Hallouche and Remini, 2004).
These quantities of deposited materials come from
accelerated erosion of watersheds due to physical,
geomorphologic, hydro-climatic and socio-economic
conditions.
Various technical means to combat this phenomenon of sedimentation have been used by services
of hydraulics since more than a century, namely: the
management of watersheds and rivers, the evacuation
of sediments by bottom valves, the filling of density currents, the elevation and the dredging of dams
(Remini et al., 2003).
A large number of dams has been recovered which
life is prolonged for several years. However, the quantities of sediment removed by dredging and deposited
299
Figure 2. Global view of the two Cheurfas and Bakhadda

deductions.
PRESENTATION OF THE TWO STUDY SITES
The two studied Northwestern Algerian deductions

are (Figure 2):
Cheurfas dam located at 15 km above the town of
Sig in the Mascara region. Its exploitation begun in
1992 and is intended to irrigate surrounding agricultural land as well as drinking water supply of the
town of Sig. This dam, which crosses the wadi El
Mebtouh is designed to regulate 45 million m3 of
water. It is built on the ruins of an old dam swept
away by a flood in 1927 (Labiod et al., 2001). The
dams major river basin has a total area estimated
of 4190 km2 .
Bakhadda dam located on the upper reaches of the
Mina, a tributary the wadi Chelif near the town
of Machraa_SFA, 25 km from west of the Tiaret
region. This dam is one of the first built in Algeria
(Serbah, 2011) rockfill dams. Its exploitation begun
in 1936 and is intended to irrigate the Relizane
region as well as drinking water supply of the Tiaret
region. This dam is designed to regulate 50 million
m3 of water. The dams major river basin is the wadi
Mina which extends over a total area estimated of
5000 km2 .
Figure 1. Location dams most threatened by siltation.
downstream of the dam can lead to long term pollution in rural areas. Thus, the recuperation and reuse
of these sediments should allow their medium-term
recovery in the field of civil engineering such as road
engineering, brick manufacturing and network grout
filling.
For an evaluation in road engineering, the purpose
of this paper, the methodology is divided into two
phases:
Characterization of the sediment: this phase allows
to identify the physico-chemical and mineralogical
sediment characteristics in its natural state and to
assess their impact on the environment.
The formulation phase: different constituted formulations will be proposed for use in pavement base
material (foundation layer and base layer) (GTR,
2000; GTS, 2000).
The objective of this study is to establish if dredged
sediments may be used for the production of pavement
layer. For this, the introduction of cohesionless soil
and binders such as granulometric correctors allows
for these materials, generally sensitive to water, the
creation of links between the particles. The effect of
added in proportions of these materials in the sediment mixtures focused on the evolution of particle
size distribution on the physico-chemical parameters
and on the densification and the bearing capacity. The
choice of the respective assays of these various components will also allow to reduce the water content
by the phenomena of hydration and addition of dry
matter (Abriak, 2007; Colin, 2003; Sfar Felfoul et al.,
2003).
PHYSICO-CHEMICAL
CHARACTERIZATION OF THE TWO
STUDY SITES
Samples in surface were carried out on the downstream

side area where the dredged sediments are accumulated in remarkable quantities for the two study sites
(Figure 3).
The physico-chemical characteristics of two soils of
study are presented in tables 1 and 2 (Bourabah, 2012;
Serbah, 2011).
These results show that the fine fraction less than
63 m (silt and clay) is dominant and represents for
almost 90%, while the presence of sand particles is
minimal.
The calculation of the coefficients of uniformity
(Cu) and curvature (Cc) shows a tight granulometry and poorly graded sediments. According to the
LPC-USCS classification (Magnan, 1998), these two
materials are classified an organic highly plastic
soil (OH).
300
Table 3. Influence of organic matter on the particle size

distribution of Cheurfas sediment.
Figure 3. Zone of accumulation of the two dredged

sediments on the downstream of Cheurfas and Bakhadda
deductions.
Table 1.
Cheurfas sediment
Bakhadda sediment
Table 2.
study.
Silt
%
Clay
%
Cu
Cc
4
9
43
77
53
14
2.89
2.20
0.84
0.98
Sand
%
Silt
%
Clay
%
Natural
Calcined
4
21
43
44
53
35
Table 4. Influence of organic matter on the physicochemical and geotechnical Cheurfas sediment parameters.
Particle size distribution of two sediments to study.

Sand
%
Cheurfas
Sediment
Cheurfas
Sediment
wL
%
wP
%
wR
%
Ip
%
VBS
Natural
Calcined
62
39
36
32
9
25
26
7
3.22
1.37
sediment. In these circumstances, the use of these sediments in road engineering requests processing to meet
the requirements of requested behaviour (Bourabah,
2012).
Physico-chemical parameters of two sediments to
Physical
Parameters
Chemical
Parameters
wn (%)
wL (%)
wP (%)
wR (%)
Ip (%)
IL (%)
Ic (%)
Ac
VBS
CaCO3 (%)
OM (%)
Cheurfas
Sediment
Bakhadda
Sediment
96
62
36
9
26
2.31
1.31
0.5
3.22
12.5
6.5
86
62
31
12
31
1.79
0.79
0.34
6.33
20
4.5
RESULTS AND ANALYSIS
The sediments have a complex structural composition and can be considered an evolutionary material
(Colin, 2003). Indeed, the presence of organic matter (6.5%) for the Cheurfas sediment results in plastic
and compressible behaviour. Physico-chemical and
geotechnical tests conducted on natural and treated
sediment of Cheurfas (Tables 3 and 4) clearly indicate a highly influential organic activity on all of the
sediment matrix.
Thus, as long as no special, simple or complex, treatment will be directed on these sediments to destroy
organic materials, they will still act in the structure of
the material.
From Table 2, we can note that there is a high natural

water content (wn ), but still characteristic of sediment
dredging (Bourabah, 2012; Abriak, 2007; Semcha,
2005; Colin, 2003). These two natural sediments are
also characterized by high plasticity and swelling
potential. This is a consequence of their fineness and
high organic matter content (% OM > 3%).
On the basis of geotechnical data (Dmax 50 mm
and passing to 80 m > 35%), the classification of
Cheurfas and Bakhadda sediments according to the
standard NF-P11-300 GTR guide allows to place these
two materials, in class F, corresponding to natural
materials containing organic matter. These two materials reposition in class A following the same guide
designated by the class of fine-grained soils. Subclass
determined by particle size analyses, the value blue
(VBS) and Atterberg limits (Ip), leading to the A2 class
for Cheurfas sediment and the class A3 for Bakhadda
4.1 Grain size of the skeleton

To be used as a layer of pavement structure, wide distribution material is advisable (Abriak, 2007; Dubois,
2006). To this effect, granulometry and organic content are key factors for the choice of the use of dredged
sediment in road engineering.After treatment, it is necessary to obtain a material that meets the following
conditions:
Reply the classification standards and dimensioning in effect (GTR, 2000);
To maintain its mechanical and structural properties
through time to satisfy the durability norms.
Sediments mixed with other granular components,
in economically attractive proportions, allow easy
application on an industrial scale and an interesting reduction of water content and the proportion
301
Table 5.
For 1
For 2
Formulations of the studied mixtures.
Table 7. Influence of addition of sand on the physicochemical characteristics of the two mixtures.
Cheurfas
Sediment
(VC)
Bakhadda
Sediment
(VB)
Career
Sand
(CS)
Dredging
Sand
(SD)
30%
---
--50%
70%
---
--50%
For 1
For 2
OM
%
wL
%
wP
%
wR
%
Ip
%
Ac
4.45
4.02
41
41
21
30
14
28
20
11
0.8
0.7
Table 6. Influence of addition of sand on the grain size of

the two mixtures.
%>63 m
For 1
For 2
%<63 m
Gravel
%
Sand
%
Silt
%
Clay
%
Cu
Cc
--14
57
42
18
28
25
16
243
28
2.72
1.12
of organic materials (Bourabah, 2012; Serbah, 2011;

Tran, 2009; Scordia, 2008; Dubois, 2006).
The addition of air and hydraulic binders, will also
allow to reduce the water content, by the phenomena
of hydration and contribution of dry matter (Boutouil,
1998; Sfar Felfoul et al., 2003).
The treatment methodology consists of two steps:
Correction of the granulometry of natural sediment
by adding a granular fraction;
Improvement of the mechanical properties by treatment with air and hydraulic binders.
4.2
Influence of the addition of a granular fraction
4.2.1
Effect on the grain size and the physical and

chemical characteristics
Granular adding selected in this study consists of two
types of sand:
Career sand (0/4) mm added to the Cheurfas sediment, with 3% lower than 80 m elements and
47% of coarse sand. It is a clean sand well graduated according to the classification LPC-USCS
(Magnan, 1998).
Dredging sand added to the Bakhadda sediment
containing a high percentage of sand (85%) to
classify the soil in a silty sand according to the
classification LPC-USCS (Magnan, 1998).
Table 5 shows the two types of formulations adopted
by using the method of experiment plans (Bourabah,
2012). The physico-chemical and geotechnical characteristics formulations studied for two sediments of
study are summarized in tables 6 and 7 below.
Note that uniformity coefficients are much higher
than the value reference 6 satisfying spread size mixtures. According to particle size analyses and the limits
of liquidity and plasticity, the mixtures of formulations
are classified according to the technical guide (GTR,
2000) class A2 for formulation 1 (For 1) and in class
A1 for formulation 2 (For 2).
Figure 4. Variation of compaction and bearing capacity with

water content to the MPO for two sediments treated granular
sand.
It is then, said that the proportion on fines and the

low plasticity means these soils to change brutally their
consistency for small changes in water content, particularly when the natural water content is close to the
optimum Proctor modified (MPO) (Abriak, 2007).
4.2.2 Effect on the densification and the bearing
capacity
A companion of compaction-capacity tests was carried
out on the two proposed mixtures, to highlight the variation of parameters of compaction (dMPO ) and bearing
capacity (IPI).
Figure 4 indicates the contribution of the particle size correction on the Optimum Proctor modified
settings (MPO) as well as on the lift on the two
types of natural sediment. Table 8 gives the values
of the optimum Proctor modified (MPO) for each
formulation.
The use of these natural sediments without treatment may be considered in road engineering. The
302
Table 8. Effects of addition of a granular sand on the

optimum Proctor modified and bearing capacity parameters.
wMPO dMPO IPIMPO IPI95%MPO
%
kN/m3 %
%
Cheurfas Sediment
(VC)
Bakhadda Sediment
(VB)
Formulation 1 (For 1)
Formulation 2 (For 2)
23.3
15.54
14
27
23
19
16
30
20
40
28
36
13.26 19.41
26
17.4
Table 9. Proportions of sand and binders added on the

sediments of study.
Formulations
VC
%
VB
%
Career
Sand
%
Lime
(0/2) mm
%
Cement
%
For 3
For 4
For 5
For 6
For 7
For 8
30
30
30
-------
------94
97
91
67
64
61
-------
3
0
3
0
3
3
0
6
6
6
0
6
formulation 1 adopted which is the treatment on natural sediment of Cheurfas (VC) by 70% of sand of
career (CS) showed that:
The optimum water content has clearly decreased
by addition of sand on the sediment (VC);
The maximum density of the material is much
improved, what characterizes a good compactness
where it obtained an increase in average 26%
dMPO ;
Moreover, we note a decrease of the immediate
bearing capacity index (IPI < 35) with a decrease in
optimum moisture content; this is compared to those
of mixture to the natural sediment. This is probably due
to occluded air bubbles generating a reversible compressibility of the interstitial fluid and thus occurrence
of the phenomenon known as Padding.
Treatment of sediment by dredging sand (SD)
exceeds a significant improvement in the bearing
capacity (an average increase of 80%) from the natural
Bakhadda sediment.
The natural sediments, as mixtures are therefore
sensitive to water content and their use without binders
is likely to be detrimental to the sustainability of the
pavement. This sensitivity is related to the presence of
organic matter, which by water retention, increase the
phenomena.
4.3
Influence of the addition of binders
The treatment with binders consists of the addition of

binder only hydraulic (cement type) or only air Binder
Figure 5. Variation of the Optimum Proctor modified for

studied formulations.
Table 10. Influence of air and hydraulic binders on the

physico-chemical characteristics for studied formulations.
Formulations
For 3
For 4
For 5
For 6
For 7
For 8
OM (%)
wL (%)
wP (%)
wR (%)
Ip (%)
Ac
Cu
Cc
3.55
32
23
25
9
0.75
127
6.6
4.56
42
33
33
9
0.75
34
0.43
3.25
48
40
38
8
2.67
71
5
3.05
59
43
34
16
0.89
5
0.51
2.95
64
52
37
12
0.40
3.2
0.91
2.18
61
47
33
14
1.40
4
0.67
(lime type) or a combination of these two binders.

The used cement is a CPJ CEM II/A 42.5 composite
Portland cement and lime is a quicklime 0/2 mm.
Studied formulations are shown in table 9.
The densification-capacity tests on sediment in
air and hydraulic binders are shown on Figure 5. Tables 10 and 11 repeat the values of the
geotechnical characteristics (particle size, physicochemical and MPO values) corresponding to each
formulation.
303
than for the untreated natural sediment. This leads

to say that the addition of binders does not affect
the granularity mixtures.
Plasticity strongly decreased after treatment by
addition of binders; this is attributed to the flocculation of particles. This flocculation which translates
a share by a significant rise of the plastic limit
without significant change to its limit liquidity and
the other the decrease of the organic fraction that
will eventually lead to a significant reduction in its
percentage of with shrinkage-swelling.
The addition of binders has no bearing on the compactness of the sediment uncorrected by comparing
it to the natural sediment. However, the bearing
capacity is significantly improved or immediate
capacity index (IPI) is greater than 35 to ensure
the movement of the gears on this layer of road
structure.
Table 11. Influence of air and hydraulic binders on the values of modified Proctor optimum for studied formulations.
No
Formulations
30% VC + 67% CS +
3% Lime
30% VC + 64% CS +
6% Cement
30% VC + 61% CS +
6% Cement + 3% Lime
94% VB + 6% Cement
97% VB + 3% Lime
91% VB + 6% Cement +
3% Lime
4
5
6
7
8
wMPO
%
dMPO
kN/m3
IPIMPO
%
13.35
18.62
34
13.28
18.83
30
13.94
18.91
28
20
25
24
19.20
17.70
18.20
36
34
38
4.3.1
Influence of binders on the sediment particle

size corrected natural based formulations
The contribution of air and hydraulic binders on the
Cheurfas sediment corrected sand of career revealed
the following findings:
Size spread for the three formulations (3, 4 and 5)
is retained after addition of binders or the values of
the coefficients of uniformity (Cu) are much higher
than the value reference of 6 (Magnan, 1998). However, the values of the coefficients of curvature (Cc)
are below the recommended values (1 < Cc < 3).
The addition of air and hydraulic binders reduced
organic matter from an average of 15% from treated
sand sediments.
A slight decrease in the density dry maximum
(dMPO ) (variation of the order of 3%) to the three
formulations binders to the treatment-treated by
addition of cohesionless granular material. These
values of density remain very close to 19 kN/m3 ,
indicating a level of acceptable compactness.
The optimal moisture content (wMPO ) increases percentage by about two points when the sediment
is treated with the binders. However this value is
relatively constant regardless the combination of
added binders. Obtained granular classes allow to
consider these formulations as sands insensitive to
water given their large particle size fraction. This
has led to a slight flattening of curves (w, d ) in
the vicinity of the MPO in particular for materials treated with cement, which is a decrease in
sensitivity to water.
Cheurfas corrected and binders-treated sediment
has a value at 9598% of near MPO of 35 for use
as a foundation layer and 45 for use as base layer
in particular for the sediment to quicklime alone or
in combination.
4.3.2
Influence of binders on the sediment particle

size not corrected natural based formulations
The treatment of the natural sediment of Bakhadda by
addition of binders alone or in combination shows that:
Particle size distribution of the formulations of
mixtures presents a very close average granularity
CONCLUSION
Treatment approach considered in this study relies on

choice of preparation and simple formulations, facilitating the use of dredging sediment of dam of Western
Algerian for road construction.
The preliminary study performed on fine dredged
sediments revealed high initial water content, high
plasticity, and a moderate amount of organic matter. According to the LPC/USCS classification and to
the guide technical (GTR, 2000), this type of material is difficult to use in road construction due to its
sensitivity to water content and its compressibility.
The analysis of the results of physico-chemical and
mechanical tests on formulations of studied mixtures
is encouraging.A particle size correction was observed
followed by an improvement of the maximum dry
density obtained by substituting these sediments to
a cohesionless granular material. Optimal water contents obtained are relatively reduced and the organic
content does not affect the compaction of the material,
however it is unfavorably on the bearing capacity.
The addition of quicklime as air binders in mixtures
of sediments allowed us to note their influence. Plasticity of the material decreases markedly and capacity
is significantly improved.
LISTE OF NOTATION
Ac: Colloidal activity

CaCO3 : Calcium carbonate content
Cc: Coefficient of curvature
Cu: Coefficient of uniformity
Dmax: Diameter for which 95% of the soil grains are
smaller than (so D95 if the size particle curve is available, otherwise visual assessment of the size of larger
elements)
Ic: Consistency index
IPI: Immediate bearing capacity index
304
IPIMPO : Immediate bearing capacity index to the

Modified Proctor Optimum (MPO)
wL : Liquidity limit
wP : Plasticity limit
wR : Shrinkage limit
IP: Plasticity index
IL : Liquidity index
VBS: Blue methylene value
OM: Organic matter content
wMPO : Moisture content at Modified Proctor Optimum
(MPO)
dMPO : Unit weight at Modified Proctor Optimum
(MPO)
REFERENCES
Abriak, N. 2007. Amliorer la valorisation des dchets industriels en BTP, Guide Technique Rgional Relatif la
mthodologie de gestion des sdiments de dragage portuaire, Dmarche PREDIS Nord Pas de Calais, Groupe
de travail n 5, Ecole Nationale Suprieure des Mines de
Douai, France.
Bourabah, M.A. 2012. Comportement mcanique des sols
fins. Application la valorisation des sdiments de barrages en technique routire, Thse de Doctorat, Facult
de Technologie, Universit Abou-Bekr Belkaid, Tlemcen,
Algrie.
Boutouil, M. 1998. Traitement des vases de dragage par solidification/stabilisation base de ciment et additifs, Thse
de doctorat de lUniversit du Havre, 245 p, France.
Colin, D. 2003. Valorisation de sdiments fins de dragage en
technique routire, Thse de doctorat de luniversit de
Caen. Octobre 2003. 147 p, France.
Dubois, V. 2006. Etude du comportement physicomcanique et caractrisation environnementale des sdiments marins-Valorisation en technique routire, Thse de
doctorat, Ecole Nationale Suprieure des Mines de Douai,
France.
GTR, 2000. Guide technique pour la ralisation des remblais
et des couches de forme, fascicules 1, principes gnraux,
100p, LCPC-SETRA.
GTS, 2000. Traitement des sols la chaux et/ou aux liants
hydrauliques en remblai et en couche de forme, Guide
technique, 240 p, LCPC-SETRA.
Hallouche, W. & Remini, B. 2004. Le dragage des retenues

de barrages: quelques exemples algriens, La houille
blanche, N 5, pp. 95100.
Hallouche, W. 2007. Prvision du transport solide et sdimentation des barrages, Thse de doctorat, Universit de
Biskra, Algrie, 105 p.
Labiod, Z. & Benamra, T. & Benrahou, H.K. & Rabah, M.
2001. Auscultation des barrages en Algrie, Sminaire
National de Gnie Civil SNGC01, Sidi Bel Abbs, 1314
mai 2001, Algrie.
Magnan, J.P. 1998. Cours de mcanique des sols et des roches,
Volume 1, Ecole Nationale des Ponts et Chausses, France.
Mekerta, B. & Semcha,A. & Rahmani, F. &Troalen, J.P. 2008.
Erosion spcifique et caractrisation de la rsistance au
cisaillement des sdiments du barrage de Fergoug, Xme
Journes Nationales Gnie Ctier-Gnie Civil, Sophia
Antipolis, 1416 Octobre 2008.
Remini, B. & Avenard, J.M. & Kettab, A. 2003. Envasement
des barrages, Les ditions CEDOC 2003.
Remini, B. & Hallouche, W. & Achour, B. 2007. LAlgrie:
Plus dun sicle de dsenvasement des barrages, Partie II,
Les eaux douces et de surface, chapitre 8.
Scordia, P-Y. 2008. Caractrisation et valorisation de sdiments fluviaux pollus et traits dans les matriaux
routiers, Thse de doctorat de lEcole Centrale de Lille,
France.
Semcha, A. 2005. Etude des proprits gomcaniques des
sdiments denvasement de barrage et leur valorisation,
Thse de doctorat dtat de lUniversit des Sciences et
de la Technologie dOran, Algrie.
Serbah, B. 2011. Etude et valorisation des sdiments de
dragage du barrage Bakhadda Tiaret, Thse de magister en Gnie Civil, Facult de Technologie, Universit
Abou-Bekr Belkaid, Tlemcen, Algrie.
Sfar Felfoul, H. & Rahali, L. & Mahmoudi, M.A. & Chedly,
S. 2003. Traitement des sols la chaux et au ciment en
technique routire tunisienne: prsentation dun exemple,
Sminaire International de Gomatriaux, Msila, 1011
mars 2003, Algrie, pp. 5566.
Tran, N.T. 2009. Valorisation de sdiments marins et fluviaux
en technique routire, Thse de doctorat de lUniversit
dArtois, France.
305

Dynamic characterization of municipal solid waste by SDMT

F. Castelli
Faculty of Engineering and Architecture, University of Enna Kore, Italy
M. Maugeri
Department of Civil and Environmental Engineering, University of Catania, Italy
ABSTRACT: The design of landfills includes stability analyses based on appropriate characterization of soil
foundation and evaluation of Municipal Solid Waste (MSW) properties. As documented by some recent earthquakes, landfills can be damaged by the seismic loading, then to analyze the stability of landfills, in addition to
the dynamic characterization of soil, the dynamic properties of MSW are also required. With this aim, both in
situ and laboratory tests can be performed, even if, due the heterogeneity of the material, the estimation of the
waste material properties for the seismic design of landfill is a very difficult task. In the paper the results of a
geotechnical investigation carried on the Cozzo Vuturo landfill in the Enna area (Sicily, Italy) are reported and
analyzed. In particular, Seismic Marchetti Dilatomer Tests (SMDT) have been carried out for the geotechnical
seismic characterization and the tests results have been used to derive shear wave velocity profiles and the
dynamic properties of the deposited waste materials.
INTRODUCTION
Municipal Solid Waste Landfills (MSW) are environmentally sensitive facilities that need to perform well
both under static and seismic conditions. Potential
failures during an earthquake may cause significant
damage to a landfill or to some of its components,
such as the base containment system, and can lead
to environmental disasters and significant financial
costs. Damage to gas collection systems or the cover
system needs also to be repaired. Deep-seated slope
failures involving the base containment system may
be difficult to repair and may require excavation and
removal of the solid waste material.
In the United States, significant effort has been
expended to provide guidance on the design of MSW
landfills under seismic loading. Federal or state regulations and guidance documents have also been developed and are regularly used in engineering practice.
According to Italian Regulation for Construction
(NTC, 2008), the design of landfills include the geological and geotechnical characterization of the site,
a detailed description of the constructive phases and
monitoring systems, taking into account the nature of
waste materials and the environmental vulnerability.
The design include static and seismic stability analyses, based on appropriate mechanical characterization
of the shear strength of soil and MSW, and the evaluation of the risks related to damages of the lining system
(Maugeri et al., 2011).
In the present paper some data relating to field tests
carried on the Cozzo Vuturo landfill (Figure 1) in the
Figure 1. Cozzo Vuturo landfill (Enna, Italy) location.
Enna area (Italy) are reported and discussed. Seismic

Marchetti Dilatomer Tests (SMDT) have been performed for the geotechnical seismic characterization
(Castelli et al., 2012) and the tests results have been
used to derive the dynamic properties of the deposited
waste materials.
It is important to recognize that MSW is a highly
variable material. As such, some considerations presented may not be applicable to all MSW. Nevertheless,
as more data is generated, our understanding of the
various factors that affect the dynamic properties of
MSW will improve.
307
DYNAMIC PROPERTIES OF MSW
During an earthquake, ground motions reaching the

site will be amplified as they go through the various
foundation soils. Additional potential amplification
will take place as the seismic waves propagate through
the waste material. The geometric features of the landfill are also likely to cause topographic amplification
effects.
The response of the landfill to shaking may cause
seismic instability that may be shallow, involving only
the cover of the landfill, or larger size instabilities
involving the waste mass, or the waste mass and the
base containment system or foundation soils.
There is a limited experience with the performance
of modern MSW landfills since most of these landfills were built in the early 1990s. No information
is reported in Europe on the performance of modern
landfills during earthquakes.
In the US, modern MSW landfills have been subjected to moderate shaking during the 1994 Northridge
earthquake. A total of 22 MSW landfills experienced significant levels of shaking (i.e., free-field rock
PGA > 0.05 g), with 12 experiencing free-field
rock PGA beyond 0.2 g and 7 experiencing free-field
rock PGA beyond 0.3 g (Augello et al., 1995).
Chiquita Canyon landfill was subjected to an estimated 0.33 g free-field rock PGA and experienced
significant damage that consisted of two tears in the
geomembrane side slope liner, down slope movement
of the waste and cracks in the cover soil. A temporary
shutdown of the gas removal system took place as a
result of the loss of power at the landfill.
Landfills subjected to significant shaking during
the 1989 Loma Prieta Earthquake did not have engineered liner systems. Minor cracking of the landfill
slopes was observed at one out of ten landfills although
it is not entirely clear whether the cracking was a
consequence of shaking. Landfill gas recovery systems were impacted by power shutdown, but were
back in operation within 24 hrs after the earthquake.
Overall performance of these landfills was very good
(Anderson & Kavazanjian, 1995).
It is important to stress that the dynamic characteristics of solid waste materials play an important role
on the seismic response of landfill (Maugeri & Sco e
Pinto, 2005).
Some properties are measured directly, such as dry
density and water contents, whereas other properties,
due the difficulties related with sampling, are obtained
from indirect methods combining with the existent
knowledge of waste properties.
Both in situ (Matasovic et al., 2011) and laboratory tests (Athanasopoulos, 2011) can be performed,
even if in this last case, the characterization of waste
material for seismic design is difficult due the heterogeneity of the material, requiring the procurement of
large samples. MSW properties that are used as input
in the seismic analyses of landfills include:
the unit weight of MSW;

the (dynamic) shear strength of MSW;
Figure 2. Landfill area dived into two disposal basin.
the shear wave velocity Vs and small-strain shear

modulus Go profile;
the shear strain-dependent shear modulus reduction
G/Go and material damping () curves; and
the dynamic Poissons ratio .
An extensive review of the state of the art on the
shear wave velocity Vs , the strain-dependent nonlinear
dynamic properties of MSW and Poissons ratio has
been performed by Zekkos et al. (2011).
3 TESTING PROGRAM
3.1 Test site
The testing program was carried on Cozzo Vuturo
landfill located at about 3.8 km from the city of
Enna (Sicily, Italy). The landfill covers an area of
about 120.000 m2 and it receives wastes from five
main waste districts (Enna, Calascibetta, Leonforte,
Villarosa, Valguarnera), including more than twenty
towns, for a total of about 180.000 inhabitants.
The landfill is located in a hilly area, geologically
made up of Numidian Flysch of Holigocene-lower age,
marly and sandy brown clay of medium Miocene age,
and river alluvium of Holocene age. There is a mixture
of humus and clay 12 m deep within the area of the
landfill and a clay layer of about 3040 m below it.
The permeability of the clay layer varies in the range
2 109 7 109 cm/s. The humus layer is an aquifer
but the recharge area is very small and the groundwater stays for only a short time in the aquifer, 3040 m
below the ground surface.
The total landfill area is dived into two disposal site
named B1 and B2 respectively. The B1 landfill activity took place from 1999 to 2006. The B2 landfill is
located in the smaller upper part of the catchments
basin, with a final volume of 330,000 m3 . The landfill was designed in order to fill a naturally occurring
valley (Figure 2).
According to the original project, the residual useful
life of the B2 waste disposal plant is expected to be
about 3 years and half (end of the disposal activity
in 2012) but an extension up to about 650,000 m3 is
currently considered.
The landfill is about 18 m high. The upper soil layers have been removed to create space for the waste
308
and to guarantee geotechnical safe conditions by the

execution of a terracing profile.
The waste is piled up in compacted layers of about
1.8 m thick. In the landfill the unit weight of waste
ranges between 7 and 8 kN/m3 . The daily covering layers consist of clayey soil. The refuse dumping plan
considers the landfill divided in 6 cells. As a consequence drainage of storm-waters was designed on
a four level basis, according to the landfill filling
sequence.
One-fourth of the landfill is used as an area for
leachate collection. The leachate extraction is carried out only from the two lower cells. The collected
leachate is treated in a external water plant. A system for biogas collection is expected to be installed
in the landfill and will be operated after the landfill
closure.
The landfill sealing at the bottom consists of numerous different layers. Specifically landfill bottom is
lined with recompacted clay (about 1.0 m depth), a
geotextile protection layer and an HDPE geomembrane to prevent any seepage of leachate into the
underlying ground.
3.2
In-situ testing
Field measurement techniques may be divided into

non-intrusive measurements that do not penetrate the
waste mass and intrusive measurements that penetrate
the waste mass.
Non-intrusive techniques range from seismological
and geophysical measurements of wave propagation
velocities and electrical resistivity, while intrusive
techniques include measurements made in borings or
soundings such as Standard Penetration Test (SPT),
Cross-Hole Test (CH), Down-Hole Test (DH) and
Pressuremeter Test, and internal measurements that
include Cone Penetration Test (CPT) and Dilatometer
Marchetti Test (DMT).
Similarly, in situ measurements can be grouped
into direct measurements of the MSW properties and
indirect measurements that rely on correlations to
evaluate properties of interest.
The Dilatometer Marchetti Test (DMT) is widely
used for in situ measurements of physical and mechanical soil properties, and the latest iteration of DMT is
the seismic dilatometer (SDMT), that is the combination of the mechanical flat dilatometer introduced by
Marchetti (1980), with a seismic module for measuring
the shear wave velocity Vs .
The seismic dilatometer test, conceptually similar to the seismic cone penetration test (SCPT), was
first introduced by Hepton (1988) and subsequently
improved at Georgia Tech, Atlanta, USA (Martin &
Mayne, 1997; 1998; Mayne et al., 1999). A new SDMT
system, described in Marchetti et al. (2008), has been
recently developed in Italy.
Besides the shear wave velocity Vs , the seismic
dilatometer provides the usual DMT parameters by
use of common correlations (Marchetti, 1980; TC16,
2001).
Figure 3. Cross section of the landfill with the location of

SDM Tests.
Figure 4. Material index (Id ) profiles.
Validations of Vs measurements by SDMT compared to Vs measurements by other in situ techniques

at various research sites are reported in Marchetti
et al. (2008). The small strain and large strain moduli
back calculated from the SDMT measurements can be
combined to identify the modulus degradation curve
for a specific soil (Mayne et al., 1999).
Despite its success in measuring soil properties,
the use of DMT for in situ measurements of MSW
properties has not yet been reported in literature
(Matasovic et al., 2011).
3.3 Test results
This section presents the results obtained by seismic dilatometer tests executed at the Cozzo Vuturo
landfill (Enna, Italy), as part of the geotechnical investigations planned for the mechanical characterization
of waste materials. A cross section of the landfill with
the location of the seismic dilatometer tests is reported
in Figure 3.
Two dilatometer tests are located in the old catchments basin named B1, two dilatometer tests are
located in the upper part of the catchments basin named
B2, one dilatometer test is located between the two disposal site B1 and B2. It can be observed that the results
deductible from tests named B1a and B1b, as well as
from those named B2a and B2b, regard the deposited
waste materials, while the results deductible from test
named Arg are referred to the soil foundation.
Figures 4 and 5 show the typical profiles obtained
by the Dilatometer Marchetti Test and in particular:
Figure 4 reports the material index (Id ) profile, while
Figure 5 reports the working strain constrained modulus (M ) profile, both obtained from the usual DMT
interpretation.
Analyzing these results it can be observed that
according to material index (Id ) derived by the SDM
309
Figure 7. Small-strain shear modulus (Go ) profiles.
Figure 5. Constrained modulus (M ) profiles.
Figure 6. Shear wave velocity (Vs ) profiles.
Tests, the old deposit waste materials located in the

catchments basin named B1 (19992006) can be classified as silt-clayey material, while deposit waste materials located in the new catchments basin (20062011)
named B2, can be classified as silt-sandy material
(Figure 4).
As concern the constrained modulus (M ), the values
seem to be affected by a consistent variation with depth
and M ranges between 2 MPa up to about 100 MPa
(Figure 5).
The particularity of the seismic dilatometer
(SDMT) is the possibility to provide, in addition to
the usual DMT parameters, the S-wave distribution
in terms of shear wave velocity (Vs ) profiles. In particular, Figure 6 shows the shear wave velocity (Vs )
profile and Figure 7 the small-strain shear modulus
(Go ) profile.
4 ANALYSIS OF THE EXPERIMENTAL

RESULTS
To analyze the engineering behavior of deposited waste
material to seismic loadings it is essential to determine
its dynamic modulus, Poissons ratio, and strength.
The small-strain shear modulus (Go ) is a key material property in the evaluation of the dynamic response
of MSW.
The small-strain shear modulus Go can be estimated
from mass density and shear wave velocity (Yoder &
Witczak, 1975). Being the shear modulus Go related
to the velocity of shear waves Vs (obtained as Go =
Vs2 ), the measurements of the shear wave velocity
provides a convenient method for measuring stiffness.
The shear wave distribution is not sensitive to the

fluid-saturation and the wave velocity Vs remains
almost constant with the saturation changes. Therefore
the S-wave velocity is most sensitive to the mechanical
skeleton properties of the waste.
In the last years, a large number of in situ methods
have been used to estimate the shear wave velocity
profile of MSW in the field. These include the DH
method, the CH method and the Spectral Analysis of
Surface Waves (SAWS) method.
According to experimental results, the profiles of
the S-wave reveal a general increase of the velocity
from the top to the bottom of the landfill. Values of Vs
are usually in the range between 100 and 350 m/s.
An extensive review of the state of the art on the
Vs , the strain-dependent nonlinear dynamic properties
of MSW and Poissons ratio has been performed by
Zekkos et al. (2011).
Early studies carried out by Sharma et al. (1990)
at a MSW landfill in Richmond, California, report an
average value of Vs about 198 m/s for a depth of 0
to 15.3 m. Kavazanjian et al. (1995) report recorded
shear waves velocities ranging between 80 m/s near
the surface, increasing to over 300 m/s at a depth of
30 m. Pereira et al. (2002) measured Vs ranging from
100 m/s to 50 m/s, while Lin et al. (2004) reported Vs
profiles in the range 50 m/s to more than 200 m/s.
Based on SDMT values of Vs generally increasing with depth were measured on the Cozzo Vuturo
landfill (Enna, Italy).
An average value of Vs ranging between 100 and
150 m/s (Figure 6) was observed, in good agreement
with the MSW shear wave velocity profiles reported
by Zekkos et al. (2011).
The values of the small-strain shear modulus Go
derived from SDMT are summarized in Figure 8.
These values of Go range from about 10 MPa up
to 55 MPa, with an average value equal to 35 MPa
(Figure 8). The scatter observed in the experimental
results can be attributed to the heterogeneity of the
deposited waste materials.
Assuming for the deposited waste materials a Poissons ratio equal to 0.3 (Jones & Dixon, 2005), the
corresponding values of theYoungs modulus Eo range
between 26 and 143 MPa, and a good correspondence
can be observed with the values reported by Del Greco
et al. (2007).
310
(Maugeri & Castelli, 2008) was carried out at the

Cozzo Vuturo landfill (Enna, Italy).
Generally a good reliability has been found in deriving the dynamic parameters of the waste materials by
SDMT. These preliminary results seems to be promising as concerns the assessment of the S-wave in terms
of shear wave velocity profile, and as concerns the
evaluation of the small-strain shear modulus.
REFERENCES
Figure 8. Small-strain shear modulus versus depth by

SDMT.
CONCLUDING REMARKS
Dynamic properties of Municipal Solid Waste are

required to perform reliable seismic analyses of MSW
landfills.
Knowledge of the shear wave velocity Vs and smallstrain shear modulus Go profile, Poissons ratio and
dynamic shear strength of MSW, is fundamental for
the assessment of landfill stability and integrity of both
geosynthetic and lining components.
The physical and mechanical characterization of
waste, nevertheless, is a complex task given the great
heterogeneity of the materials, the testing difficulties and the uncertainties to adapt the theories and
techniques used in soil mechanics.
A better understanding of the static and dynamic
properties of the waste materials can be provided by
laboratory and in situ experimental tests. Some properties can be measured directly, whereas others, due
the difficulties related with sampling, can be obtained
from indirect methods combining with the existent
knowledge of waste properties.
Among in situ tests, Dilatomer Marchetti Tests
(DMT) is used widely in measuring soil properties,
even if, the adoption of the DMT for the measurements
of MSW properties has not often been documented in
literature. To validate its applicability for the estimation of MSW static and dynamic properties, a program
of in situ and laboratory geotechnical investigation
Anderson, D.G. & Kavazanjian, E. Jr. 1995. Performance

of landfills under seismic loading. Proc. 3rd International Conference on Recent Advances in Geotechnical
Earthquake Engineering and Soil Dynamics, St. Louis,
Vol. III.
Athanasopoulos, G.A. 2011. Chapter 7: Laboratory Testing of
Municipal Solid Waste, in Geotechnical characterization,
Field Measurements, and Laboratory Testing of Municipal Solid Waste. In D. Zekkos (eds.), ASCE Geotechnical
Special Publication no.209: 112134.
Augello, A.J., Matasovic, N., Bray, J.D., Kavazanjian, Jr.,
E. & Seed, R.B. 1995. Evaluation of solid waste landfill performance during the Northridge earthquake. In
M.K.Yegian & W.D.L. Finn (eds.), Earthquake design and
performance of solid waste landfills, ASCE Geotechnical
Special Publication no.54: 1750.
Castelli, F., Lentini, V., & Maugeri, M. 2012. Static and
dynamic waste characterization. Proceedings 5th European Geosynthetics Congress, EUROGEO 5. Vol. 5:
105110.
Kavazanjian, E. Jr., Matasovic, N., Bonaparte, R., &
Schmertmann, G.R. 1995. Evaluation of MSW properties for seismic analysis, ASCE Geotechnical Special
Publication no. 46: 11261141.
Hepton, P. 1988. Shear wave velocity measurements during
penetration testing. Proceedings Penetration Testing in the
UK, ICE: 275278.
Jones, D.R., & Dixon, N. 2005. Landfill lining stability and
integrity: the role of waste settlement. Geotextiles and
Geomembranes, 23: 2753.
Lin, Y.C, Rosenblad, B., & Stokoe, K.H. 2004. Data report
on shear wave velocity profiles determined by SASW
method at: Altamont landfill, Redwood landfill, and TriCities landfill. Geotechnical Engineering Report GR04-3,
Civil Engineering Department, University of Texas.
Marchetti S. 1980. In Situ Tests by Flat Dilatometer.
Journal of Geotechnical Engineering, ASCE 106(GT3):
299321.
Marchetti, S., Monaco, P., Totani, G., & Marchetti, D. 2008.
In Situ Tests by Seismic Dilatometer (SDMT), From
Research to Practice in Geotechnical Engineering. In
J.E. Laier, D.K. Crapps & M.H. Hussein (eds.), ASCE
Geotechnical Special Publication no. 180: 292311.
Martin, G.K & Mayne, P.W. 1997. Seismic Flat Dilatometer Tests in Connecticut Valley Varved Clay. Geotechnical
Testing Journal, ASTM 20(3): 357361.
Martin, G.K & Mayne, P.W. 1998. Seismic flat dilatometer
in Piedmont residual soils. Proc. 1st Int. Conf. on Site
Characterization, Atlanta. In P.K. Robertson & P.W.
Mayne (eds.), 2: 837843. Rotterdam: Balkema.
Matasovic, N., El-Sherbiny, R., & Kavazanjian, E. Jr. 2011.
Chapter 6: In-situ measurements of MSW properties,
in Geotechnical characterization, Field Measurements,
and Laboratory Testing of Municipal Solid Waste. ASCE
Geotechnical Special Publication no. 209: 153193.
311
Maugeri M. & Sco e Pinto P.S. 2005. Seismic Design of

Solid Waste. Landfills and Lining Systems. Chapter 5
TC5 Report, September 2005.
Maugeri, M., & Castelli, F. 2008. Experimental analysis of waste compressibility. Proceedings Geocongress
2008. ASCE Geotechnical Special Publication no. 177:
208215.
Maugeri, M., Zekkos, D., Castelli, F. 2011. Seismic performance of landfills. Proceedings ISSMGE ERTC 12
Workshop Evaluation of Geotechnical Aspects of EC8,
Athens 2011, Greece.
Mayne, P.W., Schneider, J.A., & Martin, G.K. 1999. Smalland
large-strain soil properties from seismic flat dilatometer tests. Proceedings 2nd Int. Sym. on Pre-Failure
Deformation Characteristics of Geomaterials, Torino 1:
419427.
NTC 2008. Italian Technical Regulation for Constructions,
D.M. 14 January 2008.
Pereira, A.G., Sopena, L., & Mateos, T.G. 2002. Compressibility of a municipal solid waste landfill. Proc. Fourth
Brazil: 201206.
Maugeri, M., & Castelli, F. 2008. Experimental analysis of waste compressibility. Proceedings Geocongress
2008. ASCE Geotechnical Special Publication no. 177:

208215.
Sharma, H.D., Dukes, M.T., & Olsen, D.M. 1990. Field
measurements of dynamic moduli and Poissons ratio of
refuse and underlying soils at a landfill site. Geotechnics
of Waste Fills-Theory and Practice, STP 1070, ASTM,
Philadelphia.
Del Greco, O., Fassino,A., & Godio,A. 2007. Seismic investigation for the assessment of the elastic settlement in MSW
landfill. Proc. 11th International Waste Management and
Landfill Symposium, Sardinia 2007, Cagliari, Italy, CISA.
TC16 2001. The Flat Dilatometer Test (DMT) in Soil Investigations A Report by the ISSMGE Committee TC16. In
R.A. Failmezger & J.B. Anderson (eds.): 748. Proc. 2nd
Int. Conf. on Flat Dilatometer, Washington.
Yoder, E.J., & Witczak, M.W. 1975. Principles of pavement
design. John Wiley & Son, Inc., New York (ed.).
Zekkos, D., Matasovic, N., El-Sherbiny, R., Athanasopoulos,
A., Towhata, I. & Maugeri, M. 2011. Chapter 4: Dynamic
Properties of Municipal Solid Waste, in Geotechnical characterization, Field Measurements, and Laboratory Testing of Municipal Solid Waste. In D. Zekkos
(eds.), ASCE Geotechnical Special Publication no. 209:
112134.
312

Correlation between compressibility and degradation for a PTMB waste

M. Conte, J.P. Gourc & P. Carrubba
ICEA, Department of Civil, Environmental and Architectural Engineering, University of Padua, Padua, Italy
L. Oxarango
LTHE, University of Grenoble, BP, Grenoble Cedex, France
ABSTRACT: Settlement of Municipal Solid Waste (MSW) in landfill is a geotechnical key concern since it
could lead to dramatic failure of cap cover systems. On the other hand recent research developments demonstrate
that it is possible to propose robust relationship between secondary settlement and kinetics of biodegradation. It is
a matter of fact that settlement could be divided into two quasi-independent fractions, a classical time-dependent
mechanical contribution, similar to the creep phenomenon for cohesive soils, and a biodegradation contribution.
There is a practical interest in this finding: settlement monitored at the surface of the landfill cap cover could
be considered as a relevant indicator of the biodegradation level. This is specifically interesting in the case of a
landfill bioreactive cell, where the influence of leachate injection in this way could be assessed.
The present proposal includes two parts: a presentation of the bio-mechanical relationship through the relation between biogas production and compressibility, and an application to mono-directional (oedometer) tests
on MSW.
INTRODUCTION
One of the more pressing problems in the administration of urban agglomerations is undoubtedly the waste
disposal. Due to the problem of finding suitable lands
and of limiting the greenhouse gas emissions, landfill
are becoming very strategic structures, whose performance should be controlled even after many decades
from the closure of the plant.
To enlarge the capacity of the landfills, pretreatment is becoming widespread, even if no more
knowledge is still fully available about the mechanical and the biological performances of such wastes
(Grisolia & Napoleoni, 1996). The long-term performance of landfills is associated to a correct prevision
of the settlements (Bjarngard & Edgers, 1990); in fact,
the integrity of the barriers (with the two functions,
drainage and impermeability), are linked to the absolute and differential displacements of some parts of the
plant. Settlements are strictly related to the foundation
displacements and to the waste compressibility (Figure
1). The former mechanism is analysed within the practice of the geotechnical engineering, while the latter
one, occurring in very long time, is caused by the viscous behaviour of the solid matter and by the decay of
the organic substance; therefore, deeper understanding
is required about the behaviour of the MSW.
The settlement developing inside the MSW results
from the overlapping of three components different in intensity and evolution: instantaneous, primary
and secondary. The instantaneous settlement occurs
Figure 1. Differential settlements in landfill due to both

foundation and waste compressibility.
as soon as the load is applied. It derives from the

following mechanisms:
volumetric compression of skeleton and gas;

shear deformation of skeleton;
volumetric and shear deformations of solid
elements;
deformation, displacement and pullout of slender
and fibrous solid elements;
crushing of fragile solid elements.
The nature of this type of settlement is purely

mechanical and non-time-dependent; due to the
large heterogeneity of the incoming MSW, the pretreatments able to change the physico-mechanical
behaviour of the original waste and the immediate
application of the load, its prediction may be very difficult (Manassero et al., 1996). For this reason, the
assessment of this type of settlement is often made
313
together with the primary settlement, both being of

purely mechanical nature.
Primary settlement refers to a delayed compressibility induced by the presence of liquid and gas inside
the wastes. As well as the stress increases, the settlement is not instantaneous but delayed by the resistance
encountered by the mobile substance seeping through
the solids. However, if the wastes are placed with a
low degree of saturation, this component may be less
significant and generally occurs within the first few
weeks from the application of the load (Olivier &
Gourc, 2007).
The secondary settlement happens under constant
effective stress. The well known creep phenomenon,
typical of a particle assembly, is coupled with some
biological phenomena in MSW, due to the transfer of
organic matter into leachate and biogas.This biological
component is an important cause of compressibility and may involve very long time until complete
stabilization (Ivanova et al., 2008).
As a final consideration, the identification of the
settlements according to the former classification is
not easy; the single components are often coupled and
influenced by many parameters such as the composition of the refuse and the physico-chemical conditions
occurring inside the plant. Also the heterogeneity
resulting from storing, ageing and managing, may lead
to differentiate settlements of the wastes.
As a result, the correct prediction of spatial and
temporal distributions of the waste settlements is very
difficult (Sowers, 1973). Therefore, monitoring may
be a crucial aspect, if not essential, for establishing the
time at which the final cover should be placed as well as
the biological processes activate or end (Hettiarachchi
et al., 2009; Durmusoglu et al., 2005).
The purpose of this research was that of outlining and calibrating the secondary settlements of low
biodegradable materials, by considering both the creep
and the biodegradation processes. The model was calibrated on the results of long-term oedometric tests,
carried out by means of a small diameter instrumented apparatus allowable at the LTHE Laboratory
of the University of Grenoble. This study concerns the
behaviour of pre-treated refuse, where the biodegradation of the residual organic content may result in
little settlements, far lower than that associated to the
mechanical creep.
biodegradation. This discharge is correlated to the

waste permeability to both liquid and gas. The neutral overpressure (i.e. the hydraulic pressure within
the pores) is considered absent, even if a certain
hydraulic gradient is unavoidable to guarantee the
liquid discharges during creeping of the inert skeleton.
The initial state of the sample coincides with the
end of the primary compression, which is achieved at
a time t100 ; at this stage, the thickness of the waste column is h100 and the void ratio e100 . As soon as the time
t exceeds the value t100 , the secondary deformation
secondary is given by:
in which creep = secondary compression due to creep

(%); bio = secondary compression due to biological degradation (%) and tbio = time at which the
biodegradation starts (days), with tbio > t100 .
The creep component of deformation is evaluated
by means of the following expression:
The biological component of deformation starts at

tbio , i.e. the time at which the methane production is
recorded, at the unstable methanogenesis phase. The
amount of gas produced during the biological degradation may be evaluated according to a first-order
kinetic, similar to the USEPA biogas production model
in landfill (USEPA, 2005). According to this model,
the daily biogas yield Y (t) is given by:
in which Y (t) is given in normal m3 biogas/tonne dry

matter day; BP is given in normal m3 biogas/tonne
dry matter (DM) and accounts for the biogas potential;
k is the rate of biogas generation or constant decay
(days1 ); tbio is the time corresponding to the start
of biodegradation (days). The parameters k and tbio
depend on the initial composition of the waste. The
cumulative biogas production is given by:
where Q(+) = BP. The rate of biogas generation is

estimated from the following equation:
The secondary compression model for MSW considers

the creep and the biological components developing
independently, i.e. without coupling, according to a
one-dimensional scheme (Gourc et al., 2009, 2010;
Conte & Carrubba, 2011, 2012).
Generally speaking, during secondary compression of biodegradable wastes four phases interact:
gas, leachate, inert skeleton and organic matter. The
reduction in voids of the inert skeleton is associated
with some liquid and gas discharges coming from
where t1/2 is the time (days) in which the half biogas potential BP were generated. Some remarkable
values of the parameter k are resumed in Table 1. Figure 2 shows schematically the meaning of the different parameters employed to determine the biological
component.
314
Table 1.
wastes.
Scale
Some remarkable values of k for biodegradable
Type of waste
Laboratory rapidly biodegr.

test
moderately
biod
slowly biodegr.
Landfill
biodegradable
k (days1 )
Table 2. Selected values of the initial solid organic content

ci , depending on the pre-treatment (Gourc et al., 2010).
k (years1 )
Typical biodegradable
matter in MSW
Degradation
percentages (ci )
1) Fine materials (D < 20 mm) (FM)

2) Putrescible materials (PM)
3) Paper and cardboard (P&C)
4) Textiles (T)
5) Greenwaste and wood (G&W)
35.0%
50.0%
40.0%
35.0%
30.0%
>1.4 102
>5.00
4.1 103 1.4 102 1.505.00
5.5 104 4.1 103 0.201.50

5.5 105 5.5 104 0.020.20
1) FM
2) PM
3) P&C
4) T
5) G&W
No pre-treat.
pi
ci,correct
Short pre-treat. Long pre-treat.

pi
ci,correct
pi
ci,corr.
1.00 35.0%
1.00 50.0%
1.00 40.0%
1.00 35.0%
1.00 30.0%
0.25
8.8%
0.25 12.5%
0.25 10.0%
0.25
8.8%
0.25
7.5%
0.10 3.5%
0.10 5.0%
0.10 4.0%
0.10 3.5%
0.10 3.0%
Figure 2. The meaning of parameters during the biogas production employed to determine the biological component of
the deformation.
The secondary compression caused by the biological degradation of organic matter can be evaluated
according to the expression (Gourc et al., 2010):
in which c is the initial gravimetric solid organic content, depending on the composition of MSW; d0
is the initial global dry density of the waste (t/m3 );
so0 is the initial solid organic density of the waste
(t/m3 ). Therefore, the total secondary compression is
given by:
The relationship between solid organic content and the

biogas production is a crucial aspect for calculating
secondary compressibility. Once the organic fractions
fi,bio of the waste were identified, the degradation percentages of the organic fraction ci had to define. They
were taken from a literature reconnaissance (Machado
et al., 2009) and reported in Table 2. If the waste
undergo pre-treatment, the degradation percentage ci
should be correct according to:
Figure 3. Small diameter oedometer apparatus for onedimensional compression tests.
INTRODUCTIVE ANALYSES
The proposed model was applied to reproduce the

experimental results of some oedometer tests carried out at the LTHE laboratory of the University of
Grenoble. The aim is to verify how the model results,
concerning secondary mechanical and biological settlement, fit the laboratory data.
A bio-mechanically pre-treated MSW was subjected to compression tests in a cell-reactor present
at the University of Grenoble, for a period of 77
days, without any liquid injection. These tests have
allowed to highlight the dependence of bio-mechanical
parameters on the initial state of the waste.
3.1 Description of the used reactor
in which pi is the part of organic content remaining

after the pre-treatment (Gourc et al., 2010). No pretreatment implies pi = 1 and ci,correct = ci .
Finally, the total percentage c of the degradable
matter was evaluated by using the relationship:
The cell-reactor (Figure 3) is constituted by a stainless

steel cylinder bolted to a removable bottom, for the
cleaning and the extrusion of the material, and closed
on the top by a sliding cylindrical watertight piston
(maximum displacement of 15 cm).
The inner diameter of the cell is 270 mm, while
the maximum inner height is 290 mm. Two geotextiles were placed in the upper and lower extremities of
315
Table 3. Waste fractions in percentage ( fi ).

Organic fractions ( fi,bio )
Fine materials (D < 20 mm)
Putrescible materials
Paper and cardboard
Textiles
Green-waste and wood
Table 4.
waste.
Other fractions
54.9%
2.0%
13.1%
1.6%
5.1%
Plastics
Glass
Metals
Inert
18.0%
1.0%
0.2%
4.1%
the cell. The cell-reactor allows to carry out simultaneous analyses, as mechanical characterizations,
inner pressure values, and the control of the biogas
production.
The compression tests were performed considering a long term vertical pressure of 40 kPa, that was
transferred to the waste sample through the pneumatic
piston, moved by a pressure lever.A force sensor (Interface, 1210 standards, load capacity of 25 kN) allows to
evaluate the force applied and then the resulting load.
A pressure sensor provides the values of pressure at the
top of the reactor. A displacement sensor (Solartron,
S-series, measuring range: 5 mm) allows to determine
the waste settlement h. The three sensors are connected to a computer from which, using a dedicated
program, it is possible to see and record in real time
all the considered parameters.
3.2
Description of the sample and waste

characterization
The used material is a municipal solid waste (MSW)

coming from the Legnago landfill (VR, Italy),
obtained from a bio-mechanical pre-treated waste. The
pre-treatment consists of a sieving to 100 mm and,
subsequently, 60 mm diameter; the final bio-stabilized
material was obtained after aeration of 3 weeks in biotunnel, and a post-treatment in small piles aerated for 3
weeks. The constitutive composition of the pre-treated
MSW is shown in Table 3.
One-dimensional compression tests were carried
out under controlled rates of loading. Long-term tests
were performed for the pre-treated waste, without
any leachate injection. Information related to the tests
conditions are summarized in Table 4.
3.3
Characteristics of the oedometer and the tested
Procedures of the test execution
Before the storage of the pre-treated MSW into the

reactor, a small sample of this material was collected
and dried in a thermostatic oven, at the temperature
of 80 C for 72 hours, in order to calculate the initial gravimetric water content. The remaining material
was homogeneously wetted with de-aerated water, in
order to obtain, initially, a waste at its field capacity (w = FC = 52.47%). The waste was not crushed
mechanically, since most of it passed the sieve with
D < 20 mm.
During the storage phase of the test, 12.030 kg of
pre-treated and saturated MSW were introduced inside
Reactor characteristics
Initial height of the reactor
h0
Reactor surface
A
Initial volume of the waste*
Vt
Reactor diameter
D
Pre-treated MSW characteristics
Wet mass of the waste*
mh
Dry mass of the waste*
md
Water mass of the waste*
mw
Initial gravimetric water content
w0
Field capacity
FC
Initial wet density (without compaction) h
Initial global dry density
d0
Initial solid organic dry density
so0
Test characteristics
Maximal vertical stress
Test duration
0.290 m
0.0572 m2
0.0166 m3
0.270 m
12.030 kg
7.890 kg
4.140 kg
43.20%
52.47%
0.310 t/m3
0.837 t/m3
0.953 t/m3
40 kPa
2.5 months
(77 days)
*within the reactor.
the reactor. The waste was compacted mechanically in

five layers, having a thickness of 5 cm, plus a bottom
layer of 4 cm of thickness, with a mass of about 2 kg
for every layers. A geotextile was placed at the bottom
and at the top of the waste, before and after the waste
storage within the reactor; subsequently, was placed
the pneumatic piston, equipped with an o-ring, so as
to produce a waterproof system and prevent any gas
leakage. Subsequently, the pneumatic piston was activated, at the forces corresponding to the programmed
load path.
For this test, no liquid or leachate was added or recirculated. In order to collect the produced biogas, a
special system was created, that consists of an inverted
measurable cylinder, filled with a certain level of water
and immersed in a tank (Figure 4). A tube, coming
out from the top of the cell and used for the biogas
collection, passes through the tank and the inverted
cylinder, and exits into the portion of the cylinder that
is occupied by the air. The volume of biogas produced
pushes the water present in the cylinder upside down
towards the tank; therefore, it is measured by observing
the difference of water level (before-after) read on the
graduated cylinder. The gas volumes were expressed
in Normal m3 (normal conditions, at the temperature
of 25 C and at the atmospheric pressure); therefore,
they have been corrected as a function of the biogas
pressure, compared to the atmospheric pressure.
3.4 Oedometric compression tests with loading,
unloading and reloading cycles
The load path was made according to a succession of
loading, unloading and reloading phases, until reaching a maximum vertical stress of 40 kPa. This load
level was thereafter maintained constant, during the
long-term test. Under these different load steps, it
was possible to record the settlements that the waste
316
Figure 4. System used for determine the biogas production.
Table 5.
the test.
Step
Main geotechnical parameters calculated during
Stress
v
(kPa)
Settlement
h
(cm)
Cumulative
settlement
h
(cm)
Instantaneous and primary settlement

1
1
1.28
1.28
2
5
2.79
4.07
3
1
0.53
3.54
4
5
0.63
4.17
5
10
0.35
4.52
6
5
0.14
4.38
7
10
0.46
4.85
8
20
0.61
5.46
9
10
0.09
5.37
10
20
0.38
5.75
11
40
1.55
7.29
12
20
0.13
7.17
13
40
0.53
7.69
Secondary settlement
14
40
1.15
8.84
Loading
ratio
RC
Step ()
Unloading
ratio
RS
()
Reloading
ratio
RR
()
Instantaneous and primary settlement

1
2
0.138
0.026
0.031
5
0.040
0.015
0.052
8
0.070
0.010
10
0.044
11
0.176
12
0.014
13
0.060
Secondary settlement
14 Constant load of 40 kPa
Cumulative
deformation
(%)
Figure 5. Representation of the evolution of the vertical
settlement as a function of applied pressure, a) in a normal
scale; b) in a semi-logarithmic scale.
4.41
14.03
12.21
14.38
15.59
15.13
16.72
18.83
18.52
19.83
25.14
24.72
26.52
experimented under compression, and a series of other

parameters, such as:
the stress levels (v );

the settlement obtained every step and the cumulative settlement (h);
the deformation obtained every step and the
cumulative deformation ();
the loading ratio (RC);
the unloading ratio (RS);
the reloading ratio (RR);
the compression coefficient (mv );

the oedometric modulus (Eedom );
the coefficient of secondary mechanical settlement

(CM , see section 4.1).
30.49
Compr.
coeff.
mv
(kPa1 )
Oedom.
modulus
Eedom
(kPa)
0.024
0.005
0.002
0.003
0.002
0.001
0.003
0.001
41.6
184.3
413.2
314.5
473.9
763.4
376.6
1111.1
The parameters described above are calculated for

each load step and are reported in Table 5. These values
are shown in graphs as a function of vertical pressure,
in normal or semi-logarithmic scale (Figure 5a,b).
3.5
Long term compression test
The long term compression test was conducted at the

constant load. The pre-treated MSW was subjected to
the maximum vertical stress of 40 kPa for 2.5 months
(77 days). The deformation of the waste was studied by separating the secondary mechanical settlement
(due to the creep) and the biological settlement produced by degradation. This test was also conducted
317
Figure 7. Percentages of the major components of the

biogas.
Figure 6. a) Daily average rate of biogas production and

b) cumulative biogas production.
in conjunction with the analysis of biogas production

within the cell-reactor.
3.5.1 Leachate production
Given the conditions of absence of injection or recirculation of leachate, and assumed the conditions of initial
saturation of the material, the production of leachate
was slight (280 ml, in total). In particular, the waste
under compression has produced 90 ml (pH = 7.60)
during the 7th week, 90 ml (pH = 7.63) during the 8th
week, 60 ml (pH = 7.85) during the 10th week and
40 ml (pH = 9.15) during the 11th week.
3.5.2 Biogas production
The collection of the produced biogas was carried out
as described in section 3.3, starting from day 3, i.e.
from the end of the instantaneous and primary settlement. Figure 6a shows the daily average rate of
biogas production, expressed in Normal m3 /tonne of
dry mass day, while Figure 6b shows the cumulative
biogas production, expressed in Nm3 /tonne DM.
The collected gas was analyzed using a special instrument (Portable gas analyzer GA2000 Plus
GeoTech) able to define the percentages of the major
components (CH4 , CO2 , O2 and other components).
This analysis was conducted from the start of the test,
but the first three days were not counted in the determination of the biogas production, as still being in phase
of instantaneous and primary settlement. The results
are shown in Figure 7.
The beginning of the unstable methanogenic phase
(tbio ) was taken after 12 days. The maximum daily
biogas production was obtained approximately from
the day 16 to the day 20 of the test. At the day
77 (end of the test), the daily biogas production
appears to be constantly decreasing and down to about
Figure 8. Representation of the waste deformations.

Table 6. Data related to the settlements and deformations
of the pre-treated MSW.
Secondary
Instantaneous + Secondary mechanical +
primary
mechanical biological
Vertical
stress
from 0 to 40 kPa constant

(stress path)
load of
40 kPa
Duration
From day 0
From day 3
to day 3
to day 77
Initial height 29.00 cm
21.31 cm
Final height
21.31 cm
20.85 cm
Settlement
7.69 cm
0.46 cm
Deformation* 26.52%
1.59%
constant
load of
40 kPa
From day 12
to day 77
20.85 cm
20.16 cm
0.69 cm
2.38%
*considering the initial height of waste.
45 ml/h of produced biogas, equivalent to less than

0.10 Nm3 /tonne DM day. Since the daily biogas production is decreasing and almost constant at the end
of the test, the final value of the cumulative biogas
production tends, therefore, to stabilize around the
18 Nm3 /tonne DM. The final value of the parameter
is supposed to be BP (see equation 4).
3.5.3 Settlements and deformations
The settlement analysis represents the most important
phase of the test. The observed settlements can be distinguished in instant and primary settlements, which
nature is exclusively mechanical, and secondary settlements, obtained under constant load, which nature is
both mechanical and biological. The diagram of deformations is represented in Figure 8, while the magnitude
of settlements and deformations is reported in Table 6.
318
Table 7. Results of the determination of the solid organic

content c.
Figure 9. USEPA curve that modelled a) the daily average rate of biogas production and b) the cumulative biogas
production.
MODELLING RESULTS
4.1 Determination of mechanical settlements

The mechanical compression parameters, describing
the primary compression of the waste and the first
part of the secondary compression curve (Figure 8),
were evaluated by the observed settlement data:
the end of primary settlement was achieved in 3

days, obtaining a coefficient of primary compression Cc = 0.140;
the secondary mechanical compression was included
from day 3 to day 77 (from the end of the primary
settlement, t100 , to the end of the test), obtaining a
secondary mechanical coefficient CM = 0.033.
4.2 Determination of the biogas parameters
Due to pre-treatment, the production of biogas was
small. Starting from tbio = 412 days and obtaining the
value of BP = 418 Nm3 biogas/tonne of dry matter, it was possible to evaluate the parameters t1/2
and k (t1/2 = 29 days; k = 0.04 days1 ), so allowing
a reconstruction of the biogas production according to
equation (3), (4) and (5). The curves that modelled the
biogas production data are represented in Figure 9a,b.
4.3
Determination of the solid organic content c
The biodegradable fractions fi,bio (Table 3) multiplied

by the corresponding corrected biodegradable content
Organic fractions
fi,bio
ci,correct
fi,bio ci,correct
Fine materials
Putrescible materials
Paper and cardboard
Textiles
Green-waste and wood
54.9%
2.0%
13.1%
1.6%
5.1%
3.5%
5.0%
4.0%
3.5%
3.0%
1.92%
0.10%
0.52%
0.06%
0.15%
c = 2.75%
Figure 10. Results of the secondary settlement modelling

for the pre-treated waste.
for long pre-treatment ci,correct (Table 2), gave the total

value of biodegradable content c = 2.75% (equation 9
and Table 7). This value of total organic matter contributes, in the long time, to the production of biogas
and leachate.
4.4 Modelling of the total secondary settlement
The knowledge of mechanical parameters (Cc ,CM ,
t100 , d0 , so0 ) and biological parameters (k, BP, tbio )
allows modelling of the final vertical strains (Figure
10). After a primary settlement, characterized by the
compression ratio Cc , the waste is subjected, in time, to
a secondary settlement, separated into its mechanical
and biological components.
The pre-treated MSW shows a final vertical strain
about 30%. Almost all the secondary settlement seems
to be of mechanical nature, despite the biological
degradation starts at about 12 days. This can be
explained by the fact that the waste has been subjected to a long pre-treatment, therefore, much of the
organic matter has been already degraded before the
test. Hence, it seems reasonable to have an overall
secondary deformation, due both to the creep and
biological component, of about 4%.
Three important observations can be made regarding this test. The first relates to the biogas production:
since the waste was bio-mechanically pre-treated, it
has the particularity to produce non-negligible amount
of biogas, once reactivated the degradation process
under compression. This suggests that the waste,
because of the pre-treatments, has obtained a moderate
319
biological stability, but not definitive. This behaviour

is supported by the monitoring data of large scale
landfills: a non pre-treated waste, stored in landfill,
reaches the biological stability in a period of between
30 and 100 years, while a pre-treated waste reaches
this condition before.
The second observation concerns the very short
period of reactivation of the biodegradation (only 12
days). The waste has already undergone an aerobic
and anaerobic degradation phase, during the pretreatment period; therefore, unlike a fresh MSW, the
methanogenic phase begins in a very short time, starts
up again.
Finally, the third evidence concerns the biological settlement: the amount of vertical deformation
attributable to degradation of the organic matter is
2.4%, referring to a total strain of 30.5%. In this case
it is possible to highlight a strong mechanical component, of approximately 92%, because of a certain
degree of biological stability that the MSW has initially had, after the bio-mechanical pre-treatment and
before the oedometric test.
5
CONCLUSIONS
Settlement evolution in MSW landfills represents a

crucial topic because of the implications in the storage capacity of the plant as well as for the integrity of
the facilities, such as barriers or biogas and leachate
drainages.The strategy to stabilize the biological waste
entering in landfill and to enhance mechanically its
geotechnical properties, through the improvement of
pre-treatments, is presented as a viable path to follow. An analysis of MSW settlements, decoupling
mechanical and biological aspects, has been made.
The proposed model, assessing the biological and
the creep components of the secondary settlement, was
applied to laboratory test results, carried out at the
University of Grenoble. For this purpose, a pre-treated
MSW was used. This long pre-treatment consisted of a
sieving to 100 mm and, subsequently, 60 mm diameter;
the final bio-stabilized material was obtained after aeration of 3 weeks in bio-tunnel, and a post-treatment in
small piles aerated for 3 weeks. These few degradable
wastes have been prepared, in order to notice if this
model is able to fit suitably the observed secondary
settlement, in such particular conditions.
Due to pre-treatment, the amount of the remaining biodegradable content was small and the biogas
production very little. Nevertheless, the model gave
long-term settlement previsions which were in a good
agreement with the laboratory results.
REFERENCES
Bjarngard, A. & Edgers, L., 1990. Settlement of municipal
solid waste landfills. Proceedings 13th Annual Madison
Waste Conference, Madison, WI: 192205.
Conte, M. & Carrubba, P., 2011. Mechanical and biological aspects in MSW settlements. Proceedings Sardinia
2011, 13th International Waste Management and Landfill
Symposium, S. Margherita di Pula, Cagliari, Italy: paper
456.
Conte, M. & Carrubba, P., 2012. Componente meccanica e
biologica nei cedimenti secondari dei rifiuti solidi urbani:
modellazione teorica e riscontri sperimentali. IARG 2012,
Padova, 24 luglio 2012.
Durmusoglu, E., Corapcioglu, M.Y., Tuncay, K., 2005. Landfill settlement with decomposition and gas generation.
Journal of Environmental Engineering 131: 13111321.
Gourc, J.-P., Staub, M.J., Conte, M., Benbelkacem, H.,
Bayard, R., Redon, E., 2009. A biochemical model
to predict long-term secondary settlement of MSW
Validation on large-scale trials. Proceedings of Sardinia
2009, 12th International Waste Management and Landfill
Symposium, S. Margherita di Pula, Cagliari, Italy: paper
089.
Gourc, J.P., Staub, M.J., Conte, M., 2010. Decoupling
MSW settlements into mechanical and biochemical
processes Modelling and validation on large scale
setups. Waste Management 30: 15561568.
Grisolia, M. & Napoleoni, Q., 1996. Geotechnical characterisation of municipal solid waste: choice of design
parameters. Proceedings 2nd International Congress on
Environmental Geotechnics, Osaka, Japan.
Hettiarachchi, C.H., Meegoda, J.N., Hettiaratchi, P., 2009.
Effect of gas and moisture on modeling of bioreactor
landfill settlement. Waste Management 29: 10181025.
Ivanova, L.K., Richards, D.J., Smallman, D.J., 2008. The
long-term settlement of landfill waste. Waste and
Resource Management 161(3): 121133.
Machado, S.L., Carvalho, M.F., Gourc, J.P., Vilar, O.M, do
Nascimento, J.C.F., 2009. Methane generation in tropical landfills: Simplified methods and field results. Waste
Management 29: 153161.
Manassero, M., Van Impe, W.F., Bouazza, A., 1996. Waste
disposal and containment. Proceedings 2nd International
Congress on Environmental Geotechnics, Osaka, vol. 3:
193242.
Olivier, F. & Gourc, J.P., 2007. Hydro-mechanical behavior
of MSW subject to leachate recirculation in a large-scale
compression reactor cell. Waste Management 27: 4458.
Sowers, G.F., 1973. Settlement of waste disposal fills. Proceedings 8th International Conference on Soil Mechanics
and Foundation Engineering, Moscow, vol. 2: 207210.
USEPA, 2005. First-order kinetic gas generation model
parameters for wet landfills. EPA Report, June 2005.
320

A laboratory landfill simulator for physical, geotechnical, chemical and

microbial characterization of solid waste biodegradation processes
X. Fei, D. Zekkos & L. Raskin
Department of Civil and Environmental Engineering, University of Michigan, Ann Arbor, Michigan, USA
ABSTRACT: Bioreactor landfills enhance biodegradation of municipal solid waste (MSW) via active liquid
addition and/or leachate recirculation. This study presents a landfill simulator laboratory experimental setup that
was used to study biodegradation of MSW and to collect data to characterize physical, geotechnical, chemical,
and microbial changes in the MSW specimen. The simulator hada diameter of 0.3 m and a volume of 40 L and
was filled with 30 kg of well-characterized MSW obtained from a landfill.Leachate was recirculated and MSW
biodegradation was monitored for 250 days. The total volume of the specimen (Vt ) reduced 13%, whereas the
total unit weight increased steadily with time after an initial adjustment period. The long-term compression ratio
(CLT ) changed over time with a maximum value of 0.125. Biogas consisted of approximately 55% CH4 and 45%
CO2 . A total of 425 L CH4 was generated and the peak CH4 generation rate (rCH4 ) was 10.3 L/day. Maximum
concentrations of soluble and insoluble COD (chemical oxygen demand, sCOD and insCOD) in the leachate
were 4,190 mg O2 /L and 990 mg O2 /L, respectively. The concentration of microorganisms in the leachate was
estimated by measuring the deoxyribonucleic acid (DNA) concentration and the maximum observed value was
8,610 ng DNA/ml. The DNA concentration in the leachate was indicative of rCH4 and the change rate of soluble
COD (rsCOD ). They were related to the cumulative volume of CH4 (VCH4 ) and CLT . The experimental setup
and the data collected may provide a valuable basis for future studies of coupled phenomena relevant to the
degradation of MSW.
INTRODUCTION
Over 200 million tons of MSW are generated in the

United States and the European Union each year, and
over 50% and 40% respectively are disposed of in landfills (EPA, 2011a, EuroStat, 2011). Approximately
99% of the operating landfills in the US are designed
as conventional (dry tomb) landfills that aim to minimize moisture infiltration. Biodegradation of MSW
is inhibited by low moisture content in such landfills,
and thus waste may remain largely undegraded over
decades to centuries (Mehta et al., 2002, Pohland and
Alyousfi, 1994, Barlaz et al., 2010). Undegraded waste
poses a long-term risk to the environment and society, and contains untapped energy. Biogas consisting
of methane (CH4 ) and carbon dioxide (CO2 ) is generated as a product of MSW biodegradation. Since
MSW biodegradation is slow in conventional landfills,
CH4 and CO2 , both greenhouse gases, are released at
low rates, but over an extended period of time. As a
result, biogas collection and energy recovery become
economically unfavorable (EPA, 2010, EPA, 2009).
Landfills are the second largest anthropogenic source
of CH4 emission in the US (EPA, 2011b).
In contrast, bioreactor landfills aim to enhance
MSW biodegradation via active liquid addition and/or
leachate recirculation. Methane is generated at an
increased rate due to stimulated microbial activity and
can be recovered and used more efficiently. In addition, waste volume reduction is achieved, which results
in increased landfill capacity.
Microbial activity in landfills is influenced by
physical and chemical conditions of the waste. The
biodegradation process alters the characteristics of
solid waste, biogas and leachate. Kim and Pohland
(2003) and Barlaz et al. (1989) identified five phases
and four phases, respectively, of MSW biodegradation.
In this study, MSW biodegradation is divided into three
phases, the transition phase, the active biodegradation
phase, and the residual phase (Fei and Zekkos, 2012).
Numerous studies on the biodegradability of MSW
found that the quality of generated biogas and leachate
varied for different biodegradation conditions and
waste ages (Bookter and Ham, 1982, Ham and Bookter, 1982, Barlaz et al., 1989). Microorganisms in
leachate and on waste have been characterized (Staley
et al., 2011, Staley et al., 2012, Huang et al., 2004).
Several models have been proposed to describe and
predict physical, geotechnical and chemical properties of waste in bioreactor landfills (McDougall, 2007,
Gawande et al., 2010, Vavilin et al., 2004). However, although physical, geotechnical, chemical and
microbiological processes take place simultaneously
during MSW biodegradation in landfills, most previous studies have focused on selected variables and
have simplified other aspects. The relationships and
321
interactions between different processes have not yet

been addressed.
This paper presents a large-scale laboratory experimental setup that enables continuous monitoring
and analyses of physical, geotechnical, chemical and
microbiological properties of a MSW specimen and
the associated leachate during biodegradation. Experimental results from a 250-day biodegradation test are
presented.
2
EXPERIMENTAL METHODS
2.1 Waste specimen

A 3-month old MSW sample was excavated from
a landfill in Austin, Texas and was shipped to the
University of Michigan geoenvironmental engineering laboratory in sealed drums. Waste composition
was characterized per Zekkos et al. (2010) recommendation and was found to consist of 74.5% <20 mm
material, 15.0% paper, 5.5% soft plastic and 5.0%
wood by weight. The natural wet moisture content of
the 20 mm material was 21%. A 30-kg specimen was
re-constituted using the waste fractions from the field
sample. Waste was not milled or in any other way processed, although oversized particles were not included
in the specimen. Waste was placed into the simulator with minimal compaction to an initial waste height
(H0 ) of 57.5 cm.
2.2
Simulator setup, operation and sampling
The simulator consisted of a waste column, a heating

blanket, a weighing scale, a leachate tank, a leachate
recirculation system, a biogas collection and sampling
system, transducers and a data acquisition system
(Fig. 1). The waste column was 30 cm in diameter,

60 cm in height and 40 L in volume. It was sealed to
maintain anaerobic conditions throughout the test. No
external vertical stress was applied to the specimen.
Five sampling ports were installed along the height
of the column to allow for coring of solid waste samples. The temperature of the specimen was controlled
by a heating blanket wrapped around the simulator.
The optimal temperature for growth of mesophilic
methanogens, 40 3 C, was selected. (Zinder, 1993).
The simulator was placed on a scale for continuous
measurement of the total weight of the specimen. A
valve was installed at the bottom of the column to
facilitate saturation of the waste specimen and gravity drainage of the leachate to a leachate tank. The
tank had a port at the bottom for liquid sampling and
included a magnetic stirrer. Leachate was recirculated
by a peristaltic pump that delivered leachate to the top
of the column. A perforated acrylic plastic plate and
a 25-mm layer of gravel at the top allow for uniform
distribution of the recirculated leachate. Another layer
of gravel and a piece of non-woven geotextile were
placed beneath the MSW to prevent waste flushing-out
from the simulator. Generated biogas was collected in
a gas sampling bag through the top cap.A gas sampling
port with septum was installed on the top cap to allow
for direct sampling of biogas from the headspace. A
temperature transducer was placed in the specimen at
the center of the column to continuously measure the
temperature. A cable extension transducer was used to
measure waste settlement and was connected to a data
acquisition system.
Ten days after specimen preparation (day 10), the
leachate drainage valve was closed and deionized
water was added to the simulator until the waste specimen was saturated. Leachate was drained after 30
minutes. The collected leachate was thoroughly mixed
using the magnetic stirrer each time before recirculation. Recirculation occurred three times a week
throughout the test period.
Biogas in the headspace of the simulator was sampled immediately before leachate recirculation. A
leachate sample was collected from the leachate tank
after complete drainage of the recirculated leachate
from the column. Both samplings were performed
three times a week until day 100, and once per week
thereafter.
2.3
Biogas quality and quantity
Biogas was analyzed for CH4 and CO2 concentrations (%CH4 and %CO2 ) by gas chromatography with
a Thermal Conductivity Detector. The volume of biogas (Vgas ) was measured by a mass flow meterand was
adjusted to a 20 C temperature. The volume of CH4
(VCH4 ) was the product of %CH4 and Vgas . Methane
generation rate was calculated per equation (1).
Figure 1. Schematic of laboratory bioreactor landfill

simulator.
where VCH4,t1 , VCH4,t2 and t1 , t2 were the values

and times of two VCH4 measurements.The cumulative
volume of CH4 was calculated by summing VCH4 .
322
2.4 Physical and geotechnical properties

Total waste weight was measured in saturated (Wt,sat )
and gravity-drained (i.e., at field capacity) states
(Wt,fc ). The waste height (H) was calculated by subtracting waste settlement and height of gravel layers
from the total height of the simulator. The total volume of the waste (Vt ) equaled waste height times the
cross-sectional area of the simulator (710 cm2 ).
Total unit weights of the specimen were calculated
as Wt,sat and Wt,fc divided by Vt , respectively (t,sat and
t,fc ). Daily and 30-day long-term compression ratios
were calculated according to equation (2).
where Ht1 , Ht2 and t1 , t2 are the heights and times of

two consecutive measurements.
2.5
Leachate quality
Leachate samples were analyzed for pH, soluble COD

and insoluble COD. (APHA, 2005). Change rate of
soluble COD was calculated according to equation (3).
where sCODt1 , sCODt2 and t1 , t2 are concentrations of

sCOD and times of two consecutive measurements.
2.6
Leachate biomass concentration
DNA of the microorganisms in the leachate was

extracted on days 23, 34, 46, 83, 109 and 178 and
the DNA concentration was measured (ng DNA/ml
leachate) by the method described in Pinto and Raskin
(2012). Future work will use this DNA for microbial
community characterization. Solid waste was sampled
from the simulator through sampling ports and visually assessed by Environmental-Scanning Electron
Microscopy (E-SEM).
3
3.1
RESULTS
Physical and geotechnical properties
The total weight of the MSW in saturated state (Wt,sat )

and at field capacity (Wt,fc ) are shown for the 250 days
of the experiment in Fig. 2a. When the waste specimen was saturated at the start of the experiment, Wt,sat
increased from 30 kg to its maximum, 47.4 kg. Wt,sat
decreased steadily to 43.4 kg until day 25 and then to
39.4 kg by day 250. After initial drainage, Wt,fc was
35.0 kg.Wt,fc reached its maximum value of 36.1 kg
on day 25, and then gradually decreased to 35.8 kg by
day 250.
After initial saturation with leachate and drainage,
the total volume of the waste specimen gradually
decreased from 36.9 L to 32.1 L by day 250 (Fig. 2b).
This corresponded to a volume reduction of 13%.
Total unit weights of the specimen in saturated state

and at field capacity are plotted in Fig. 2c. Initial
t was 7.9 kN/m3 . Maximum t,sat (12.6 kN/m3 ) was
achieved after initial waste saturation, and decreased
to 11.5 kN/m3 by day 46. Then it slowly increased
to 12.0 kN/m3 by day 250. On the other hand, t,fc
increased from 7.9 kN/m3 to 10.9 kN/m3 over the test
period.
Daily and 30-day averaged CLT are shown in Fig.
2d. The maximum 30-day averaged CLT was 0.125
0.023 (mean standard deviation) between day 30 and
day 60, whereas CLT between day 150 and day 180 was
0.044 0.037. It subsequently increased, for reasons
that remain unclear and are presently being investigated. A building power failure event on day 204 may
have affected the results.
3.2 Biogas composition and generation rate
Changes in the biogas composition over the operational time period are shown in Fig. 2e. The carbon
dioxide concentration increased from 50% to 65%
between day 10 and day 15. It then dropped to 45%
by day 40 and fluctuated around this value thereafter.
The methane concentration was 20%on day 10and
climbed to 62% within 20 days. A stable concentration
of approximately 52% was measured after day 40.
The methane generation rate peaked on day 33
(10.3 L/day). It decreased to around 2.0 L/day by day
70 and was less than 1.0 L/day after day 100 (Fig. 2f).
The VCH4 increased sigmoidally. The slope of the
curve changed significantly around day 33 and day
60. The value was 425 L by day 250 (Fig. 2g).
3.3 Leachate quality
The highest concentration of sCOD in the leachate
(4,190 mg O2 /L) was measured on day 23. Concentration of sCOD decreased quickly from day 23 to day
44 (990 mg O2 /L), and then gradually decreased to
660 mg O2 /L by day 250. The change of insCOD was
less dramatic; the highest concentration was 990 mg
O2 /L on day 46 and the lowest concentration was
120 mg O2 /L on day 250 (Fig. 2h).
The rsCOD was plotted in Fig. 2i. Positive and negative values represented accumulation and consumption
of sCOD, respectively. Zero values indicate a balance
between production and removal of sCOD. The maximum rsCOD was measured on day 10 immediately after
initial waste saturation, then the rate decreased to 0 by
day 23. The minimum rsCOD was observed on day 33.
After day 46, rsCOD remained constant around 0.
Leachate pH was initially around 6 and fluctuated
around 6.5 after day 50 (data not shown). No pH
adjustment was performed.
3.4
Leachate biomass concentration
The maximum DNA concentration, 8,610 ng DNA/ml

leachate, was measured on day 34. The DNA
323
Figure 3. E-SEM image of microbial cells (indicated by

arrows) on waste paper fibers.
DISCUSSION
4.1 Physical and geotechnical properties
Figure 2. Long term behavior of the specimen: (a) Wt,sat

and Wt,fc ; (b) Vt ; (c) t,sat and t,fc ; (d) CLT and 30-day CLT ;
(e) %CH4 and % CO2 ; (f) rCH4 ; (g) VCH4 ; (h) sCOD and
insCOD; (i) rsCOD ; (j) DNA concentration.
concentration decreased to 927 ngDNA/mlleachateby

day 178 (Fig. 2j). An example of an E-SEM image of
a MSW sample is shown in Fig. 3. Microbial cells can
be seen on fibers of paper.
The total volume of the specimen was reduced due

to long-term compression of the waste. Physical processes (raveling, particle movement and deformation,
and moisture-induced softening) are the major contributors to long-term compression immediately after
specimen preparation (Wall and Zeiss, 1995, Bareither
et al., 2012). The total weight of the specimen in the
saturated state consists of weight of solids and weight
of leachate. In the early days, limited biodegradation
occurred, as discussed below, but Wt,sat was reduced by
4 kg between day 10 and day 25 (Fig. 2a), because less
leachate was needed to saturate the volume of voids.
This reduction to the volume of voids is attributed to
the physical processes described above.
Even before day 25, the contributions of the physical mechanisms diminished gradually, and biodegradation of the specimen intensified and became the
major contributor to long-term compression. Decrease
of Wt,sat after day 25 was influenced by biodegradation of waste,leading to weight loss and a change
in the volume of voids (McDougall, 2007). Overall,
Wt,sat decreased by 4 kg between day 25 and day 250
(Fig. 2a).
After initial saturation on day 10, Wt,fc reached its
maximum and then decreased very gradually (Fig. 2a).
This process was affected by both weight loss due to
biodegradation of the waste and weight change of the
waste-retained leachate at field capacity.
A total strain of 13% was achieved on day 250 (Fig.
2b). Peak 30-day averaged CLT was 0.125 0.023
between day 30 and day 60 (Fig. 2d). These values are
consistent with the statistically analyzed values of 26
anaerobic tests with no external vertical stress, where
recorded total strain was on average 10.3%, and CLT
on average 0.147 (Fei and Zekkos, 2012).
Total unit weights of the specimen in both saturated and fully drained states increased with long-term
biodegradation. Between day 25 and day 250, t,sat
increased from 11.8 kN/m3 to 12.0 kN/m3 . Also, t,fc
324
increased from 9.8 kN/m3 to 10.9 kN/m3 over the

same period (Fig. 2c). These t,fc values are consistent with field values, which typically range between
5.0 and 15.0 kN/m3 measured at shallow depths with
the upper-bound values being expected for soil-rich
waste (Zekkos et al., 2006).
4.2
Biogas composition and generation rate
The methane generation rate (rCH4 ) is indicative of the

level of methanogenesis during MSW biodegradation
(Vavilin et al., 2004). The maximum rCH4 was observed
on day 33 (Fig. 2f), shortly after the %CH4 in the
biogas exceeded the %CO2 on day 25 (Fig. 2e), indicating that methanogens were most active around that
time. Between day 33 and day 250, the rCH4 decreased
but biogas composition remained stable, suggesting
microbial community continued to contain a stable
population of methanogens, despite a decrease in readily biodegradable organics. The cumulative VCH4 (Fig.
2g) is consistent with these observations. The more
CH4 was generated, the less CH4 generation potential or readily biodegradable organics remained in the
specimen. The slope of the VCH4 curve changed significantly around day 33 and day 60, corroborating
that phase transition occurred at these times.
4.3
Leachate quality
The changes in the sCOD concentrations in the

leachate (Fig. 2h) provide an indication of changes
in the waste specimen composition. The peak sCOD
value in the leachate was indicative of the quantity
of readily biodegradable waste in the specimen and
could be further correlated to VCH4 and rCH4 (Cossu
and Raga, 2008). The sCOD concentration remaining
at the end of the experiment (residual sCOD) mainly
constitutes recalcitrant soluble organics, e.g., humic
substances, phenolic compounds, etc. (Kjeldsen et al.,
2002). The change rate of sCOD may be indicative
of the level of hydrolytic and methanogenic activity
during MSW biodegradation (Batstone, 2002).
The major source of insCOD constituted large
organic particles flushed out by leachate from the
waste specimen, before they were hydrolyzed. Some
of these large particles were retained in the waste
matrix during recirculation. They were gradually disintegrated and hydrolyzed either in the laboratory
simulator or in the leachate.
Since CH4 and sCOD were generated, MSW
biodegradation was likely minimally inhibited by the
slightly acidic pH of 6.06.5, consistent with previously reported MSW biodegradation studies (Stroot
et al., 2001). The laboratory simulator sustained anaerobic biodegradation of MSW without pH adjustment
for an extended time period, contrary to some previous studies (Bareither et al., 2012, Wang et al., 1994).
One possible explanation was that the waste provided
sufficient buffering capacity to overcome the initial
acidification due to rapid hydrolysis and fermentation
and volatile fatty acid buildup. This allowed the slower
syntrophic bacteria and methanogens to become active

and resulted in an efficient biodegradation process.
4.4 Leachate biomass concentration
The leachate was thoroughly mixed each time it was
recirculated to saturate the waste specimen and before
sampling. Also, the leachate volume in the simulator system remained relatively constant throughout
the experiment. Therefore, the concentration of DNA
in the leachate can be used to estimate the biomass
concentration in the leachate.
The DNA concentration reached its maximum on
day 34, indicating that microbial biomass increased
and that microbial growth was substantial during the
first month of operation (Fig. 2j). The DNA concentration increased in accordance with the increase of rCH4
(Fig. 2f) and the decrease of rsCOD (Fig. 2i). The maximum DNA concentration, the maximum rCH4 , and the
minimum rsCOD occurred at approximately the same
time. These data suggest that the leachate DNA concentration can be correlated to the levels of hydrolytic
and methanogenic activities before day 34.
The change of rCH4 after day 34 corresponded well
with the decrease in the leachate DNA concentration,
presumably due to the net decay of microbial biomass.
Approximately 11% of the maximum DNA concentration observed on day 34 remained on day 178, whereas
the rCH4 on day 178 was 7% of the corresponding value
on day 34. Therefore, the DNA concentration in the
leachate appears to be a good indicator of the level
of methanogenesis in MSW biodegradation. However,
the DNA concentration did not reflect the rsCOD trend
during this period.
4.5
Relationships between measured variables
As discussed above, changes in three variables,the generation rate of CH4 (Fig. 2f), the change rate of sCOD
(Fig. 2i), and the long-term compression ratio (Fig.
2d), were indicative of the three long-term biodegradation phases as identified by Fei and Zekkos (2012).
Their relationships are presented in Table 1.
The transition phase started on day 0 and lasted to
day 33. In this period, rsCOD dropped from its maximum
to its minimum value, indicating accumulation and
subsequent consumption of sCOD hydrolyzed from
biodegradable waste. Meanwhile, rCH4 increased from
0 to its peak value, indicating a rapid increase in
methanogenesis. The maximum rCH4 coincided with
the minimum rsCOD , indicating that sCOD in the
leachate was the substrate for methanogenesis. Given
a relative low VCH4 during this time period and a
still growing methanogen population, methanogenesis
was assumed to be the rate-limiting step. At the same
time, the reduction of rsCOD accompanied an increase
of CLT . It was concluded that hydrolysis of biodegradable waste was the cause of biodegradation-induced
long-term compression of the specimen. However,
hydrolytic and methanogenic microorganisms were
still proliferating and most of the biodegradable waste
325
Table 1. Relationships between rCH4 , rsCOD and CLT in three

biodegradation phases.
Transition
phase
Period
rCH4
Active
biodegradation
phase
Residual phase
Day 1 to 33
Day 33 to 70
Day 70 to 250
Increase from
0 to max.
value
Decrease from
max. value to
almost 0
Remain
constant
around 0
rsCOD
Decrease from
max. value to
min. value
Increase from
min. value to
almost 0,
then remain
constant
Remain
constant
around 0
CLT
Increase from
0 to 50% of
max. value
Increase to
max. value,
then decrease
Generally
decrease
Rate
Methanogenesis Hydrolysis or
limiting
biodegradable
factor
waste
Biodegradable
waste
had not been hydrolyzed yet, and thus, the highest CLT
had not been reached.
The system entered the active biodegradation phase
around day 33 and this phase continued to approximately day 70. The change rate of sCOD varied from
its minimum value to zero and fluctuated afterward
around the zero value, suggesting that accumulated
soluble biodegradable substances from the transition
phase were completely consumed, and any further
generated sCOD was quickly converted to CH4 . In
addition, both the rCH4 and DNA concentration in the
leachate decreased to less than 20% of their peak
values by day 70, indicating insufficient sCOD to
sustain peak methanogenic activity. Because of that,
hydrolysis of MSW to sCOD was the rate-limiting
step and the highest hydrolysis rate of this test was
expected between day 33 and day 70 (Barlaz et al.,
2010). Corresponding to the accelerated hydrolysis of
biodegradable waste, high CLT values were measured
during this period, confirming that hydrolysis activity promoted long-term compression and caused an
increase in CLT . By the end of the active biodegradation phase on day 70, most of the VCH4 measured on
day 250 had been generated already, suggesting a high
degree of biodegradation during this period.
After day 70, the specimen entered the residual
phase of biodegradation. Both rsCOD and rCH4 were
reduced to negligible values. The availability and
accessibility of biodegradable waste were considered
to be the controlling factors of the biodegradation process during this phase. The CLT decreased gradually
in response to the very low hydrolysis of waste until
day 204.
4.6
here may not be representative of universal field

conditions. More specimens with initial and operational conditions are being tested. The comprehensive datasets generated by current and future work
is intended to improve our knowledge of the MSW
biodegradation process in bioreactor landfills.
Future works
A single biodegradation test of a MSW specimen was

analyzed in this study, and thus the results presented
CONCLUSIONS
MSW biodegradation in a bioreactor landfill was simulated in a laboratory experimental setup that allowed
the characterization of physical, geotechnical, chemical and microbial changes of MSW. Leachate was
recirculated and the biodegradation process was monitored for 250 days. MSW biodegradation resulted
in changes in physical, geotechnical and biochemical
properties. The total volume of the specimen reduced
13%, whereas the total saturated unit weight increased
steadily with time after adjustments due to physical
processes. The long-term compression ratio varied
over time with a maximum value of 0.125. Biogas
consisted of 55% CH4 and 45% CO2 . 425 L CH4
was generated cumulatively and the peak rCH4 was
10.3 L/day.The maximum concentrations of sCOD
and in sCOD in the leachate were 4,190 mg O2 /L
and 990 mg O2 /L, respectively. The maximum DNA
concentration in the leachate was 8,610 ng/ml.
DNA concentration in the leachate was indicative of the levels of hydrolysis and methanogenesis.Hydrolysis of biodegradable waste, which was
described by two parameters that were readily obtained
by monitoring biogas and leachate, rCH4 and rsCOD ,
induced long-term compression of the waste specimen, as reflected by changes in CLT . These three
variables were found to likely be indicative of the
specimen degradation state.
ACKNOWLEDGEMENT
The authors would like to thank Savannah Tibbetts
for assisting with biogas and leachate measurements,
Andhika Sahadewa for assisting with MSW characterization and specimen preparation, and Adam Smith and
Tara Clancy for their suggestions on analyses methods.
REFERENCES
APHA (2005) Standard Methods for the Examination of
Water and Wastewater, Washington, DC, American Public
Health Association.
Bareither, C. A., Benson, C. H., Edil, T. B. & Barlaz, M.
A. (2012) Abiotic and Biotic Compression of Municipal
Solid Waste Journal of Geotechnical and Geoenvironmental Engineering, 138, 877888.
Barlaz, M. A., Schaefer, D. M. & Ham, R. K. (1989) Bacterial population development and chemical characteristics
of refuse decomposition in a simulated sanitary landfill.
Applied and Environmental Microbiology, 55, 5565.
Barlaz, M. A., Staley, B. F. & de los Reyes, F. L.
(2010) Anaerobic Biodegradation of Solid Waste. IN
326
MITCHELL, R. & GU, J. (Eds.) Environmental Microbiology. Hoboken, NJ, Wiley-Blackwell.

Batstone, D. J., J. Keller, I. Angelidaki, S. V. Kalyuzhnyi, S.
G. Pavlosthathis, A. Rozzi, W. T. M. Sanders, H. Siegrist,
and V. A. Vavilin (2002) Anaerobic Digestion Model No.
1. IWA Publishing, Scientific and Technical Report No. 13.
Cornwall, UK.
Bookter, T. J. & Ham, R. K. (1982) Stabilization of solidwaste in landfills. Journal of the Environmental Engineering Division-ASCE, 108, 10891100.
Cossu, R. & Raga, R. (2008) Test methods for assessing
the biological stability of biodegradable waste. Waste
Management, 28, 381388.
EPA (2009) Project Development Handbook Landfill
Gas Energy Basics. United States Environmental Protection Agency, Landfill Methane Outreach Program,
Washington, DC 20460.
EPA (2010) Green power from landfill gas helping build
a sustainable energy future while improving the environment. EPAs Landfill Methane Outreach Program,
December 2010.
EPA (2011a) Municipal Solid Waste Generation, Recycling, and Disposal in the United States: Facts and Figures for 2010. United States Environmental Protection
Agency Solid Waste and Emergency Response (5306P).
Washington, DC 20460.
EPA (2011b) U.S. Greenhouse Gas Inventory. United States
Environmental Protection Agency Solid Waste and Emergency Response. Washington, DC.
EuroStat (2011) Municipal waste statistics. http://epp.
eurostat.ec.europa.eu/statistics_explained/index.php/
Municipal_waste_statistics.
Fei, X. & Zekkos, D. (2012) Factors Influencing Long-Term
Settlement of Municipal Solid Waste in Laboratory Bioreactor Landfill Simulators. Journal of Hazardous, Toxic,
and RadioactiveWaste, published online ahead of printing.
Gawande, N. A., Reinhart, D. R. &Yeh, G. T. (2010) Modeling
microbiological and chemical processes in municipal solid
waste bioreactor, part I: Development of a three-phase
numerical model BIOKEMOD-3P. Waste Management,
30, 202210.
Ham, R. K. & Bookter, T. J. (1982) Decomposition of solidwaste in test lysimeters. Journal of the Environmental
Engineering Division-Asce, 108, 11471170.
Huang, L. N., Zhou, H., Zhu, S. & Qu, L. H. (2004) Phylogenetic diversity of bacteria in the leachate of a full-scale
recirculating landfill. Fems Microbiology Ecology, 50,
175183.
Kim, J. & Pohland, F. G. (2003) Process enhancement in
anaerobic bioreactor landfills. Water Science and Technology, 48, 2936.
Kjeldsen, P., Barlaz, M. A., Rooker, A. P., Baun, A., Ledin,
A. & Christensen, T. H. (2002) Present and long-term
composition of MSW landfill leachate: A review. Critical Reviews in Environmental Science and Technology,
32, 297336.
McDougall, J. (2007) A hydro-bio-mechanical model for
settlement and other behaviour in landfilled waste.
Computers and Geotechnics, 34, 229246.
Mehta, R., Barlaz, M. A.,Yazdani, R., Augenstein, D., Bryars,
M. & Sinderson, L. (2002) Refuse decomposition in the
presence and absence of leachate recirculation. Journal of
Environmental Engineering, ASCE, 128, 228236.
Pinto, A. J. & Raskin, L. (2012) PCR Biases Distort Bacterial
and Archaeal Community Structure in Pyrosequencing
Datasets. PLoS ONE, 7, e43093.
Pohland, F. G. & Alyousfi, B. (1994) Design and operation of
landfills for optimum stabilization and biogas production.
Water Science and Technology, 30, 117124.
Staley, B. F., de los Reyes, F. L. & Barlaz, M. A. (2011) Effect
of Spatial Differences in Microbial Activity, pH, and
Substrate Levels on Methanogenesis Initiation in Refuse.
Applied and Environmental Microbiology, 77, 23812391.
Staley, B. F., de los Reyes, F. L. & Barlaz, M. A. (2012) Comparison of Bacteria andArchaea communities in municipal
solid waste, individual refuse components, and leachate.
Fems Microbiology Ecology, 79, 465473.
Stroot, P. G., McMahon, K. D., Mackie, R. I. & Raskin, L.
(2001) Anaerobic codigestion of municipal solid waste
and biosolids under various mixing conditions - I. Digester
performance. Water Research, 35, 18041816.
Vavilin, V. A., Lokshina, L. Y., Jokela, J. P. Y. & Rintala, J. A.
(2004) Modeling solid waste decomposition. Bioresource
Technology, 94, 6981.
Wall, D. K. & Zeiss, C. (1995) Municipal landfill
biodegradation and settlement. Journal of Environmental
Engineering-Asce, 121, 214224.
Wang, Y. S., Byrd, C. S. & Barlaz, M. A. (1994) Anaerobic biodegradability of cellulose and hemicellulose in
excavated refuse samples using a biochemical methane
potential assay. Journal of Industrial Microbiology, 13,
147153.
Zekkos, D., Bray, J. D., Kavazanjian, E., Jr., Matasovic,
N., Rathje, E. M., Riemer, M. F. & Stokoe, K. H., II
(2006) Unit weight of municipal solid waste. Journal of
Geotechnical and Geoenvironmental Engineering, 132,
12501261.
Zekkos, D., Kavazanjian, E., Bray, J. D., Matasovic, N. &
Riemer, M. F. (2010) Physical Characterization of Municipal Solid Waste for Geotechnical Purposes. Journal of
12311241.
Zinder, S. H. (1993) Physiological ecology of methanogens.
IN FERRY, J. G. (Ed.) Methanogenesis: Ecology, physiology, biochemistry, and genetics. New York, Chapman &
Hall.
327

Leachate recirculation using horizontal trenches considering

the effect of MSW settlement
S.J. Feng & X. Zhang
Department of Geotechnical Engineering, Tongji University, Shanghai, China
ABSTRACT: Leachate recirculation in municipal solid waste (MSW) landfills offers economic and environmental benefits. Due to the weight of MSW and its biodegradation, a bioreactor landfill will experience more
rapid and complete settlement, which may cause the decrease of MSW void ratio. A new calculated model
of landfill leachate recirculation consisting of a horizontal trench considering MSW settlement is developed.
Based on the developed analysis method, the effects of MSW settlement on the leachate recirculation volume
and influence zone are presented. The influence zone of leachate recirculation considering the effect of MSW
settlement is smaller than the zone without considering the effect.
INTRODUCTION
Leachate recirculation (LR) was pioneered in the

United States in the 1970s as a means to enhance
degradation of landfilled wastes, degrade or immobilize harmful compounds within the waste mass,
and store excess leachate (Pohland, 1975). LR offers
many environment and economical benefits to MSW
landfills (Doran 1999, Mehta et al. 2002). The common leachate recirculation techniques are broadly
divided into surface and subsurface applications. Surface applications consist of: (1) direct application of
leachate or spray irrigation of leachate on the landfill
surface; and (2) surface ponding of leachate. Conventional subsurface application methods are (1) vertical
wells; and (2) horizontal trenches (Koerner 2000, Qian
et al. 2002, Haydar & Khire 2005).
For a leachate recirculation system design, an estimate of required leachate volume is necessary. Thus,
the flow of injected leachate in bioreactor landfills
should be analyzed. McCreanor (1998) used SUTRA2D model to simulate leachate hydrodynamics using
leachate recirculation system consisting of a single
horizontal trench with 2-m-wide by 1-m-deep. Dora
(1999) recirculated leachate using leachate recirculation system consisting of horizontal trenches at
the Crow Wing County MSW landfill (CWCL). The
leachate flux estimated by HYDRUS-2D was compared to that obtained from monitoring laterals (the
laterals which were used to recirculate leachate under
gravity drainage) in a leachate recirculation system.
Haydar & Khire (2005) presented a numerical study of
leachate recirculation consisting of horizontal trenches
using the finite-element saturated/unsaturated flow
model HYDRUS-2D. However, the effect of waste
structure on transfers, in particular, its high compressibility, was not considered in these studies. Thus, the
calculated recirculation volume will be significantly

overestimated.
In this paper, a calculated model of landfill leachate
recirculation consisting of a horizontal trench considering MSW settlement developed. Based on the
developed analysis method, the effects of MSW settlement on the leachate recirculation volume and
influence zone are presented.
2
2.1
NUMERICAL MODELING METHODOLOGY

Leachate flow in saturated zone of landfill
In saturated zone of landfill, it is assumed that MSW is

anisotropic and homogeneous. The Richards equation
for saturated water flow is adopted:
where hp is the pressure head, s is the specific storage, t is the time, Kx , Ky and Kz are the saturated
hydraulic conductivity of MSW in x, y and z directions,
respectively.
2.2
Leachate flow in unsaturated zone of landfill
In unsaturated zones of landfill, assuming that MSW

is anisotropic and homogeneous, the modified Darcys
law for leachate is used to describe the vertical flow
in MSW. Based on the law of conservation of mass,
yielding
329
where is the volumetric water content, only the vertical deformation of MSW is considered and the
variation of the density of leachate is ignored, thus
the right part of Eq. (2) can be rewritten as
According to the compression model of MSW

developed by Elagroudy (2008), the relationship
between the vertical deformation of MSW and time
is developed as follows:
where A and B are the calculation parameters, substituting Eq. (4) into Eq. (3) yields
According to the unsaturated water flow law, the left

part of Eq. (2) can be rewritten as
where H is the total pressure head, hp is the pressure

head, K(hp )x , K(hp )y and K(hp )z are the unsaturated
hydraulic conductivity of MSW in x, y and z directions,
respectively.
Through Eqs. (2,56), the new flow governing
equation of leachate is developed as follows:
Figure 1. Schematic of modeled a horizontal trench system

with boundary and initial conditions.
where k(hp ) is the relative permeability of MSW, when

MSW is saturated, hp 0, k(hp ) = 1, = 1 and = 0;
when MSW is unsaturated, hp 0, k(hp ) = 0, = 0
and = 1.
Figure 1 schematically illustrates the landfill and its
relevant initial and boundary conditions. In a homogeneous medium, the resulting liquids distribution
around the source would be symmetrical to the xz
plane across the middle of the source, thus only half
of the horizontal source was modeled to reduce computational time. In this calculated model, the initial
moisture content of MSW and the initial pressure head
is assumed as 0 and hp0 , respectively. Hence, the initial
condition can be written as
Simulations were conducted for a l w (1 1 m2 )

source for modeling a trench system, l is the horizontal
trench width and w is the horizontal trench height.
In this study, pressure-specified liquids addition was
simulated by specifying hydrostatic pressure at nodes
corresponding to the horizontal trench. Thus
where C(hp ) is the specific moisture capacity.

2.3 Two-dimensional model of leachate
recirculation using a horizontal trench
The lateral boundary of the landfill is assumed as
In this paper, the method of leachate recirculation is the

horizontal trench. Due to the main two-dimensional
migration for leachate in a landfill, the 3D form governing equations (1) and (7) can be simplified as
follows:
The upper boundary conditions used for the top

surfaces hypothesized to initial pressure head.
The lower boundary condition is assumed as
330
The influence of decomposition and the weight of

waste on hydraulic conductivity, through its impact
on void volume, can be accommodated as follows
(McDougall 2007).
where Ksat is the saturated hydraulic conductivity of

MSW, e is the void ratio of MSW, a and b are the
calculation parameters. According to the settlement
model of MSW developed by Elagroudy et al. (2008),
the relationship between the void ratio of MSW and
time is presented as follows:
where e0 is the initial void ratio of MSW, substituting

Eq. (15) into Eq. (14) yields
For the unsaturated zone of the landfill, the relationship between the unsaturated hydraulic conductivity
of MSW and the water pressure head is given by Van
Genuchten (1980).
Combining the initial and boundary conditions
[Eqs. (913)], the formulas of permeability coefficients [Eqs. (1416)], the governing equation (Eq. 8)
can be solved using the finite differential.
3
Figure 2. Pressure head distribution in the landfill at steady

state: (a) considering the effect of MSW settlement; (b)
without considering the effect of MSW settlement.
RESULTS
In order to analyze the results from the method presented, detailed calculations were conducted for a
bioreactor landfill with different variables. The case
for the analysis adopts the following parameters:
the Van Genuchten parameters (n = 4, m = 1 1/n,
= 10), porosity s = e/(1 + e), residual moisture
content r = 0, a = 108 , b = 4.9, A = 0.25, B = 0.045,
e0 = 4 and Ha = 5 m.
Figure 2 shows the pressure head distribution in the
landfill at steady state with effect of the MSW settlement. Each contour plot in Fig. 2 shows a series
of isoclines that correspond to pressure head ranging from initial pressure head to a maximum value
equal to injection pressure head (5 m). The added
leachate moved away from the trench with time under
the influence of injection pressure, gravity, and capillary forces. It can be noticed that the influence zone of
leachate recirculation considering the effect of MSW
settlement in Fig. 2(a) is smaller than the value without considering the effect in Fig. 2(b). This is mainly
attributed to the effect of MSW settlement on hydraulic
properties, and with the increase of time, the void
ratio and hydraulic conductivity of MSW gradually
decrease.
Figure 3 shows the water content distribution in the
landfill at steady state with effect of the MSW settlement. Each contour plot considering the effect of
MSW settlement in Fig. 3(a) shows a series of isoclines
that correspond to moisture content ranging from initial moisture content (0.15) to a maximum value equal
to waster porosity (0.8). While each contour plot
without considering the effect of MSW settlement in
Fig. 3(b) shows a series of isoclines that correspond to
moisture content ranging from initial moisture content
(0.16) to a maximum value equal to waster porosity
(0.77). Further analyses show that the liquids front
continually expanded and moved away from the trench
under the influence of gravity, injection pressure, and
capillarity as the leachate were progressively added
to the trench. The leachate, for the most part, moved
as a sharp wetting front, and the individual isoclines
were close to each other primarily because of the
use of unsaturated properties. The saturation isoclines
indicate that the influence zone has a range of saturations propagating outwards from a saturated zone
near the trench. A similar observation was reported
by McCreanor (1998) who used unsaturated hydraulic
properties determined by Korfiatis et al. (1984) in
laboratory-scale experiments.
The horizontal spacing between adjacent horizontal trenches is one of the key parameters in leachate
recirculation system. Jain et al. (2010) regard lateral extent measured along the horizontal plane of
the bottom of the trench as the horizontal spacing.
As a matter of fact the maximum influence width
occurred below the bottom of the trench. So in this
331
Figure 4. Variation of influence zone with time.
Figure 3. Water content distribution in the landfill at steady

state: (a) considering the effect of MSW settlement; (b)
without considering the effect of MSW settlement.
paper the wetted zone corresponding to degree of saturation, S 90%, would be regarded as influence zone.
In the influence zone, the maximum influence width
and influence deep was taken as the influence width
and deep.
Figure 4 presents the variation of impact zone with
time. It can be found that the influence zone of leachate
recirculation gradually increases with the increase of
time, and the influence deep is greater than the influence width. After the simulated leachate injection was
started, the influence width increased until a steadystate was reached after which the influence width did
not increase. The time required to reach the maximum influence width is about 30 days once continuous
leachate injection was started at a constant injected
pressure head. For a given injected pressure head,
considering the settlement of MSW the steady-state
is attained more quickly compared to when without
considering the settlement of MSW.
Figure 5 illustrates the variation of leachate recirculation volume with time. It can be seen that the
leachate recirculation volume gradually increases with
the increase of time, and the leachate recirculation volume without considering the effect of MSW settlement
is larger than that considering the effect. Consequently,
Figure 5. Variation of leachate recirculation volume with

time.
the design engineer is prepared to design leachate

recirculation system, the effect of MSW settlement
cannot be neglected.
4
CONCLUSIONS
A new calculated model of landfill leachate recirculation consisting of a horizontal trench considering
MSW settlement developed. The time required to
reach the maximum influence width is about 30 days,
and this steady-state is attained more quickly considering the settlement of MSW compared to when without
considering the settlement of MSW. The influence
zone of leachate recirculation considering the effect
of MSW settlement is smaller than the one without
considering the effect.
REFERENCES
Doran, F. 1999. Lay leachate lay. Waste Age, April: 7479.
Elagroudy, S. A., Abdel-Razik, M. H., Warith, M. A. &
Ghobrial, F. H. 2008. Waste settlement in bioreactor
landfill models. Waste Management, 28(11): 23662374.
332
Haydar, M. M. & Khire, M. V. 2005. Leachate recirculation

using horizontal trenches in bioreactor landfills. Journal of Geotechnical and Geoenvironmental Engineering,
131(7): 837847.
Koerner, R. 2002. Leachate recycling loading to bioreactor landfills for the rapid degradation of municipal
solid waste. Proc., ESD solid waste management conf.,
Mich., 39.
Korfiatis, G., Demetracopoulos, A., Bourodimos, E. & Nawy,
E. 1984. Moisture transport in a solid waste column.
Journal of Environmental Engineering, 110(4): 789796.
McCreanor, P. 1998. Landfill leachate recirculation systems:
mathematical modeling and validation. Ph.D. Dissertation, University of Central Florida, Orlando, FL.
McDougall, J. 2007. A hydro-bio-mechanical model for

settlement and other behaviour in landfilled waste.
Computers and Geotechnics, 34: 229246
Mehta, R., et al. 2002. Refuse decomposition in the presence and absence of leachate recirculation. Journal of
Environmental Engineering, 128(3): 228236.
Pohland, F. O. 1975. Accelerated solid waste stabilization and
leachate treatment by leachate recycle through sanitary
landfills. Progress in water technology 7: 753765.
Qian, X., Koerner, R. & Gray, D. 2002. Geotechnical aspects
of landfill design and construction, Prentice-Hall, N.J.
Van Genuchten, M. Th. 1980. A closed-form equation for
predicting the hydraulic conductivity of unsaturated soils.
Soil Science Society of American Journal 44: 892898.
333

2011 East Japan earthquake and tsunami Geoenvironmental challenges

T. Katsumi, T. Inui & A. Takai
Kyoto University, Kyoto, Japan
K. Endo & H. Sakanakura

National Institute for Environmental Studies, Tsukuba, Ibaraki, Japan
T. Yasutaka
National Institute of Advanced Industrial Science and Technology, Tsukuba, Ibaraki, Japan
Y. Otsuka
Okumura Corporation, Morioka, Iwate, Japan
H. Suzuki
Nippon Koei, Co., Ltd., Tokyo, Japan
H. Sakamoto
Fujita Corporation, Tokyo, Japan
M. Okawara
Iwate University, Morioka, Iwate, Japan
H. Imanishi
Tohoku Institute of Technology, Sendai, Miyagi, Japan
ABSTRACT: This paper summarizes the main geoenvironmental issues caused by the 2011 earthquake off
the Pacific coast of Tohoku, and presents several approaches against these issues. Major geoenvironmental
issues include (1) generation of disaster wastes and tsunami deposits, (2) contamination with salt, (3) land
subsidence, and (4) geoenvironmental contamination with nuclides. Countermeasures against soils and wastes
contaminated with nuclides require the approach from several geoenvironmental viewpoints, such as the design
and performance evaluation of containment system for contaminated materials. Utilization of disaster debris for
the recovery works after proper treatment have been proposed and conducted at the disaster affected areas, in
particular at the areas subsided by this disaster. Characterization of waste mixed soils obtained from the disaster
debris has been conducted to evaluate the applicability to construction materials for disaster recovery.
INTRODUCTION
A massive earthquake of magnitude 9.03 (Mw),

referred to as the 2011 earthquake off the Pacific coast
of Tohoku, occurred at 14:46 on March 11, 2011, as
the most powerful known earthquake ever to have
hit in Japan, as well as one of the five most powerful earthquakes in the world since modern recordkeeping started. The earthquake triggered tsunami
that reached heights of up to 40.5 m at maximum
in Miyako city, Iwate Prefecture and travelled up to
10 km inland in the Sendai plane in Miyagi Prefecture.
More than 15,800 deaths and 2700 people missing
across twenty prefectures were reported by the national
government. Totally and half collapsed buildings
counted more than 129,000 and 254,000 respectively.
The recovery from this catastrophic disaster is a crucial issue not only for the affected areas but for all over
Japan.
This earthquake and subsequent tsunami caused
several serious geoenvironmental problems mainly
in the coastal area of the Tohoku and North-Kanto
Regions in Japan. These geoenvironmental problems
may include (1) generation of disaster wastes and
tsunami deposits, (2) contamination with salt, (3)
land subsidence, and (4) geoenvironmental contamination with nuclides. This paper summarizes these
geoenvironmental issues and challenges which have
been conducted by the several researchers and engineers including the authors. These challenges require
the knowledge about several coupled phenomena of
physics, chemistry, and biodegradation.
335
Figure 2. Composition of disaster debris in Iwate and

Miyagi Prefectures (Iwate Prefecture 2011, Miyagi Prefecture 2011, Inui et al. 2012).
Figure 1. Location of Prefectures of concern (Inui et al.

2012).
GEOENVIRONMENTAL ISSUES
Approximately 20,000 Gg (20,000,000 ton) of disaster

debris was generated, through the 2011 earthquake off
the Pacific coast of Tohoku and subsequent tsunami.
In addition, about 10 million m3 of tsunami deposits
transported by the tsunami requires proper treatment
as well. These numbers are comparable to those generated at the previous catastrophic disasters such as 2010
Haiti earthquake, 2008 Sichuan earthquake in China,
etc. (Brown et al. 2011). Since it is geographically and
economically unrealistic to construct new waste disposal facilities having a sufficient capacity to accept
these wastes, which corresponds to several times of
the annual generation of municipal solid waste in each
local municipalities, utilization of these materials is
required. Since the national government decided that
treatment of debris and tsunami deposits should be
completed by 2014 March, proper treatment and utilization of the materials treated from these disaster
wastes have strongly been expected.
Treatment of the disaster wastes generated by this
earthquake is a big challenge, since such a large
amount of mixed wastes had never been generated in Japan before. A significant fraction of these
wastes corresponds to the soils which are the tsunami
deposits. Figure 2 shows the fractions of generated
waste materials in Iwate and Miyagi prefectures. The
important points are not only the amount of wastes
but also the mixed types of wastes and spatial dependency. Figure 3 indicates a typical flow of treatment
of mixed wastes. Currently, several private companies
have proposed their own treatment systems, and have
been engaged in the treatment of such mixed wastes at
each local municipality. Evaluation of these treatment
systems from the viewpoint of treatment efficiency
Figure 3. Location of Prefectures of concern and

Fukushima Daiichi Nuclear Power Plant (Inui et al. 2012).
and applicability to construction materials will be a

next challenge.
At the affected areas, ground subsidence has
occurred due to the earthquake. For example, maximum lateral and vertical movements were recorded
5.3 m and 1.2 m respectively at Oshika Peninsula
in Miyagi prefecture. Many other places along the
Pacific coast from Iwate to Fukushima prefectures suffered more than 0.5 m ground subsidence (Geospatial
Information Authority of Japan 2011). These ground
subsidence are considered permanent, and caused secondary problems such as difficulty in recovery and
resettlement and sanitary problems because the stored
water was not able to go out from the area. Therefore,
embankment is necessary for the recovery both for
residential, commercial, and green areas. Utilization
of the materials treated from the disaster wastes and
tsunami deposits have been strongly expected. In particular, the geotechnical utilization of the soil fraction
in disaster debris and tsunami deposits has become a
big challenge for geotechnical engineers since temporal and spatial variations in the geotechnical properties of waste-mixed soils should be considered if a
large amount of waste-mixed soil are used for constructing embankments and levees against tsunami in
reconstruction projects.
336
Figure 4. Green park construction for disaster recovery

using waste materials (Katsumi et al. 2012).
Table 1. Criteria for the materials treated from disaster
wastes used for embankment (MLIT 2012b).
Item
Criteria
Maximum grain size

Cone index
Salt content
Electrical conductivity
pH
Swelling (at CBR soaking)
<300 mm
>400 kN/m2
<1 mg/g (in principle)
<200 mS/m
6.09.0
>3%
Figure 5. Histogram of the salinity of the tsunami deposits

collected at farmlands in Fukushima Prefecture (Takai et al.
2012).
To promote the utilization of disaster debris in

recovery works, the Japanese Ministry of Land, Infrastructure, Transports and Tourism (MLIT) has established two technical guidelines to construct 1) parks
and green spaces as a redundancy zone against huge
tsunamis in which embankments be constructed using
disaster wastes (MLIT 2012a) as shown in Figure 4
and 2) fill embankments at the areas where ground
subsidence occurred significantly due to the earthquake (MLIT 2012b). Green areas had a positive
effect on reducing the energy of tsunami and trapping the flowing obstacles such as cars, while some
trees did not have sufficient root depth, which resulted
in insufficient resistance against tsunami. Therefore,
embankment construction for green areas is considered advantageous to provide a sufficient distance
from groundwater table to the ground surface.
The material criteria proposed by the above guideline are shown in Table 1. There have been several
discussions about these criteria. As for the salt content, to prevent the corrosion of underground steel
materials such as steel piles and adverse impacts on
vegetations, a salt content lower than 1 mg/g is generally required. However, this technical guideline might
consider the situation to construct embankments at the
tsunami affected areas, in which the salt concentration is already high because they are located close to
the ocean, although significant portion of the materials such as tsunami deposits may exceed this criteria
of salt content as shown in Figure 5 indicating the
results of salt content of tsunami deposits byTakai et al.
(2012). It was anticipated that the strict application
of this criteria might limit the utilization of materials. Therefore, this technical guideline also expects
the utilization of the materials having a salt content
higher than 1 mg/g. Another concern of such materials is the potential of degradability which may cause
Figure 6. Relationship between the fine fraction content and

ignition loss of tsunami deposits (Inui et al. 2012, based on
Katsumi et al. 2011).
the generation of gas and leachate and ground settlement. In the field of waste management, ignition loss
has been used to evaluate the degradability. However,
the ignition loss is not only reflected by the materials
which will be degradated but also other components
such as hydration products. For example, if the materials (soils) are rich in fine fraction, they will exhibit
higher ignition loss as shown in Figure 6. Therefore,
it is required to establish the criteria to evaluate the
intactness of materials.
Another serious geoenvironmental issue is the management of the radioactive contamination of surface
soils caused by the accident at Fukushima Daiichi
Nuclear Power Plant, which includes fall-out radioactive materials such as 134 Cs and 137 Cs. The tsunami
generated by the earthquake caused great damage
to widespread coastal areas in the Tohoku Region
including Fukushima prefecture where the Fukushima
Daiichi Nuclear Power Plant is located. On March 14,
2011, hydrogen explosions occurred in both the No. 3
and No. 1 reactors. As a result of the explosions, large
amounts of radioactive materials including 131 I, 134 Cs,
and 137 Cs were released into the atmosphere.
Sewage and waste materials may contain radioactive substances concentrated through natural and
337
Figure 7. Leaching ratio of radioactive cesium from MSW

incinerator ashes (NIES 2012).
Figure 9. Landfill of MSW fly ash lower than 8000 Bq/kg

with higher dissolution (edited from NIES 2012).
Figure 8. Landfill of MSW bottom ash lower than

8000 Bq/kg with lower dissolution (edited from NIES 2012).
artificial processes in some areas in the Tohoku

and Kanto regions in Japan. The wastes which contain radioactive materials lower than 8,000 Bq/kg are
allowed to be disposed of at existing MSW (municipal
solid waste) landfill sites, according to regulations.
However, these regulations only consider the exposure of radioactive substances to workers, but not the
fate and transport of these substances in the subsurface, including landfill sites. For example, the leaching
ratios are very different between MSW bottom ash and
fly ash, as shown in Figure 7, even though they may
contain the same level of concentration of radioactive cesium. While radioactive cesium in MSW bottom
ash may be stable in terms of the leaching (only 1.6%
leaching ratio), 79.8% cesium from fly ash, and 62.5%
from granulated fly ash, will leach out (NIES 2012).
In addition, it should be noted that the basic concept
of Japanese MSW landfills is to allow the infiltration
of rainfall into the waste layer, and to accelerate the
biodegradation under semi-aerobic condition, as well
as to wash out the contaminants that may be dissolved.
At the landfill sites designed following such basic concept, if the leaching ratio of radioactive cesium in the
disposed wastes is high, it will be important to evaluate its fate and transport, because the existing landfill
leachate treatment facilities will not be able to treat
the radioactive cesium. Installing soil layers to act as a
sorption layer against radioactive cesium is required
for MSW bottom ash as shown in Figure 8, while
soil layers both for hydraulic barrier and for sorption
are required for MSW fly ash which exhibits higher
leaching potential as shown in Figure 9. Since these
sorption layers may be constructed over the existing
Figure 10. Disposal scenario of radioactive contaminated

soils and wastes (MOE 2011).
waste layer, differential settlement and chemical compatibility should be taken into account. Geosynthetic
reinforcement such as geogrids and/or geosynthetic
barriers such as GCLs may be considered for the safe
disposal of such waste materials. Researches on the
applications of such geosynthetics against differential
settlement and chemical compatibility are expected
(Katsumi et al. 2012).
For the wastes higher than 100,000 Bq/kg and waste
soils and plants discharged through decontamination
works, the scenario consisting of storage at a temporary yard, then at an interim storage facility, and
finally at a permanent disposal site, has been decided
by the national government (Figure 10). At some temporary yards, GCLs are used as a containment barrier.
At the interim storage facilities, as well as at the permanent disposal sites, establishment of design method
including the use of geosynthetics for reinforcement,
hydraulic, and chemical barrier, filtration, and other
functions, is strongly anticipated.
338
Figure 11. Disaster waste piled up in a temporal storage site

(Inui et al. 2012).
Figure 12. Typical treatment procedure of waste mixed soil

(Morita et al. 2012).
CHARACTERIZATION OF WASTE
MIXED SOIL
Evaluating the engineering properties of tsunami

deposits and soils separated and/or obtained from disaster waste mixture is an important issue to ensure
the performance such as stability of the earthen structures if they are used in geotechnical applications in
recovery works. However, no experience or knowledge
on the treatment and geotechnical properties of such
waste mixed soils shown in Figure 11 has been accumulated. Evaluation of the engineering properties of
such soils obtained from the disaster wastes have therefore started to be carried out by some researchers and
engineers including the JGS (Japanese Geotechnical
Society) Geoenvironmnetal Technical Committee on
the 2011 East Japan Earthquake and Tsunami chaired
by the first author.
Typical treatment procedure of waste mixed soil
implemented at the disaster waste treatment facilities
was shown in Figure 3. Wastes with large sizes are
removed from the mixture at a first temporary storage site (rough separation), and further separation is
conducted using trommels, vibrating screens, or other
machineries/equipments at a secondary storage site
(secondary separation). Morita et al. (2012) sampled
the soils and soil-waste mixtures from several temporary storage sites, and conducted the experiments to
evaluate their basic properties such as particle density,
particle size distribution, ignition loss, waste composition, compaction, and cone index. Non-separated soil
(sample A-1, A-3) and roughly separated soil (A-2)
were taken at a first temporary storage site in Town
A, and non-separated soil (B-1, B-2) were collected
at a storage site in Town B. Secondary separated soil
(C-1) was taken at a secondary storage site in Town
C. Roughly separated soil (D-1), which was passed
through a 20-mm opening screen, was taken at a temporally storage site in Town D. The samples used in
Morita et al. 2012) are indicated in Figure 12 in terms
of the treatment system.
Figure 13. Compaction curves of the soils obtained from

disaster waste. Numbers are the content of combustible
materials in percentage (Morita et al. 2012).
One of the important considerations is the effect of

combustible substances on the engineering properties,
since these combustibles may be deteriorated, resulting in the emission of gas and leachate and ground
settlement. Ignition loss test conducted on the sample
sieved by a 2 mm opening screen according to JIS A
1226 does not consider all the organic matters in the
soil-waste mixture mainly consisting of combustible
wastes such as wood chips, paper scraps or plastic, etc.
Because most of these combustible materials exhibit
the particle sizes larger than 2 mm. Therefore, Morita
et al. (2012) conducted manual separation test in which
the fraction over 2 mm was separated into combustible,
noncombustible, and soil particles.
Compaction characteristics of each sample are
shown in Figure 13. The numbers listed in left side
of each compaction curve mean the combustible content larger than 2 mm of each sample. From this
result, the samples of high combustible content such as
A-1 and D-1 exhibited a high optimum water content
for obtaining the maximum dry density. Besides, the
maximum dry density values of such samples are low
339
(1.1 to 1.4 g/cm3 ) compared to other samples (1.7 to

2.0 g/cm3 ). This is because the samples which contain
large amounts of combustible substances cannot be
properly compacted, and the densities of combustible
substances are low. The samples which contain smaller
volumes of combustible substances exhibited lower
optimum water contents and higher maximum dry densities, and have clear peaks of compaction curves.
Therefore, it is expected that such samples can be
used as geo-materials with their strength values greatly
improved by the compaction. Further investigation
should evaluate how the waste and combustible matters will affect the engineering properties for a long
term by their decomposition if the waste mixed soils
be used in geotechnical applications.
4
CONCLUDING REMARKS
The geoenvironmental issues caused by the 2011 East

Japan earthquake and tsunami, such as generation of
disaster wastes and tsunami deposits, contamination
with salt, land subsidence, and geoenvironmental contamination with nuclides, are briefly summarized in
this paper. Countermeasures against soils and wastes
contaminated with nuclides require the geoenvironmental approaches, for which the accumulated knowledge is utilized as well as new efforts are required for
the design and performance evaluation of containment
system for the contaminated materials. Utilization of
disaster debris for the recovery works after proper
treatment has been proposed at the disaster affected
areas, in particular at the areas that suffer from subsidence. Characterization of waste mixed soils obtained
from the disaster debris that has been conducted
to evaluate the applicability to construction materials for disaster recovery requires further research in
understanding the coupled phenomenon of physical,
chemical, and biochemical effects.
ACKNOWLEDGEMENT
Contributions and discussions provided by the members of JGS (Japanese Geotechnical Society) Geoenvironmental Technical Committee on the 2011 East
Japan Earthquake and Tsunami chaired by the first
author are greatly appreciated. The authors would like
to express appreciation to Dr H. Sasaki (Iwate Prefectural Government) and Mr S.Yamagiwa (ditto) for their
great support to the works presented in this paper, as
well as Mr. K. Morita (Kyoto University) for his great
contribution conducting experimental works.
Inui, T., Yasutaka, T., Endo, K., and Katsumi, T. (2012):

Geo-environmental issues induced by the 2011 Off the
Pacific coast of Tohoku Earthquake and Tsunami, Soils
and Foundations (accepted for publication).
Iwate Prefecture (2011): Detailed Plan for the Disaster Debris
Treatment in Iwate Prefecture, http://www.pref.iwate.jp/
download.rbz?cmd=50&cd=34456&tg=3 (access on
2012 March 20, in Japanese).
Geospatial Information Authority of Japan (2011): Land subsidence caused by 2011 Tohoku earthquake and tsunami,
http://www.gsi.go.jp/sokuchikijun/sokuchikijun40003.
html (accessed on 2013 January 20, in Japanese).
Katsumi, T., Takai, A., Omine, K., Kazama, M., and Inui,
T. (2011): Physical and chemical properties of tsunami
deposits collected in the flooded area, Miyagi Prefecture, Proceedings of the Annual Symposium on Groundwater and Geo-environment, The Research Committee
of Groundwater and Geo-environment, pp. 8390 (in
Japanese).
Katsumi, T., Kotake, N., and Viswanadham, B.V.S. (2012):
Geosynthetics for environmental protection Compatibility and integrity, Geosynthetics Asia 2012 5th Asian
Regional Conference on Geosynthetics, D.T. Bergado
(ed.), pp. 4768.
Miyagi Prefecture (2011): Action Plan for the Disaster
Debris Treatment in Miyagi Prefecture (First Draft),
http://www.pref.miyagi.jp/haitai/shinsai/pdf/20110804
shorikeikaku-1.pdf (access on 2012 March 20, in
Japanese).
Ministry of Land, Infrastructure, and Transport (2012a):
Technical Guideline on the Development of Parks
and Green Spaces in the Reconstruction from the
off the Pacific coast of Tohoku Earthquake Disaster,
http://www.mlit.go.jp/common/000205823.pdf (access on
2012 March 30, in Japanese).
Ministry of Land, Infrastructure, and Transport (2012b):
Basic Concepts on the Utilization of Recycled Materials in Constructing Fill Embankment to Accelerate the Revitalization from the Earthquake Disaster,
http://www.mlit.go.jp/common/000205731.pdf (access
on 2012 March 30, in Japanese).
Ministry of Environment (MOE) (2011): Guidelines on the
Treatment of Waste Contaminated by Accidental Nuclides
(in Japanese).
Morita, K., Katsumi, T., Takai, A., and Inui, T. (2012):
Characterization of waste mixed tsunami sediments generated by the 2011 East Japan earthquake and tsunami,
Geo-Environmental Engineering 2012 Proceedings
of the 11th Japan-Korea-France-Canada Joint Seminar on Geoenvironmental Engineering, Caen University,
pp. 3342.
National Institute for Environmental Studies (2012): Proper
waste treatment and disposal considering the fate and
transport of radionuclide Ver. 2 (in Japanese).
Takai, A., Endo, K., Yasutaka, T., and Katsumi, T. (2012):
Distribution and characterization of tsunami sediments
in Fukushima Prefecture, Proceedings of the 47th
JGS Conference on Soil Mechanics and Geotechnical
Engineering, pp. 19511952 (in Japanese).
REFERENCES
Brown, C., Milke, M., and Seville, E. (2011): Disaster
waste management: A review article, Waste Management,
Vol. 31, pp. 10851098.
340

Composition and degradation effects on the compaction characteristics of

municipal solid wastes
H.F. Pulat & Y. Yukselen-Aksoy
Celal Bayar University, Manisa, Turkey
ABSTRACT: In order to protect environment and human health, MSW must be stored regularly and confidently.
For safe and economic MSW storage, significant processes are initialized (dumping, spread out, burying and
compacting). Compaction is the one of the most important stages of the storage process. Well compacted MSW
renders less volume than the same weighted uncompacted MSW. Furthermore, well compaction provides safer
storage area. In this study, the effect of composition, degradation, and energy on the compaction behavior of
artificially prepared and natural fresh and aged MSW samples were determined. The standard and modified
proctor tests were performed on the fresh and aged samples. Artificial samples were prepared in different
compositions to examine the effect of composition. In addition to the synthetic MSW samples, natural MSW
samples were provided from municipal landfill area of Manisa city.
INTRODUCTION
According to the significant growth in population

and economy, the amount of municipal solid waste
(MSW) is increasing enormously. In order to protect environment and human health, MSW must be
stored regularly and confidently. The use of disposing of trash in landfills is by far the most common
form of municipal waste disposal. After the MSW
are brought the landfill area, significant storage processes are initialized (dumping, spread out, burying,
and compacting). Compaction is the one of the most
important stages of the storage process. Compaction is
the densification of material by mechanical energy and
may include impact, vibratory and static compaction
methods (Wong 2009). The compaction process is
helpful in the development of the engineering properties of municipal solid waste. In parallel with the
compaction degree increment, cohesion and internal
friction angle values increase and hence, shear strength
of MSW increases. Increase in the mobilized shear
angle of MSW was investigated by Manassero et al.
(1996). Figure 1 presents the differences between shear
angles for compacted and non-compacted MSW. The
more dense compacted solid waste, the steeper MSW
slope can be designed at the landfill (Fakharian &
Taherzadeh 2004, Dixon & Jones 2005, Blight 2008,
Turer & Turer 2011). As a result of this, service life of
the landfill will be longer and more economical (Ham
et al. 1978). At the end of the compaction procedure,
due to the decrease of void ratio, the permeability of
MSW also decreases (Staub et al. 2009). Controlled
compaction process reduces the compressibility of
municipal solid waste thus; uncontrolled settlements
could be prevented by an amount (Bareither 2010).
Figure 1. Effect of compaction on the mobilized shear angle

of municipal solid waste (Manassero et al. 1996).
Compaction affects shear strength, slope stability,

hydraulic conductivity and compressibility properties
of MSWs. However, many parameters also affects the
compaction behavior of MSWs. Compaction characteristics of MSW is extremely dependent on water
content, waste structure (anisotropy), degree of degradation, particle size distribution, layer thickness and
waste composition (Langer 2005, Dixon & Jones
2005).
It can be generally accepted that the more compression energy provides, the bigger degree of compaction.
The bigger hammer than the standard hammers can
be used to obtain greater energy for laboratory experiments (Fakher 2006, Taufiq 2010, Hanson et al.
2010). The passes number can be increased in order
to obtain better compaction degree for field experiments (Oweis & Khera 1998, Dixon & Jones 2005,
Fakher 2006). Another important parameter affects
341
Table 1. Water content values of municipal solid wastes

(Hossain 2002).
Reference
Water content (%)
Sowers (1973)
Gifford et al. (1990)
Landva and Clark (1990)
Blight et al. (1992)
Huitric (1981)
Tchobanoglous et al. (1993)
Coumoulos et al. (1995)
Gabr and Valero (1995)
1050
1468
15125
10100
1540
1545
20125
30130
the compaction performance is organic matter content. Due to the degradation process, organic waste
is able to reposition itself so, degree of compaction
could change. Water content value of MSW, which
significantly higher than most soils, can show broad
variety (Hossain 2002). Table 1 shows water content
values of the MSWs obtained from literature. Lubrication effect of the water provides the tighter and denser
structure for MSW until the optimum moisture content
is reached.
Composition effect plays an important role to determine compression characteristic of MSW. However,
composition of solid waste will vary depending on
many factors. Social surroundings, average income
level, industry level, climatic conditions and owners
equity of country shapes the solid waste composition. Therefore, solid waste compositions vary widely
from country to country. For example metal and plastic waste takes large share in industrialized countries;
low income countries have abundant in organic matter
content (Aykol 2008).
Composition of MSW not only affects the compaction characteristic but also so many engineering
parameters are affected by composition. For example,
large settlements were observed on the MSWs with
high percentage of organic matters. Moreover, fibrous
structure compositions have higher shear strength values. Although composition is a significant parameter
that shapes the engineering behavior of MSWs, limited
number of studies were obtained about this subject.
Hanson et al. (2010) conducted laboratory and
field tests to determine compaction characteristics of
MSWs. Water content, compaction effort and seasonal effects were chosen as the decisive parameters. Laboratory and field test were conducted by
using USEPA (United States Environmental Protection Agency) (2008) MSW composition. Two different compaction methods which called modified and
4X modified (224 blows for per lift) efforts were
used for laboratory experiment. Maximum dry density values were obtained 5,2 kN/m3 for modified
test and 6,0 kN/m3 for 4X modified test. Field test
maximum dry density values were ranged between
5,7 kN/m3 and 8,2 kN/m3 . Hanson et al. (2010) stated
that moisture addition to wastes during compaction
increased the workability, the unit weight, and the
amount of incoming wastes disposed, and reduced the
compaction time.
Hyun et al. (2011) investigated the engineering

properties of old municipal solid waste sampled in
Whamyung MSW landfill site, Korea. According to
their results, the compaction tests optimum moisture
content values ranged between 21% and 24% and the
maximum dry unit weight ranged between 13 and
14 kN/m3 . Hyun et al. (2011) reported that the values
of the unit weight of the specimens fall in the range
suggested by Kavazanjian et al. (1995) for old MSW,
mainly because the samples in this study include a high
ratio of soil materials in comparison with other wastes.
Reddy et al. (2009) carried out a laboratory study
to designate the geotechnical properties of fresh MSW
collected from the working phase of Orchard Hills
Landfill, USA. Shredded and oven-dried fresh MSW
samples were used in the standard proctor test. Four
different initial target moisture contents (44%, 60%,
80% and 100%) were prepared for compaction tests.
It was stated that dry unit weight value 420 kg/m3
corresponded to 71% optimum water content. Reddy
et al. (2009) emphasized that the maximum size and
size distribution should be taken into account when
laboratory results are interpreted or compared.
Hettiarachchi (2005) used synthetic municipal solid
waste at the standard proctor tests. MSW composition
ratio was generated by considering USEPA investigation reports. For fresh waste a maximum dry density of
525 kg/m3 was observed corresponding to an optimum
moisture content (gravimetric) of 60%.
The studies about the compaction characteristics
of MSW show that, there are a lot of differences
between the reported optimum water content and maximum dry density values, because of the composition
and degradation effects. In this study, composition
and degradation effects on the compaction behavior
of artificially prepared and natural fresh and aged
MSW samples were determined. In order to determine
the composition effect the three different composition
types United States of America (USA), average of four
European countries (EUR) and Turkey (TR) were used
in the experiments.
2
2.1 Materials
In this study, synthetic (lab-prepared) and natural
(obtained from landfill area) MSW samples were used.
Synthetic (lab-prepared) municipal solid waste samples were prepared in different compositions which
are summarized in Table 2. Before preparation of the
synthetic MSW sample, materials were sieved from
the pre-determined sieves. Following this, materials
were mixed at defined proportions. In this way, all the
materials used during the study will have the same texture. Particle size characteristics of used materials can
be seen in Table 3. At the end of comprehensive literature study, the specific gravity values of the used
materials were obtained. Table 4 shows the specific
gravity values of used materials obtained from literature. The specific gravity of the compositions was
342
Table 2. Composition ratios (CR) of MSW for USA, EUR

and TR.
MSW composition ratios (%)
Used material
Material type
Organic
Glass
Plastic
Garden
waste
Metal
Ash
Paper
Potatoes
Mowed Grass
Broken Glass
Plastic particles
Wood shavings
USA
EUR
TR
16
16
7
12
7
17,5
17,5
6
8
6
22,5
22,5
4
10
4
5
0
28
7
3
2
2
15
12
3
5
0
Textile
Soil
Metal dust-chips
9
Ash
0
Paper pieces
22
Cardboard pieces 6
Textiles
5
Fine-grained soil
0
Table 3.
Particle size characteristics of the used materials.
No
Materials
Sieve number
Particle size
(mm)
1
2
3
4
5
6
7
8
9
10
11
Potatoes
Broken Glass
Mowed Grass
Plastic particles
Wood shavings
Metal dust-chips
Ash
Paper pieces
Cardboard pieces
Textiles
Fine grained soil
No. 4
No. 10
No. 4
No. 4
No. 10
1/4
No. 40
1/2
1/2
1
No. 40
4,75
2,00
4,75
4,75
2,00
6,30
0,425
12,50
12,50
25,4
0,425
Figure 2. Synthetic aged MSW samples (at outdoor

conditions).
Table 4.
2.2
No
Materials
Specific gravity (Gs )
1
2
3
4
5
6
7
8
9
10
11
Potatoes
Broken Glass
Mowed Grass
Plastic particles
Wood shavings
Metal dust-chips
Ash
Paper pieces
Cardboard pieces
Textiles
Fine grained soil
1,15
2,60
0,90
0,93
0,52
7,19
0,66
1,20
0,69
1,31
2,65
Table 5.
determined according to percentage of each material.

As a result, the specific gravities of TR, EUR and
USA compositions were determined as 1.20, 1.52 and
1.73, respectively. In addition to particle size characteristics and specific gravity values of used materials,
natural water contents of the used materials were also
determined. In order to avoid loss of organic matter,
materials were dried at 65 C for 24 hours. In addition
to fresh synthetic MSW samples, also aged synthetic
MSW samples were used during the experiments. The
synthetic prepared samples were aged for 30 days at
the outdoor conditions in the plastic containers (Figure
2). The fresh and aged (34 years) natural MSW samples were provided from municipal landfill area of
Manisa city.
Methods
At the beginning of the experiments, the materials were

weighted according to composition percentages and
mixtures were generated in plastic containers (Figure 3). Samples were prepared different initial water
content (gravimetric) values (0%20%40%60%
80%100%120%). It should be noted that in addition
to initial water contents, each organic material has
own water content. Hence, the actual water content of
Specific gravity of the used materials (Wong 2009).
Initial water content values of the used materials.
Sample no
Material
Water content (%)
1
2
3
4
5
6
7
8
9
10
11
Potatoes
Broken Glass
Mowed Grass
Plastic particles
Wood shavings
Metal dust-chips
Ash
Paper pieces
Cardboard pieces
Textiles
Fine grained soil
620
0,1
468
0
8
0,1
1
3
2
0
1,5
the mixture will be different from these values. Each

water content value was added to the separate samples
and mixed homogenously. After adding the required
quantities of tap water, synthetic mixtures were stored
for 24 hours in the closed containers for equilibrium. The standard Proctor and modified Proctor
compaction tests were performed in order to define
compaction energy effect on the MSW.
Standard Proctor test was performed to determine
the optimum moisture content (wopt ) maximum dry
density (drymax ) relationships for MSW (ASTM-D698
(2007)). The prepared MSW sample was placed into
standard proctor mold in three layers and each layer
was impacted 25 times. In order to determine the
composition effect on the compaction behavior of
343
Table 6.
Organic matter content of the MSW samples.
No
Sample type
OM content (%)
1
2
3
4
5
6
7
8
Synthetic Fresh TR
Synthetic Fresh EUR
Synthetic Fresh USA
Synthetic Aged TR
Synthetic Aged EUR
Synthetic Aged USA
Natural Aged
Natural Fresh
65,0
64,0
58,0
62,0
55,0
57,0
41,0
61,0
Figure 3. Synthetic fresh MSW mixtures for USA, EUR

and TR.
Figure 5. Standard Proctor compaction curves and zero air

trend lines.
Figure 4. Extrusion process of the compacted MSW

sample.
the three different synthetic fresh MSW compositions

(TR, EUR, USA), total 18 samples were generated and
compacted by using standard proctor hammer. After
the compaction test, the compacted samples and mold
were weighted and then the samples were extruded
from mold to moisture content determination. Figure 4
presents the extrusion process of the compacted MSW
sample.
In addition to standard proctor test, modified proctor test (ASTM-D1557 (2007)) was also conducted
on the synthetic fresh MSW samples. Compaction
energy produced by using modified proctor test is
much higher than the standard proctor test (approximately 4.5 times). All process applied for standard proctor test to determination of optimum water
content maximum dry density was also applied for
modified proctor test.
Organic matter content influences many of the
physical, chemical and engineering properties of
MSW. The organic matter (OM) contents of the synthetic and natural MSWs were determined according
to the ASTM D2974. The MSW samples were hold
in the furnace at 440 C until the specimen was completely ashed. Table 6 shows the determined OM
contents of the MSW samples.
3.1 Compaction behavior of synthetic fresh and

synthetic aged MSW
The compaction behavior of synthetic fresh MSW
samples was investigated on the three different compositions. While the MSW composition of TR includes
higher percentage organic matter than EUR and USA,
USA composition contains higher percentage plastic and metal wastes than the other compositions.
Moreover, parallel to the higher literacy rate, MSW
composition of EUR comprises higher paper wastes
than the USA and TR. While the fresh synthetic USA
and EUR samples were prepared at six different initial
water content values (20%40%60%80%100%
120%), TR sample was prepared at eight different
initial water content values (0%10%20%40%
60%80%100%120%). However, the actual water
content of the samples was different from these added
values, because organic content has its own water
content.
Standard and modified proctor tests were conducted, so as to determine compaction energy effect
on fresh synthetic MSW samples. Figure 5 shows the
compaction curves of the EUR, TR, and USA synthetic
fresh MSW compositions.
The results have shown that the highest maximum
dry density value was obtained with USA composition (dry max = 3,80 kN/m3 ). The optimum moisture
content value corresponds to maximum dry density
was found approximately 158% for this composition. The maximum dry unit weight and optimum
344
Figure 7. Compaction curves for synthetic aged MSW

compositions (TR, USA and EUR).
Figure 6. Modified Proctor compaction curves and zero air

trend lines.
moisture content values of EUR composition was

obtained 3,72 kN/m3 and 162%, respectively. The
lowest maximum dry density and optimum water
content values were provided from TR composition
(dry max = 3,67 kN/m3 wopt = 138%). According to
these results, it can be concluded that the TR composition has the highest organic matter and ash content,
hence the TR composition cause lower max. dry unit
weight and wopt than the other compositions. The
optimum water contents are very high of the all compositions. Similar to these very high optimum water
contents, the optimum moisture contents are reported
in the literature for MSW are noticeably higher than
for many soils, ranging between 10% to 130% (Gabr &
Valero 1995, Hettiarachchi 2005, Itoh et al. 2005).
Additionally, modified proctor test producing
approximately four times more compaction energy
than standard proctor test, was conducted on the fresh
synthetic MSW samples. The Figure 6 represents the
modified proctor test results of the three compositions.
In contrast to the standard proctor test, the highest
maximum dry unit weight value was obtained for TR
composition (dry max = 4,96 kN/m3 ) and the optimum
water content was wopt = 73%. The maximum dry unit
weight and optimum moisture content values of EUR
composition are found 4,90 kN/m3 and 111%, respectively. Eventually, for USA composition maximum
dry unit weight of 4,84 kN/m3 was achieved at optimum moisture content of 104%. As energy increases,
the maximum dry unit weight increases and optimum
water content decreases for all three compositions.
In order to determine the aging effect on the compaction behavior of MSW samples, the synthetic MSW
samples were left in the outdoor conditions for 30 days
and then the modified proctor tests were conducted
on these samples. Due to the inadequate energy production, standard proctor test was not performed on
the aged MSW samples in contrast to the fresh MSW
samples. While fresh and aged MSW samples were
compared, it can be easily seen that the aged MSW
samples reached the higher dry unit weight values than
the fresh samples. The aging effect is more significant
on the TR composition. The TR composition has the
highest max. dry unit weight, 6,24 kN/m3 . The maximum dry unit weight values of USA and EUR samples were approximately 5,68 kN/m3 and 4,97 kN/m3 ,
Figure 8. Modified Proctor compaction curves for fresh and

aged natural MSW samples.
respectively. The maximum dry unit weight of the TR

sample increased from 4,96 kN/m3 to 6,24 kN/m3 due
to aging effect. The degradation of organic materials
in the MSW causes an increase in max. dry unit weight
and decrease in optimum water content (Figure 7).
3.2
Compaction behavior of natural fresh and

natural aged MSWs
In addition to synthetic MSW samples, the compaction

characteristics of the natural fresh and aged MSW samples were examined by modified Proctor tests. The
results of these tests are shown in Figure 8. The natural
aged MSW sample has significantly higher maximum
dry unit weight and lower optimum water content than
the natural fresh sample, because of the lower organic
content of the aged sample (Table 6).
4
CONCLUSIONS
The compaction behavior of the EUR, TR and USA

compositions were determined with standard and
modified proctor tests. The composition, energy, and
aging effect were determined on the samples.
The results have shown that the highest maximum
dry unit weight was observed with the USA composition which has the lowest organic content and highest
metallic content. As energy increased from standard
Proctor test to modified Proctor test maximum dry unit
345
weight increases and optimum water content decreases

for all three compositions. When the synthetic fresh
samples were aged at outdoor conditions the max.
dry unit weight increased and optimum water content
decreased for all samples. The aging effect was investigated on the natural samples. The aged MSW sample
(34 years) is significantly higher max. dry unit weight
than the fresh natural MSW sample because of its low
organic content. According to results of this study,
the composition especially the percentage of organic
waste is the most important factor on the compaction
behavior of MSWs. In addition to this, higher metallic
content cause higher max. dry unit weight for the USA
composition.
REFERENCES
ASTM D2974-00. 2000. Standard Test Methods for Moisture,
Ash and Organic Matter of Peat and Other Organic Soils,
ASTM International, West Conshohocken, PA.
ASTM D698. 2007. Standard Test Methods for Laboratory
Compaction Characteristics of Soil Using Standard Effort,
ASTM International, West Conshohocken, PA.
ASTM D1557. 2007. Standard Test Methods for Laboratory Compaction Characteristics of Soil Using Modified
Effort, ASTM International, West Conshohocken, PA.
Aykol, K. 2008. Modelling geotechnical properties of MSW,
M.Sc. Thesis, Istanbul Technical University, Turkey.
Bareither, C.A. 2010. Compression behavior of solid waste.
Ph.D. Thesis, University of Wisconsin-Madison, USA.
Blight, G.E., Ball, J.M. & Blight, J.J. 1992. Moisture and
suction in sanitary landfills in semiarid areas, Journal of
Environmental Engineering, 118 (6): 865877.
Blight, G.E. 2008. Slope failures in municipal solid waste
dumps and landfills: a review. Waste Management and
Research, 26: 448463.
Dixon, N., & Jones, D.R.V. 2005. Engineering properties of
municipal solid waste. Geotextiles and Geomembranes,
23(3): 205233.
Fahker, A. 2006. Experimental Study on the Compaction of
Waste. Proc. 5th ICEG, Cardiff, 1: 445452. Edited by
H.R.Thomas.
Fakharian, K., & Taherzadeh., R. 2004. Seismic response and
stability analysis of an area method MSW landfill. 13th
World Conference on Earthquake Engineering Vancouver,
B.C., Canada August 16, 2004 Paper No. 2402.
Gabr, M., & Valero, S. 1995. Geotechnical Properties of
Municipal Solid Waste, Geotechnical Testing Journal,
ASTM, 18 (2): 241251.
Gifford, G.P., Landva, A.O., & Hoffman, V.C. 1990. Geotechnical Consideration when Planning Construction on a
Landfill, Geotechnics of Waste Fills Theory and Practice: ASTM STP 1070 (pp. 4156): American Society for
Testing and Materials, Philadelphia, Pa.
Ham, R.K., Reinhardt, J.J., & Sevick, G.W. 1978. Density of
milled and unprocessed refuse. Journal of the Environmental Engineering Division, 104 (EE1): 109125.
Hanson, J.L., Yesiller., N., Von Stockhausen.,S.A. & Wong,
W.W. 2010. Compaction Characteristics of Municipal
Solid Waste. Journal of Geotechnical and Geoenvironmental Engineering, 136(8): 10951102.
Hettiarachchi, C.H. 2005. Mechanics of Biocell Landfill
Settlements, Ph.D. Dissertation, Department of Civil &
Environmental Engineering. New Jersey Institute of
Technology, Newark NJ.
Hossain, M. S. 2002. Mechanics of compressibility and

strength of solid waste in bioreactor landfills. Ph.D.
Thesis, NC State University, Raleigh, NC.
Huitric, R. 1981. Sanitary landfill settlement rates, Verlangerung Der Nutzungsdauer Von Deponien, Technische
Universitat Berlin, Berlin.
Hyun Il, P., Borinara, P. & Hong, K.D. 2011. Geotechnical Considerations for End-Use of Old Municipal Solid
Waste Landfills. International Journal of Environmental
Research 5(3):573584.
Itoh, T., Towhata, I., Kawano, Y., Kameda, M., Fukui, S., &
Koelsch, F. 2005. Mechanical Properties of Municipal
Waste Deposits and Ground Improvement, Proceedings of
the Sixteenth International Conference on Soil Mechanics and Geotechnical Engineering, Volume 4, Millpress Science Publishers, Rotterdam, the Netherlands,
22732276.
Kavazanjian, E.Jr., Matasovic, N., Bonaparate, R. & Schmertmann, G.R. 1995. Evaluation of MSW Properties for
Seismic Analysis. Proc. International Conference Geoenvironment 2000, ASCE GSP n. 46, New York.
Landva, A.O. & Clark, J.I. 1990. Geotechnics of waste fill.
Geotechnics of waste fills-theory and practice, ASTM
STP 1070. In: Landva A, Knowles D (eds) American society for testing and materials. Philadelphia, Pennsylvania,
86103.
Langer, U. 2005. Shear and Compression Behaviour of
Undegraded Municipal Solid Waste, Ph.D. Thesis at
Loughborough University.
Manassero, M., Van Impe, W., & Bouazza, A. 1996. Waste
Disposal and Containment, Proceedings of the 2nd
Preprint of Special Lectures, A.A. Balkema, Osaka, Japan,
14251474.
Oweis, I.S., & Khera, R.P. 1998. Geotechnology of waste
management. Second edition, PWS Publishing Company,
Boston.
Reddy, K.R., Hettiarachchi H., Parakalla N.S., Gangathulasi
J., Bogner J. E. 2009. Waste Management. Geotechnical
properties of fresh municipal solid waste at Orchard Hills
Landfill, USA. 29(2): 952959.
Sowers, G. 1973. Settlement of Waste Disposal Fills, Proceedings of the International Conference on Soil Mechanics and Foundation Engineering, International Society for
Soil Mechanics and Foundation Engineering, Montreal,
Quebec, Canada, 207210.
Staub, M., Galietti, B., Oxarango, L., Khire, M.V. &
Gourch, J.P. 2009. Porosity and hydraulic conductivity
of msw using laboratory-scale tests. Third International
Workshop, Hydro-Physico Mechanics of Landfills,
Braunschweig, Germany.
Taufiq, T. 2010. Characteristics of Fresh Municipal Solid
Waste. MSc. thesis. The University of Texas at Arlington.
Tchobanoglous, G., Theisen, H., and Vigil, S. 1993. Integrated Solid Waste Management: Engineering Principles
and Management Issues, Irwin/McGraw-Hill, Boston, MA.
Turer, D., & Turer, A. 2011. A simplified approach for slope
stability analysis of uncontrolled waste dumps. Waste
Management & Research, 29(2): 146156.
United States Environmental Protection Agency, 2008.
Municipal Solid Waste, Wastes. US EPA. Last accessed:
29 June 2009, from United States Environmental
Protection Agency.
Wong, W.W. 2009. Investigation of the geotechnical properties of municipal solid waste as a function of placement conditions. MSc. thesis, California Polytechnic State
Univ., San Luis Obispo, USA.
346

Shear strength of municipal waste materials from two landfills in Serbia
c
D. Rakic, L. Caki
& S. Cori
University of Belgrade, Faculty of Mining and Geology, Belgrade, Republic of Serbia
ABSTRACT: In this paper the results obtained by laboratory testing of municipal waste materials from two
different sanitary landfills in Serbia are presented. For defining parameters of shear strength, linear shear strength
equation was used, which is determined from mobilized shear stresses at horizontal displacement of l =
14 mm for each of normal stresses (n = 25, 50 and 100 kPa). Also, an interpretation of the test results was
made for nonlinear shear strength envelope with logarithmic and hyperbolic functions. Thus obtained results
are compared with the proposed linear Coulomb-Mohr-Terzaghi strength equations and nonlinear forms of the
logarithmic equation proposed by Zekkos (2005). It was concluded that the proposed nonlinear shear strength
envelope with hyperbolic shape, compared to the envelope of logarithmic form, is in better agreement with the
linear Coulomb-Mohr-Terzaghi envelope.
1
INTRODUCTION
Table 1.
Shear strength or the shear stress parameters for

municipal solid waste have been obtained in different
ways that are divided into three groups: laboratory tests
on small or large samples using standard or specialdesign equipment, in situ tests and assessments of
the shear strength parameters based on back analyses
of the landfill slope stability. Published approaches
to the interpretations of test results still differ, based
on various assumptions. Shear strength of municipal
solid waste has been much studied and the reported
results are mainly those of laboratory tests on samples of different sizes: Landva & Clark (1986), Gabr
& Valero (1995), Manassero et al. (1996), Eid et al.
(2000), Pelkey et al. (2001), Dixon & Jones (2005),
Langer (2005), Zekkos (2005), Zekkos et al. (2007),
Kavazanjian (2006), Athanasopoulos et al. (2008),
Bray et al. (2009), Stark et al. (2009), and many others.
Some authors are of the opinion that the waste shear
strength also considers the tensile strength activated
by shearing. Klsch (1996) and later Athanasopoulos
et al. (2008) tried to explain shear strength of solid
waste including reinforcing elements.
Direct shear tests on undisturbed samples in situ
certainly give more realistic values than the measured
shear strengths of the municipal artificially prepared
samples.Tests in situ by Houston et al. (1995), Withiam
et al. (1995), Mazzucato et al. (1999), Thomas et al.
(1999), Caicedo et al. (2002) made worthy contributions to the description of the municipal waste shear
strength.
2
LABORATORY TESTS SAMPLES

PREPARATION
Municipal waste tested in laboratory for the shear

strength was sampled by drilling or excavating from
Composition of analyzed municipal waste.

Mass (%)
Kind and denotation of

waste by S.W.A.
Tool catalog
Ada Huja landfill, Landfill for

Belgrade
Novi Sad
Wood W2
1.0
Paper and Cardboard PC3 3.7
Plastics PL4
5.6
Glass G5
4.9
Textiles T6
2.3
Metals M7
1.9
Complex Products C9
1.1
Soil IN10 01
34.1
Ceramics IN10 02
6.1
Unclassified (fines) F12 39.3
2.9
4.2
6.4
6.3
1.8
2.4
1.3
29.4
5.3
40.0
two landfills in Serbia (active landfill in Novi Sad and

old Ada Huja landfill in Belgrade closed 40 years
ago). The composition of waste (Tab. 1) is based on
the materials sorted and classified following instructions of the S.W.A. Tool (European Commission,
2004).
Table 1 shows the old waste predominantly composed of unsorted and soil materials characteristic
of Serbian municipal landfills older than thirty years
(Rakic et al. 2011a).
Preparation of waste samples consisted of reducing
the particle size (d) to suit the shear box (L) i.e. L/d
5. Depending on the apparatus used, the maximum particle size (d) was 1220 mm. A small proportion only
(not higher than 5%), mainly of plastics, textiles and
paper particles, were somewhat coarser (max. 40 mm).
Samples were artificial, prepared and separated into
four series (A, B, C, D), each series of three specimens
(36 test samples) different in unit weight and natural
moisture (Tab. 2).
347
Table 2. Basic identification and classification data for test

samples.
Series
A
Sample
labels
w (%)
(kN/m3 )
U-1
U-2
U-3
U-4
U-5
U-6
U-7
U-8
U-9
U-10
U-11
U-12
37.2
36.4
27.2
39.8
35.0
36.9
30.9
33.9
32.9
28.7
29.7
31.8
10.6
15.0
17.5
11.1
12.1
13.4
10.0
12.1
14.3
10.0
11.6
14.0
Gs
2.2
1.848
1.000
0.599
1.519
1.231
1.043
1.683
1.268
0.905
1.638
1.292
0.930
2.0
2.05
2.05
Waste samples were prepared from the Belgrade

(Series A) and Novi Sad (Series B) and the mixture
from the two landfills (Series C and D). Samples from
Series A, B and C were compressed in horizontal layers, similarly to the waste deposition on the landfills.
Samples of Series D had reinforcing particles oriented
at an angle = 90 to the horizontal plane of shearing
in the shear box.
3
Figure 1. Characteristic relation of the shearing

stress and displacement for loose compacted samples
( 12.1 kN/m3 ).
DIRECT SHEAR TEST RESULTS
Shear strength of the waste was determined from

the direct shear tests in boxes 60 mm 60 mm and
100 mm 100 mm. For consolidation of samples, normal stresses (n ) were selected of 25, 50 and 100 kPa.
In most test samples, subjected to shearing parallel
with the reinforcing particles (Series A, B and C), the
relation shearing stress-displacement indicated different behaviour of waste in relation to compaction or to
unit weight.
For less compacted samples with unit weight
12.1 kN/m3 , the stress-displacement relation was closest to the behaviour of soil with strain hardening, so
that the strains were inadequate to mobilize the peak
shearing resistance of the waste (Fig. 1).
The stress-displacement relation of the dense samples with 13.4 kN/m3 indicated waste behaviour
similar to the waste with plastic failure (Fig. 2).
The shearing stress-displacement relation for samples including reinforcing particles oriented normal to
the shear plane (Series D) had a different behaviour,
because all samples behaved similarly to the soil with
strain hardening. The similarity of the relation indicated that different densities didnt have great effect
on the shape of the stress-strain behaviour for these
samples. The test results, for a characteristic sample
from series D, are represented in Figure 3.
Vertical deformations of samples also were measured during the shearing. It was noted that the
upper frame of the box-shear apparatus had heaving
Figure 2. Characteristic relation of stress and displacement

during shearing for dense samples ( 13.4 kN/m3 ).
with most samples. The heaving was greatest with

the samples consolidated at the lowest normal stress
n = 25 kPa (Fig. 4).
After the test completion and the sample extraction,
the shear surface remained horizontal, with only the
upper frame front heave.
4
4.1
SHEAR STRENGTH DETERMINING

Linear shape of the shear strength equation
For the shear strength of municipal solid waste the

authors of reference literature mainly recommend the
348
Figure 5. Unit weight versus shearing resistance for

municipal waste.
Figure 3. Characteristic relation of the shearing stress and
displacement for samples with the particles oriented normal
to the shearing plane.
Figure 4. Characteristic heaving of the box-shear apparatus

upper frame and a view of the sample after shearing.
linear Coulomb-Mohr-Terzaghi equation, using mobi

lized angle of internal friction (mob
) and mobilized

cohesion (cmob
). In the case of unexpressed failure
(which is a common case), the horizontal displacement
(l), which depends on the size of the test sample,
is taken to vary between 7 mm and 150 mm. With
this criterion applied, the mobilized shear strength for
displacement (l) of 14 mm was used in the interpretation of the test results. This failure criterion is not
perfectly adequate for each normal stress level in view
of the fact that for some samples, at a normal stress
n = 25 kPa, the peak shearing resistance was clearly
expressed. Where this was the case, the shear strength
parameters were combined from the mobilized and
peak resistances. The interpretation was based on the
linear shape of the Coulomb-Mohr-Terzaghi equation for shear strength and on the nonlinear failure
envelope, using relations for the logarithmic and the
hyperbolic shapes.
In view of different unit weights of the prepared
samples, shear strengths were determined for loose
( 12.1 kN/m3 ) and for dense ( 13.4 kN/m3 )
waste (Fig. 5).
Figure 6. Relation of shearing stress and displacement for

loose compacted samples.
Minimum and maximum shear strengths were

determined for both conditions and their average values taken to be characteristic of the dense and the loose
landfills in Serbia.
Samples from Series D were not considered,
because of their different particle orientation. They

had much higher angle of internal friction (mob
)

and cohesion (cmob ) than samples of the other three
series, which varied in relation to the unit weight

within the ranges from mob
= 32 51 and from

cmob
= 3157 kN/m2 , respectively.
It was concluded for both considered conditions
that maximum shear strength was preferential in samples of the Series B (municipal waste from Novi Sad
landfill), which supports the general assumption that
shearing strength decreases with the degradation of the
waste.
Average values of the shear strength parameters for
the Coloumb-Mohr-Terzaghi equation are plotted in
Figure 6 and Figure 7 (Rakic et al. 2011b).
Straight lines of the failure envelopes were compared with those recommended by other authors and
extensively used. Shear strength of the loose waste
(best fit to the published data) seems to be between
349
Figure 7. Relation of shearing stress and displacement for

dense samples.
Figure 9. The recommended and the reference straight-line

failure envelopes.
Figure 8. Comparative illustration of the shear strength

parameters from reference sources.
the upper and lower limits, which confirm a comparatively good agreement with the published values. As
the shear strength parameters were obtained for dense
waste, the range of their values, recommended by
Sanchez-Alciturri et al. (1993) and by Gabr & Valero
(1995), could be widened and the lower limit proposed
(Fig. 8).
Figure 9 shows the linear strength envelope for
loose compacted and well compacted municipal waste,
which are compared with the proposals of other
authors, which are widely used in practice.
Shearing strength parameters are computed for all
the measured horizontal displacements (Fig. 10).
In most of tested samples, which had particle
orientation parallel to the shear plane, cohesion
reached maximum value in the zone of comparatively small horizontal displacements (l = 25 mm)
and depended on the sample compacted. For greater
displacements (l 12 mm), it was found that compression didnt have great effect on the cohesion, as the
increments for all tested samples were within the range

cmob
= 25 kN/m2 . Orientation of the reinforcing
particles, however, had a notable effect on cohesion, as
the Series D examples, which had reinforcing particles
normal to the shearing plane, demonstrated maximum
Figure 10. Mobilised cohesion in relation to shearing displacement and unit weight Series A samples.
cohesion values of l = 1315 mm (Fig. 11) in the

zone of large horizontal displacements (Rakic, 2013,
unpubl.).
Unlike cohesion, the angle of internal friction grows
with the horizontal displacement (Fig. 12).
Jessberger & Kockel (1993) and Gabr & Valera
(1995) came to similar conclusions that mobilized
angle of internal friction in the old degraded waste
increased to a maximum at certain displacement
and remained constant at subsequent displacements.
Unlike cohesion, however, unit weight can not
markedly change the angle of internal friction within a
small range of displacements (l = 34 mm), because
at l = 3 mm the value of the mobilized angle of

internal friction varies only within mob
= 2 .
350
Figure 11. Mobilised cohesion in relation to the shearing

displacement and unit weight Series D samples.
Figure 13. Mobilised angle of internal friction in relation to

the shearing displacement and unit weight Series D samples.
shear test results for sixteen solid waste samples

( = 10.215.1 kN/m3 ) and on the results of other
researchers, he recommended the following logarithmic shear strength equation:
Figure 12. Mobilised angle of internal friction in relation to

the shearing displacement and unit weight SeriesA samples.
Like with cohesion, the orientation of reinforcing

particles has a marked influence on the mobilized
angle of internal friction values and the angle increases
with the horizontal displacement (Fig. 13).
4.2
The same author compared the recommended logarithmic nonlinear failure envelope with the previous linear envelopes by Kavazanjian et al. (1995),
Manassero et al. (1999), Eid et al. (2000) and found
them fundamentally similar, noting that an advantage
of the nonlinear failure envelope was its being based
on a larger number of the direct shear tests performed
on wastes from all over the world.
Nonlinear shear strength equation, both logarithmic and hyperboles shapes also were used in the
interpretation of the test results. Samples from the
group of loose compacted waste ( 12.1 kN/m2 )

were tested for mobilized cohesion cmob
= 0, 5 and
10 kPa. The linear envelope obtained by the mobilized

cohesion cmob
= 10 kPa that best fitted the nonlinear
shear strength equation is:
A nonlinear envelope was similarly developed for

dense waste:
Nonlinear shape of the shear strength equation
On the basis of the published and own research data,

Zekkos (2005) states that linear failure envelope may
be a good fit of the obtained data, but notes that a
nonlinear envelope is more accurate and recommends
the use of the nonlinear envelope of logarithmic shape.
An equation of shear strength in the logarithmic shape was first published by Nobari & Duncan
(1972) to determine shearing resistance of a rockfill. Zekkos (2005) recommended its use for shear
strength of the municipal solid waste. Based on direct
The results for both groups of samples are represented in Figure 14 and Figure 15.
The nonlinear failure envelope may be determined using hyperbolic shape relation proposed by
Maksimovic (1989), Maksimovic (1993), who considers that a change in the effective secant angle of the
351
Figure 14. Nonlinear shear strength equation of logarithmic

shape for loose compacted waste.
Figure 16. Geometric representation of the hyperbolic

shape of a nonlinear envelope (Maksimovic, 1989).
Figure 17. A mode of representing elements of a hyperbolic nonlinear failure envelope, = f (log n ) (Maksimovic,
1989).
Figure 15. Nonlinear shear strength equation of logarithmic
shape for dense waste.
For dense waste it is:
shearing strength may be expressed in relation to the

normal shearing stress written as:
Figures 1617 are geometric representations of the

parameters in the above equation.
This form of the nonlinear failure envelope was
considered using the same values of the mobilised
cohesion. Also, the envelope that best agreed with
the linear shear strength equation for loose compacted
waste was the envelope obtained with the mobilized

cohesion cmob
= 10 kPa, i.e.:
The results for both groups of samples are given in

Figures 1819.
The resulting nonlinear failure envelopes are compared with the linear Coulomb-Mohr-Terzaghi equation and the logarithmic equation proposed by Zekkos
(2005) and shown in Figure 20.
The logarithmic shape of the nonlinear envelope fitted in the range from 25 kPa to 100 kPa of the applied
normal effective stresses (n ), but differed for stresses
higher than n 100 kPa in giving lower values of the
angle of internal friction, or of the shearing resistance.
The proposed hyperbolic shape of the nonlinear
failure envelope fitted a somewhat wider range of normal stresses (n ) from 20 kPa to 150 kPa. Moreover,
it clearly indicated that compared with the logarithmic envelope it agrees much better with the linear
Coulomb-Mohr-Terzaghi envelope due to the wider
range of normal stresses.
352
Figure 18. Nonlinear strength equation of hyperbolic shape

for loose compacted waste.
shearing resistance parameters are within the ranges

found in published literature. Published data, however,
are not to be directly used without preliminary knowledge of the composition of waste, pretreatment of the
components, sample preparation and the component
identification procedure.
The research resulting relations, whether linear
(Coulomb-Mohr-Terzaghi) or nonlinear (logarithmic
or hyperbolic shape), give satisfactory values for the
shear strength that may be used to analyse slope stability of 1015 m high landfills in Serbia for normal
effective stresses n 150 kPa. For higher landfills,
the hyperbolic form of the nonlinear failure envelope
is recommended, which gives somewhat lower shearing resistance values than the usual linear envelope. In
view of the always lower than real shear strength values
based on field data and back-analyses, use of nonlinear
shearing resistance equation is still justifiable.
REFERENCES
Figure 19. Nonlinear strength equation of hyperbolic shape

for dense waste.
Figure 20. Comparative review of proposed linear and nonlinear shear strength envelope of municipal waste material in
Serbie.
CONCLUSIONS
Conventional methods used in soil investigations are

practical for research of the highly heterogeneous
municipal solid waste. The values obtained for the
Athanasopoulos, G., Grizi, A., Zekkos, D., Founta, P. &

Zisimatou, E. (2008). Municipal solid waste as a reinforced soil: Investigation using synthetic waste, Proc.,
Geo-Institute Geocongress 2008, The Challenge of Sustainability in the Geoenvironment, Geotechnical Special
Publication, ASCE, No. 177, 168175.
Bray, J.D., Zekkos, D., Kavazanjian, E.Jr., Athanasopoulos,
G.A. & Michael, F.R. (2009). Shear strength of municipal
solid waste, Journal of Geotechnical and Geoenvironmental Engineering, ASCE, Vol. 135, No. 6, 709722.
Caicedo, B., Giraldo, E., Yamin, L. & Soler, N. (2002).
The landslide of Dona Juana landfill in Bogota. A case
study, Proceedings of the Fourth International Congress
on Environmental Geotechnics, Rio de Janeiro, Brazil,
171175.
Dixon, N. & Jones, D.R.V. (2005). Engineering properties
of municipal solid waste, Geotextiles and Geomembranes
23, 205233.
Edincliler, A., Benson, C.H. & Edil, T.B. (1996). Shear
strength of municipal solid waste, Interim Report year
1, Environmental Geotechnics Report 96-2, Department
Wisconsin Madison, 65.
Eid, H.T., Stark, T.D, Douglas, W.D. & Sherry P.E. (2000).
Municipal Solid waste slope failure 1: Waste and foundation properties, Journal of Geotecnical and Geoenvironmental Engineering, ASCE, Vol. 126(5), 397407.
European Commission [EC], (2004). Project: S.W.ATool,
Development of a Methodological Tool to Enhance the
Precision & Comparability of Solid Waste Analysis Data,
5th Framework Program, EU, Austria, 57.
Gabr, M.A. & Valero, S.N. (1995). Geotechnical properties
of municipal solid waste, Geotechnical Testing Journal,
ASTM, Vol. 18(2), 241251.
Grisolia, M. & Napoleoni, X. (1996). Geotechnical characterization of municipal solid waste: Choise of design parameters, Proc., 2nd Int. Cong. On Environmental Geotechnics,
Osaka, Japan, Balkema, Rotterdam, 641646.
Houston, W.N., Houston, S.L., Liu, J.W., Elsayed, A. &
Sanders, C.O. (1995). In situ testing methods for dynamic
properties of MSW landfills. In: Procedings of specialty
conference on earthquake design and performance of solid
waste landfills, Geotechnical Special Publication, ASCE,
No.54, San Diego, 7382.
353
Howland, J.D. & Landva, A.O. (1992). Stability analysis of a

municipal solid waste landfill, Stability and Performance
of Slopes and Embankments, II: Proceedings of a Specialty Conference, Berkeley, California, USA, American
Society of Civil Engineers. 2, 12161231.
Jessberger, H.L. & Kockel, R. (1993). Determination and
assessment of the mechanical properties of waste materials, Proc., International Symposium. Green93, Bolton,
U.K. A.A. Balkema, Rotterdam, 313322.
Jones, R., Taylor, D. & Dixon, N. (1997). Shear strength
of waste and its use in landfill stability analyses, British
Geoenvironmental Engineering Conference, Cardiff, UK,
British Geotechnical Society, 343350.
Kavazanjian, E.Jr., Matasovic, N., Bonaparte, R. &
Schmertmann, G.R. (1995). Evaluation of MSW properties for seismic analysis, Geoenvironment 2000, Geotechnical Special Publication No.46, 11261141.
Kavazanjian, E.J., Matasovic, N. & Bachus, R.C. (1999).
Large-Diameter Static and Cyclic Laboratory Testing of
Municipal Solid Waste, Proceedings of the 7th International Landfill Symposium, Sardinia 99, CISA, Italy,
437444.
Kavazanjian, E.Jr. (2006). Waste mechanics: Recent findings and unanswered questions, Geotechnical Special
Publications, ASCE, 148, 3454.
Klsch, F. (1996). Der Einfluss der Faserbestandteile auf
die Scherfestigkeit von Siedlungsabfall (The Influence
of Fibrous Constituents on Shear Strength of Municipal Solid Waste), Ph.D. Thesis, Technische Universitt
Braunschweig, Brauschweig, Germany.
Landva, A.O. & Clark, J.I. (1986). Geotechnical testing
of wastefill, Proceedings, 39th Canadian Geotechnical
Conference Ottawa, Ontario, 371385.
Langer, U. (2005). Shear and Compression Behaviour of
Undergraded Municipal Solid Waste, Ph.D. thesis, Loughborough University, 292.
Maksimovic, M. (1989). Nonlinear failure envelope for soils,
Journal of Geotechnical Engineering,American Society of
Civil Eng., Vol. 115 (4), 581586.
Maksimovic, M. (1993). Nonlinear failure envelope for the
limit state design, Limit state design in geotechnical engineering, Interbational symposium, Danish Geotechnical
Society, Bul. 10, Copenhagen, 131140.
Manassero, M., Van Impe, W.F. & Bouazza, A. (1996).
Waste disposal and Containment, Proc. 2nd International Congress on Environmental Geotechniques, Osaka,
Japan, 14251474.
Mazzucato, N., Simonini, P., Colombo, S. (1999). Analysis of
block slide in a MSW landfill, Proceedings Sardinia 1999,
Seventh International Waste Management and Landfill
symposium, Cagliari, Italy.
Nobari, E.S. & Duncan, J.M. (1972). Effect of reservoir filling on stresses and movements in earth and rockfill dams.
Report T-72-1, Office of Research Services, University of

California, Berkeley, Calif.
Pelkey, S.G., Valsangkar, A.J. & Landva, A.O. (2001). Shear
displacement dependent strength of MSW and its major
constituents, Geotechnical Testing Journal, ASTM, 24(4),
381390.
Rakic, D., Caki,

L. & Ljubojev, M. (2011a). Deformable characteristics of the old municipal solid waste sites with Ada
Huja, Belgrade Serbia, Technics Technologies Education
Management, Volume 6(1), No 1, 5260.
c, S. (2011b). Shear strength of

Rakic, D., Caki,
L. & Cori
municipal solid waste (in Serbian), Proceedings, Fourth
Scientific Symposium, Geotechnical aspects of civil engineering, Society of Civil Engineers of Serbia, 457466.
Rakic, D. (2013, unpubl.). Constitutive relations of landfill municipal waste in Serbia (in Serbian), Ph.D. thesis,
University of Belgrade, Faculty of Mininag and Geology,
Department of Geotechnics, Belgrade.
Snchez-Alciturri, J.M., Palma, J., Sagaseta, C., & Canizal,
J. (1993). Mechanical Properties of Waste in a Sanitary
Landfill, Waste Disposal by Landfill; Symposium entitled
Geotechnics related to the environment, 193199.
Singh S. & Murphy, J. (1990). Evaluation of the stability of
sanitary landfills, Geotechnics of Waste Fills Theory
and Parctice, American Society for Testing and Materials
ASTM, vol. 1070, 240258.
Singh, M.K. (2008). Characterization of stress-deformation
behaviour of municipal solid waste, Ph.D. thesis, University of Saskatchewan, Saskatoon, Canada, 292.
Stark, T.D., Huvaj-Sariham, N. & Li, G. (2009). Shear
strength of municipal solid waste for stability analyses,
Environ Geol 57(8), 19111923.
Thomas, S., Aboura, A.A., Gourc, J.P., Gotteland, P., Billard,
H., Delineau, T., Gisbert, T., Ouvry, J.F. & Vuillemin, M.
(1999). An in situ waste mechanical experimentation on
a french landfill, 7th International Landfill Symposium,
S. Margherita di Pula, Cagliari, Italy, CISA. 445452.
Withiam, J.L., Tarvin, P.A., Bushell, T.D., Snow, R.E. &
German, H.D. (1995). Prediction and performance of
municipal landfill slope. In: Proceedings of speciality
conference geoenvironment 2000, Geotechnical Special
Publication, ASCE, 46(2), New Orleans, 10051019.
Zekkos, D.P. (2005). Evaluation of static and dynamic properties of municipal solid waste. Ph.D. thesis, Department
California, Berkeley.
Zekkos, D., Bray, J.D., Reimer, M., Kavazanjian, E. &
Athanasopoulos, G.A. (2007). Response of municipal
solid-waste from Tri-Cites landfil in triaxial compression,
Proc., Sardinia, 11th Int. Waste Management and Landfill
Symp., Paper No. 447.
354
Stability and settlement analysis of

landfills and natural slopes

HydroMechanical coupled model for unsaturated phenomena

in a rainfall-induced landslide
W. Arairo, I. Djran-Maigre & F. Prunier
Universit de Lyon, INSA de Lyon, Laboratoire de Gnie Civil et dIngnierie Environnementale (LGCIE),
Villeurbanne Cedex, France
A. Millard
CEA/DM2S/LM2S, CEN Saclay, Gif-sur-Yvette, France
ABSTRACT: A heavy rainfall is one of the most factors which induce shallow landslides. This type of landslides
usually takes place in partial saturated slopes.The main slip mechanisms can be explained by the different physical
key processes in variably saturated soils during rainfall events. Those processes are related to water effect on
the mechanical behaviour, water flow through the porous media, and the soil water retention behaviour. In this
paper, an elasto-plastic constitutive model that captures the main features of the behaviour of unsaturated soils
is used in a fully coupled hydromechanical finite element analysis to catch the destabilizing processes in a steep
slope which may occur on the construction phase of landfills during rain infiltration.
INTRODUCTION
Landslides exhibit an initial failure stage followed by

a run-out. A central problem in classification is what
weight to give to each of these two, often contrasting stages. Landslides can be initiated by rainfall,
earthquakes, volcanic activity, changes in groundwater, disturbance and change of a slope by manmade construction activities, or any combination of
these factors. Rainfall-induced landslides affect many
mountain valleys, and landfill slopes. Groundwater
table is mostly absent in such slopes and soils are
usually in a partial saturation state.
Such kind of landslides is characterized by a low
probability of evolution into a catastrophic event but
can have large impacts on population, infrastructures
and the environment. This impact is becoming more
and more pronounced especially during the excavation
of landfills. Such landslides involve the spontaneous
failure, involving volumes measured in hundreds or
thousands cubic meters (Dai et al., 1999). Actually,
landslides may exhibit not only under the action
of positive pore pressures variations which is well
documented and most recurrent, but also in unsaturated conditions where the contribution of capillary
forces to slope stability is substantial (Godt et al.,
2009).
Numerical models taking into account the involved
physical key processes in variably saturated soils during different phases of rainfall events could help in
understanding the main slip mechanisms. Those processes are related to water flow through the solid
matrix, soil water retention behaviour and the effects
of matric suction on the mechanical behaviour. In

this paper, a hydromechanical constitutive model is
presented to predict the hydraulic and stress-strain
behaviour of unsaturated soils.
The proposed model is built with a non associated
elastoplastic model for the solid phase and an original
formulation for water retention curves.
The Bishop definition of effective stress tensor has
been used (Bishop 1959) with a new definition for the
hydromechanical parameter.
A smooth variation of degree of saturation as function of suction has been adopted for the soil-water
retention curve. Two sets of equations have been developed respectively for the boundary retention curves
and for the scanning curves.
The effect of porosity has been introduced into the
water retention curve: the Air Entry Value and the Air
Expansion Value depend on the porosity.
The hydraulic model has been coupled to an elastoplastic model based on the yield surface of Van
Eekelen.
The suction effect is described using the effective Bishop stress. Due to the new formulation for
the parameter (described below), the model is able
to reproduce the loading collapse phenomenon upon
wetting, in oedometer device.
Model predictions of the stressstrain and waterretention response are compared with those obtained
from experimental triaxial tests on unsaturated soil
during isotropic and shear loading. The comparisons
indicate that the model can quantitatively reproduce
the hydraulic and mechanical behaviour of unsaturated
soils.
357
PHYSICAL MECHANISMS IN
UNSATURATED SOIL SLOPES
Usually in natural slopes, two different flow types

can be found: a deep flow regime and a surface flow
regime. The deep flow is most often parallel to the
slope surface and the shallow flow is controlled by
rainfall which affects the capillary compressive pore
water pressures. The slopes response to a rainfall event
depends on rainfall intensity and duration, material
parameters (shear strength, density, porosity and plasticity index) and the topography (the thickness of
the sliding mass). One of the most-cited factors for
landslides initiation is hydraulic predisposition factors. The initiation mechanism depends also on slope
angle and mechanical characteristics of soil. During
the rainfall, the degree of saturation of the upper soil
layer increases, which reduces the capillary tension
between the soil particles and weakens the slope. At
high degrees of saturation, the wetting pore collapse
may take place what can be a cause for debris flow initiation. Water is consequently carried to the toe region.
Upon that, due to the mobilized fluid flow inside the
soil matrix, the fluid exerts a destabilizing, downhill
frictional drag.
All these physical processes may take place as
response of soil slope to a rainfall event. Their identification and integration in the modeling process
is necessary in order to perform an analysis of the
onset of failure. Hence, beside the improvement of the
mechanical constitutive model for capillary effects, an
additional constitutive relation is required to describe
the liquid retention of the pores. Therefore, it is important to introduce a coupling between the mechanical
and the water retention constitutive relations.
Barnichon (1998) for the deviatoric hardening behavior. A smooth compression cap has been added in
order to describe the main inelastic coupling between
deviatoric and volumetric behaviors. The strain rate is
partitioned into elastic and plastic components:
The elastic part is linked to the stress tensor according

to Hookes law:
Where the stiffness elastic tensor is defined by:
The limit between the elastic and plastic domain is

represented by the yield surface of Van Eekelen in
the principal stress space. The cone of the yield surface depends on the current friction angles c and e
(indices c and e stand for compression and extension,
respectively) and on current cohesionc.
The yield surface of Van Eekelen is given by:
where c is the cohesion, c is the triaxial compression friction angle, I1 = tr( ) and I2 = tr(h2 ) are
the first and second invariants of effective stress tenI
sor respectively, with h = 13 I. Coefficient m is
defined as:
HYDROMECHANICAL COUPLED MODEL

FOR UNSATURATED SOILS
A complete description of unsaturated soil behavior,

should take into account:
is the Lode angle, parameters a and b are defined as

functions of the friction angles:
The hydraulic behavior (retention features)

The mechanical behavior (stress-strain response)
The water flow in porous media.
In order to capture the main features of soil slope
behavior in unsaturated conditions for modeling purpose, an elasto-plastic constitutive model including the
soil water retention behavior developed by Arairo et al.
(2012) is used in the framework of hydromechanically
coupled porous media. The principal model concepts
are presented in this section.
3.1
with rc and re respectively the reduced radii in compression and in extension:
Elastoplasic mechanical model
The stress-strain behavior of unsaturated soils at the

macro-scale has been described in the framework of
elasto-plasticity theory. The elastoplastic model used
here is based on the elastoplastic model developed by
controls the convexity of the deviatoric yield surface.

The hardening law is defined by a hyperbolic evolution
of friction angles and cohesion, as a function of the
358
total equivalent plastic strain eq . The initial values

(c0 ; c0 ; e0 ) give the elastic yield while the maximum
values (cf ; cf ; ef ) give the plastic limit:
Bp and Bc are two hardening constants used to calibrate the shape of the hardening. A non-associated
flow rule has been introduced to correctly describe the
volumetric behaviour of soil.
The flow rule has been defined with the same equ
tion as the yield surface, but using the dilatancy angle
instead of friction angle in the definition of
parameter m. The variation of is assumed to be:
The integration algorithm is based on a backward Euler

integration of the rate constitutive equations, which
give rise to an elastic predictor and a plastic-corrector
(return map) stress update algorithm.
The deviatoric plastic mechanism is coupled to an
isotropic one, defined by a compression cap. The hardening cap is a deformed sphere in accordance with
the Van-Eeckelen surface. It intersects the deviatoric
envelope in a smooth fashion to avoid a singular corner point, a state which has been recognized as giving
rise to both numerical and constitutive instabilities.
The compression cap surface moves along the I1 axis
with incremental changes in plastic volumetric strain.
The yield function of the compression cap surface
is given by:
with
Figure 1. (a) Shear yield surface with compression cap on

the deviatoric plan. (b) Yield surface of Van Eekelen with
compression cap on the stress tensor plan.
The hardening law for the cap surface is defined by

an exponential variation of the consolidation pressure
p
pc as a function of the volumetric plastic strain v :
where = soil parameter and n0 = initial porosity.
3.2
Hydraulic behavior: water-retention features
Modeling the variation of degree of saturation Sr with

respect to the suction s is also taken into account in the
model (Arairo et al. 2012). This constitutive model
takes into consideration the hysteresis between drying and wetting curves. The relationship between the
degree of saturation and suction takes several forms,
depending on the incremental way to reach the suction
value. Since the water retention curves are classified
into 4 types (boundary drying curve, boundary wetting
curve, wetting scanning curves and drying scanning
curves) an analytical expression has been adopted for
each curve.
For the boundary water retention curve, the Van
Genuchten-Mualem equation, with two sets of empirical parameters (aw , nw ) and (ad , nd ) for the boundary
wetting and drying curves respectively, has been used:
where m is given by Equation (5), h is shown in figure

1(a). The radius of the cap part is calculated:
359
With Srres and Srsat being the residual and full saturated
degree of saturation respectively.
The atmospheric pressure Patm is introduced in
the equation to make the ad and aw parameters
dimensionless.
In order to take into consideration the porosity effect
on the hydraulic behavior, ad and aw parameters have
been expressed as function of air entry value PAEV
which represents the suction limit where air starts to
enter into the largest pores in the soil.
A modified exponential variation of the PAEV
depending on the porosity is adopted:
where = material parameter, n = porosity and

PAEV 0 = air entry value for the porosity value n0 .
Once the air entry value is updated, the ad and aw
parameters are recalculated as function of the PAEV :
The index v in (av ; nv ) is identical to w in the wetting

case and identical to d in the drying case.
Additionally, in order to capture the phenomenological behaviour of unsaturated soils, the non-linear and
hysteretic characteristics of the water retention curve
should be determined, including the so-called scanning curves. A new approach inspired by the work of
Mualem (Mualem 1974) is adopted. Smooth interpolation equations are developed relating the degree of
saturation on scanning curves to those on the boundary
curves. The drying scanning curve, which starts from
suction s1 , is given by:
where Srw (s) = the water content on the boundary wetting curve at the suction value s; Srd (s) = the water
content on the boundary drying curve at suction s.
Similarly, the wetting scanning curve, which starts
from suction s2 , is given by:
Srd and Srw correspond to the offset given by the

two sets of scanning equations at the suction value of
cycle changing.
The generalized Darcys law is used to describe the
relative velocity of the fluid with respect to the solid
skeleton, for the flow problem. In general, the permeability tensor Kw depends on the suction value s.
Assuming hydraulic isotropic conditions, the water
relative permeability coefficient Kw is defined according to the following relationship, where k and nk
being material parameters, and K is the intrinsic
permeability:
3.3 Coupled behavior using the Bishop effective

stress
The stress-state variables used in this model are effective stress tensor ij and suction s multiplied by the
porosity n, while the dual variables are the soil skeleton
strain tensor ij and the degree of saturation Sr .
According to Equations (14) to (18) the degree of
saturation Sr is an incrementally non linear function
of suction s and porosity n.
This proposed stress framework seems to be the best

choice, provided that most of the existing finite element codes are programmed for saturated soils. Consequently, a natural extension to partially saturated states
is provided, by simply exchanging Terzaghi effective
stresses with Bishops generalized stresses.
In the incremental description of the model, the
incremental form of Bishops stress tensor is written as:
The most common formulation for factor is

(Bishop 1959; Lewis and Schrefler 1987; Nuth and
Laloui 2008): = Sr . For sake of simplicity in the
derivative, the parameter has been taken as function of the suction similar to the Van-Genuchtens
equation (Equation (14)) with (a ; n ) parameters.
These parameters are defined in a way to keep the
value between the two boundary water retention
curves (lower than the main drying curve and higher
than the main wetting curve). Using this definition,
the phenomenon of loading collapse of a soil when it
is re-wetted under constant load can be predicted by the
effective Bishop stress (Arairo et al. 2012). Assuming
that the effective Bishop stress depends on the product
s and not only on the suction value s, the product s
depends on the variation of the factor. When is
not constant, the product s, may not be monotonous.
360
Table 1. Material parameters used to describe the mechanical and hydraulic behaviour of unsaturated silt.
Mechanical
parameters
E
c0
cf
c0
e0
Bp
Bc
cf
ef
cf
ef
pc0
69 MPa
0.15
15 kPa
23 kPa
6.9
6.9
0.001
0.0085
26
26
2
2
482 kPa
0.015
Water retention
ad
nd
aw
nw
Srsat
Srres
K
1.6
1.38
100
1.38
1
0.1
4.106 m/s
Coupling
parameters
a
n
k
nk
n0
2.0
2.15
2.0
2.15
0.1
0.4
Figure 3. Finite element mesh, boundary and initial conditions for the first and second steps of calculation.
parameters for the coupled model were calibrated on

the basis of triaxial, oedometer and retention data for
a silty sand (Cui et al. 1996). All the parameters are
summarised in Table 1. The geometry of the studied
slope is presented in Figure 2. The steep slope of 10
is composed of silty sand.
4.1
Figure 2. Finite element mesh of the studied slope.
Qualitatively, function s can exhibit a peak depending

on the values of parameters a and n .
Therefore, for soils exhibiting a peak in the s versus s plane, the shrinkage observed after a rewetting
cycle can be described with the effective stress concept of Bishop, since the product s can be increased
by decreasing suction. Thus, the effective mean pressure will increase as well, which leads to the observed
shrinkage, when it is re-wetted under an oedometric or
isotropic path.
The constitutive model presented above is implemented into the French finite element code called
Cast3M (Verpeaux et al. 1989).
NUMERICAL MODELLING OF AN
UNSATURATED SOIL SLOPE SUBJECTED
TO RAINFAL INFILTRATION
The main objective of this study is to show the effect

of rain infiltration on the stability of a partially saturated slope and its failure mechanisms. Despite the fact
that this study is completely numerical, the hydromechanically coupled analysis is inspired from realistic
geometrical and material parameters. The material
Definition of the coupled HydroMechanical

Model and the calculation Procedure
The 3D finite element mesh presented in Figure 2

is composed of twenty-noded quadratic cubes. The
hydromechanical coupled model is applied here using
the parameters of Table 1. A fine mesh is used
throughout the soil mass in order to properly model
the hydromechanical processes and the development
of plasticized zones during rain infiltration.
Two successive calculation steps are run. They consist of a gravity loading phase and a coupled water
pressure and deformation analysis to determine the
water table position and distribution of matrix suction
in the unsaturated zone. The calculation starts from
fully saturated soil mass.
On the first step, a constant negative water pressure
(positive suction) was applied on the soil top surface
(See Figure 3). During all the calculation steps, the displacement of the lower surface has been blocked and
the lateral surfaces were considered impervious. The
resulting water table runs inclined approximately parallel to the slope (Johnson and Sitar 1990). The matrix
suction is distributed hydrostatically in the partially
saturated zone above the water table and the degree of
saturation reaches Sr = 0.754 along the slope surface,
see Figure 4.
In a final step, rain infiltration is simulated by
means of an imposed boundary positive water pressure
which increases up to 22 kPa.
The water saturation distribution in Figure 5 shows
clearly that the slope surface has been quasi completely
saturated. The degree of saturation has increased to
reach the maximum saturation upon wetting or a near
value.
361
Figure 4. Spatial distribution of the matrix suction and

degree of saturation after the first step.
4.2
Figure 5. Spatial distribution of the matrix suction and

degree of saturation along the rain infiltration through on
the slope surface.
Plastic Mechanisms induced by Rainfall

Infiltration
Adding water on a partially saturated slope affects negatively on its stability. By increasing the water content,
the capillary forces prevail between the soil particles.
The de-bonding effect of wetting is taken into account
in the model through the effective stress (Equation
(21)), which decreases with rain infiltration.
Figure 6 shows the deformed slopes after rainfall
infiltration. The shape of slope deformation shows that
the maximum shear stress is reached on the top of the
slope surface. This factor is significant for debris flow
initiation mechanisms, which could take place on the
upper part of the slope where the shear stress reaches
= 25.4 kPa. The maximum shear stress value according to the Mohr-Coulomb criterion is max = 26.8 kPa
with an actual cohesion of c = 18.26 kPa and friction
angle = 21.2 .
Figure 6. Deformed slope after rainfall infiltration (100
times amplified).
CONCLUSION
Several hydromechanical processes act in a destabilizing way on a slope. The degree of saturation of the
upper soil layer increases, thereby reducing the capillary tension between the soil particles which weakens
the slope. Additionally, due to the mobilized fluid
flow inside the soil matrix, the fluid exerts a destabilizing, downhill frictional drag. In order to model
numerically the onset of rainfall-induced landslides

in variably saturated-unsaturated conditions, a consistent stress framework needs to be chosen. For this
purpose, a theoretical framework (Arairo et al. 2012)
to couple hydraulic (water retention) and mechanical
(stress-strain) behavior has been used. Due to adequate constitutive law, the model gives encouraging
362
results and detailed information about the active mechanisms during rain infiltration. The applicability of
these studies for natural hazards requires a further
effort to enhance the description and the modeling
of rainfall, which is responsible for landslide triggering. In particular, it is fundamental to assess the role
of antecedent rainfall in different geo-environmental
settings, to achieve a better spatialization of rainfall
data at large-scale. Then, static liquefaction and rapid
changes in pore volume should be analyzed together
with influence of unsaturated conditions.
REFERENCES
Arairo, W. & Prunier, F. & Djran-Maigre, I. & Darve, F. 2012.
A new insight into modelling the behaviour of unsaturated
soils. International Journal for Numerical and Analytical
Methods in Geomechanics. DOI: 10.1002/nag.2151.
Barnichon, J. D. 1998. Finite element modelling in structural
and petroleum geology. (PhD thesis). University of Liege.
Bishop, A. W. 1959. Principle of effective stress. Teknisk
Ukeblad. 106(39): 859863.
Bonnard, C. & Forlati, F. & Scavia, C. (eds) 2004. Identification and mitigation of large landslide risks in Europe:
Advances in risk assessment, IMIRILAND Project,
Leiden, Balkema.
Cui, Y. J. & Delage, P. 1996. Yielding and plastic behaviour

of an unsaturated compacted silt. Gotechnique. 46(2):
291311.
Dai, C.F. & Wang, S. 1999. Analysis of rainstorm-induced
slide-debris flows on natural terrain of Lantau Island,
Hong Kong. Engineering Geology. 51 (4): 279290.
Godt, J.W. & Baum, R.L. & Lu, N. 2009. Landsliding in partially saturated materials. Geophysical Research Letters.
36, L02403, 15.
Johnson, K.A. & Sitar, N. 1990. Hydrologic conditions
leading to debris flow initiation. Canadian Geotechnical
Journal. 27 (6): 789801.
Lewis, R.W. & Schrefler, B.A. 1987. The Finite Element
Method in the Deformation and Consolidation of Porous
Media, 15th edition. Wiley Interscience publication,
99139.
Mualem, Y. 1974. A conceptual model of hysteresis. Water
Resources Research. 10: 514520.
Nuth, M. & Laloui, L. 2008. Effective stress concept in
unsaturated soils: clarification and validation of a unified framework. International Journal for Numerical and
Analytical Methods in Geomechanics. 32:771801.
Verpeaux, P. & Millard, A. & Charras, T. & Combescure, A.
1989. A modern approach of large computer codes for
structural analysis. Proceedings of the 10th conference
on structural mechanics reactor technology, Los Angeles.
363

A model for anaerobic degradation of municipal solid waste

Y.M. Chen, R.Y. Guo, Y.C. Li & L.T. Zhan
MOE Key Laboratory of Soft Soils and Geoenvironmental Engineering, Department of Civil Engineering,
Zhejiang Univ., Hangzhou, China
ABSTRACT: A two-stage anaerobic digestion model for degradable components of municipal solid waste is
presented in this paper. The aim of developing this model is to describe the decomposition process for coupled
modeling of municipal solid waste landfills. The hydrolysis/acidogenesis stage is assumed to follow the firstorder kinetics while the methanogenesis stage is represented by the Monod kinetics. Volatile fatty acid is regarded
as the primary variable correlating these two stages. The methanogenesis rate is dependent on the growth and
decay of methanogenic biomass. Five formulas are obtained to describe the changes of solid degradable fraction,
volatile fatty acid concentration, methane/carbon dioxide generation, water content and methanogenic biomass
concentration with time. The effects of moisture, volatile fatty acid and pH on hydrolysis and methanogenesis
are taken into consideration. A laboratory batch experiment is simulated by the presented model, and the results
are in a good agreement with experimental data.
INTRODUCTION
Degradable components of municipal solid waste

(MSW) are decomposed after being placed in landfills. The decomposition process is subject to coupled mechanical, hydraulic, bio-chemical and thermal
effects, accompanied by occurrence of landfill settlement and release of biogas and leachate. In order to
describe the complex behaviors in landfills, a comprehensive consideration of all these behaviors is needed.
This paper presents a degradation model for simulation
of coupled phenomena in MSW landfills.
Degradation of MSW is variable and complicated.
Researchers have identified the process into three to
seven or even more phases, based on varied criteria. A
typical five-phase characterization (Parker, 1983; The
Interstate Technology & Regulatory Council, 2006)
includes a lag phase (aerobic phase), a transition phase,
an acid phase, methane production phase and stabilization phase. As shown in Figure 1, each phase is
characterized by the quality and quantity of leachate
and landfill gas produced, representing a change in
the biochemical processes within the landfill. In terms
of the amount of degraded MSW and gas production, anaerobic reactions (phases II to V) are dominant,
and the aerobic phase is ignorable in most cases. So
the degradation models below only consider the last
four phases. Methane production rate, concentration of
volatile fatty acid (VFA), pH and mass of degradable
solid waste per total volume are important indicators
of decomposition evaluation.
A degradation model including many phases is difficult to be incorporated with other equations in the
coupled models of MSW landfills. As a result, existing
Figure 1. Schematic representation of waste decomposition

phases (Parker, 1983).
coupled MSW landfill models simplified the degradation process to some extent. Empirical first-order
models (e.g. Findikakis et al. 1988; Xue et al. 2007;
Chen et al. 2011), which simplify the degradation into
a stage with relatively few parameters, are able to
365
Figure 2. Schematic of two-stage anaerobic MSW degradation.
give the rates of biodegradation and methane production.The ignorance of intermediate products makes the
first-order models incapable of considering important
landfill behaviors, such as the migration of substances
in leachate. Some researchers (e.g. El-Fadel et al.
1996a; Haarstrick et al. 2001; White et al. 2003, 2004)
adopted multi-phases degradation models, and used
Monod kinetics to describe biodegradation reactions.
However, a great number of parameters are required
by these models. And two-stage degradation models
(e.g. McDougall et al. 2007; Gholamifard et al. 2008)
are chosen for compromise. Hydrolysis/acidogenesis
and methanogenesis describe the transformation of
degradable solid to biogas. VFA is regarded as the
primary substance correlating these two stages.
In this paper, an advanced two-stage anaerobic
degradation model is presented. It considers inhibition caused by rapid hydrolysis of easily hydrolysable
solid organics and release of intracellular water. The
inhibitions are both on further hydrolysis and methanogenesis, and may even cause the process to be totally
stopped. The acid tomb phenomenon is common in
landfills of developing countries, whose MSW has a
high percentage of kitchen waste (major source of easily hydrolyzed matter). The model is finally applied to
simulation of a laboratory batch experiment.
MODEL FRAMEWORK
The kinetic scheme of the presented anaerobic degradation model is shown in Figure 2. Following
assumptions are made: (a) based on hydrolysis rate,
degradable solid is divided into two fractions: easily
hydrolysable fraction and slowly hydrolysable fraction, and both of them use C6 H10 O5 as formula for
simplicity; (b) hydrolysis and methanogenesis are
rate-limiting steps, which means that acidogenesis
(including acetogenesis) reacts instantly; (c) all transformation pathways of intermediates to methane are
lumped together as a single process. The degradation process can be expressed by two stoichiometric
equations (McDougall et al. 2007):
Hydrolysis of 1 mole of degradable solid (162 g)

produces 1 mole of intermediate product acetic acid
(60 g), a representative of VFA, and 3 moles of carbon
dioxide (132 g) and 3 moles of methane (48 g).
HYDROLYSIS
Hydrolysis means that both the solubilization of

insoluble particulate matter and the biological decomposition of organic polymers turn to monomers or dimmers, which can pass the cell membrane (Gavala et al.
2003). MSW consists of different types of substances:
kitchen waste, papers, wood, leaves, plastic, metals
and so on. The hydrolysis rates of these substances
are dependent on their chemical compositions. The
fast hydrolysis/acidogenesis of some organics brings
accumulation of VFA, which may inhibit hydrolysis of
slowly hydrolysable solid.
Intracellular water is released as cell membrane
breaks due to compression or degradation. Since this
model includes no mechanical effects, it is assumed
that intracellular water is released at a rate proportional to hydrolysis rate and that hydrolytic reactions
are not infected by intracellular water.
3.1 Kinetics of hydrolysis
A modified first-order reaction is developed to represent the hydrolysis/acidogenesis process, with consideration of the influence of water content and VFA
inhibition:
where Si is degradable solid fraction, i.e. the remaining

mass of degradable solid i (dry weight) per unit total
volume of MSW (kgm3 ): i = 1, easily; i = 2 slowly;
t is time (day); fW () describes the influence of water
content (w/w) on hydrolysis; fIh (c) is the inhibition
function of VFA on hydrolysis, c is the concentration of VFA (kgm3 ); khi is first-order hydrolysis rate
constant of degradable solid i (day1 ).
Hydrolysis cause changes of water content by consuming water and releasing intracellular water as
follows:
where is volumetric water content; W is density of

water (kgm3 ); is the mass ratio of released intracellular water to correspondingly hydrolyzed degradable
solid; W is the mass ratio of consumed water to correspondingly hydrolyzed solid, and is 18/162 according
to degradation stoichiometric equation (see Equ. (1)).
Water content can be written by volumetric water
content and degradable solid fraction Si :
VFA is the product of hydrolysis/acidogenesis, and

its accumulation rate, rg (kgm3 day1 ):
366
exist. Veeken et al. (2000) concluded that no inhibition by ionized VFA or non-ionized VFA can be
measured at pH values between 5 and 7, and the firstorder hydrolysis rate constant has a linear relation
with pH in this range. Pan et al. (2006) designed a
set of experiments to elucidate that both pH and VFA
(ionized and unionized) can prevent organics from
being hydrolyzed. A function proposed by Vavilin et al.
(2006), which considers the combining effects of pH
and VFA, is adopted in the presented model:
Figure 3. Influencing function of water content.
where VFA is the mass ratio of VFA produced to correspondingly hydrolyzed solid, and is 48/162 according
to degradation stoichiometric equation (see Equ. (1)).
where Kh is inhibition constant (kgm3 ) and nh is

inhibition degree index. It is noted that both easily
and slowly hydrolysable solid are influenced by VFA
accumulation and subject to the same function.
3.2 Influence of water content
Water is one of the most important influencing factors

on MSW degradation. Hydrolysis, microbial growth
and migrations of microorganisms and chemical substances depend on water and its transport. Based on
experimental results, Rees et al. (1980) described that
under anaerobic conditions, gas production rate is a
log-linear function of water content (w/w) within the
range of 25%75%. El-fadel et al. (1999) mentioned
that microbial activity in MSW has doubled from 26%
to 46% water content (w/w). Gholamifard et al. (2008)
indicated that degradation would stop when moisture
is under a certain value. The increase of water content
has two main positive effects on hydrolytic step: providing sufficient water for reactions and diluting VFA
inhibition.
Based on works done by Pommier et al. (2007) and
Meima et al. (2008), an influencing function is used
in the presented model:
Methanogenesis is the conversion of intermediate products to methane by metabolic activities of

methanogenic organisms. Hydrogenotrophic and acetoclastic methanogenesis are two primary pathways,
utilizing hydrogen and VFA (acetate) as substrate
respectively. They are integrated into a single step in
the presented model, described by the same biomass
and kinetics.
where min is minimal water content, below which

hydrolysis stops; max is maximum water content,
above which moisture is no more a limiting factor.
Figure 3 shows the image of this funtion.
3.3 VFA inhibition on hydrolysis
VFA is the primary intermediate product during
biodegradation of MSW, and a balance between the
rates of hydrolysis/acidogenesis and methanogenesis should be achieved (Vavilin et al. 2003). If
VFA hydrolyzed cannot be consumed by methanogen,
VFA will accumulate and impede further hydrolysis.
Regarding this inhibitory effect, contradictory views
METHANOGENESIS
4.1 Kinetics of methanogenesis

The depletion of substrate and growth of methanogenic
biomass (MB) is described by Monod kinetic, accounting for VFA inhibitory effect, hence for MB accumulation, rj (kgm3 day1 ):
where fI (pH) and fIm (c ) are functions of pH and unionizedVFA inhibitions on microbial growth respectively;
c is the concentration of VFA unionized (kgm3 ).
The combination of both factors is the total VFA
retarding effects, which is discussed at length in section 4.2; km is the maximum specific growth rate of
methanogenic biomass (day1 ); KS is the half saturation constant (kgm3 ); m is the concentration of
methanogenic biomass (kgm3 ). The consumption
rate of VFA, rk (kgm3 day1 ), is directly correlated to MB accumulation with a cell/substrate yield
coefficient, Y :
Methane is produced at the rate, vm (kgm3 day1 ):
367
Table 1.
Figure 4. Inhibition function of pH on MB growth.
where m is mass ratio of methane produced to correspondingly consumed VFA, and is 48/60 based on
degradation stoichiometric equation (see Equ. (1)).
The MB decay is described by the first-order kinetics,
rd (kgm3 day1 ):
where kd is the MB death rate (day1 ).

The changes of MB concentration (m) andVFA concentration (c) can then be obtained using Equs. (8) and
(11) and Equs. (5) and (9), respectively:
Composition of experimental waste
Materials
Mass percentage
Classification
Kitchen waste
Fruits/vegetables
Leaves
Papers
Slag
Coal ash
Soil
15.7
61.9
3.1
3.2
4.2
8.1
3.8
Organics
Organics
Organics
Organics
Inorganics
Inorganics
Mixtures
neutral environment, and in this way its ability of producing methane is undermined. A function proposed
by Vavilin et al. (2003) is used herein to represent such
effect:
where Km is inhibition constant (kgm3 ) and nm is

inhibition degree index.
Leachate in MSW landfills is a natural buffering
system. The pH value can be calculated based on ionization balance of acids and charge balance, and the
concentration of unionized VFA can then be deduced
as well.
5
SIMULATION
In order to verify the effectiveness and applicability

of the presented model, a laboratory batch experiment (Zhou et al. 2012) is simulated herein. The
presented model is solved by iterative computations
via Microsoft Excel.
4.2 VFA inhibition on methanogenesis
5.1 Experimental Setup
Methanogenic bacteria are extremely sensitive to pH

fluctuations. VFA accumulation will lower pH and
then damage environment for MB growth. Clark et al.
(1970) carried out a series of experiments and found
out the relationship between MB growth rate and pH,
which is shown in Figure 4. It can be seen that MB has a
favorite grow condition when pH is near 7. The expression proposed by Angelidaki et al. (1992) as inhibition
function fI (pH) is adopted in the presented model:
The waste was typical Chinese MSW, featuring in high

water content and high proportion of easy degradable organics. The specific composition is listed in
Table 1. The waste was shredded into pieces less than
2 cm, filled into the container (47.1 cm inside diameter) and compacted. The refuse weighed 112.0 kg,
and was 80 cm in height. The initial dry weight of
the waste was 28.47 kg, and 62.4% (mass ratio) of it
could be decomposed. During filling process, anaerobic activated sludge was added when certain heights
were reached, assuring a high initial concentration of
anaerobic bacteria. The degradation was processed at
room temperature and lasted for 360 days. A number of environmental factors and mass of solid, liquid
and gas phases were monitored and measured periodically. Every 30 days, all leachate collected at bottom
was recirculated from the top without treatment.
where pKl and pKh are the lower and upper pH values
where the MB growth rates are approximately 50%
of the uninhibited rate, and are 5.25 and 8.75 respectively. As shown in Figure 4, this expression fits well
experimental data.
Zoetemeyer et al. (1982) pointed out that unionized VFA could enter the cells of MB, lowering the
pH inside. MB has to consume more energy to keep a
5.2 Model Parameters

The degradation process was regarded to take place
in 1.0 m3 waste. The initial values for mass of dry
368
Table 2.
Biological parameters for simulation
Variable
Value (Unit)
kh
km
kd
w
w
VFA
m
min
max
Kh
Km
nh
nm
KS
Y
pKh
pKl
0.004 (day1 )
0.02 (day-1)
0.02 (day1 )
1000 (kgm3 )
1.0 (/)
18/162 (/)
48/162 (/)
48/60 (/)
0.12 (/)
0.50 (/)
18 (kgm3 )
10 (kgm3 )
4.0 (/)
3.0 (/)
4.0 (kgm3 )
0.08 (/)
8.75 (/)
5.25 (/)
Figure 5. Relationship between degradable solid fraction

and time.
waste and degradable solid (no inclusion of intracellular water) per unit total volume were 204.3 kg and
127.5 kg respectively. From table 1, we can see that
kitchen waste, fruits, vegetables and leaves accounted
for the majority of organics. All these components
were wet and biological, which means a large
amount of intracellular water existed in the waste.
It is assumed that the weight of initial intracellular
water was 127.5 kg per unit total volume, i.e. its mass
ratio to degradable solid was 1.0, considering some
water release in compaction. Accordingly, free water
weighed 471.8 kg per unit total volume at the beginning. Initial volumetric water content () was 47.2%
and water content (, w/w) 58.7%. A high value for
initial concentration of MB, 0.18 kgm3 , is adopted
as activated sludge was added in the experiment. The
initial concentration of VFA in leachate is assumed
to be 0.01 kgm3 . Other biological parameters have
been chosen following Fadel et al. (1996), Vavilin et al.
(2002, 2006) and an literature review accomplished by
Meima et al. (2008), and are listed in Table 2.
Figure 6. Relationship between pH value and time.
5.3 Results and discussion

A comparison of degradable solid fraction, concentration of VFA and pH of leachate between batch
experimental data and the results obtained by the presented model are made here. As shown in Figures 57,
the results obtained have a good correlation with the
experimental results (see Figures 5 and 6).
Figure 6 shows that the obtained pH value, based
on biochemical reactions and buffering system within
leachate, is in a good agreement with results of laboratory batch experiment. The validity of pH in a long
term is uncertain since the limitation of experimental
time. Given the production of ammonia at the later
reaction stage, the obtained value might be lower than
the real situation.
Figure 7. Relationship between VFA concentration and

time.
The simulated VFA has a longer period at high concentration level (see Figure 7) though the time of peak
concentration and the time of low concentration are
captured by the presented model. Further modification
and verification are required for the presented model.
369
CONCLUSIONS
This paper presents an advanced two-stage anaerobic digestion model for degradable components of
MSW. The hydrolysis/acidogenesis stage is assumed
to follow the first-order kinetics while the methanogenesis is described by the Monod kinetics. The effect
of water content on hydrolysis rate and the effects of
VFA concentration on hydrolysis and methanogenesis
are considered in the presented model. Simulation of
a laboratory batch experiment demonstrates the effectiveness and applicability of the presented model. This
model is helpful to improve the coupled modeling of
MSW landfills if it is incorporated.
For developing countries, MSW has a high percentage of kitchen waste, which means it has more easily
hydrolysable waste and intracellular water. Classification of easily and slowly hydrolysable organics allows
the model to consider varied degradation rates of MSW
components. Intracellular water release, which accompanies hydrolysis process, is included in the model
to account for the MSW having large amount of this
type of water. The modifications above enable this presented model to simulate the degradation process of
MSW in developing countries well.
ACKNOWLEDGEMENT
The financial supports received from the National
B-asic Research Program of China (973 Program)
via Grant No. 2012CB719802 and from the National
Natural Science Foundation of China (NSFC) via
Grant Nos. 51009121, 51010008 and 10972195 are
gratefully acknowledged.
REFERENCES
Angelidaki, I., Ellegaard, L. & Ahring, B. (1992). Compact
automated displacement gas metering system for measurement of low gas rates from laboratory fermentors.
Biotechnology and bioengineering, 39, 351353.
Chen, Y. M., Xu, X. B. & Zhan, L. T. (2011). Analysis
of solid-liquid-gas interactions in landfilled municipal
solid waste by a bio-hydro-mechanical coupled model.
SCIENCE CHINA Technological Sciences, 19.
Clark, R. & Speece, R. (1970). The pH tolerance of anaerobic
digestion.
El-Fadel, M., Findikakis, A. & Leckie, J. (1996). Numerical modelling of generation and transport of gas and heat
in landfills I. Model formulation. Waste management &
research, 14, 483504.
El-Fadel, M. (1999). Leachate recirculation effects on settlement and biodegradation rates in MSW landfills.
Environmental Technology, 20, 121133.
Findikakis, A., Papelis, C., Halvadakis, C. & Leckie, J.
(1988). Modelling gas production in managed sanitary
landfills. Waste management & research, 6, 115123.
Gavala, H., Angelidaki, I. & Ahring, B (2003). Kinetics and
modeling of anaerobic digestion process. Biomethanation
I, 5793.
Gholamifard, S., Eymard, R. & Duquennoi, C. (2008).

Modeling anaerobic bioreactor landfills in methanogenic
phase: Long term and short term behaviors. Water
Research, 42, 50615071.
Haarstrick, A., Hempel, D., Ostermann, L., Ahrens, H. &
Dinkler, D. (2001). Modelling of the biodegradation of
organic matter in municipal landfills. Waste management & research, 19, 320331.
McDougall, J. (2007). A hydro-bio-mechanical model for
settlement and other behaviour in landfilled waste. Computers and Geotechnics, 34, 229246.
Meima, J., Naranjo, N. M. & Haarstrick, A. (2008) Sensitivity
analysis and literature review of parameters controlling
local biodegradation processes in municipal solid waste
landfills. Waste Management, 28, 904918.
Pan, X. J., He, P. J., Lv, F. & Shao, L. M. (2006). The inhibition of the ph and acetate on the anaerobic hydrolysis of
biodegradable organic waste. Environmental Chemistry,
25(4), 449453.
Pommier, S., Chenu, D., Quintard, M. & Lefebvre, X. (2007).
A logistic model for the prediction of the influence
of water on the solid waste methanization in landfills.
Biotechnology and bioengineering, 97, 473482.
Rees, J. F. (1980). The fate of carbon compounds in the
landfill disposal of organic matter. Journal of Chemical
Technology and Biotechnology, 30, 161175.
Shao, L. M., He, P. J. & Qu, X. (2006). Impact of
VFA concentration of recirculated leachate on stabilized
methanogenetic landfill layer. Research of Environmental
Sciences, 19(5), 6569.
Vavilin,V.A., Rytov, S.V., Lokshina, L.Y., Pavlostathis, S. G. &
Barlaz, M. A. (2003). Distributed model of solid waste
anaerobic digestion: effects of leachate recirculation and
pH adjustment. Biotechnology and bioengineering, 81,
6673.
Vavilin, V. A., Jonsson, S., Ejlertsson, J. & Svensson, B. H.
(2006). Modelling MSW decomposition under landfill
conditions considering hydrolytic and methanogenic inhibition. Biodegradation, 17, 389402.
Veeken, A., Kalyuzhnyi, S., Scharff, H. & Hamelers, B.
(2000). Effect of pH and VFA on hydrolysis of organic
solid waste. Journal of environmental engineering, 126,
10761081.
White, J. K., Ren, Q. & Robinson, J. P. (2003). A framework
to contain a spatially distributed model of the degradation
of solid waste in landfills. Waste management & research,
21, 330345.
White, J., Robinson, J. & Ren, Q. (2004). Modelling the biochemical degradation of solid waste in landfills. Waste
Xue, Q., Zhao, Y., Liu, L. & Lu, H. J. (2011). Study
of thermo-hydro-mechanical-chemical coupling effect of
catastrophe process of landfill. Chinese Journal of Rock
Mechanics and Engineering, 30(10), 19701988.
Zacharof, A. & Butler, A. (2004). Stochastic modelling
of landfill leachate and biogas production incorporating
waste heterogeneity. Model formulation and uncertainty
analysis. Waste Management, 24, 453462.
Zhou, X. Z., Sang, S. X., Cao, L. W. & Ji, X. Y. (2012).
Three-phase evolution and its control methods of waste
biomass in landfill. Journal of Central South University,
43(2), 769775.
Zoetemeyer, R., Matthijsen, A., Cohen, A. & Bolhouwer,
C. (1982). Product inhibition in the acid forming stage
of the anaerobic digestion process. Water Research, 16,
633639.
370

Water retention and shrinkage properties of a cemented dredged material

A.C.F. Chiu, X.P. Hu, C.W.W. Ng & G.H. Lei
Key Laboratory of Ministry of Education for Geomechanics and Embankment Engineering,
Hohai University, Nanjing, China
ABSTRACT: Cement-based solidification/stabilization is one of the popular technologies to remediate dredged

material (DM).A laboratory study is presented to investigate the effects of cement content and initial water content
on the water retention and shrinkage properties of a cemented DM. It is found that the air-entry value of water
retention curve increases significantly with increasing cement content, but it decreases with increasing initial
water content. The amount of shrinkage decreases with increasing cement content. For weakly cemented DM,
the shrinkage increases with increasing initial water content. It is postulated that the modifications of the water
retention and shrinkage behavior of cemented DM are related to the change in the pore-size distribution and the
cement bonds at the interparticle contacts, respectively.
INTRODUCTION
The demand for dredging operations in China has

increased tremendously in recent years, for instance
the volume of dredged materials (DM) involved in the
ocean disposal follows an increasing trend in the last
decade and the annual volume reached around 160
million m3 in 2011 (SOA 2012). Thus, utilization of
the DM for beneficial use is one of the options under
consideration to reduce the need for new disposal sites.
Cement-based solidification/stabilization is one of the
popular technologies for the treatment of DM and
hazardous materials (Connor 1990). Cement is used
as a binding agent and mixed with the soft or hazardous material to produce a mechanically stable soil
matrix that exhibits higher strength and stiffness, and
can also encapsulate possible contaminants (Mitchell
1981, Meegoda et al. 2000). The mechanical behavior
of cement-treated soils has been widely studied in the
past (Tremblay et al. 2001, Chew et al. 2004, Lee et al.
2005, Horpibulsuk et al. 2005, Zhu et al. 2007 & 2009,
Chiu et al. 2009).
Water retention curve (WRC) represents a relationship between water content and suction of an
unsaturated soil. It is one of the hydraulic properties
for numerical modeling of contaminant transport in
unsaturated soils. Thus, the WRC of cement-treated
material is an essential input in the numerical simulation for long-term performance of cement-based
solidification/stabilization for encapsulating contaminants under different climate conditions, e.g. wet-dry
cycle, However, it has been rarely reported in the literature (Puppala et al. 2006, Hoyos et al. 2007). The
volume change of specimen is normally neglected in
the conventional measurements of WRC. However,
soils normally shrink during drying. Recent studies
(Nuth & Laloui 2008, Sheng & Zhou 2011, Chiu &
Ng 2012) have shown that the water retention and
volumetric behaviors are coupled and the WRC is
influenced by the amount of shrinkage. As a result,
the shrinkage curve is also important in evaluating the
water retention properties of a material.
The main objective of the paper is to study the
effects of cement-based treatment on the water retention and shrinkage behavior of DM. A series of WRC
and shrinkage tests conducted on cemented and uncemented DM are first presented. The test results of
WRC are then evaluated by van Genuchtens (1980)
equation, from which the effects of cement content,
initial water content and shrinkage volume on the
shape of WRC are discussed.
2 TESTED MATERIALS AND TESTING
PROGRAMME
2.1 Tested materials
The tested soil used in the study was a river sediment
taken from Yangtze river, China. The basic physical properties of the soil are summarized in Table 1.
According to the Unified Soil Classification System
(USCS), the soil is classified as clay of high plasticity (CH). Type I Ordinary Portland cement (OPC) was
used as the cementing binder.
2.2 Specimens preparation
The initial water content of the soil slurry was maintained around 52%104%, i.e. between 12 times of
liquid limit. The cemented samples were formed by
mixing the soil slurry with dry OPC powder to achieve
371
Table 1.
A compact centrifuge (Hitachi CR21) with an operable radius of 98 mm was used to measure the WRCs
for suction ranging from 0300 kPa. The diameter and
height of the specimen holder is 50 mm and 51 mm,
respectively.The trial run showed that a duration of 180
minutes is sufficient for the specimen to attain equilibrium condition under the centrifugal acceleration. All
tests were conducted at a controlled temperature of
20 C. For an angular velocities of 6,000 rpm, the corresponding applied suction is around 316 kPa. After
each round of centrifugal acceleration, the dimensions
of specimen were measured by the caliper to evaluate
the volume change induced by the target suction.
The contact filter paper method was conducted
in accordance with the procedures recommended in
ASTM (2010). Whatman No.42 filter paper was used
and the following calibration equations suggested by
Leong et al. (2002) were used to convert the equilibrium gravimetric water content of filter paper (w in %)
to suction ( in kPa):
Basic physical properties of tested soil.
Specific gravity
Percentage of sand
Percentage of silt
Percentage of clay
Liquid limit (%)
Plastic limit (%)
Plasticity index (%)
2.6
1
62
37
52
24
28
Table 2. Testing conditions.
Specimen ID
Cement
content
(kg/m3 )
Initial water
content
(%)
Void ratio
before
drying
C50_1L
C200_1L
C50_1.5L
C200_1.5L
C50_2L
C200_2L
50
200
50
200
50
200
52
52
78
78
104
104
1.36
1.12
1.97
1.55
2.50
2.06
cement content (CC) of 50 kg/m3 and 200 kg/m3 . The

mixing was carried out as fast as possible inside a
domestic mixer to avoid hardening of the cement-soil
mixture. After mixing thoroughly, the mixture was
placed into stainless steel molds 50 mm in diameter
and 80 mm in height.The mixing and placing was completed at room temperature and took approximately
20 minutes. The specimens were cured inside an environmental chamber for 56 days, where the ambient
temperature and relative humidity were maintained at
202 C and higher than 90%, respectively.
2.3 Testing programme and procedures
The testing conditions of WRCs and shrinkage curves
are summarized in Table 2. It should be noted that
the initial void ratio of specimens at CC = 200 kg/m3
is smaller than that at CC = 50 kg/m3 because more
cement is added to the soil slurry for a given initial
water content.
The WRCs were measured by the centrifuge method
for suction ranging from 0300 kPa (Gardner 1937,
ASTM 2008) and filter paper method for suction
greater than 300 kPa (ASTM 2010). The suction (
in kPa) generated under centrifugal force is evaluated
by the following equation:
where r1 is the radial distance to the free water surface

(cm), i.e. the bottom of the specimen in this study, r2
is the radial distance to the midpoint of the specimen
(cm) and n is the angular velocity (rpm). Equation
1 shows that is a function of the angular velocity.
After reaching the equilibrium of target suction, the

dimensions of specimen were measured by the caliper.
3 WATER RETENTION CURVES
3.1
Effect of cement content
Figures 1 to 3 depicts the WRCs of cemented DM at

CC = 50 and 200 kg/m3 for an initial water content
of 52, 78 and 104%, respectively. The WRC of uncemented DM at an initial void ratio of 0.9 is also shown
in the figure for comparison. It should be noted that
the slurry is a highly compressible material. Hence,
the uncemented sample was first consolidated in a
consolidometer to a smaller void ratio, i.e. a higher
strength, such that the disturbance of specimens during the WRC tests can be minimized. It is seen from the
figures that for a given initial water content, the WRC
of cemented DM shifts to the right for a higher CC and
that of uncemented DM lies on the furthest left. In other
words, the cemented DM can hold more water in the
high suction regime (suction above 1500 kPa) and the
water storage capacity increases with increasing CC.
Air-entry value (AEV) is the suction beyond which
air commences to enter a saturated soil. AEV is estimated from the empirical procedures recommended by
Fredlund and Xing (1994), which is shown as closed
square in the figures. It can be seen that the AEV
increases with CC for a given initial water content.
Table 3 summarizes the AEVs of cemented and uncemented DM. As the AEV is related to the size of largest
pores from which the saturated soil commences to
desaturate, it seems that the size of largest pore for
uncemented DM is greater than those of cemented DM
and it also decreases with increasing CC.
372
Figure 4. WRCs of cemented DM at a cement content of

200 kg/m3 .
Figure 1. WRCs of cemented DM at an initial water content

of 52%.
Table 4. Curve fitting parameters of van Genuchtens equation (1980).

of 78%.
C50_1L
C200_1L
C50_1.5L
C200_1.5L
C50_2L
C200_2L
Untreated
0.0004
0.00015
0.0009
0.0004
0.0009
0.0004
0.0006
1.3
1.3
1.1
1.1
1.1
1.1
0.8
shifts to the right for a lower initial water content.

In other words, the WRC of lower initial water content exhibits higher water storage capacity in the high
suction regime (suction above 1500 kPa). It can be
seen that the difference between water content of
78 and 104% is negligible. However, as the initial
water content reduces further to 52%, the difference
can be substantial, for example the AEV increases
from 400 kPa to 1400 kPa for decreasing initial water
content from 78% to 52%.
The measured WRCs are curved-fitted by the following empirical equation (van Genuchten 1980):
Table 3. AEV of cemented and uncemented DM.

C50_1L
C200_1L
C50_1.5L
C200_1.5L
C50_2L
C200_2L
Uncemented
Specimen ID
500
1400
200
400
200
400
140

of 104%.
3.2 Effect of initial water content

Figure 4 depicts the WRCs of cemented DM at
CC = 200 kg/m3 . It is found that the initial water content also influences the position of WRC. The WRC
where = volumetric water content, s = saturated

volumetric water content, r = residual volumetric
water content, and n are two non-negative curve fitting parameters. In this study, it is assumed that r
is equal to zero. Figures 1 to 3 show the curve-fitted
WRCs and the corresponding curve fitting parameters
are summarized in Table 4.
It seems that for a given initial water content,
decreases with increasing CC. On the other hand, n
does not change with CC for the range of CC studied. It should be noted that and n is related to
the AEV and gradient at the inflection point on the
WRC, respectively. decreases with increasing AEV.
n increases with increasing gradient at the inflection
point, i.e. loses more water for a given change of suction. It is generally recognized that the shape of WRC
373
Figure 5. Shrinkage curves of cemented DM at an initial

water content of 52%.
depends on the pore-size distribution (Ayra & Paris

1981, Fredlund et al. 2002, Simms & Yanful 2002,
2005). It may suggest that and n should be related
to the size of largest pore and the spreading of the
pore-size distribution, respectively.
The principal chemical reactions involved in the
cemented DM are cement hydration and pozzolanic
reactions. The resulting products are crystalline
hydrates (e.g. calcium hydroxide and ettringite) and
amorphous hydrates (e.g. calcium silicate hydrates,
CSH), which contribute differently to the soil structure of cemented DM. CSH has a relatively high
surface area which may fill up the larger pores of
the DM. As a result, its pore size distribution would
be changed. Despite the void ratio of cemented DM
is higher than that of uncemented DM, a smaller
in Equation (3) may suggest that the sizes of large
pores in the cemented DM are smaller than those in
the uncemented DM.
SHRINKAGE CURVES
Figures 5 to 7 shows the corresponding shrinkage

curves measured from the drying tests presented in
Figures 1 to 3, respectively. It is found that the shrinkage curves of the cemented specimens lie above that
of the uncemented specimen. In other words, the void
ratio of the cemented DM is higher than that of the
uncemented DM for a given suction. Similar phenomenon is observed in the compression curves of
cement-treated soils (Chiu et al. 2009). It is postulated
that the newly formed cement bonds at the interparticle contacts provide the mechanical stability for the
loose structure (or high void ratio) of cemented DM.
As expected the amount of shrinkage decreases with
increasing CC. For CC = 200 kg/m3 , the shrinkage
is relatively small and the corresponding volumetric
strain is less than 3% for the range of initial water
content studied. For CC = 50 kg/m3 , (see Figure 8) a
well-defined yield point is observed on the shrinkage
curve, beyond which substantial volume change commences. The corresponding suction at the yielding is
independent of initial water content and it is around 10


to 20 kPa. However, the amount of post-yield shrinkage increases with increasing initial water content.
The corresponding volumetric strains are 18, 21 and
30% for an initial water content of 52% (C50_1L),
76% (C50_1.5L) and 104% (C50_2L), respectively.
Previous studies on chemically treated materials
(Tremblay et al. 2001, Chiu et al. 2009) have shown
that the yield stress of chemically treated materials observed in compression tests can be attributed
to the structures developed from chemical reactions
like cement hydration and pozzolanic reactions. Thus,
yielding may be considered as the onset of bond
breakage at the interparticle contacts. The yielding
observed in drying test is similar to that observed
in compression test. Hence, it may also suggest that
the yielding is related to the onset of breakage of
cement bonds as the result of increase in suction. Figure 8 depicts that yielding occurs in the lower suction
regime (suction below 100 kPa) for weakly cemented
DM, e.g. CC = 50 kg/m3 . Beyond the yield point, there
are substantial breakages of cement bonds within the
soil matrix, a looser structure should exhibit higher
volumetric strain for a given change of suction.
The shrinkage curves shown in Figures 5 to 7
indicate that the amount of shrinkage of cemented
DM is influenced by both cement content and initial water content. For strongly cemented DM, e.g.
CC = 200 kg/m3 , the volume change during drying is
negligible and the initial void ratio before drying is
374
Figure 8. Shrinkage curves of cemented DM at a cement

content of 50 kg/m3 .
sufficient to account for the shape of WRC. On the contrary, for weakly cemented DM, e.g. CC = 50 kg/m3 ,
the amount of shrinkage can be substantial leading
to the modification of pore-size distribution. Hence,
apart from the initial void ratio, the amount of shrinkage should be considered to evaluate the shape of WRC
for weakly cemented DM.
5
CONCLUSIONS
This paper presents a laboratory study that investigates the influences of cement-based solidification/
stabilization on the water retention and shrinkage
properties of DM. A series of WRC and shrinkage
tests conducted on DM treated with different cement
contents and initial water contents are reported. The
key findings are summarized as follows:
(i) The air-entry value of WRC increases with
increasing cement content, but it decreases with
increasing initial water content. It is postulated that the newly formed amorphous cement
hydrates (e.g. CSH) alter the pore-size distribution of the cemented DM leading to smaller pore
sizes.
(ii) The amount of shrinkage decreases with increasing cement content. Furthermore, for weaklycemented DM, the shrinkage also increases with
increasing initial water content. Beyond yielding,
specimen at higher initial water content shrinks
more because of looser structure.
ACKNOWLEDGEMENT
This study is sponsored by the National Basic
Research Program (973 Program) project (grant no.
2012CB719804), National Natural Science Foundation of China (grant nos. 51178161 and 51278171)
and the 111 Project (grant no. B13024).
REFERENCES
ASTM 2008. Standard test methods for determination of
the soil water characteristic curve for desorption using
hanging column, pressure extractor, chilled mirror

hygrometer, or centrifuge. ASTM standard D6836-02,
ASTM International.
ASTM 2010. Standard test method for measurement of soil
potential (suction) using filter paper. ASTM standard
D5298-10, ASTM International.
Arya, L.M. & Paris, J.F. 1981. A physicoempirical model to
predict the soil moisture characteristic from particle-size
distribution and bulk density data. Soil Sci. Soc. Amer. J.
45(6): 10231030.
Chiu, C.F. & Ng, C.W.W. 2012. Coupled water retention and
shrinkage properties of a compacted silt under isotropic
and deviatoric stress paths. Can. Geotech. J. 49(8):
928938.
Chiu, C.F., Zhu, W. & Zhang C.L. 2009. Yielding and
shear behaviour of cement-treated dredged materials.
Engineering Geology 103(1-2): 112.
Chew, S.H., Kamruzzaman, A.H.M. & Lee, F.H. 2004.
Physicochemical and engineering behavior of cementtreated clays. J. Geotech. Geoenviron. Engrg. 130(7):
696706.
Connor, J.R. 1990. Chemical fixation and solidification of
hazardous wastes. New York: Van Nostrand Reinhold.
Fredlund, M.D., Wilson, G.W. & Fredlund, D.G. 2002.
Use of the grain-size distribution for estimation of the
soil-water characteristic curve. Can. Geotech. J. 39(5):
11031117.
Fredlund, D.G. & Xing, A. 1994. Equations for the soil-water
characteristic curve. Can. Geotech. J., 31(4): 521532.
Gardner, R.A. 1937. The method of measuring the capillary
tension of soil moisture over a wide moisture range. Soil
Science 43: 277283.
Horpibulsuk, S., Miura, N. & Nagaraj, T.S. 2005. Claywater/cement ratio identity for cement admixed soft clays.
J. Geotech. Geoenviron. Engrg. 131(2): 187192.
Hoyos, L.R., Thudi, H.R. & Puppala, A.J. 2007. Soil-water
retention properties of cement treated clay. In Problematic
soils and rocks and in-situ characterization. GSP 162, 18.
Lee, F.H., Lee, Y., Chew, S.H. & Yong, K.Y. 2005. Strength
and modulus of marine clay-cement mixes. J. Geotech.
Geoenviron. Engrg. 131(2): 178186.
Leong, E.C., He, L. & Rahardjo, H. 2002. Factors affecting the filter paper method for total and matric suction
measurements. Geotech. Test. J. 25(3): 322332.
Meegoda, J.N., Partymiller, K., Richards, M.K., Kamolpornwijit, W., Librizzi, W., Tate, T., Noval, B.A., Mueller,
R.T. & Santora, S. 2000. Re-mediation of chromium
contaminated soils a pilot scale investigation. ASCE
practice periodical of hazardous, Toxic Radioactive Waste
Manage 4(1): 715.
Mitchell, J.K. 1981. Soil improvement state of the art report.
Proc. 10th International Conference on Soil Mechanics
and Foundation Engineering, Stockholm, 4:509565.
Nuth, M. & Laloui, L. 2008. Advances in modeling
hystereticc water retention curve in deformable soils.
Computers & Geotechics 35(6): 835844.
Puppala, A.J., Konnamas, P., & Vanapalli, S.K. 2006. Soilwater characteristic curves of stabilized expansive soil.
J. Geotech. Geoenviron. Engrg. 132(6): 736751.
Sheng, D. & Zhou, A.N. 2011. Coupling hydraulic with
mechanical models for unsaturated soils. Can. Geotech.
J. 48(5): 826840.
Simms, P.H. & Yanful, E.K. 2002. Prediction soil-water characteristic curves of compacted plastic soils from measured
pore-size distributions. Gotechnique 52(4): 269278.
Simms, P.H. & Yanful, E.K. 2005. A pore-network model
for hydromechanical coupling in unsaturated compacted
clayey soils. Can. Geotech. J. 42(2): 499514.
375
SOA 2012. China Marine Environmental Quality Report

2012. State Oceanic Administration Peoples Republic of
China, Beijing. (In Chinese)
Tremblay, H., Duchesne, J., Locat, J. & Leroueil, S. 2002.
Influence of the nature of organic compounds on fine soil
stabilization with cement. Can. Geotech. J. 39: 535-546.
Tremblay, H., Leroueil, S. & Locat, J. 2001. Mechanical
improvement and vertical yield stress prediction of clayey
soils from eastern Canada treated with lime or cement.
Can. Geotech. J. 38: 567579.
van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils.
Soil Sci. Soc. Amer. J., 44(5): 892898.
Zhu, W., Zhang, C.L. & Chiu,A.C.F. 2007. Soil-water transfer
mechanism for solidified dredged materials. J. Geotech.
Geoenviron. Engrg. 133(5): 588598.
Zhu, W., Chiu, C.F., Zhang, C.L. & Zeng, K.L. 2009. Effect of
humic acid on the behavior of solidified dredged material.
Can. Geotech. J. 46(9): 10931099.
376

J -integral as a useful fracture parameter for analysis of desiccation

cracking in clayey soils
S. Costa
School of Applied Sciences and Engineering, Monash University, Australia
J. Kodikara
Department of Civil Engineering, Monash University, Australia
J. Xue
School of Applied Sciences and Engineering, Monash University, Australia
ABSTRACT: Fracture behaviour of clay soils during desiccation is an important area in soil mechanics that
needs further development. The current approach is to use Linear Elastic Fracture Mechanic parameters, such as
fracture energy and fracture toughness, measured from load tests for analysis and modelling of crack propagation.
However, the importance of taking plasticity of the material into consideration has been highlighted by several
researchers. This is especially true for soft and/or slurry clay. The recently developed double ring test is the only
test available up to date which enables determination of Elasto-Plastic Fracture Mechanic parameter, J -integral.
This is an important parameter, which accounts for the change in potential energy with fracture propagation.
Currently, there is no data for J -integral in literature as applicable to desiccation cracking. This paper fills the
said gap by presenting a set of J -integral data for Churchill clay and Kaoline clay measured from the double
ring test. Values have been calculated for both slurry and compacted clay with corresponding coefficients of
linear shrinkage. The challenging task of calculating strains and stresses has been dealt with the use of an image
analysis technique. In addition, the path independence for J -integral calculation and the behaviour of J -integral
with the moisture content are also discussed.
INTRODUCTION
Tensile fracture in cohesive soils is an important

consideration in many geotechnical engineering applications. In the recent past, desiccation induced tensile
cracks have gained a special attention of researchers
(Costa et al. 2013, Lakshmikantha et al. 2012,
Nahlawi & Kodikara 2006, Peron et al. 2009), because
of the unnecessary problems they cause in stability
of slopes, earth dams, road pavements, landfill clay
liners etc.
Theoretical understanding and the modeling of desiccation cracking is less developed compared to the
other areas of soil mechanics. Traditional way of
approaching analysis of tensile fracture in soils is
through Linear Elastic Fracture Mechanics (LEFM).
Several researchers have used this approach in the past
and conducted tests to evaluate LEFM fracture parameters such as fracture toughness and fracture energy
(Amarasiri et al. 2011, Wang et al. 2007, Nahlawi
et al. 2004, Konrad & Ayad 1997, Nichols & Grismer
1997, Harrison et al. 1994). All these tests were performed with external loading and in many cases with
compacted clay specimens. A concise history of different types of tests and modelling approaches related
to fracture in soil was presented by Kodikara & Costa

(2012).
Despite the prevalence of LEFM as a tool
to describe soil fracture, some have highlighted
(Hallet et al. 1995) the insufficiency of LEFM to
capture the complete process when the material is predominantly plastic. This is truer for very soft and wet
clays. Hallet & Newson (2005) attempted to incorporate Elasto-Plastic Fracture Mechanics (EPFM) to
analyse fracture in deep-notched clay beams. Nevertheless, the use of EPFM has been very limited
due to the difficulty of evaluating relevant fracture
parameters such as crack tip opening angle.
Recently, Costa & Kodikara (2012) presented a new
ring test method to determine elasto-plastic fracture
properties in the form of J -integral. This parameter
was first introduced by Rice (1968). It is defined as
the change in potential energy during crack growth.
J -integral has the capacity to capture the change in
both elastic and plastic energy, thus making it more
suitable for analysis of desiccation cracking where
the state of soil changes significantly from plastic to
elastic.
The key advantages of the test proposed by
Costa & Kodikara (2012) are its simplicity and
377
Table 1.
Basic properties of Churchill clay and Kaoline clay.
Property
Churchill clay
Kaoline clay
Liquid limit
Plastic limit
Plasticity index
Linear shrinkage (%)
OMC (modified)
Colour
90.0
24.0
66.0
21.5
19.0
Grey
54.8
26.0
28.8
6.9
24.0
White
applicability to desiccation cracking since no external

load is applied during the test. However, the authors
did not provide sufficient experimental data to further
validate the new test method and to demonstrate its
versatility. This paper fills that gap by establishing a
set of J -integral values for both slurry and compacted
clay. Furthermore, the paper discusses the relationship
of J -integral with the change in moisture content.
MATERIALS
Churchill clay, a locally sourced expansive clay from

Churchill, Victoria, Australia, and commercially available Kaolin clay were used in the experiments. Basic
properties of these clays are given in Table 1.
Figure 1. Churchill clay specimen prepared for testing.
specimens, without trapped air and flaws, was a challenging task.According to the stress analysis presented
by Costa et al. (2013) and Costa & Kodikara (2012),
crack initiation must occur at the inner edge in the
radial direction. However, after few trials, it was possible to obtain specimens which delivered satisfactory
results.
3.2 Determination of coefficient of linear shrinkage
3
3.1
EXPERIMENTAL PROCEDURE
Ring test
Specimens were prepared following the steps

described in Costa & Kodikara (2012) with some
minor modifications. The inner and outer radii of the
specimen were 55 mm and 104 mm. The same base
treatment was applied to ensure zero or negligible friction at the soil-base interface. Specimens were placed
on an electronic balance and a digital camera was
mounted above the setup. Both the balance and the
camera were operated by a computer and they automatically recorded the weight of the specimen and the
images at regular intervals. A typical specimen ready
for testing is shown in Figure 1.
Churchill clay was processed by drying, crushing
and sieving before preparing specimens. For compacted specimens moisture content of 35%, which is
slightly wet of optimum moisture content (32.5%),
was used. For slurry tests, specimens were prepared
at 100% moisture content, slightly above the liquid
limit (90%).
Kaolin clay was already processed and ready for
use at the time of purchase. Specimens were prepared
at moisture contents of 55% for slurry tests and 35%
for compacted tests.
Compaction of the specimens for both clays was
achieved by hand compaction using a spatula. Unlike
with slurry clay, obtaining homogenous compacted
Coefficient of linear shrinkage () is required to

compute J -integral. Standard linear shrinkage moulds
were used for these tests. Clays were allowed to dry
under the room temperature and humidity. Two sets of
specimens were prepared for each clay type, one starting from the liquid limit and the other starting from half
of the liquid limit. This arrangement was followed to
study the behavior of the coefficient of linear shrinkage throughout the drying process. The idea was to
ensure that a single value can be applied at any point
of the drying process.
4
CALCULATIONS
The calculation of J -integral is based on the definition

introduced by Rice (1968). Creating a zero friction
interface at the base of the specimen is essential to use
this concept as the crack growth needs to be the primary mode of energy dissipation from the specimen.
Riceconcept consists of using a closed contour around
the crack tip which is described in mathematical form
by Equation 1 below.
where is a closed contour followed counterclockwise

direction encompassing the crack tip on the stressed
soil body, W is the strain energy per unit volume,
378
Figure 2. Rectangular contour selected for calculating

J -integral.
Figure 3. Change in moisture content with time.
T is the traction vector perpendicular to in the outside direction, u is the displacement in x direction and
ds is an element of .
Luo et al. (2003) showed that Equation 1 can be
written as the algebraic sum of four line integrals if
is taken to be a rectangle. Figure 2 shows such a
contour encompassing the crack tip selected for one of
the specimens. Strain energy can be calculated using
incremental strains and stresses in a non-linear elastic
manner. Details of calculations are not presented in
this paper as a step by step description of them can be
found in Costa & Kodikara (2012).
GeoPIV program with Particle Image Velocimetry
(PIV) technique was used to analyse the images and
to measure the displacements of clay body between
the images. These displacements were then used to
determine the strains and thereby the stresses. For the
conversion of strains to stresses, values of Youngs
modulus were assumed: 0.005 MPa for slurry clay and
0.5 MPa for compacted clay. Both these values were
incorporated into calculations of stresses in slurry tests
as the Youngs modulus changes with the removal of
water.
5
More than one day was taken for the slurry clay specimens to produce the first crack whereas compacted
clay specimens started cracking less than a day. Almost
all the cracks initiated from the inner edge in radial
direction as expected according to the stress analysis presented by Costa & Kodikara (2012). Drying
process was stopped before the initiation of a second
crack since subsequent cracking can disturb the stress
field around the first crack resulting in discrepancies in
strain and stress measurements. As stated earlier, this
was a hard task to achieve for compacted specimens.
However, after several trials, satisfactory specimens
were obtained.
Desiccation of some selected specimens is shown in
Figure 3. All the specimens were dried at a reasonably
uniform drying rate though room temperature and the
humidity were not controlled. Information presented
Figure 4. Variation of J -integral as desiccation progress in

compacted clay specimens.
in Figure 3 is for the first 48 hours of the tests. Actual

test durations varied depending on the crack initiation
time.
From the J -integral values calculated from compacted clay beam bending tests, Costa (2010) suggested that J -integral increases with the decreasing
moisture content. Test results presented by Costa &
Kodikara (2012) also suggested a similar trend for
slurry specimens. Figure 4 and Figure 5 illustrates
the behavior of J -integral with moisture content for
compacted and slurry tests respectively for the current investigation. These results verify the trend of
J -integral proposed by Costa & Kodikara (2012) and
Costa (2010). Values of J -integral for Churchill clay
in Figure 4 clearly show this decreasing trend.
Being a highly plastic material, Churchill clay possesses a relatively high J -integral than Kaolin under
compacted conditions (Fig. 4). A different picture is
obtained for slurry state (Fig. 5) where Kaolin acquires
higher J -integral values than Churchill clay. However, Kaolin, not being an expansive clay, does not
shrink adequately to produce cracks until the specimen has dried close to its plastic limit, hence the higher
J -integral values of Kaolin for slurry tests. It can also
379
Figure 5. Variation of J -integral as desiccation progress in

slurry clay specimens.
Figure 7. Change of J -integral with void ratio
Figure 6. Comparison of J -integral values calculated using

two different contours for the same specimen.
dries and become more brittle, the size of the crack tip
plastic zone and thereby the plastic energy release rate
are reduced.
However, elastic energy release rate become dominant in the J -integral values at this stage. The energy
needed to open a crack, or in other words to create
two new surfaces by tearing the existing surfaces is
relatively large for brittle materials. This explains the
pattern of variation of J -integral values with moisture
content.
The results of the linear shrinkage tests suggested
that coefficient of linear shrinkage () remains constant throughout the shrinking process when the initial moisture content is close to the liquid limit or
lower. Thus the values of were obtained as 0.39 for
Churchill clay and 0.37 for Kaolin and were used in
these calculations.
be seen that these higher values are attributed to lower

moisture content values than the Churchill clay.
According to the original definition postulated
by Rice (1968) for J -integral (Eq. 1), the value of
J -integral for a given stress field is independent of
the shape and the size of the contour () that is used to
calculate strain energy. This criterion was used to verify the strain and stress calculation procedures as well
as the reliability of the image analysis technique used
in the current test method. Figure 6 shows the comparison of J -integral values calculated using two different
contours for the same specimen at a given time. Both
contours were rectangles in shape for the sake of convenience but different in size. It can be seen (Fig. 6)
that the numbers from the two calculations matches
reasonably well.
Figure 7 shows how J -integral varies with void ratio
for compacted Churchill clay specimens. As it can be
expected after the patterns suggested in Figures 4 & 5,
J -integral increases with the decreasing void ratio.
As the soil mass becomes drier, the material
becomes more elastic. J -integral consists of both elastic and plastic components of energy release rates.
Plastic component of the energy release rate depends
on the size of the crack tip plastic zone. When soil
CONCLUSION
Values of J -integral for both compacted and slurry

clay were successfully determined from the study presented in this paper following the new ring test method
proposed by Costa & Kodikara (2012). It is clear that
the ring test can be conveniently used to determine the
strain energy release rate in the form of J -integral at
different moisture contents.
The key advantage of these values is that they can be
directly used in analysis of desiccation cracks in soils
where the state of the material can change from plastic
to elastic within a short period of time. Unlike the fracture parameters measured from load controlled tests,
J -integral values measured from the ring test deem
more suitable for the desiccation cracking analysis.
Authors believe that the introduction of the new
ring test and the J -integral values will pave the opportunity for researchers to incorporate elasto-plastic
fracture mechanics in to the analysis of desiccation
cracking phenomenon in soil consequently complementing the modeling of desiccation induced cracking
in compacted clay liners, mine tailings, embankments,
pavements etc.
380
REFERENCES
Amarasiri, A., Kodikara, J. & Costa, S. 2011. Numerical modelling of desiccation cracking, International Journal for
Numerical and Analytical Methods in Geomechanics 35:
8296.
Amarasiri, A. L., Costa, S. & Kodikara, J. 2011. Determination of cohesive properties for Mode I fracture from
compacted clay beams. Canadian Geotechnical Journal
38: 11631173.
Costa, S. & Kodikara, J. 2012. Evaluation of J integral for clay
soils using a new ring test. Geotechnical Testing Journal
35(6).
Costa, S. 2010. Study of desiccation cracking and fracture properties of clay soils, PhD dissertation, Monash
University, Australia.
Costa, S., Kodikara, J. & Shannon, B. 2013. Salient factors controlling desiccation cracking of clay in laboratory
experiments. Geotechnique 63(1): 1829.
Hallet, P. D. & Newson, T. A. 2005. Describing soil crack
formation using elastic plastic fracture mechanics.
European Journal of Soil Science 56: 3138.
Hallet, P. D., Dexter, A. R. & Seville, J. P. K. 1995. The application of fracture mechanics to crack propagation in dry
soils. European Journal of Soil Science 46: 591599.
Harison, J. A., Hardin, B. O. & Mahboub, K. 1994.
Fracture toughness of compacted cohesive soils using
ring test. Journal of Geotechnical Engineering 120(5):
872891.
Kodikara, J. & Costa, S. 2012. Desiccation cracking in clayey

soils: Mechanisms and Modelling. Multiphysical Testing
of Soils and Shales, Springer Series in Geomechanics and
Geoengineering 2013, 2132.
Konrad J. M. & Ayad, R. 1997. An idealized framework
for the analysis of cohesive soils undergoing desiccation.
Canadian Geotechnical Journal 34:477488.
Lakshmikantha, M. R., Pract, P.C. & Ledesma, A. 2012.
Experimental evidence of size effect in soil cracking.
Canadian Geotechnical Journal 49:264284.
Luo, P. F., Chuang, C. K. & Chao, Y. J. 2003. Measurement
of J integral by shadow spot generated from out-of-plane
displacement. Optical Engineering 42(12): 34333442.
Nahlawi H., Chakrabarti S., Kodikara J. 2004. A Direct
Tensile Strength Testing Method for Unsaturated Geomaterials. Geotechnical Testing Journal 27(4):16.
Nahlawi, H. & Kodikara, J. K. 2006. Laboratory experiments
on desiccation cracking of thin soil layers, Geotechnical
and Geological Engineering 24: 16411664.
Nichols, J. R. & Grismer, M. E. 1997. Measurement of fracture mechanic parameters in silty clay soils. Soil Science
162: 309322.
Peron, H., Hueckel, T., Laloui, L. & Hu, L. B. 2009. Fundamentals of desiccation cracking of fine grained soils:
Experimental characterization and mechanisms identification. Canadian Geotechnical Journal 46:11771201.
Rice, J. R. 1968. A path independent integral and the approximate analysis of strain concentrations by notches and
cracks. Journal of Applied Mechanics 379386.
381

Soil mass stability analysis using Euler-Lagrange coupling

M.S. Serbulea,
A. Andronic, D.M. Manoli & A. Priceputu

Department of Geotechnical Engineering and Foundations, Technical University of Civil Engineering,
Bucharest, Romania
ABSTRACT: The numerical modelling of the propagation phase of landslides surpasses the limits of classical
Soil Mechanics, entering the Fluid Mechanics field. For modelling the displacements of fluid behaviour continuous masses, using the Finite Element Method, two approaches can be used: (1) Complete 3D formulations,
where the mass equilibrium equations and the linear moment equations are solved in a time-dependent domain.
These kinds of models require high computational resources and their usage is restricted to the cases where the
fine structure of the flow, structure forces etc. are known. (2) Using the linear flow on the vertical direction
under a conservative forces field hypothesis, it is possible to integrate the equilibrium equations on the vertical
direction of the differential hydrostatic level, obtaining simple 1D models. The depth integration equations
can be written using Euler or Lagrange formulation. The current paper presents the numerical modelling of
the solid-liquid interaction and establishes a comparison between this method and the methods proposed by
Ito-Matsui and Beer-Carpentier for calculating the optimum distance between the piles axes.
INTRODUCTION
Classical numerical analysis of landslides uses the

Lagrange model, in which the mesh and the material are interconnected bi-univocally, which leads to
a macro mesh deformation through the summation
of deformations of elements with continuous material properties. Consequently, large deformations lead
to large nodal displacements, which lead to excessive
distortion in some elements. This makes the use of
this method inadequate for modelling phenomena in
which large deformations occur such as yielding
type of landslides. In addition, the material properties
remain evenly distributed in the elements throughout
the analysis, no matter the mechanical transformations
suffered by the model.
The Euler formulation is characterized by the use
of a fixed and undeformable mesh and acts as a flow
medium for the geometrical model in question, allowing large deformations. In this way, new free surfaces
are generated allowing one element to contain more
than one material; in the same time, however, a reference material has to be created-vacuum, for instancefor establishing the exterior free surface. In the models
boundary zones the material properties are represented
by linear combinations of the materials that are in contact. The boundary isnt an actual physical limitation
but is rather represented by elements containing the
materials in the pure, initially defined state.
Numerical modelling using Euler-Lagrange couplings has been used in describing different phenomena, but no references have been found pointing to its
use in case of sliding unstable soil masses.
Landslides are caused by changes in the state of

effective stress, variations of the material properties or
changes in geometry. The changes in the state of stress
can be induced directly, as a result of the variation of
exterior forces (earthquakes, anthropical actions), or
indirectly, by the change of the pore water pressure
induced by rain, for instance. The changes in geometry can have natural causes, such as anthropic actions
(excavations, buildings, changes of the angle of slopes
etc.) or erosion.
The study of landslides and of their consequences
has become the object of many fields of activity such
as geography, pedology and urban planning. The engineering aspect of the problem, though, is to understand
and analyse the triggering mechanism and propagation
of the landslides, which are essential in understanding
the reasons of the actual event but, also for preventing
future ones. Concerning the propagation, once triggered, it is important to know the flow rate, the distance
a given landslide could travel and the path of propagation. This way defence strategies, based on mass
flow can be proposed, as well. The instruments used in
the determination of landslides are mathematical and
constitutive models for which there are only a limited
number of analytical solutions available, so obtaining coherent results commands the use of numerical
models such as the finite element method.
The process can be described at length from triggering to propagation using one mathematical model.
There are, however, difficulties in case the landslide
evolves into a flow like phenomenon, because the
phase describing the transition of a material from solid
to fluid lacks a coherent mathematical model. This is
383
why it is advisable to use different approaches for the

two phases.
2
2.1
where 0 being a material constant and 0 being the

reference density.
Hugoniot energy EH is linked to the Hugoniot
pressure through the following equation:
STATE EQUATIONS USED IN EULER

MODELLING
Energy equation and the Hugoniot curve
The equation of energy conservation links the increase

of the ratio between internal energy and mass unit Em
to the increase rate of the mechanical work done by
the forces and increase rate of the heat. In the absence
of the heat conduction equation, the energy equation
can be written as:
where p is the pressure, defined as positive for compression, pbv is the pressure due to volumetric viscosity. S is the deviatoric stress tensor, e being the
deviatoric component of the strain rate, Q being the
rate of thermal increase on the unit of mass.
In case of the equation of state pressure is expressed
as a function of current density, and internal energy
on the unit of mass, Em :
which defines al the states of equilibrium that can be

found in a material.
The internal energy can be eliminated from equation (2) in order to obtain a p-V relationship (where
V represents the current volume), or equivalently a
relationship of the type p 1/, which is unique for
the material described by the equation of state of the
model. This unique relationship is called Hugoniot
curve and it is the geometrical representation of the
p-V states they can result from a shock.
The Hugoniot pressure, pH , is only dependent of the
density and can be generally defined by experimental
data. We can say that an equation of state is linear when
it can be written like:
where = 1 0 is the nominal volumetric strain.

By taking and EH out of the equations (4), (5) and
(6) it yields to:
The equation of state and the one of energy represent the coupled equations of the internal pressure and
of the energy.
2.3 The linear Hugoniot form: Us -Up
A normal adjustment of the Hugoniot relation is given
by the following equation:
where c0 and s define the linear relation between linear

shock velocity Us and particle velocity Up as follows:
Considering the above assumptions, the linear

Hugoniot form Us Up is rewritten as:
where 0 c02 is the equivalent volumic elasticity

modulus for small strains.
3
where f () & g() are functions of density and depend

on a specific equation of state of the model.
COMPUTATION OF DISCONTINUOUS
STRUCTURES LOADED BY ARCHING
EFFECT
3.1 Ito-Matsui method
2.2 Mie-Grneisen equations of state

A Mie-Grneisen equation of state is linear in terms
of energy. The most common form of this equation is:
where pH and EH are Hugoniot pressures and specific energies (on unit of mass), depending only on
the density being the Grneisen ratio, defined as:
Briefly, the proposed method starts from examining

the way in which the arch effect could manifest in soil
between the stabilization piles arranged in one line.
Finally, we get to determine the maximum reactive
force R on one of the piles being in sliding when the
soil between the piles reaches the plastic state. The
value of the reactive force R depends on the depth
H of the instable soil layer, of the distance between
axes D considered between the piles, of the diameter d
of the piles, respectively to the physical mechanical
properties of the instable soil, respectively: the volume
weight , internal friction angle , and cohesion c.
384
According to the Ito-Matsui method the researches

made on theoretical way for estimating the lateral force
that acts on the piles, are very closed to, as results,
to the field tests made in zones where the loosing
stability of the slopes phenomena is very active. It
is enough to analyse the behaviour of the soil layer
between two consecutive piles. For analysing the lateral force that acts on the piles, we use the theory of
the plastic deformation.
The lateral force that acts on the unit thickness of a
pile, on X direction is obtained through the difference
between the pressures that act on BB and AA planes:
Figure 1. The assembly of elements.
3.2 Beer-Carpentier method

The practical method of Beer and Carpentier for determining the horizontal force that is acting upon a pile
differs from Ito and Matsui method regarding the
assumption made about the stresses acting on the
failure planes.
The Ito and Matsui method considers that the
stresses acting on the failure planes are principal
stresses, while the Beer and Carpentier method takes
into consideration any certain stresses that act on these
planes.
According to this method, the soil is considered in
the equilibrium limit state.
The total lateral force that acts on the unit thickness
of a pile on X direction has the following expression:
3.3 Euler-Lagrange coupling method

For this analysis model, the piles, the raft and the walls
surrounding the earth were imported as Lagrange parts
and the reference part and the medium defining the
earth were imported as Euler parts.
For the materials used in the analyses, the structure is modelled as concrete with elastic behaviour,
governed by the density , Youngs modulus E and
Figure 2. The mesh of the model.
Poissons ratio . The soil, defined as Euler part,

is modelled by the density , viscosity , DrukerPrager plastic criterion and the Us Up Hugoniot
linear equation of state.
The contact type is general contact with tangential
behaviour penalty friction and a normal hard contact
behaviour that allows post-contact material separation.
This sort of contact corresponds to a dynamic-type
explicit interaction in the case of the coupling.
The only load taken is the gravity. The boundary
conditions are applied to the raft and the concrete box,
limiting its displacement on the z direction at the basis
and the displacements on x and y directions on the
lateral faces.
On one opening of the box piles fixed at the base
(where the shearing occurs) were set at different distances between axes allowing the soil to pass through
them. The effect of soil passing in a plastically rheological way through the isolated retaining structures loads
them with a pressure corresponding to the mobilised
soil mass thrust. The main difference between this
385
Figure 3. Piles displacement due to soil action.
Figure 6. Active thrust, Ito-Matsui and Beer-Carpentier

capable curves and Euler-Lagrange thrust versus pile
inter-axes distance.
The results of soil thrust obtained from the variation of the distance were plotted next to the results
obtained from Ito-Matsui and Beer-Carpentier. It may
be noticed that the optimum distance given by the
classical theories are unconservative with respect to
the Euler-Lagrange coupling computation and the
observed actual behaviour of soils.
Figure 4. Soil thrust on piles.
REFERENCES
Figure 5. Soil thrust upon piles versus inter-axes distance.
analysis and the classical models (Ito-Matsui and BeerCarpentier) is that the piles are allowed to displace
(Fig. 3). The thrust of soil on the piles is computed
starting from the contact pressure on piles side surfaces
(Fig. 3).
CONCLUSIONS
It may be noticed that the variation of thrust with the

piles inter-axes distance has a minimum corresponding
to the value usually provided by experience.
Abaqus Analysis Users Manual, 2012.

Alonso, E. et al. (2010). Geomechanics of Failures. Advanced
Topics. Springer; 1st Edition.
Arrigada, M.C.S. et al. (2008). Numerical Modeling of
Coupled Phenomena in Science and Engineering: Practical Use and Examples (Multiphysics Modeling). Taylor
& Francis; first edition.
De Beer E. & Carpentier R. (1977). Discussions: Methods to
Estimate Lateral Force Acting on Stabilizing Piles. Soils
and Foundations.
Chapman, D.J. et al. (2006). The behaviour of dry sand under
shock-loading. American institute of Physics.
Chirica, A. et al. (2012). Modelarea numerica a propagarii
maselor de pamnt instabile n cazul alunecarilor de
teren folosind metoda cuplajului Euler-Lagrange. A XIIa Conferinta Nationala de Geotehnica si Fundatii, Iasi,
Romnia.
Ito T. & Matsui T. (1975). Methods to Estimate Lateral Force
Acting on Stabilizing Piles. Soils and Foundations.
Ito T. & Matsui T. (1978). Discussions: Methods to Estimate Lateral Force Acting on Stabilizing Piles. Soils and
Foundations.
Pastor, M. et al. (2003). Modelling of Landslides: (I) Failure
Mechanisms. Centro de Estudios y Experimentacion de
Obras Publicas, Madrid, Spain.
Pastor, M. et al. (2003). Modelling of Landslides: (I) Propagation. Centro de Estudios y Experimentacion de Obras
Publicas, Madrid, Spain.
386

Piggyback liner detection by means of spatially constrained

inversion of SWM data
Stefano Stocco
Gamut S.r.l.
Cesare Castiglia
Golder Associates S.r.l.
Davide Mainero
Acea Ambiente
ABSTRACT: Acea Pinerolese Industriale (Acea) has presented a change to the final closure design of the
Torrione 4 MSW landfill in order to be able to fill the additional volume freed by waste settlements developed
during construction. Since Torrione 4 landfill is a piggybacking expansion of a previous landfill, separated by a
clay liner,Acea has been requested to estimate the settlements induced on the clay liner by the waste deposition and
thus be able to evaluate the total occupied volume. As non-invasive methods were needed, seismic investigation
based on SWM (Surface Waves Method) has been carried out in order to confirm analytical predictions of the
settlements at liner level, calculated using common design approaches such as Sowers (1973). This methodology
has allowed the contact between the expansion and the previous landfill to be identified, confirming the analytical
predictions.
INTRODUCTION
The Municipal Solid Waste (MSW) facility located

in the Torrione area in Pinerolo (Torino, Italy) comprises to date three main landfills and their subsequent
piggyback extensions.
In order to be able to fill the additional volume freed
by waste settlements developed during construction of
one of the newest extensions (Torrione 4), the owner
(Acea Pinerolese Industriale, now Acea Ambiente) has
presented a change to the final closure design to the
local authorities.
To support the decision upon the new landfill configuration, the approvers have requested that the total
waste volume disposed in the Torrione 4 would be
calculated, function of the settlements of the underlying Torrione 1 developed after the construction of
the intermediate clay liner. A field indirect verification
procedure has also been requested.
The following paragraphs describe the design theoretical calculations and the adopted field verification
procedure.
From the geomorphological point of view, the area

is located in the plain at the base of three mountain
incisions originated by the Lemina, Chisone and
Pellice torrents.
The plain has been originated by various depositional cycles and is slightly tilted in ESE direction
(11,5% slope), with small river terraces.
Torrents Chisone e Lemina flow from WNW to
ESE respectively 500 m south and 1200 m north of
the facility area.
Groundwater level depth varies between 10 m
and 15 m below ground level, with ample seasonal
oscillations.
The facility comprises the following landfills:
the first sector, Torrione 1, active between 1976 and
1986 and between 1989 and 1990 and its piggyback
extension, Torrione 4, where the final capping is
currently under completion;
the second sector, Torrione Bis, active between
1987 e 1988, its piggyback extension, Torrione
3, active between 2001 and 2002 and now
completed;
the third sector, Torrione 2, active between 1990
and 2001, and its piggyback extension, Torrione 5,
currently under construction.
SITE DESCRIPTION
The facility area is surrounded by forest on the S and

SW limits, by the SS23 highway on the N side and by
the Doria property on E and SE.
Prior to the construction of the piggyback extensions, new clay liners and drainage layers have been
built to separate the old and new landfills.
387
2.1
Description of Torrione 4 building sequence
2.1.1 First upraise (2001)

The final design for the first extension was completed
in 2001 and included the following items:
construction of an intermediate embankment
between Torrione 2 and Torrione 1;
preparation of the liner base and construction of a
clay liner;
construction of a leacheate drainage, collection and
storage system
modifications to the existing biogas collection
network
construction of concrete ditches for water control.
The bottom area was divided in 4 subsectors
(B1 B4) to be constructed independently and separated by intermediate embankments.
The bottom liner comprises the following layers:
1 m thick compacted clay layer, with in situ permeability lower than 107 cm/s;
bentonite geotextile, 6 mm thick;
HDPE membrane, 2 mm thick.
The equivalent confinement capacity of the three
layers fits the requirements of Italian regulation
(D.Lgs 36/2003).
The liner package has been protected by the following overlying layers:
geocomposite drain;
drainage fill (clean sand and gravel, with fraction
passing 200 ASTM sieve lower than 5%) 30 cm
thick.
Figure 1. Scheme of a piggyback extension and of the

comparison between theoretical and actual profiles.
(about 50000 m3 ) freed by the waste settlements developed after the completion of the disposal phase.
A further survey carried out in April 2010 validated
and confirmed the information.
Therefore, a change to the landfill closure plan
has been developed, introducing the following modifications:
additional waste disposal in the residual spaces;
optimization of the waste disposal operations and
of the capping construction sequence;
optimization of the final profile to facilitate the
construction of the capping layers.
3
In order to be able to approve the 2010 expansion, local

authorities has requested an estimate of the following
waste volumes (Fig. 1):
2.1.2 Second upraise (2004)

A further modification to Torrione 4 final waste profile
was approved in 2004. The change design comprised
the following items:
excavation and movement of waste;
extension of the leachate collection system;
construction of new perimeter cemented clay
embankments;
internal waterproofing of the new embankments by
means of HDPE membrane and bentonite geotextile connected to the bottom layers;
extension of the geocomposite drain over the internal walls;
construction of a new surface water collection
system.
2.1.3 Third upraise (2010)
In January 2010, after the completion of the second
upraise in November 2009, a detailed topographic
survey has been carried out to measure the waste volume disposed on top of Torrione 1. Three-dimensional
numerical models have been prepared to allow for
automatic volume calculations.
The comparison of the 2010 survey with the
as-built topographical survey of the bottom layer and
with the approved theoretical waste profile has highlighted the possibility of exploiting residual volumes
CALCULATION OF BOTTOM LINER

SETTLEMENTS
residual free volume (volume difference between

theoretical approved profile and 2010 surveyed
profile);
hidden disposed volume due to the deformation
of the bottom liner following due to the settlement
of underlying waste volume.
total waste volume for the new authorization, sum
of the previous authorized volume, of the residual
free volume and of the hidden volume.
Since the hidden volume evaluation is strictly correlated to the settlement of Torrione 1 waste due to
static (stress-dependent) mechanical compression and
to long term creeping (non stress-dependent) waste
consolidation, analytic calculations of both settlement
components have been carried out.
3.1 Static compression settlement
(Primary settlement)
The static compression due to the weight of the
waste disposed in the first and second extensions
(Torrione 4) has been calculated assuming linear elastic deformation in oedometric conditions (zero lateral
expansion):
388
where s = settlement; q = total surcharge; H =

thickness of waste layer and Mo = confined oedometric modulus, which can be evaluated, according to
elastic theory, from field plate load tests:
The time dependent settlement calculated at the end

of the piggyback extension construction is therefore
equal to:
The total settlement stot at liner level can be estimated

as direct sum of the two components s and ws (2010):
The following have been assumed:

average thickness of Torrione 4 waste = 20 m
average waste volume weight after compaction =
10 kN/m3
average thickness of bottom liner layers = 2 m
average weight of the bottom liner layers = 18 kN/m3
q = 20 10 + 2 18 = 236 kPa
H
= 13 m
ME = waste compressibility from plate load
tests = 4 Mpa
= Poisson ratio = 0.25 0.35
The calculation leads to an average value of static
compression settlement after the construction of the
piggyback extension equal to:
The calculation of stot has been repeated on various

points of the landfill, with different waste thicknesses
both for the Torrione 1 and the Torrione 4 layers,
also keeping in account the local load contribution of
intermediate embankments.
The process has allowed for a deformed configuration of the piggyback liner to be modeled and to
be used as reference for the calculation of hidden
volume and total waste volume as requested by local
authorities.
4
3.2 Waste consolidation settlement

(Secondary settlement)
The settlement due to waste consolidation has been calculated with the method proposed by Sowers (1973),
which expresses the time-dependent settlement ws as
a direct function of the consolidation coefficient C :
where ws (t) = secondary settlement; H = thickness of

the settling layer; C = waste consolidation coefficient; tref = reference time (end of primary consolidation); t = calculation time > tref .
Sowers was one of the first to propose a transposition of mechanical behavior relationships of the
compressible soils to MSW waste. This transposition was limited to negligible lateral strain conditions,
sufficiently far from the edges of the cell. The proposed method is simple and based on a reduced set
of parameters which can be calibrated based on field
observations on similar waste deposits.
Based on experience acquired from monitoring
other relevant waste landfills in the same region, where
waste with similar compositions are disposed, verified
during the construction of the other extensions, the
following hypotheses have been assumed:
C = 0.03 0.06
H = 13 m
tref = 1 year after closure
t = period between calculation date and closure
(1990)
FIELD DETECTION OF CLAY LINER DEPTH
The SWM seismic data are acquired with a 48 channels geophone land-streamer, 4.5 Hz frequency, 1.25 m
spaced apart. The points of acquisition are placed
according to a designed grid of about 10 10 m,
acquiring 118 profiles in 2 working days. The accurate position of each measured point is ensured by
a double-frequency GPS/GLONASS receiver in real
time kinematic (RTK) configuration.
The dispersion curves have been retrieved through
a processing tool properly developed to manage the
data acquired by the acquisition chain.
After this first phase, the quality of the estimated
dispersion curves has been assessed and the dataset,
constituted by the estimated SW fundamental modes
along the line, was ready for the inversion phase.
The inversion has been performed starting from
a preliminary Monte Carlo Inversion (MCI) and the
a priori information of the bottom position, which
provides a consistent 1D S-wave velocity reference
model.
After the MCI, the dispersion curves have been
inverted simultaneously adopting a linearised Spatially
Constrained Inversion (SCI). The obtained results are
pseudo 3D S-wave velocity models.
In the following, the method for the processing (estimation of the surface wave dispersion curves) and
the inversion (the estimation of the S-wave velocity
model) is outlined.
4.1
Method
The applied analysis can be schematically described

as follows:
1. Estimation of the dispersion curves;
2. Dispersion curve quality evaluation;
389
Figure 2. Fundamental modes of the dispersion curve of the

whole dataset.
3. Setting of the reference model using the evaluation

of the bottom liner settlement and the Monte Carlo
Inversion (MCI) of the dispersion curves;
4. Spatially Constrained Inversion (SCI) of the dispersion curves to supply a pseudo-3D velocity
model.
4.2
Processing
We retrieve the dispersion curves by picking the energy

maxima on the f-k spectrum (Gabriels et al., 1987;
Tselentis and Delis, 1998) at each position of the
streamer. For each shot gather, the estimated dispersion curve is spatially located in the middle of the
array. In this way the dispersion curve dataset to be
inverted is constituted by a series of dispersion curve
estimated along the seismic lines. In Figure 2 we show
the cleaned dispersion curves of the whole dataset. We
have good quality data in the frequency range 225 Hz.
The fundamental modes can be shown as pseudodepth, in the LMP (Line Mid Point) /2.5 domain
in order to evaluate the investigated depth (Fig. 3).
The information content in the data are adequate to
construct a reliable model up to 25 m depth.
4.3 The inversion strategy
The inversion strategy can be summarised a following
and it is similar to the one proposed by Socco et al.
(2009):
1. Setting of the reference model using the evaluation
of the bottom liner settlement and the Monte Carlo
Inversion (MCI) of the dispersion curves;
2. Spatially Constrained Inversion (SCI) of the fundamental modes.
The different steps start from a broad band model
parameter space exploration and focus on the high
probability density (low misfit) model regions with
successive refinements performed through the SCI.
The MCI supplies global information about the site
velocity model. It disregards eventual lateral variations
and is also used to assess the level of information about
Figure 3. Dispersion curve as pseudodepth.
the presence of eventual velocity decreases within the

velocity profile. The MCI is also used as a preliminary inversion to define a consistent initial model for
the further SCI. The SCI supplies the account for eventual lateral variations and supplies the local properties
of the site.
In the inversion algorithms the unknown parameters are the shear wave velocity and the thickness
of the layers. The densities of the layers and the
Poisson ratios (or P wave velocity) are assumed a priori because dispersion curves are poorly sensitive to
their variations.
4.4 Initial model
In this work, as we know the initial geometry of the
bottom liner, we use both the a priori information and
the MC for the definition of the initial model.
We use the MC for the definition of the number of
the layers and the average S wave velocities (Wisn
et al., 2010). We use the evaluation of the bottom liner
settlements (Sowers, 1973) for the definition of the
thickness of the layers (Del Greco et al., 2007).
4.5 MCI
The inversion process of the Surface Wave dispersion
curves suffers from non uniqueness solution and it is
hence strongly biased by the initial model. The Monte
Carlo approach can handle the non uniqueness evidencing the local minima but it is inefficient for high
dimensionality problems and makes use of subjective
criteria, such as misfit thresholds, in order to interpret the results. An innovative Monte Carlo procedure
has been used for the inversion of experimental surface
wave dispersion curves (Socco and Boiero, 2008). The
algorithm uses scale properties of Rayleigh wave propagation (Socco and Strobbia, 2004, Socco and Boiero,
2008) to efficiently explore the model parameter space
390
reducing the required number of simulations.The analyses for the present study have been performed using
a misfit function that compares the experimental and
synthetic dispersion curves accounting also for problem dimensionality (number of unknown parameters
with respect to available data points).
The misfit is evaluated for each profile of the population and it is used to select acceptable models
according to a statistical test. The statistical one tailed
Fisher test (Socco and Boiero, 2008) selects all profiles
that are equivalent according to a given confidence
level. The selected models hence represent a set of
possible solution which may be considered equally
probable, given the experimental data and their uncertainties, the chosen parameterization and the level of
confidence. Socco and Boiero (2008) showed that the
solutions of deterministic inversions performed with
different initial models fall in a wider model region
with respect to the models selected by the Monte Carlo
inversion. The set profiles can therefore be assumed as
a picture of the non uniqueness solution for the specific dataset. More details on the algorithm used for
the inversion are reported in Socco and Boiero (2008).
Figure 4. Identification of the bottom in the 2002 (bottom

mesh).The upper mesh shows the actual surface of the landfill
with the measured points.
4.6 SCI
The Spatially Constrained Inversion (SCI) is a deterministic inversion in which each 1D model is linked
to its neighbours with a mutual constraint to provide a
single pseudo-2D or 3D model. The lateral constraints
can be considered as a priori information on the geological variability in the investigated area: the smaller
the expected variation of a model parameter, the more
rigid the constraint. Moreover, it is possible to use any
available a priori information, to further constrain the
inversion. The concept applied to surface wave data
using an acquired dataset with a walk away scheme
was originally presented by Wisn and Christiansen
(2005) and further applied by Socco et al. (2009).
The local dispersion curves extracted along the
line are inverted through a least squares algorithm,
based on the implementation given by Auken and
Christiansen (2004), to get a final pseudo 3D result.
The inversion result is built up of a set of 1D models, in
which each separate model is linked to an experimental dispersion curve. These neighbouring 1D models
are then connected sideways by lateral constraints that
claim equality between neighbouring model parameters of the same type. In this way, all the dispersion
curves along the line are inverted simultaneously, minimizing a common objective function, and the number
of output models is equal to the number of 1D soundings located along the seismic line. The misfit function
can be written as:
In eq. (5) the model parameters m are the 1DVS profile

for all the soundings. The ms are linked to the observed
dataset dobs with the associated observational covariance matrix Cobs. R is the lateral regularisation matrix
Figure 5. Identification of the bottom in the 2010 (bottom

mesh).The upper mesh shows the actual surface of the landfill
with the measured points.
Figure 6. Difference of the bottom of the actual landfill

between 2012 and 2002 as resulted by the seismic survey.
for the model parameters. The effectiveness of the

R matrix depends on the strength of the constraints
described in the covariance matrix Cp (Auken and
Christiansen 2004).
391
The forward response g(m) uses the Haskell (1953)

and Thomson (1950) approach modified by Dunkin
(1965) and implemented by Herrmann (1996) to solve
the forward problem for a stack of homogeneous elastic isotropic layers, assuming that the experimental
curves are fundamental modes for the pseudo Rayleigh
waves.
In this scheme, the constraints and the data are both
part of the inversion. Information from one model
will spread to neighbouring models through the lateral
constraints; the result is a smoothly varying pseudo 2D
model. Consequently, the output models form a balance between the constraints, the physics and the data.
Model parameters with little influence on the data will
be controlled by constraints.
4.7 Results
The following figures show the results of the inversion
as the identification of the bottom of the actual landfill.
5
CONCLUSIONS
We show a methodology for the identification of the

bottom of a piggybacking expansion of a previous
landfill with non-invasive methods. The survey consists in the acquisition of several surface waves data on
the landfill surface using a seismic land-streamer. The
processing of the data give several dispersion curve
to be inverted. The initial model is settled using the a
priori information of the bottom of the actual landfill
and an improved Monte Carlo inversion over the whole
dataset. The inversion of all the fundamental modes of
the dispersion curves is done simultaneously using a
spatially constrained inversion method.
The analytic predictions assume an average settlement of about 0.81.5 m. The seismic survey indicates
an average settlement of about 0.51 m. The differences fall within the experimental uncertainties and
resolution depth of the SWM. So, the results of the
seismic survey confirm the analytic predictions.
REFERENCES
Auken, E., and A. V. Christiansen, 2004, Layered and laterally
constrained 2D inversion of resistivity data: Geophysics,
69, 752761.
Bjarngard A., Edgers L. 1990, Settlement of municipal
solid waste landfills. Proc. 13th Annual Madison Waste
Conference, Madison (USA).
Del Greco, O., Fassino, A., and A. Godio, 2007, Seismic
investigation for the assessment of the elastic settlement
in MSW landfill, Sardinia Symposium, Cagliari, October,
2007, Vol. I, 16.
Dunkin, J.W., 1965. Computation of modal solutions in layered, elastic media at high frequencies. Bulletin of the
Seismological Society of America, 55, 335358.
Gabriels, P., R. Snieder, and G. Nolet., 1987, In situ measurements of shear-wave velocity in sediments with highermode Rayleigh waves: Geophysical Prospecting, 35, no. 2,
187196.
Gourc,J.P. , Nousheen Arif, K. & Olivier, F., 2007, Long term
settlement of domestic waste in landfill: ISPM method,
18eme Congrs Franais de Mcanique, Grenoble, 2731
aot 2007.
Haskell, N., 1953, The dispersion of surface waves on
multilayered media: Bulletin of seismological society of
America, 43, no. 1, 1734.
Herrmann, R.B., 1996, Computer programs in seismology:
an overview on synthetic seismogram computation: Users
manual, StLouis University, Missouri.
Ourry, J.F. & Page, B., 2005, Settlement of waste packaged in
chaffs in a domestic waste disposal. In: Proc. of International workshop Hydro Physico Mechanics of Landfills,
Grenoble 1 University, France.
Pezzetti G., Bonnardel F., Vacca O., Campione C.,
Castiglia C., 2007, Design and construction of the upper
lot of the basse di stura landfill in Torino: a new independent landfill above closed lots. Eleventh International Waste Management and Landfill Symposium, 15
October 2007, S. Margherita di Pula Cagliari, Sardinia,
Italy.
Socco, L.V., and C. Strobbia, 2004, Surface Wave Methods
for near-surface characterisation: a tutorial: Near Surface
Geophysics, 2, 165185.
Socco, L.V., and D. Boiero, 2008, Improved Monte Carlo
inversion of surface wave data: Geophysical Prospecting,
56, 357371.
Socco, L.V., D. Boiero, S. Foti, and R. Wisn, 2009, Laterally constrained inversion of ground roll from seismic
reflection records: Geophysics 74, 6, G35G45.
Sowers G.F., 1973, Settlement of waste disposal fills. Proc.
8th ICSMFE, Mosca, Russia.
Thomson, W. T., 1950, Transmission of elastic waves through
a stratified solid medium: Journal of Applied Physics, 21,
no. 89.
Tselentis, G.A., and G. Delis, 1998, Rapid assessment of
S-wave profiles from the inversion of multichannel surface wave dispersion data: Annali di Geofisica, 41, 115.
Van Impe W.F., Manassero M., Bouazza A., 1996, Waste disposal and containment, Second International Congress on
Environmental Geotechnics, Osaka, Giappone.
Van Impe W.F., Squeglia N., 1997, Propriet geotecniche dei
rifiuti solidi urbani, Argomenti di ingegneria geotecnica
N 6, Hevelius Edizioni.
Wisn, R., and A. V. Christiansen, 2005, Laterally and mutually constrained inversion of surface wave seismic data and
resistivity data: Journal of Environmental & Engineering
Geophysics, 10, 251262.
Wisn R., Boiero D., Maraschini M., Socco L.V., 2010,
Shear Wave Velocity Model from Surface Wave Analysis
A Field Case Example. In: Proceedings of 72nd EAGE
Conference & Exhibition incorporating SPE EUROPEC
2010, EAGE (NLD), 72nd EAGE Conference & Exhibition incorporating SPE EUROPEC 2010, Barcelona,
1417 June 2010.
392

Three-dimensional rotational-translational mechanism

for the stability analysis of landfill
Haoran Wang & Maosong Huang
Department of Geotechnical Engineering, Tongji University, Shanghai, China
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education,
Tongji University, Shanghai, China
ABSTRACT: The poor strength of the liner system may enhance the instability of the solid waste landfill. Based
on the upper bound limit analysis method, a new three dimensional failure mechanism consisting of rigid blocks
undergoing impending rotational or translational movements is proposed to study the landfill stability with the
liner system. And this combined mechanism can be expanded longitudinally through the plane insert, and the
method could be used to analysis the landfill stability with different width. The proposed rotational-translational
mechanism is finally verified by comparisons with the elasto-plasticity finite element method. The comparisons
show that the proposed mechanism improves the efficiency of the previously-developed block mechanism to
account for the presence of the liner system in the landfill and can be used as a simple evaluation method for the
engineering design.
INTRODUCTION
The landfill is by far a popular and effective method for

disposal of solid waste, and the liner system is widely
used in landfill for protecting solid waste polluting the
environment.The liner system consists of clay, geosynthetic materials, or their composition. However, these
protective materials have poor strength and may form
sliding surface between the protective system and solid
waste. So the liner system may increase the likelihood
of failure of landfill though it is in the prevention of
waste. Once the landfill fails, in addition to causing
human casualties, it will result in the proliferation of
solid waste, leachate spillage and exhaust emission
that can cause serious pollution of the surrounding
environment. Therefore, regardless of the size of the
landfill, its stability is the most important factor to
consider during the design, construction and operation. And the stability analysis of landfill is of great
significance.
Over the past decades, the limit equilibrium method
has been widely used to analyze the landfill stability.
Chen et al. (2000) adopted the Bishop slice method to
determine the safety factor of landfills. Based on the
wedge theory, Giroud et al. (1990), Koerner & Hwu
(1991) evaluated the safety factor of landfills. Qian
et al. (2003) proposed the double wedge method to
analyze the translational failure of landfill along the
underlying liner system. Qian & Shi (2011) applied
the double wedge method to studying the stability of
the landfill with multilayer geosynthetic liner system.
Limit analysis can provide an accurate upper bound
solution or lower bound solution for slope stability
analysis. Chen (1975) analyzed the stability of the

slopes on the basis of rotational mechanism. Considering the effect of weak interlayer, Huang et al. (2012)
proposed a two dimensional rotational-translational
mechanism to analyze the stability of the slope with
weak interlayer. However, these studies were mostly
for general slope and there is little literature about
the stability of landfills with liner system using
upper bound limit analysis method, especially threedimensional stability analysis. The landfill, due to
the impact of the underlying liner system, may slide
along the liner interface. That is to say the failure
mechanism of the landfill is different with the traditional mechanism of general slopes and similar to
the mechanism of the slope with weak interlayer. In
this paper, considering the effect of the liner system, a
three dimensional rotational-translational mechanism
is proposed. In combination with the strength reduction technique, the stability of the landfill with liner
system is studied by limit analysis, and analysis results
are compared with the results of FEM.
2
2.1
UPPER BOUND LIMIT ANALYSIS METHOD

Failure mechanism
The key of the upper bound limit analysis is the

construction of failure mechanism. According to the
flow rule associated with the Mohr-Coulomb yield
condition, the angle between sliding surface and
velocity is equal to the soil friction angle . At
present, most scholars use a rotational or a translational failure mechanism in the limit analysis. In the
393
Figure 3. Rotational-translational failure mechanism.
Figure 1. Rotational failure mechanism.
Figure 4. Velocity hodograph.
Figure 2. Translational failure mechanism.
two-dimensional analysis, the logarithmic spiral simulates the rotational failure surface, and the straight
line simulates the translational failure surface. Huang
et al. (2012) combined the logarithmic spiral and the
straight line to construct a two dimensional rotationaltranslational mechanism to analyze the stability of the
slope with weak interlayer.
In the three-dimensional analysis, Michalowski &
Drescher (2009) proposed a three-dimensional rotational mechanism. As shown in Fig. 1, the mechanism
has the shape of a curvilinear cone with apex angle 2
and rotates around the point O. This surface is smooth,
and has one symmetry plane. The trace of the mechanism on the symmetry plane is described by two logarithmic spirals, r = r0 e(0 )tan and r = r0 e(0 )tan .
A three-dimensional translational mechanism is shown
in Fig. 2. The mechanism has the shape of a cone with
apex angle 2 and moves along the axis.
Similar to the slope with weak interlayer, due to the

existence of the underlying liner system, the landfill
may slide along the liner interface. That is to say that
the sliding surface may partially locate in the landfill body and the rest of it locates on the liner system.
A single rotational or translational mechanism is not
suitable for this case, therefore, taking into account the
impact of the liner system, a new three-dimensional
mechanism is constructed through the combination of
rotational and translational mechanisms, as shown in
Fig. 3. The mechanism is constituted by one rotational
block and two translational blocks. The rotational
block a is a curvilinear cone as shown in Fig. 1. A
translational block c (CMK) is a cone as shown in
Fig. 2, and slides along the liner system. CM is located
on the liner system and OM has an angle of 1 p
with the liner system. FNMK is a translational block
b, which locates between blocks a and c. Because the
cross-section of blocks a is circular, if blocks a and c
are combined directly, the combined mechanism will
not be flexible. The introduction of block b is to make
the combined mechanism more general.
It is assumed that the velocity of block c is
vc , and its direction has an angle of p with CM.
Fig. 4 shows the hodograph for the failure mechanism given in Fig. 3. Thus the velocity of block b is
vb = vc cos(1 + p )/cos(1 p ), and the interfacial
velocity jumps MK is vrb = vc sin(21 )/cos(1 p ).
The linear velocity of a is a function of radius r and
394
Figure 5. Schematic diagram of the 3D mechanism (a)

mechanism without plane insert; (b) mechanism with plane
insert.
, its direction is perpendicular to radius r, and its

magnitude is given by
where is the angle between the radius and horizontal line, and a is the angular velocity about the axis
passing through the point O. The velocities of point N
in both a and b are the same, and hence a = vb /rh ,
where rh is the radius at point N .
Previous experience (e.g. Duncan 1996) has shown
that plane-strain mechanisms of failure are more critical than three-dimensional ones. Calculations for the
mechanism in Fig. 3 have shown, however, that the
mechanism is at some finite width L (Fig. 5(a)), even
if no constrains are placed on the width of landfills. To
allow transition to plane-strain mechanism, the threedimensional failure patterns (Fig. 5(a)) were modified
with a plane insert, by splitting and separating laterally the halves of the 3D surface, as illustrated in
Fig. 5(b). The width of the insert is l, and when l ,
the plane mechanism results can be retrieved. Therefore, the mechanism can be applied to analyze landfills
with any width.
Figure 6. Flow chart for computing safety factor.
the left-hand side of equation (2) represent the external

rate of work E, and the term on the right-hand side represents the internal rate of plastic energy dissipation
D. In this paper, the force acting on the slope surface
is not taken into account, thus the external rate of work
E is mainly produced by gravity, and can be written as
And the internal rate of plastic energy dissipation D

mainly occurs at the velocity discontinuity surface S
and the rotational mechanism Va , and can be written as
2.2 Work-energy balance equation

In limit analysis, when the landfill just reaches the
critical state, the external rate of work equals to the
internal rate of plastic energy dissipation. This can be
expressed as (Drucker et al. 1952)
where Ti is the vector of traction acting on the boundary ; Xi is the vector of body force acting on the
volume V; vi is the kinematically admissible velocity field; ij is the strain rate compatible with vi ; ij
is stress tensor associated with ij . The two terms on
2.3 Factor of safety

When the landfill reaches the critical state, the rate
of external work should equal to the rate of internal
plastic energy dissipation. The factor of safety Fs can
then be defined as
where c and are the cohesion and friction angle of the

soil respectively, and cm and m are the values which
make the landfill to maintain energy balance in an
395
Table 1.
Soil parameters.
Unit weight
Cohesion
Friction angle
Poissons ratio
Youngs modulus
Landfill
Bedrock
13 kN m3
25 kPa
20
0.3
5 MPa
27 kN m3
27200 kPa
27.8
0.3
20 MPa
Figure 7. Schematic illustration of landfill.
admissible failure mechanism. Denoting = D/E, the

search for Fs is equivalent to the search for = 1.0.
In the landfill analysis, a coordinate system (x, y) is
introduced at the symmetry plane, the origin of coordinate is at the toe of the landfill (point C), the positive
direction of axis x is to the right, and the positive direction of axis y is to the up. In the rotational-translational
mechanism, three unknown terms, which are angle a0 ,
ratio r/r0 and the location of point M (xM ), need to be
optimized. Because point M is in the liner system, only
one coordinate value is required to be determined for
its location. In addition, the width of insert plane l also
needs to be optimized.
Finding the value of Fs requires a numerical procedure in which the geometry of the mechanism is varied
in the search for the minimum value of . In the procedure, the reduction factor F1 is assumed to be 1.0 at
first, as shown in Fig. 6. And then, to find the minimum
of objective function , the independent variables ( a0 ,
r/r0 , xM and l) were changed sequentially by a small
increment in a single computational loop. If = 1.0,
the value of F1 is the safety
factor of the landfill. If not,
the reduction factor Fi = Fi1 is chosen to calculate repeatedly until = 1.0. The value of Fi will be
the safety factor, and the surface determined by a0 ,
r/r0 , xM and l is supposed to be the most probable
failure surface.
Figure 8. Critical failure mechanism determined by limit

analysis (a) mechanism without plane insert; (b) mechanism
with plane insert.
CASE STUDY
In this section, analyses for landfill with liner system

were conducted on a 40 m high landfill with a width of
400 m, as shown in Fig. 7.The friction angle of the liner
system p = 7 and the strength parameters of landfill
are listed in Table 1. The rotational-translational mechanism are applied to analyze the stability of the landfill.
The finite element method with strength reduction
technique (SSRFEM) is also used for comparison.
Firstly, the mechanism shown in Fig. 3 is applied
to analyze the stability of the landfill, and the failure
surface obtained is shown in Fig. 8(a), which corresponds to the safety factor 1.82. The width of failure
surface is only 86.4 m, however, the width of the case

is 400 m. Therefore, the mechanism shown in Fig. 5(a)
is obviously not suitable and the mechanism shown in
Fig. 5(b) should be applied. A modified failure surface obtained is shown in Fig. 8(b), and its width
is 400 m with an insert plane of width 320.6 m. The
corresponding safety factor is 1.19.
The SSRFEM is then used to analyze the stability
of the landfill. Due to the particularity of the liner system, it should be simulated by setting the tangent of
friction angle as the friction coefficient. The landfill
body is assumed as an elastic-perfectly plastic material satisfying the Mohr-Coulomb failure criterion.The
396
Figure 9. Three dimensional FEM mesh.
Figure 11. Critical failure mechanism of (a) overall;

(b) slice planes.
Figure 10. Relationship between reduction coefficient and
displacement (3D result 1 without plane insert; 3D result
2 with plane insert).
element mesh is shown in Fig. 9. In SSRFEM, the factor of safety is determined according to the relationship
between the reduction coefficient and the displacement at a specific point. If a sudden increase of the
displacement at the specific point is found, overall
failure is said to have occurred. The corresponding
reduction factor can be therefore taken as the safety
factor of the landfill. Fig. 10 shows this relationship
of the landfill toe on the symmetry plane, and the
safety factor determined from the SSRFEM analysis
can be taken as 1.05. The failure mechanism obtained
by the SSRFEM is shown in Fig. 11. In order to view
the failure mode inside the landfill body clearly, the
slice planes from the border to the symmetry plane are
shown in Fig. 11(b). Above studies are based on the
three-dimensional analysis, if l , it corresponds
to the two-dimensional case. The symmetry plane is
taken to be analyzed, and the safety factors evaluated
by 2D limit analysis and 2D SSRFEM are 1.09 and
1.02 respectively. The values are less than the safety
factors from the three-dimensional analyses. This is
because the safety factor of three-dimensional landfill
decreases with an increase of the width, and the result
of two-dimensional stability analysis is a minimum
limit value of the three-dimensional stability analysis.
In what follows, the critical failure mechanisms
obtained by the limit analysis and SSRFEM are analyzed. In the upper bound limit analysis (Fig. 8(b)), a
Figure 12. Critical failure mechanism on symmetry plane.
plane strain block with a width l = 320.6 m is inserted

to the symmetry plane, and the failure mode extends
to the landfill face at the boundary. The similar situation in the SSRFEM can be seen in Fig. 11(b). The
failure modes of the cross sections near the symmetry plane are substantially the same (Fig. 12), and the
range of equivalent plastic strain gradually decreases
near the boundary. The whole failure mode shows an
ellipsoid shape. Comparisons show that the shapes of
the failure surface obtained form two methods are
similar, and the safety factors are relatively close.
Therefore, the proposed three-dimensional rotationaltranslational mechanism is valid for the stability
analysis of solid waste landfills.
CONCLUSIONS
Considering the effect of the liner system on the

stability of landfills, the three-dimensional rotationaltranslational mechanism is proposed and applied to
study the stability of landfills. Through the case study,
the solutions are compared with the results of numerical method to verify the rationality of the proposed
397
mechanism. The results have shown that the traditional

rotation mechanism cannot reflect the effect of the
liner system, but the proposed rotational-translational
mechanism has the superiority to describe the effect of
the liner system on landfill stability.The failure surface
is much clearer than that obtained by the SSRFEM.
The propose mechanism can be applied for landfills
with any width through a plane insert, and used as a
simple evaluation method for the engineering design.
ACKNOWLEDGEMENTS
This work was financially supported by National
Basic Research program of China (Grant No.
2012CB719803) and National Fund for Distinguished
Young Scholars of China (Grant No. 50825803).
These supports are gratefully acknowledged.
REFERENCES
Chen Yun-Min, Wang Li-Zhong, Hu Ya-Yuan, et al. 2000.
Stability analysis of a solid waste landfill slope. China
Civil Engineering Journal, 23(1): 9297. (in Chinese)
Chen W F. 1975. Limit analysis and soil plasticity. New York:

Elsevier Scientific Publishing Company.
Duncan J M. 1996. State of the art: limit equilibrium and
finite-element analysis of slopes. Journal of Geotechnical
Engineering, ASCE, 122(7): 577596.
Giroud J P, Beech J F, Khatami A. 1990. Stability of Solid
Waste on Geo-synthetic Lining System. Proceedings of
the Fourth International Conference on geo-textiles: Geomembrane sand Related Products. Rotterdam: Balkema.
Huang Mao-Song, Wang Hao-Ran, Liu Yi-Lin. 2012.
Rotation-translation combined mechanism for the stability analysis of slopes with weak interlayer under rainfall
condition. Chinese Journal of Geotechnical Engineering,
34(9): 2025. (in Chinese)
Koerner R M. & Hwu B L. 1991. Stability and tension considerations regarding cover soil on Geo-membrane lined
slopes. Geotextiles and Geomembranes, 10(2):335355.
Michalowski R L & Drescher A. 2009. Three-dimensional
stability of slopes and excavations. Gotechnique, 59(10):
839850.
Qian X, Koerner R M, Gray D H. 2003. Translational failure analysis of landfill. Journal of Geotechnical and
Geoenvironment Engineering, ASCE, 129(6):506519.
Qian Xue-De, Shi Jian-Yong. 2011. Stability Problems For
Landfills With Multilayer Geosynthetic Liner System.
Chinese Journal Of Geotechnical Engineering, 33(11):
16761682. (in Chinese)
398
Linings for radioactive waste and capping systems

for landfills and polluted sites

Environmental restoration with deep roots herbaceous species

Nicola Era & Tiziana Verrascina
Prati Armati srl, Opera (MI), Italy
ABSTRACT: The use of deep rooting perennial grasses represents an innovative solution for soil protection and
renaturation in severe environmental situations. It allows interventions also in areas where climatic conditions
and soil characteristics were until a few years ago considered prohibitive for the development of vegetation.
An attractive application would be in abandoned landfills and dumps of mining scraps where the presence of
heavy metals (As, Cd, Pb, Zn etc.) usually inhibits the vegetation survival.
Germination tests were conducted on 9 deep roots grass species and 7 highly contaminated soil samples.
By monitoring the 63 resulting test pots it was possible to select 4 best performing species that could also
demonstrate to develop up to 1,80 m of root system after 1 year in same soil. These results are accompanied
by the presentation of 3 full-scale cases (an urban wastes landfill and two abandoned quarries) where similar
grasses allowed for a quick environmental recovery and protect the soil from erosion risks. The Authors are
therefore confident that in similar situations renaturation may be achieved despite the presence of mentioned
phytotoxic agents, thus making this innovative technology a unique and effective approach specifically where
time constraints and high costs often prevent the use of traditional remediation procedures.
EROSION PHENOMENA AND

ENVIRONMENTAL RESTORATION
The erosion on the Italian territory is quite widespread

and closely related to weather characteristics of our latitudes, where rain falls represent the more aggressive
agent (water erosion).
The severity of erosion action depends on various factors, such as intensity and duration of rainfall,
length and angle of slope, soil permeability and degree
of saturation, presence of vegetation, intrinsic soil
erodibility.
The intense erosion prevents the formation of
humus in the soil and accelerates the leaching of nutrients making very difficult the growth of vegetation.
Some pioneer species, particularly frugal, sometimes
succeed in taking root on bare eroded slopes, but often
a meteoric event stronger than usual is sufficient to
eradicate them.
Nevertheless, recent studies have highlighted the
ability of some deep rooting herbaceous species (prati
armati) to vegetate in prohibitive climatic conditions
where plant species typically used for greening interventions cannot grow. These plants behave as pioneer
organisms vegetating even on sterile and contaminated
soils. The dense vegetation blanket that originates
improves the soil structure and its fertility, preventing
erosion very effectively (see Cecconi et al. (2013) and
Rettori et al. (2010)). The soil would therefore become
more suitable to colonization by most demanding
plant species (such as shrubs and trees) accelerating the process of naturalization and environmental
restoration.
Several approaches for quantitative evaluation of

erosion (soil loss) have been proposed, like those based
on theoretical models, physical models at reduced
scale and empiric models. Among the last ones is
evidenced the Universal Soil Loss Equation USLE
(see Wischmeier & Smith (1965)) empirical equation
adopted by United States Department of Agriculture
for assessment of hydraulic erosion. Such equation is
generally diffused in the following form:
where:
A: specific soil loss [t/ha year], associated to phenomena of rill and interill erosion;
R: Rainfall-Runoff Erosivity Factor: climatic factor
relevant to the intensity and duration of precipitations [MJ mm/ha h year];
K: Soil Erodibility Factor: pedologic factor expressing the erodibility of the ground [t h/MJ mm];
LS: Geometrical factor function of the steepness and
length of the slope;
C: Cover-Management Factor: reduction factor
depending on the vegetation.
P: Supporting Practices Factor: reduction factor
taking into account possible interventions of
protection, control and conservation;
The anti-erosion techniques todays most popular date from 195060. All involve the use of synthetic products such as geocells, geomats, geonet,
or biomats, fascinate, wicker, etc. which, however,
in difficult climatic conditions, may not completely
401
solve the problem of erosion and do not allow rapid

renaturation.
In recent years, considerable interest is attributed
to anti-erosion interventions that use natural systems,
such as certain types of herbaceous species which
combine high resistance to phytotoxic conditions with
good geotechnical properties.
Their presence helps to reduce the factors P and C in
the reported USLE equation with the result of reducing
the specific soil loss.
The role of vegetation in protecting slopes from
erosion has long been studied and documented by
experimental research (see Cecconi et al. (2012)).
To reduce water erosion phenomena, and thus the
soil loss due to water run-off, the use of deep rooting
herbaceous plants looks promising for the following
reasons:
grass leaves dissipate most of raindrops kinetic
energy, thus mitigating their erosive effects (splash
erosion);
in case of heavy rainfall, a major fraction of meteoric water streams above the flattened vegetation
layer, even when the leaves are dried, waterproofing
the slope and substantially reducing the infiltration
of water;
the presence of vegetation may reduce the speed
of runoff water on the ground and therefore the
intensity of erosion;
there will be a delay in meeting the conditions of
complete saturation of the soil due to the mechanism of plant transpiration, as liquid water is
absorbed from the soil through the roots and transferred to the atmosphere as vapour (see Richards
(1931), Rassam and Cook (2002));
the soil is mechanically reinforced by the presence
of roots (see Waldron (1977)) when their quality
and strength match specific criteria (see Bischetti
et al. (2001), Bischetti et al. (2009), Bonfanti and
Bischetti (2001));
the vegetal blanket retains the soil particles, resulting in containment actions, filter and contrast of
grains sweeping away phenomena;
the content of organic matter in the soil increases,
which in turn leads to a reduction of the intrinsic
soil erodibility thanks to the formation, in the soil
itself, of more stable structural aggregates;
the increase in soil organic matter results in a significant increase of its water retention capacity
(see Celi (2010));
the retention of nutrients is also increased by both
the biological uptake by the vegetation and the
chemical withholding operated by the humus.
1.1
Improvement of fertility, structure and water

holding capacity of the soil due to the presence
of vegetation
organic matter may be considered as a mixture of

compounds derived from plants and micro-organisms
at different stages of degradation, starting from fresh
organic scraps up to products almost transformed into
humus. Humus is the fraction of organic matter most
active in terms of chemistry and physics and is derived
from its decomposition and elaboration. Its content
of organic matter varies from less than 1% in desert
soils to 115% in forest soils and to more than 90%
in the peat. The presence of organic matter in the
soil significantly increases its holding capacity of
water and nutrients: humus can hold water amounts
up to 20 times its own weight and may chemically
retain nutrients such as potassium, calcium, magnesium, phosphorus. It also determines a slowing-down
of the process of phosphorus retrogradation, the protection of trace elements from insolubilization and the
increase of buffering capacity.
The presence of organic matter in soil also reduces
its susceptibility to erosion because it leads to the
formation, in the soil itself, of more stable structural aggregates. It also favours the development of
soil fauna and microorganisms, since these will use it
as foodstuff substrate. The organic matter also plays
an important role in inactivating, by adsorption, various organic compounds bearing biotoxic action, both
from biological origin (polyphenols) and synthetic
(herbicides and pesticides in general). The soils rich
in organic matter are therefore important systems
to dispose of organic compounds bearing biotoxic
properties, thus are capable to reduce pollution of
groundwater.
2 TESTING GERMINATION AND ROOT
DEPTH IN CONTAMINATED SITES
To verify the ability of prati armati to vegetate in
contaminated soils, germination test have been carried out on soil samples taken from landfill mining
district of Montevecchio (south-western Sardinia).
These sterile materials contain residues from mines
tracking and extraction of minerals such as galena
(lead sulphide) and blende or sphalerite (zinc sulphide). Quantitative analysis of pollutants in different
soil samples were conducted using optical emission
plasma spectrometry ICP-OES (Inductively Coupled
Plasma-Optical Emission Spectrometer). The principal pollutants detected were: arsenic, cadmium, cobalt,
chromium, copper, mercury, nickel, lead, antimony,
selenium and zinc, in concentrations sometimes over
ten times higher than the upper limits admitted by
law. As an example the concentrations detected in soil
sample N 14 are reported in Table 1.
2.1 The germination test
The presence of vegetation causes a significant

increase in soil organic matter and this in turn improves
the soil structure, fertility and water retention. Soil
The investigation was conducted on 9 deep root grass

species and 7 soil samples taken from different dumps
of the same mining district. Each soil sample was used
to fill 9 pots of 16 cm diameter, thus allowing the 9
402
Table 1. Content of pollutants in contaminated soil samples

used in germination tests of prati armati.
Pollutants
Analysis date
Unit
Found val.
Ref. val.
As
Cd
Pb
Zn
01 June 2010
01 June 2010
01 June 2010
01 June 2010
mg/kg
mg/kg
mg/kg
mg/kg
544.4
140.3
9263.0
20216.5
50
15
1000
1500
Figure 1. The test pots after 1 month.
Figure 3. Plexiglas tube used for roots development testing.
In order to be acceptable for practical applications

the grass sample must therefore be verified against
the ability of its root system to develop and to deeply
penetrate in the contaminated soil.
2.2 The taking root test
Figure 2. Development of the root system within the test

pots, after 1 month.
grass species to be tested on every different soil. A

total of 63 pots were therefore monitored. The pots
were subjected to watering cycles simulating rainfall.
One month after sowing the following results were
obtained (see Fig. 1 and 2): out of the 9 herbaceous
species tested, at least 4 were able to germinate in all
soil samples, developing at the same time a root system
that embraced the entire volume of the pot.
Unfortunately germination is only the first stage
of the plant life as phytotoxic substances may kill the
sprout or affect its development by roots dwarfing.
Among the pots of the species that passed the first germination test, 4 were taken, one for each kind. Each
pot content was transplanted into a plexiglas tube with
a length of 2 m and diameter of 20 cm, filled with the
same type of contaminated soil of the pot (see Fig. 3).
The test tubes were equipped with drip irrigation system. Thanks to the transparency of the tube walls, it
was possible to monitor over time the root growth of
the 4 herbaceous species.
About a year after seeding it was clear that the root
growth was remarkable in all tested species, exceeding one meter depth in 50% of cases, with one specie
exceeding 1,80 m of root length. The study demonstrated therefore that the tested grass species are not
only capable to germinate on contaminated soils, but
may vegetate and increase the depth of their root system (Fig. 4). It was then possible to point out which
species are best suitable for the treatment of sites
according to the specific phytotoxic agent (arsenic,
403
Figure 5. The MSW landfill Ozieri (Sardinia) in November

2005, before the intervention.
Figure 4. Detail of the root system.
cadmium, cobalt, chromium, copper, mercury, nickel,

lead, antimony, selenium, zinc).
3 APPLICATIONS FOR RENATURALIZATION

OF QUARRIES, MINES AND LANDFILLS
The application fields of these technologies, such as
the one developed in Italy by Prati Armati srl. (see
Napoli (2011)), is quite wide: ridges and embankments
of roads and railways, abandoned quarries and mines,
sea facing areas, embankments protection of rivers,
streams, artificial waterways and landfills, including
those for Municipal Solid Waste (MSW).
In particular, when used in contaminated sites and
old landfills:
coverage and water erosion protection of slopes is
obtained in a short time;
wind erosion, generating polluting clouds of dust,
is reduced;
the plants isolate the waste material from the environment and improve the visual impact of the
treated area;
leachate generation in landfills is strongly reduced,
thanks to the intense transpiration capacity of these
plants. Moreover, rainwater infiltration is highly
reduced as the slope waterproofing brings the major
fraction of meteoric waters to stream above the
flatten grassy layer;
downstream dragging of emerging waste is reduced,
as they will be wrapped within the vegetation cover.
The treatment of contaminated soil areas with deep
rooting herbaceous plants does not replace the remediation, but may provide a fast, affordable solution for
the emergency securing of landfills slopes. In addition,
these interventions (see Italian Decree DM 471/99) fall
within the category of in situ techniques that do not
foresee handling or removal of polluted soil and waste.
Figure 6. The MSW landfill in May 2006, after the intervention of re-naturalization with prati armati.
3.1 An example of intervention for re-naturalization

of a MSW landfill
A typical example of installation of deep rooting herbaceous plants, aimed at the re-naturalization of a MSW
landfill, was carried out in Sardinia and represented
in Figures 5 and 6. A few months after intervention, the sown herbaceous species could completely
re-naturalize the site despite the unfavourable climatic
conditions. The deep root system could moreover protect the superficial portion of the slope, contrasting
at the same time the erosion and reducing leachate
production (see Fig. 6).
3.2 An example of intervention for the
re-naturalization of an abandoned quarry
An example of re-naturalization intervention on an
abandoned quarry through the use of deep rooting
herbaceous species was carried out in Sicily, near
Catania, in an area currently used for industrial
separation of MSW.
The intervention is shown in Figures 7 and 8. A few
months after seeding, planted herbaceous species have
completely re-naturalized the slope, blocking erosion
(see Fig. 8).
404
Figure 7. An abandoned quarry near Catania (Sicily), currently used for industrial waste processing: situation in
February 2010, before the intervention.
Figure 9. The abandoned limestone quarry in Spoleto:

situation in October 2010, prior to intervention.
Figure 8. The abandoned quarry in April 2011, after the

intervention of re-naturalization.
Figure 10. The limestone quarry in May 2011 showing the

first evidence of re-naturalization.
3.3 Another example of intervention for

re-naturalization of a quarry
fractured rocks, soils treated with addition of lime up

to 5% by weight and polluted by wastes and hydrocarbons may be successfully treated. Three interesting
cases show pictures of wastes landfill and abandoned quarries before and after treatment. This paper
illustrates the preliminary investigations in view of
extending the application to sites where high levels of
phytotoxicity further worsen the possibility of vegetation survival. Laboratory-scale tests were conducted
on a number of grass species in order to assess their
capability to germinate and take roots in mining scrap
samples where heavy metals concentration reaches 10
times the limits admitted by law. The very positive
results encourage to proceed with full scale deployment in contaminated sites and old landfills where
this innovative technology may represent a unique and
effective approach considering that time constraints
and high costs often prevent the use of traditional
remediation procedures.
Another example of a re-naturalization of an abandoned limestone quarry through the use of deep
rooting herbaceous species, is the one carried out in
the Umbria region near Spoleto (see Fig. 9).
After about 7 months from intervention, despite the
prohibitive climatic conditions, the sown herbaceous
species began to colonize the slope, triggering the
process of re-naturalization. This process is still active
and Figure 10 shows the first visible results of the
ongoing re-naturalization.
4
CONCLUSIONS
An emerging innovative solution based on planting

natural herbaceous perennials species with deep rooting system proved to be effective in contrasting soil
erosion and favouring renaturation. Careful selection of proper species allows to operate in areas
where climatic conditions and soil characteristics were
until a few years ago considered prohibitive for the
development of vegetation. Barren lands, altered or
REFERENCES
Bischetti, G., F. Bonfanti, & M. Greppi (2001). Misura della
resistenza a trazione delle radici: apparato sperimentale
405
e metodologia danalisi. Quaderni di Idronomia Montana 21/1, 349360.

Bischetti, G., E. A. Chiaradia, & T. Epis (2009). Prove di
trazione su radici di esemplari di piante PRATI ARMATI.
Rapporto conclusivo. Dipartimento di Ingegneria Agraria,
Universit degli Studi di Milano.
Bonfanti, F. & G. Bischetti (2001). Resistenza a trazione delle
radici e modello di interazione terreno-radici. Rapporto
interno. Istituto di Idraulica Agraria, Milano.
Cecconi, M., V. Pane, P. Napoli, & E. Cattoni (2012). Deep
roots planting for surface slope protection. Electronic
Journal of Geotechnical Engineering 17U, 28092820.
Cecconi, M., V. Pane, P. Napoli, & C. Zarotti (2013, July).
Mechanical and hydraulic effects of deep roots planting on
slope stability. In Coupled Phenomena in Environmental
Geotechnics Torino 2013. Accepted for publication.
Celi, L. (2010). La sostanza organica del suolo. Internet page:
Environnement, Ambiente e Territorio in Valle dAosta.
Napoli, P. (2011). Studio degli effetti meccanici di impianti
radicali profondi nel terreno. Tesi di laurea, Universit
Degli Studi Di Perugia.
Rassam, D. & F. Cook (2002). Predicting the shear strength

envelope of unsaturated soils. ASTM geotechnical testing
journal 25(2), 215220.
Rettori, A., M. Cecconi, V. Pane, & C. Zarotti (2010, March).
Stabilizzazione superficiale di versanti con la tecnologia PRATI ARMATI: implementazione di un modello di
calcolo per la valutazione del coefficiente di sicurezza.
In Accademia Nazionale dei Lincei-X Giornata mondiale
dellacqua, Convegno:Frane e dissesto idrogeologico.
Richards, L. (1931). Capillary conduction of liquids through
porous mediums. Physics 1(5), 318333.
Waldron, L. (1977). The shear resistance of root-permeated
homogeneous and stratified soil. Soil Science Society of
America Journal 41(5), 843849.
Wischmeier, W. & D. Smith (1965). Predicting rainfallerosion losses from cropland east of the Rocky Mountains:
Guide for selection of practices for soil and water conservation, Volume 282. Agricultural Research Service, US
Department of Agriculture.
406

Micropore vs. macropore flow: Implications for landfill final cover design
M.V. Khire & D.S. Saravanathiiban
Michigan State University, E. Lansing, USA
ABSTRACT: Alternative final covers made up of only soils are relatively common in arid, semi-arid, and subhumid regions of the United States. However, long term performance of earthen covers containing fine-grained
soils can be jeopardized due to the formation of preferential flow pathways. In this paper, field data collected
from an instrumented earthen cover located in a sub-humid climate is presented. In addition, preliminary water
balance modeling results of the cover using the numerical model UNSAT-H are included. During the first year
after the construction, flow through the cover is dominated by capillary pores. Whereas, during the second year,
there is additional contribution to the flow due to the formation of macropores. UNSAT-H is able to simulate the
percolation relatively accurately for the first year but not for the second year because Richards equation based
models only simulate flow through capillary pores.
1
1.1
INTRODUCTION
1.2 Water balance modeling
Landfill cover
Due to economical and sustainability reasons, the use

of earthen covers as an alternative to composite landfill
final covers (which contain synthetic geomembrane
liner), has been extensively tested during the last two
decades. As a result, many earthen covers have been
permitted for new landfills and old waste dumps in
the United States. Several field-scale demonstration
projects were carried out across the United States to
assess the performance of earthen covers for a wide
range of geoclimatic conditions. A majority of these
projects were carried out under the Alternative Cover
Assessment Program (ACAP) of the United States
Environmental Protection Agency (USEPA).
Lysimeters are built at field demonstration sites to
quantify percolation through the landfill cover. Percolation is the infiltrated water from the cover that is
assumed to enter the underlying waste. Nests of sensors that measure soil suction and volumetric water
content are installed along the depth of the proposed
alternative earthen cover to monitor changes in the soil
water contents and meteorological data is collected
to assess and model the long-term percolation from
earthen covers before they are permitted.
Field data from such demonstration projects is
collected over a period of only few years. However, if the earthen cover is permitted for permanent
use at the site, it is required to perform satisfactory
beyond the landfills 30-year post-closure monitoring
period. To extrapolate and predict long-term performance of earthen covers, water balance models are
routinely used.
Commonly used numerical models for earthen covers

are UNSAT-H, HYDRUS, and VADOSE/W wherein
independently measured hydrological data and soil
hydraulic properties are used as input parameters
(Khire et al. 1999, Benson et al. 2001). These models are called water balance models because they
simulate surface runoff, soil water storage, infiltration/percolation, and evapotranspiration from soil
covers.
Several studies have evaluated the ability of these
models to predict the field water balance (Khire et al.
1997, Benson et al. 2005). The numerical predictions of these studies were reasonably accurate for soil
water storage, surface runoff and evapotranspiration.
However, they were not always within the acceptable
level of accuracy for percolation. Percolation is typically the smallest quantity among the water balance
parameters and requires the highest accuracy due to
its environmental implications. These commonly used
state-of-the-art models are based on Richards Equation for flow through saturated-unsaturated porous
materials that exhibit capillarity. These models are
not designed to simulate flow in macropores where
the liquid flow is predominantly downward, similar
to in a conduit. Figure 1a shows a schematic of an
as-built compacted clay landfill cover where the flow
is dominated by micropores. The transient liquid flow
in micropores is downward or upward depending on
the hydraulic gradient due to the soil capillarity and
gravity. Figure 1b shows a schematic of the landfill
cover after the formation of macropores where liquid flow is through micropores as well as through the
macropores.
407
Figure 2. Measured and simulated percolation from a

compacted clay cover shows the inability of conventional Richards equation-based numerical models to predict
macropore flow.
Figure 1. Schematic of flow through an earthen cover

containing only micropores (a) versus micropores and
macropores (b).
Because most earthen covers exhibit various

degrees of preferential flow during the several decades
of service life (Benson et al. 2007), inability to
model preferential flow is one of the key limitations
of existing numerical water balance or vadose zone
models.
Figure 2 shows percolation measured for a clay
cover located in semi-arid climate where the Richards
equation based model, UNSAT-H (Fayer et al. 1992)
was not able to capture the relatively large pulse of
percolation measured during the third year after the
cover was built.
1.3
Macropore flow models
Shrinkage or desiccation cracks due to drying, wetting, and freezing, and thawing, vegetation root
penetrations, rodent holes, and worm holes trigger
preferential flow in soil systems. Beven and Germann

(1982) presented an overview of the relevant processes and conditions that cause the preferential flow
in soils. Macropore flow reduces the surface runoff and
increases the total infiltration (Workman and Skaggs,
1990). Daniel (1984) found 5 to 10,000 times higher
saturated hydraulic conductivity in the field than laboratory measured values for compacted fine-grained
soils. Miller and Mishra (1989) published a conceptual
framework for a numerical model to simulate micro
and macropore flow through clay liners. Khire et al.
(1997) and Albright et al. (2006) published field evidence for occurrence of preferential flow in the field.
Current preferential flow models that can simulate
water balance can be viewed under two categories. One
is dual-permeability/porosity models based on capillary flow and the other one is two-domain models. The
dual permeability/porosity models use the Richards
equation for the macropore flow system with higher
conductivity or porosity for the macropore region.
Thus, flow in the macropore system can be upward
during evapotranspiration (Kohne et al. 2009). One of
such models is HYDRUS-1D (Simunek et al. 1998,
2008, Simunek & van Genuchten, 2008). Two-domain
models implement gravity flow in macropore flow
and Ritchards equation for capillary flow. Root Zone
Water Quality (RZWQM) is a two-domain water balance model developed by Agricultural Research Service (ARS), United States Department of Agriculture
(USDA) (Ahuja et al. 1995, Ahuja et al. 2000).
Poiseulles equation is utilized to simulate the flow
in macropores in RZWQM. In this 1D vertical multilayer flow model, flow through cylindrical macropores
in surface layer and cylindrical or in plane macropores
in the layers beneath the surface layer can be simulated.
For saturated flow, a numerical model based on the
Lattice Boltzmann Method (LBM) (Succi 2001, WolfGladrow 2000) has been used. LBM-based model
requires high resolution pore structure geometry that
is generally developed from X-ray imaging. The LBM
was studied and validated to measure the saturated
hydraulic conductivity of porous media by many
researchers (Succi et al. 1989, Ferreol and Rothman
408
1995, Kutay et al. 2006, Sukop et al. 2013). However,

solid matrix was considered impermeable which is a
reasonable assumption for aggregates and rocks.
Table 1.
1.4 Objective
One of the key objectives of this study is to quantify
the influence of macropore on percolation. This paper
presents field data collected over a period of two years
(from November 2009 to end of October 2011) from
an instrumented earthen cover located in a sub-humid
climate. During the first year, flow was predominantly
through micropores and during the second year, flow
was through micropores and macropores. In addition,
results of preliminary water balance modeling of the
cover using the numerical model UNSAT-H are also
presented. Simulations were carried out for the two
years to evaluate the accuracy of the model predictions.
2
Physical properties of soils.
Soil Property
Topsoil
Storage Layer
Unified Soil
Classification
System classification
D10 (mm)
D50 (mm)
D60 (mm)
Coefficient of
uniformity, Cu
Coefficient of
curvature, Cc
Liquid limit
Plasticity index
Optimum water content
for Standard Proctor
(SP) effort (%)
Maximum dry unit weight
for SP effort (kN/m3 )
Low-plasticity
silt
Low-plasticity
clay
9.2 104
7.0 102
9.1 102
99
2.0 104
6.5 103
1.4 102
70
11
0.70
25
3
15.7
29
13
13.4
17.4
18.5
FIELD SCALE TEST SECTION
2.1 Test section construction

A compacted 1.8 m thick (6 ft.) instrumented earthen
cover was constructed at a landfill located in Detroit,
Michigan in late October 2009. A lysimeter pan was
used to measure the percolation. The test section was
constructed using a 1.5 m thick native glacial till (clay)
which was overlain by a 0.3 m thick topsoil layer. The
clay was compacted at dry of optimum water contents in 30 cm thick lifts using a sheepsfoot compactor.
The lysimeter consisted of a trapezoidal excavation of
the test section having plan view dimensions of about
8.5 m 8.5 m and 0.6 m deep. Time Domain Reflectometry (TDR) sensors and Heat Dissipation Matric
Potential (HDMP) sensors were used at six different
depths in three vertical nests along the test section profile to measure the volumetric water content and matric
suction, respectively. More details on the field test
section can be found in Mijares et al. (2012) and
Mijares and Khire (2012). Physical properties of soils
used for the test section are summarized in Table 1.
2.2
Figure 3. Recorded air temperatures.
Field data
The site had a weather station to record air temperature,

precipitation, solar radiation, barometric pressure, and
relative humidity and sensors to monitor in-situ volumetric water contents, soil water potentials, soil
temperatures, gas pressures, and level of water in the
lysimeter pan to monitor percolation. The daily maximum and minimum temperatures recorded at the site
are shown in Figure 3. In winter of year 1, the lowest minimum temperature was about 16 C whereas
during the winter of year 2 the lowest temperature
reached 20 C. During the summer of year 1 the highest maximum temperature was 35 C while the highest
temperature during the summer of year 2 was 38 C.
Figure 4 shows the recorded net solar radiation at the
site. The average net daily solar radiation during year
1 was 320 langleys and during year 2 it dropped to 289
langleys.Year 2 was the record wettest year in 50 years.
Figure 4. Recorded net solar radiation.
The cumulative percolation and precipitation

recorded at the site are presented in Figure 5. Annual
precipitation within 5 km of the site recorded by
National Oceanic and Atmospheric Administration
409
NUMERICAL MODELING
3.1 UNSAT-H
The 1-D finite-difference water balance model
UNSAT-H Ver. 3.0 (Fayer 2000) was used to numerically simulate the water balance of the test section.
UNSAT-H numerically solves the modified form of
Richards partial differential equation to simulate
water flow through both saturated and unsaturated soil.
UNSAT-H can simulate water flow as well as heat
flow processes in 1D. Also, the model is capable of
simulating steady-state and transient conditions. This
model has been widely used for water balance modeling of landfill earthen covers (Khire et al. 1997, Khire
et al. 2000, Benson et al. 2005, Bohnhoff et al. 2009,
Mijares and Khire 2012). The simulations were carried
out separately for each year. This approach allowed
to reinitialize the model with field measured initial
conditions for years 1 and 2, separately.
Figure 5. Measured percolation and precipitation.
3.2 Input parameters
Figure 6. Historical precipitation.
(NOAA) weather station during the 50 years is presented in Figure 6. During the first year, precipitation
recorded at the site was about 730 mm while the
cumulative percolation was 7 mm. During the second
year, the precipitation was 990 mm, while percolation was about 70 mm. The reason for about an order
increase in percolation during the second year is
partly due to record high precipitation during year 2.
Year 2 (2010) was the wettest year in the 50 year history (Fig. 6). Precipitation received during the first
year (2009) is equal to the average precipitation (50
percentile) for the site. During the second year (2010),
precipitation received at the site corresponds to about
98 percentile.
Macropores, shrinkage cracks and animal barrows,
were noticed on the surface of the test section during
the second year. An initial measurement of the depth of
the cracks suggested that some of the observed cracks
could be up to 0.6 m deep. These macropores may have
contributed to the much higher percolation observed
in year 2.
The top boundary of the model was assigned as a

flux boundary which considers infiltration and evapotranspiration as specified fluxes. Precipitation, daily
values of maximum and minimum air temperature,
dew point temperature, total solar radiation, average
wind speed, average cloud cover are required for the
prediction of infiltration and evaporation. Measured
data were used as input. Daily dew point temperatures
were collected from NOAA weather station. Because
net solar radiation was measured at the site, average
cloud cover was assumed to be zero. Hourly precipitation was applied as it was recorded by the tipping
bucket located at the site. However, the precipitation
recorded at the site did not include snow. Hence, snow
received during the winters recorded by the NOAA
weather station located about 5 km from the site was
used as input. The lower boundary was assigned unit
gradient to simulate the vented lower boundary of the
lysimeter.
Saturated and unsaturated soil properties input to
the model are presented in Table 2. The saturated
hydraulic conductivity of the compacted clay test section and the vegetative layer were measured in the
field using a 1 m by 1 m single-ring infiltrometer
(ASTM D5126, Guyonnet et al. 2000) and the average
value of saturated conductivities were 4.4 106 cm/s
and 1.0 102 cm/s, respectively (Mijares and Khire
2012). Brooks-Corey function was used to represent
soil water characteristic curve (SWCC) and the unsaturated hydraulic conductivity function. SWCC and
unsaturated hydraulic conductivities were estimated
using the water contents and suctions measured in the
field from the co-located sensors (Mijares and Khire
2010). Temperature depended calibration was carried
out to correct the effect of temperature on the TDR
measured time average. Instantaneous profile method
was used to calculate the unsaturated hydraulic conductivity (Meerdink et al. 1996). The same saturated
410
Table 2.
Unsaturated and saturated hydraulic properties.
Parameters
Topsoil
Storage Layer
s (cm3 /cm3 )
r (cm3 /cm3 )
Air entry pressure (cm)
SWCC function
Conductivity function B
SWCC function
Conductivity function
Pore interaction term
Hydraulic Conductivity (cm/s)
0.41
0.00
0.39
0.03
63
63
3.33
3.33
2.0
1 102
30
12
8.3
2
0.84
4.4 106
Figure 8. Measured and simulated percolation for year 2.
4.2 UNSAT-H results for year 2

Figure 8 presents precipitation and measured and
simulated percolation for year 2. No irrigation was
applied during year 2. Cumulative precipitation during the year was 1,038 mm. The simulated cumulative
percolation is about 38 mm which about half of the
measured percolation which was 70 mm. In addition,
the slope of the field cumulative percolation was
greater than that simulated by UNSAT-H. Possible
reason for this discrepancy is that during the second year, flow through the cover occurred not only
through the capillary pores but also through macropores. Macropores usually carry more flow when the
soil matrix approaches its water retention capacity.
Year 2 was the wettest year in the 50 year history.
Hence, the soil water storage of the cover was at its
peak for a relatively long duration of time when macropores carried a major fraction of the flow. This flow
could not be simulated by UNSAT-H.
Figure 7. Measured and simulated percolation for year 1.
hydraulic conductivity and hydraulic functions were

used to simulate years 1 and 2 in UNSAT-H.
3.3 Initial conditions

UNSAT-H requires initial suction values along the
entire cover profile to be provided as an input. Field
measured suctions on October 31st 2009 and October
31st 2010 were used as the initial conditions for the
simulation of years 1 and year 2, respectively. The initial soil water storages for years 1 and 2 were 425 mm
and 440 mm, respectively.
4.1 UNSAT-H results for year 1

Figure 7 presents the measured and simulated cumulative percolation for year 1 and precipitation with
irrigation. During year 1, cumulative water applied to
the model is 993 mm. The simulated cumulative percolation exceeds the measured percolation only by a
few mm. However, the model was not able to show the
effect of irrigation on percolation which was observed
in the field data. In addition, UNSAT-H predicted significant fraction of the applied irrigation as runoff.
However, field runoff was minimal.
CONCLUSION
Earthen covers containing fine-grained soils will

undergo structural changes during the service life.
These structural changes will include macropore formation due to shrinkage, freeze-thaw, plant root penetrations, and animal or worm burrows. Macropores
will results in additional percolation through the cover
especially during relatively wet years. Water balance models that do not account for macropore flow
will under estimate percolation. In order to simulate long-term percolation from earthen covers, water
balance models capable of simulating flow through
macropores should be used.
ACKNOWLEDGEMENT
This project has been funded by the Environmental
Research and Education Foundation, Waste Management, Inc. and the National Science Foundation (Grant
411
No. CMMI-1100020). The opinions expressed in this

paper are those of the authors. This paper has not been
reviewed by the sponsors.
REFERENCES
Ahuja, L.R., Johnson, K.E. & Heathman, G.C. 1995.
Macropore transport of a surface-applied bromide tracer:
Model evaluation and refinement. Soil Science Society of
America Journal 59: 12341241.
Ahuja, L.R., Rojas, K.W., Hanson, J.D., Shaffer, J.J. &
Ma, L. (Eds.) 2000. The Root Zone Water Quality Model.
LLC, Highlands Ranch, Colorado: Water Resources
Publications.
Albright, W., Benson, C., Gee, G., Abichou, T., Tyler, S. &
Rock. S. 2006. Field Performance of Three Compacted
Clay Landfill Covers. Vadose Zone Journal 5: 11571171.
Albright, W., Benson, C., Gee, G., Roesler, A., Abichou, T.,
Apiwantragoon, P., Lyles, B. & Rock, S. 2004. Field water
balance of landfill final covers. Journal of Environmental
Quality 33(6): 23172332.
ASTM D5126. 2010. Standard Guide for Comparison of
Field Methods for Determining Hydraulic Conductivity in
Vadose Zone. West Conshohocken, Pennsylvania: ASTM
International
Benson, C.H., Abichou, T., Albright, W.H., Gee, G.W., &
Albright, W.H. 2001. Field evaluation of alternative
earthen final covers. International Journal of Phytoremediation 3(1): 105127.
Benson, C., Bohnhoff, G., Ogorzalek, A., Shackelford, C.,
Apiwantragoon, P. & Albright, W. 2005. Field data and
model predictions for an alternative cover. Waste Containment and Remediation, GSP No. 142: 112. Reston,
Virginia: ASCE
Benson, C., Sawangsuriya, A., Trzebiatowski, B., & Albright,
W. 2007. Post-construction changes in the hydraulic properties of water balance cover soils. Journal of Geotechnical and Geoenvironmental Engineering 133(4): 349359.
Beven, K. J. & Germann, P. F. 1982. Macropores and water
flow in soils. Water Resources Research 18: 13111325.
Bohnhoff, G., Ogorzalek, A., Benson, C., Shackelford, C. &
Apiwantragoon, P. 2009. Field data and water-balance predictions for a monolithic cover in a semiarid climate.
Journal of Geotechnical Engineering 135(3): 333348.
Daniel, D. E. 1984. Predicting Hydraulic Conductivity of
Clay Liners. Journal of Geotechnical Engineering ASCE
110(2): 285300.
Fayer, M. 2000. UNSAT-H version 3.0: Unsaturated soil
water and heat flow model Theory, user manual, and
examples. PNNL-13249. Richland, Washington: Pacific
Northwest Laboratories
Fayer, M., Rockhold, M. & Campbell, M. 1992. Hydraulic
modeling of protective barriers: Comparison of field data
and simulation results. Soil Science Society of America
Journal 56: 690700.
Ferreol, B. & Rothman, D.H 1995. Lattice-Boltzmann
simulations of flow through Fontainebleau sandstone.
Transport Porous Media 20: 320.
Guyonnet, D., Amraoui, N. & Kara, R. 2000. Analysis of
transient data from infiltrometer tests in fine-grained soils.
Ground Water 38 (3): 396402.
Khire, M.V., Benson, C. & Bosscher, P. 1999. Field data from
a capillary barrier and model predictions with UNSAT-H.
Journal of Geotechnical and Geoenvironmental Engineering 125: 518528

Khire, M.V., Benson, C. & Bosscher, P. 1997. Water balance
modeling of earthen final covers. Journal of Geotechnical
Khire, M.V, Benson, C. & Bosscher, P. 2000. Capillary barriers: Design variables and water balance. Journal of
695708.
Kohne, J.M., Kohne, S. & Simunek, J. 2009. A review of
model applications for structured soils: a) Water flow
and tracer transport. Journal of Contaminant Hydrology:
435.
Kutay, M.E., Aydilek, A.H. & Masad, E. 2006. Laboratory validation of lattice Boltzmann method for modeling
pore-scale flow in granular materials. Computers and
Meerdink, J., Benson, C. & Khire, M.V. 1996. Unsaturated
hydraulic conductivity of two compacted barrier soils.
Journal of Geotechnical Engineering 122(7): 565576.
Mijares, R. G. & Khire, M. 2010. Soil water characteristic
curves of compacted clay subjected to multiple wetting
and drying cycles. ASCE Geotechnical Special Publication No. 199:Advances inAnalysis, Modeling and Design,
Reston, VA., 400409.
Mijares, R. & Khire, M.V. 2012. Field Data and Numerical
Modeling of Water Balance of Lysimeter versus Actual
Earthen Cap. Journal of Geotechnical and Geoenvironmental Engineering 138(8): 889897.
Mijares, R., Khire, M.V. & Johnson, T. 2012. Field-scale
evaluation of lysimeters versus actual earthen covers.
Geotechnical Testing Journal 35(1): 3140.
Miller, C. J. & Mishra, M. 1989. Modeling Leakage
Through Cracked Clay Liners II: A new Perspective. Water
Resources Bulletin 25(3): 551656.
Simunek, J., Sejna, M., Saito, H., Sakai, M. & van Genuchten,
M.Th. 2008. The HYDRUS-1D software package for simulating the one-dimensional movement of water, heat,
and multiple solutes in variably-saturated media. Version
4.0, HYDRUS Software Series 4. University of California
Riverside, Riverside, CA, USA.
Simunek, J., Sejna, M. & van Genuchten, M.Th. 1998. The
HYDRUS-1D software package for simulating the onedimensional movement of water, heat, and multiple solutes
in variably saturated media. Version 2.0. IGWMC-TPS70, School of Mines, Golden, Colorado: International
Ground Water Modeling Center
Simunek, J. & van Genuchten, M.Th. 2008. Modelling
nonequilibrium flow and transport with HYDRUS.
Vadose Zone Journal 7 (2): 782797.
Sukop, M.C., Huang, H., Alvarez, P.F., Variano, E.A. &
Cunningham, K.J. 2013: Evaluation of permeability and
non-Darcy flow in vuggy macroporous limestone aquifer
samples with lattice Boltzmann methods.Water Resources
Research 49: 216230.
Succi, S. 2001. The Lattice Boltzmann Equation for Fluid
Dynamics and Beyond, Oxford: Clarendon Press.
Succi S., Foti, E. & Higuera, F. 1989. Three dimensional
flows in complexgeometries with the lattice Boltzmann
method. Europhysics Letter 10: 433438.
Workman, S.R., & Skaggs, R.W. 1990. PREFLO: a water
management model capable of simulating preferential
flow. Transactions of the ASAE 33(6): 19391948.
Wolf-Gladrow, D.A. 2000. Lattice-Gas Automata and Lattice
Boltzmann Models: An Introduction. Berlin: Springer.
412

Coupled thermo-hydraulic pulse tests on two reference

Belgian clay formations
A. Lima, E. Romero, A. Gens & J. Vaunat
Universitat Politcnica de Catalunya, Barcelona, Spain
X.L. Li
EURIDICE/SCK.CEN, Mol, Belgium
ABSTRACT: Boom and Ypresian clays are being studied in Belgium in connection with the design of a repository for radioactive waste. Within this context, thermal impact may play an important role on the behaviour of
these low-permeability clays. To evaluate this impact, heating pulse tests on intact borehole samples were carried
out using an axi-symmetric and constant volume heating cell with controlled hydraulic boundary conditions.
Attention is focused on the time evolution of temperature and pore water pressure changes along heating and
cooling paths i.e., pore pressure build-up during quasi-undrained heating and later dissipation to the applied
hydraulic boundary conditions. A coupled thermo-hydro-mechanical finite element program was used in a first
stage to determine thermal parameters by back-analysis and then to simulate selected experimental results on
Boom clay.
2 TESTED MATERIALS AND EQUIPMENT
INTRODUCTION
Thermal impact may play an important role on the

behaviour of low-permeability saturated clayey host
formations in connection with the design of a repository for High-Level Radioactive Waste. Boom and
Ypresian clays are currently the subject of extensive
research on thermal, hydraulic and mechanical phenomena that may possibly affect its performance as
potential geological host formations.
There are a number of laboratory results concerning
the saturated hydro-mechanical behaviour of Boom
clay under a constant temperature field and studies
on this area are described -to cite but a few of them- in
Horseman et al. (1987), De Bruyn (1999), Le (2008)
and Lima (2011). Nevertheless, there is less information on clay hydro-mechanical response on heating
and cooling paths under controlled small-scale laboratory condition. To this end, the paper presents results
from a comprehensive testing program performed on
Boom and Ypresian clays to determine thermal and
hydraulic parameters using an axi-symmetric heating
cell with controlled-power heater and measurement of
temperature and pore pressure at different positions.
The thermal response on both clays has been simulated
with the finite element program CODE_BRIGHT
to calibrate by back-analysis some thermal properties (thermal conductivity of the soil and convection
coefficient that controls thermal flux at the cell interface). The data was further used to calibrate selected
hydraulic parameters on Boom clay by analysing joint
thermal and hydraulic results.
2.1 Tested materials

The Boom clay is the youngest sea sediment, which
was deposited 36 to 30 million years ago during the
Rupelian stage (Northwest European Tertiary Basin).
At Mol (Belgium) it is located between 160 and 270
m depth. The studied sample has been retrieved at a
depth of 223 m using a horizontal borehole. Table 1
summarises the main properties of this clay (20%30% kaolinite, 20%30% illite, 10%20% smectite,
25% quartz and feldspar), which is slightly overconsolidated (Coll 2005, Lima 2011).
TheYpresian clay is also constituted by marine sediments deposited during the Ypresian stage (Eocene
series, 54 to 49 million years ago). Smectite is the dominant phase (25%), followed by illite (15%) and small
amounts of kaolinite and chlorite (ONDRAF/NIRAS
2005). Non-clayey fraction is constituted predominantly by quartz (45%) and feldspar (5%). Ypresian
clay at Kallo (Belgium) is located between 300 and
450 m depth, and the tested material has been retrieved
at depth of 350 m using a vertical borehole. Table 1
summarises the basic characterisation and the main
volumetric and gravimetric properties of this clay,
which is also lightly overconsolidated. As observed
in the table, Ypresian clay displays a larger liquid
limit, which is consistent with the important smectite content. The slightly larger water permeability
of Ypresian clay is a consequence of its higher
void ratio.
413
Table 1.
Main properties of Boom and Ypresian clays.
Property
Boom clay
Ypresian clay
Density (Mg/m3 )
Dry density (Mg/m3 )
Gravimetric water content (%)
Density of soil solids (Mg/m3 )
Void ratio
Liquid limit
Plastic limit
Vertical water permeability at
22 C (m/s)
Vertical water permeability at
80 C (m/s)
Horizontal water permeability
at 22 C (m/s)
Small-strain shear
modulus (MPa)
Poissons ratio
2.05
1.651.67
25
2.67
0.600.62
1.00
56
29
2.3 1012
1.972.02
1.551.60
26 to 26.6
2.76
0.720.78
0.940.98
142158
26
6.6 1012
6.7 1012 2.0 1011

4.5 1012 7.9 1012
347
150
0.20
0.25
Figure 1. Axi-symmetric heating cell and transducers

(Muoz et al. 2009, Lima et al. 2010, Lima 2011).
2.2
as shown in the figure: two miniature pore water

pressure transducers (Pw1 and Pw2 ), and three thermocouples (T1 near the heater, and T2 and T3 , placed
in the cell wall and on the opposite side). The cell is
equipped with top and bottom valves to apply controlled hydraulic boundary conditions (uu and ub ).
The heater with controlled power supply remains
switched on for 24 hours during the heating stage and
later it is switched off to perform the cooling phase.
The initial and external temperatures are regulated by
submerging the cell inside a temperature controlled
water bath at temperature T4 .
Boom clay sample has been oriented with bedding
planes parallel to the axis of symmetry of the cell,
whereas Ypresian clay has been tested with bedding
planes orthogonal to the axis of symmetry.
Experimental setup
Figure 1 shows a scheme of a constant volume and axisymmetric heating cell (Muoz et al. 2009, Lima et al.
2010, Lima 2011), which has been used to perform
heating pulse tests with controlled power supply and
controlled hydraulic boundary conditions. Soil sample size is 75 mm in diameter and 100 mm high. A
controlled-power heater (H in the figure) is installed
along the axis of the sample in the lower part of the cell.
Different transducers monitor the sample response,
EXPERIMENTAL RESULTS
Attention has been focused on the time evolution of

temperature and pore water pressure changes during heating and cooling paths (i.e., pore pressure
build-up during quasi-undrained heating and later dissipation to the applied hydraulic boundary conditions).
Throughout the course of the heating/cooling paths,
the bottom drainage is maintained open at constant
water backpressure (1 MPa in tests with Boom clay and
1.2 MPa with Ypresian clay) using an automatic pressure/volume controller, while the upper valve is kept
closed. This backpressure is important since it allows
measuring the pore pressure drop during the cooling phase without reaching the negative range (below
atmospheric conditions).
Figure 2a shows the time evolution of temperature for different locations and along a heating and
cooling cycle performed on Boom clay up to a maximum heater temperature of 54 C. Figure 2b presents
the corresponding time evolution of pore water pressures at different locations. During heating, pore water
pressure increases due to the larger thermal expansion
coefficient of water and its discrepancy with the thermal expansion of the pore volume. The magnitude of
the water pressure change depends on the rate of temperature increase/decrease, on soil compressibility and
thermal-expansion coefficient, on water permeability
and porosity, as well as on the hydraulic boundary
condition applied. After the heating path, pore water
pressure dissipates at constant temperature towards the
hydraulic value applied at the boundary. Pore water
pressure Pw2 at the top dissipates more slowly due to
the larger distance to the draining boundary. An opposite pore pressure evolution is observed on cooling: an
initial pore pressure drop followed by pressure recovery to the applied boundary condition (again, Pw2
recovers more slowly).
Figure 3 shows a zoom of the early stage evolution of pore water pressure Pw1 and temperature T2
changes during another heating phase. These sensors
are located close to the draining boundary (Figure 1).
It can be observed that pressure changes develop at a
414
Figure 3. Zoom of time evolution of temperature and pore

water pressure during heating (Lima et al. 2010, Lima 2011).
Figure 2. Temperature and pore water pressure evolutions

during heating and cooling paths on Boom clay (Lima et al.
2010, Lima 2011).
faster rate compared to temperature changes. In fact,

pore pressure starts to dissipate towards the applied
hydraulic boundary condition well before the temperature reaches its maximum value. The maximum
thermally induced pore fluid pressurisation coefficient
under quasi-undrained conditions reached a value
Pw /T = 0.192 MPa C1 during heating.
Figures 4a and 4b show the corresponding time evolutions of temperature and pore pressure for different
locations during the heating and cooling phases performed on Ypresian clay. A maximum temperature of
56 C has been attained in the heater T1 and a maximum
pressure of 1.42 MPa measured at Pw2 (see Figure 4b).
During this heating stage, a larger pore pressure buildup is detected at Pw2 , which dissipates at a slower rate
due to the larger distance to the draining boundary.
The maximum thermally induced pore fluid pressurisation coefficient under quasi-undrained conditions
reached a lower value Pw /T = 0.066 MPa C1 during heating compared to Boom clay. This lower value
is consistent with the higher compressibility of the
pore-volume of Ypresian clay during pressurisation
(refer to the lower small-strain shear modulus of
Ypresian clay reported in Table 1). Ghabezloo &
Sulem measured thermal pressurisation coefficients
between 0.05 and 0.25 MPa C1 during undrained
heating tests performed on sandstone.
INTERPRETATION OF THERMAL RESULTS
In the interpretation of the test results, it was assumed

that temperatures and heat flux were not influenced by
water pressure and flow, which means that heat convection was assumed to be negligible. The driving process
for temperature change during the test is thus conduction only. This assumption is justified by the condition
of constant overall volume prevailing in the heating
cell that makes the change in porosity and the velocities of the solid phase very small. Moreover, the low
permeability of the material prevents the existence of
high velocities for the liquid phase. The flux of heat
convected by the solid and liquid phases is, therefore,
extremely low. In contrast, water pressure and flow
were assumed to be influenced by temperature: as a
consequence, while the thermal problem was decoupled from the hydraulic one, the hydraulic problem
was coupled to the thermal one.
The test was then interpreted in two separated
stages. First, a back-analysis of temperature measurements was carried out by performing uncoupled
thermal simulations using the finite element program CODE_BRIGHT (Olivella et al. 1996); only
the balance equation for energy was solved. Heat
exchanged by the highly conductive stainless steel cell
with the controlled water bath was accounted for as a
convection-type boundary condition of the problem.
This heat flux was assumed to be proportional to the
difference between the temperature of the cell and the
temperature of the water bath at each boundary node,
through a convection coefficient h. Thermal optimisation was then aimed at identifying the values of the
saturated thermal conductivity and the convection
coefficient h. Calculations were performed for different combinations of and h. For each of them,
a measure of the least squares difference between
temperature simulation results and temperature
experimental measurements was computed for different elapsed times. The three-dimensional plot in
Figures 5 and 6 display the least squares differences
between simulation results and experimental observations of Boom and Ypresian clays, respectively.
The best agreement for Boom clay was obtained at
415
Figure 6. Three-dimensional graph showing the differences

in temperature between observations and calculations in the
back-analysis of the heating pulse test performed onYpresian
clay. Determination of thermal conductivity and convection
coefficient h (Lima 2011).
Figure 4. Temperature and pore water pressure evolutions

during heating and cooling paths on Ypresian clay (Lima
2011).
Back-analysed thermal parameters were used to study

the coupled thermal and hydraulic results on Boom
clay. Water permeability and elastic soil parameters
used in the simulations, which are reported in Table 1,
were obtained from independent tests. Controlledgradient tests at different temperatures and constant
volume conditions for water permeability, as well as
small-strain shear moduli with resonant column and
bender element tests, have been reported by Lima
(2011). Figure 7 displays the time evolution of temperature and pore water pressure (experimental and
simulated results) during the same heating and cooling paths presented in Figure 2. A good agreement is
observed in the pore water pressure response, which
shows the consistency between the back-calculated
and directly measured parameters from independent
laboratory tests.
Figure 5. Three-dimensional graph showing the differences

in temperature between observations and calculations in the
back-analysis of the heating pulse test performed on Boom
clay. Determination of thermal conductivity and convection
coefficient h (Lima 2011).
parameters = 1.6 Wm1 K1 and h = 24 Wm2 K1 .

In the case of Ypresian clay, these values were
= 1.9 Wm1 K1 and h = 11 Wm2 K1 , as shown in
Figure 6. The higher thermal conductivity of Ypresian
clay can be associated with the higher quartz content
reported in section 2.1.
SIMULATION OF SELECTED BOOM CLAY

RESULTS
CONCLUSIONS
A series of heating and cooling paths were performed on Boom and Ypresian clays (reference host
formations for potential geological disposal of HighLevel Radioactive Waste in Belgium) to study the
impact and consequences of thermal loads on this lowpermeability clay formation. Tests were performed in
a fully-instrumented heating cell (with several thermocouples and pressure transducers) under constant
volume and controlled hydraulic boundary condition:
constant water pressure at the bottom drainage and top
end with no flow condition. Selected results of a comprehensive experimental programme on intact borehole samples were presented and discussed in terms
of the joint measurements of temperature and pressure changes during the application of heating-cooling
cycles.
416
ACKNOWLEDGEMENTS
The authors acknowledge the financial support provided by EIG-EURIDICE/SCK.CEN and ONDRAF/
NIRAS (Belgium) through PhD and post-doc collaborations with International Centre for Numerical
Methods in Engineering (Spain). The contribution of
Yessenia Pia to the experimental work with Ypresian
clay is also greatly acknowledged.
REFERENCES
Figure 7. Time evolution of temperature and pore water

pressure: experimental and simulated results (Lima 2011).
Thermal and hydraulic results were calibrated and

simulated using coupled thermo-hydro-mechanical
analyses performed with a finite element code
(CODE_BRIGHT). In particular, the thermal conductivity of the two clays was determined by back-analysis
of the thermal response. The coupled thermal and
hydraulic results on Boom clay were also successfully simulated using parameters that had been backcalculated from previous heating pulse tests and also
directly from independent laboratory tests. An overall
examination of the results obtained allows the identification of the main features of the hydro-thermal
coupling under the test conditions.
Coll C. 2005. Endommagement des roches argileuses et permabilit induite au voisinage douvrages souterrains.
PhD Thesis, Universit Joseph Fourier, Grenoble, France.
De Bruyn D. 1999. Influence of a temperature increase on
the physical and mechanical behaviour of the Boom Clay
in the framework of the HLW geological disposal. PhD
Thesis, UCL, Louvain-La-Neuve, Belgium.
Ghabezloo S. and Sulem J. 2010. Temperature induced pore
fluid pressurization in geomaterials. Rivista Italiana di
Geotecnica XLIV (1), 3549.
Horseman S.T., Winter M.G., Entwistle D.C. 1987. Geotechnical characterization of Boom Clay in relation to the
disposal of radiactive waste. Commission of the European
Communities, Report EUR 10987.
Le T.T. 2008. Comportement thermo-hydro-mcanique de
largile de Boom. PhD Thesis, CERMES, Ecole Nationale
des Ponts et Chausses, Paris, France.
Lima A., Romero E., Gens A., Muoz J. and Li X.L. 2010.
Heating pulse tests under constant volume on Boom clay.
Journal of Rock Mechanics and Geotechnical Engineering
2 (2), 124128.
Lima A. 2011. Thermo-hydro-mechanical behaviour of two
deep Belgian clay formations: Boom and Ypresian clays.
PhD Thesis. Universitat Politcnica de Catalunya, Spain.
Muoz J.J., Alonso E.E., Lloret A. 2009. Thermo-hydraulic
characterization of soft rock by means of heating pulse
tests. Gotechnique 59(4), 293306.
Olivella S., Gens A., Carrera J., Alonso E.E. 1996. Numerical formulation for a simulator (CODE_BRIGHT) for the
coupled analysis of saline media. Int. Journal Engineering
Computations 13 (7), 87112.
ONDRAF/NIRAS 2005. The Ypresian clays as possible
host rock for Radioactive Waste Disposal: an evaluation.
Publication NIROND 2005-01.
417

Improved correction factor for indirect determination

of permeability function
Priono
Interdisciplinary Graduate School, Nanyang Technological University, Singapore
H. Rahardjo, E.C. Leong & J.Y. Wang

School of Civil and Environmental Engineering, Nanyang Technological University, Singapore
K. Chatterjea
National Institute of Education, Nanyang Technological University, Singapore
ABSTRACT: Capillary Barrier System has been widely utilized in geo-environmental engineering. Basically,
the system consists of a fine-grained soil layer overlying a coarse-grained soil layer and requires good understanding of unsaturated soil mechanics, particularly on the unsaturated permeability characteristics of capillary barrier
materials. This paper focuses on the estimation of permeability function from measured soil-water characteristic
curve (SWCC). A statistical method proposed by Childs & Collis-George (1950), which was modified by Kunze
et al. (1968), is usually used to calculate unsaturated permeability function from SWCC. However, Agus et al.
(2005) suggested using Mualem (1976) statistical model with Fredlund & Xing (1994) SWCC equation based
on the earlier works by Agus et al. (2003). In this paper, Mualem (1976) equation was modified to produce more
accurate results with physically meaningful parameters for coarse-grained soils. The modified Mualem (1976)
equation was verified against experimental data from the literature and the results indicated good agreement
between the estimation and the experimental data.
INTRODUCTION
Climate change contributes significantly towards one

of the most frequent geo-environmental problems
i.e. rainfall-induced slope failures around the world
(Rahardjo et al. 2012a). Climate changes triggered
longer dry season and more intense precipitation during the wet season than ever before in United Kingdom
(Mendes 2011). Springman et al. (2003) suggested
that rainwater infiltration and snowmelt were the most
detrimental factors for slope failures in Switzerland.
Evangelista et al. (2008) and Papa et al. (2008) showed
the significance of rainwater infiltration on slope
failures in Naples, Italy.
Climate conditions in Singapore are characterized
by abundant rainfalls due to the fact of high humidity and small temperature range in tropics. Global
warming and urban heat island play a significant role
in increasing volume and intensity of the precipitation as recorded by National Environmental Agency
(2010). Previous research by Chatterjea (1989), Toll
et al. (1999) and Rahardjo et al. (2001) showed that
many natural slopes in Singapore failed during heavy
rainfalls.
As slope failures often cause serious damages,
prevention methods such as Capillary Barrier System (CBS) become more essential (Tami et al. 2004,
Krisdani et al. 2006, Rahardjo et al. 2012b). CBS

consists of a fine-grained soil layer overlying a coarsegrained soil layer which systematically prevents rainwater infiltration into slope. Rainwater infiltrates
through the unsaturated zone of slope and then builds
up pore-water pressure in the slope which may ultimately cause slope failure. Understanding rainfallinduced slope failure mechanism and its preventive
measures requires good understanding of unsaturated
soil mechanics, particularly on the hydraulic characteristics of unsaturated soil.
Hydraulic characteristics of unsaturated soil refer
primarily to soil-water characteristic curve (SWCC)
and permeability function. SWCC indicates variation of water storage capacity within macropores and
micropores of soil subjected to variation in matric
suctions. One of the most important parameters of
SWCC is air-entry value (AEV) where the soil begins
to lose water and it becomes unsaturated (Fredlund &
Rahardjo 1993).
Water coefficient of permeability of an unsaturated
soil is a function of water content or matric suction. Direct measurement of permeability function is
a very time-consuming process (Leong & Rahardjo
1997b). The duration of the test increases as the water
content in the soil decreases. Therefore, indirect measurements of permeability are frequently performed
419
through the use of SWCC (Leong & Rahardjo 1997b).

Mualem (1986) suggested three types of indirect methods to determine permeability function, specifically
empirical, macroscopic, and statistical methods. These
methods differ on the degree of theoretical sophistication, with statistical models being the most rigorous
methods.
A statistical method proposed by Childs & CollisGeorge (1950), which was modified by Kunze
et al. (1968), is usually used to calculate unsaturated permeability function from SWCC. Agus et al.
(2003) performed statistical assessment on combination of well-known best-fit SWCC and permeability prediction models. Best-fit SWCC models used
included Gardner (1958), van Genunchten (1980) and
Fredlund & Xing (2004) whereas prediction permeability models used included Childs & Collis-George
(1950), Burdine (1953) and Mualem (1976) models. Combination of Mualem (1976) with Fredlund &
Xing (1994) yielded the least error amongst all combinations (Agus et al. 2005).
Mualem (1976) statistical model uses a correction
factor for improving prediction of permeability function. This correction factor is supposed to account
for tortuosity. Tortuosity can have various meanings
when used by physicists, engineers or geologists to
describe fluid flow processes within a porous material. Geotechnical engineers usually define tortuosity
either geometrically or hydraulically in resolving geoenvironmental problems. In principle, both geometrical and hydraulic tortuosity compare the shortest path
length of fluid flow through available pores to its corresponding minimum length which is a straight line
distance inside a porous media (Clennell 1997). The
main difference is in the geometrical definition-where
paths taken by fluid flow are either straight lines or
close to tangents of soil grains whereas in the hydraulic
definition, fluid flows are smooth curves following
axes of the flow channels.
Some researchers from other disciplines have investigated and argued that tortuosity should not be treated
as a constant value (Wyllie & Rose 1950, Wyllie &
Gregory 1955, Knackstedt & Zhang 1994, Scholes
2007). They suggested that tortuosity can be related to
porosity of the medium. In Mualem (1976) permeability prediction model, tortuosity is treated as a constant
correction factor. Therefore, this paper attempts to
treat correction factor related to tortuosity in Mualem
(1976) statistical model as a non-constant term for
obtaining more accurate permeability function with
the use of more physically meaningful parameters
i.e. porosity and air-entry value of a soil.
2
2.1
DEVELOPMENT OF THEORY
Scope of works
In indirect determination of permeability function,

coefficient of permeability inferred from drying and
wetting SWCC will be different. In drying SWCC, the
soil starts at saturated condition and the matric suction
is gradually increased leading to a reduction in water

content of the soil, and vice versa for wetting SWCC.
This paper presents permeability functions calculated
from drying SWCC only since results from drying
SWCC measurement were more readily available in
the literature.
There are limited literatures which have a set of
measured SWCC and permeability data of the same
soil specimen. Generally, only experimental SWCC
and saturated permeability data are available. This
paper focuses only on coarse-grained soils for both
theory development and verification purpose due to
limited data available in the literature.
In this research, nine published data sets from literatures were used. Six data sets were used for theory
development, specifically two data sets from Agus
et al. (2005), two data sets from Soeiro (1964), one
data set from Mbonimpa et al. (2004) and one data set
from Leij et al (1996). Three data sets were used for
verification, specifically from Nemes et al. (2001).
2.2 Methodology
After obtaining measured SWCC and permeability
data from literature, units of SWCC and permeability data were made consistent for the analysis. In this
research, kPa and m/s were selected for the unit of
matric suction and coefficient of permeability, respectively. Equation 1 was used to find equivalent matric
suction from centimeter head of water to kPa.
where = matric suction, w = density of water,

g = gravitational constant and hw = head of water.
The published SWCC data were then best-fitted
using Fredlund & Xing (1994) equation from matric
suction of 102 to 106 kPa unless when the SWCC
fitting parameters of Fredlund & Xing (1994) were
given in the original publication. Fredlund & Xing
(1994) equation is shown in Equations 2 and 3.
where w = volumetric water content, C() =

correction function to ensure zero water content
at matric suction of 106 kPa, s = saturated volumetric water content, e = natural logarithmic number, a = parameter related to air-entry value of soil,
n = parameter related to inflection point of SWCC
and m = parameter related to slope of SWCC. In this
420
where v = average flow velocity, r = hydraulic

radius, = kinematic viscosity, C = shape constant
of the flow system, g = gravitational constant and
(d/dx) = hydraulic gradient.
research, C() was treated as unity following suggestion from Leong & Rahardjo (1997a) to simplify
Fredlund & Xing (1994) equation yet still produce
accurate SWCC fitting.
Since Mualem (1976) prediction model requires
matric suction to be expressed in terms of water content, Fredlund & Xing (1994) was then rearranged in
such a way as shown in Equation 4.
where (w ) = matric suction at particular volumetric

water content.
Mualem (1976) equation is illustrated in Equation
5 as follows:
where kw (w ) = predicted water coefficient permeability at particular volumetric water content,

= normalized volumetric water content = w /s ,
= correction factor related to tortuosity = 0.5,
w = particular volumetric water content, r = residual
volumetric water content. In this paper, s and r were
set to correspond to matric suction of 102 and
106 kPa, respectively in order for the analytical integration term to have finite boundaries.The integration was
performed using commercially available Mathcad 15
software (Parametric Technology Corporation 2010).
2.3 Proposed equation
In the proposed equation, three fundamental assumptions of the statistical models are applicable as follows:
1. The porous medium consists of a set of randomly
distributed interconnected pores characterized by a
pore radius r and its statistical distribution is given
by f (r). The areal pore distribution is the same for
all cross sections and is equal to f (r).
2. The soil-water characteristic curve is considered
analogous to the pore-size distribution function
using Kelvins capillary law
In this theory development, correction factor related

to tortuosity is defined using hydraulic tortuosity
concept as follows:
whereT = hydraulic tortuosity = Leh /L, Leh = effective

hydraulic path length for water flow and L = straight
path length.
Porosity of a soil indicates total amount of voids
within a soil. A larger amount of voids generally represents more alternative paths for water to flow through
water-filled pores inside a soil. The most effective
hydraulic path length is then used for water flow as
fluid prefers path with the least resistance. Tortuosity
values then become smaller as a consequence. Therefore, there is direct relationship between soil porosity
and tortuosity.
Air-entry value of a soil as generally implied by
parameter a in Fredlund & Xing (1994) can be
related to mean particle size of a soil. A higher a
commonly indicates a larger air-entry value of a soil
which corresponds to smaller mean particle sizes.
As water can only flow through water-filled pore,
smaller particle size reduces the resistance for water
flow inside a soil and effective path length for water
flow is then shortened. Tortuosity values then become
smaller as a consequence. Hence, there is plausible
relationship between air-entry value of a soil and
tortuosity.
From the relationships explained above, statistical
analysis was then carried out to express relationship between correction factor related to tortuosity,
porosity and air-entry value of soil in a mathematical form. Correction factor related to tortuosity is
treated as an optimized fitting parameter in the analysis and multi-parameter regression analysis was then
carried out. From the analysis, the following non-linear
relationship was obtained:
where e = natural logarithmic number, a = fitting

parameter of Fredlund & Xing (1994) best-fit SWCC
related to air-entry value of a soil and = porosity.
Equation 9 was then inserted into Equation 5 as
follows:
where r = pore radius, Ts = surface tension of water

and (ua uw ) = applied matric suction.
3. The Hagen-Poiseuille equation is valid to estimate
the permeability of a pore channel and the total
permeability is determined by integration over the
contributions from the filled pores.
Tables 1 and 2 display the results of analyzed data
for the proposed equation.
421
Table 1. SWCC fitting parameters of Fredlund & Xing

(1994) for data used in developing equation.
No
Data
a (kPa)
SSEnorm
1
2
3
4
5
6
NTU
LA
AS
FB
EO
C2221
440.797
51.673
4.500
4.000
5.500
5.615
0.633
0.508
33.86
7.757
33.579
3.862
2.048
0.886
0.445
0.924
0.555
0.973
0.111
0.315
0.227
0.027
Figure 1. Soil-water characteristic curve of C1460 data.
Table 2. Permeability parameters for data used in developing equation.
No
Data
modified
AREinitial
(%)
AREfinal
(%)
1
2
3
4
5
6
NTU
LA
AS
FB
EO
C2221
0.440
0.401
0.333
0.351
0.320
0.328
12.15
1.14
1.63
1.63
1.63
1.62
16.9
1.35
7.51
4.69
8.21
4.86
4.1
0.6
7.25
4.36
6.36
3.47
3
3.1
Figure 2. Permeability functions of C1460 data.
EVALUATION OF THE PROPOSED

EQUATION
Evaluation criteria
There are two evaluation criteria used in this research,

specifically Sum Square Error (SSEnorm ) and Average Relative Error (ARE). SSEnorm was applied in
best-fitted SWCC data from Fredlund & Xing (1994)
model unless it had already been given in the original
publication. SSEnorm is defined as follows:
where: yi = actual value of ith data, yp = predicted

value of ith data and n = total number of data. The
smaller the SSEnorm is, the better is the best-fit SWCC
from Fredlund & Xing (1994) model.
ARE was used in comparing the original Mualem
(1976) model and the proposed modified model to
the measured permeability data. ARE is defined as
follows:
Goh et al. (2010) suggested good agreement is

obtained when ARE is smaller than or equal to 20%,
and vice versa.
3.2
Results of evaluation
For evaluation purpose, Figures 1 to 6 show graphical

plot of SWCC and permeability functions of C1460,
422
The effect of air-entry value on correction factor

related to tortuosity can arguably be more significant than that of porosity since air-entry value of soil
can change at a considerably large range as compared
to porosity. For example, correction factor related to
tortuosity of NTU data differs significantly from that
of LA data representing a significant difference in the
air-entry values of soils.
The combined Fredlund & Xing (1994) best fit
SWCC and Mualem (1976) model yields ARE less
than 20% for all data. It shows that that combined
model can indeed give reasonably accurate prediction
for coefficient of permeability. The proposed equation
yields even smallerARE which shows improvement on
accuracy with the use of more physically meaningful
parameters i.e. porosity and air-entry value.

Table 3. SWCC fitting parameters of Fredlund & Xing
(1994) for data used in evaluation.
No
Data
a (kPa)
SSEnorm
1
2
3
C1460
C1461
C1463
2.516
2.611
3.298
3.114
14.644
33.47
1.036
0.539
0.557
0.153
0.000
0.148
Table 4. Permeability parameters for data used in

evaluation.
No
Data
modified
AREinitial
(%)
AREfinal
(%)
1
2
3
C1460
C1461
C1463
0.297
0.373
0.399
1.60
1.62
1.63
6.28
18.57
9.66
2.51
13.00
3.74
C1461 and C1463 data, respectively. The parameters

are also tabulated in Tables 3 and 4.
4
DISCUSSIONS
There are two kinds of coarse-grained soils analyzed

in this research, specifically clean and residual coarsegrained soils. All data are clean coarse-grained soils
having saturated coefficient of permeability in the
order of 105 m/s except the data from NTU and LA.
NTU and LA are residual soils in Singapore which
have 50% and 34% fines, respectively (Agus et al.
2005).
For clean coarse-grained soils, correction factor
related to tortuosity is observed in the range 1.60 to
1.64 with an average value of 1.63. Clean coarsegrained soils generally have small range of porosity
and air-entry value.Therefore, it is reasonable for clean
coarse-grained soils to have small range in correction
factor related to tortuosity, as observed.
For residual soils, the proposed equation of correction factor related to tortuosity yields negative value.
This phenomenon is believed due to the fact that additional grains from fine-grained particles create pore
configuration that differ significantly from the cylindrical configuration as assumed in the fundamental
assumption of statistical models.
CONCLUSIONS AND RECOMMENDATIONS
The proposed equation showed that there was nonlinear relationship between correction factor related
to tortuosity, porosity and air-entry value of a soil
with the air-entry value has the most significant influence. The modified Mualem (1976) equation was
verified against experimental data from literature and
the results indicated good agreement between the prediction and the experimental data. This study supports
a point of view where correction factor related to tortuosity should not be treated as constant term for all
soil types. For clean coarse grained soils, correction
factor related to tortuosity equal to 1.5 is suggested
for practical purpose.
Tortuosity remains an arguable concept until today
since there is no apparatus capable of measuring
tortuosity directly. The development of this device,
albeit difficult, is recommended. Tortuosity in finegrained soils is likely to differ from coarse-grained
soils. Therefore, this aspect can be explored in further study. Lastly, permeability anisotropy is closely
related to tortuosity, but the relationship has not been
understandable very well since there are limited available data on direct permeability measurements in
horizontal direction.
ACKNOWLEDGEMENT
The first author would like to acknowledge Interdisciplinary Graduate School research scholarship
provided by Nanyang Technological University and
Nanyang Environment & Water Research Institute,
Singapore. This work was supported by a research
grant from the competitive research project (CRP) on
Developing Low-Cost Landfill Capping Technologies, National Research Funding (NRF), Singapore.
REFERENCES
Agus, S.S., Leong, E.C., Schanz, T. 2003. Assessment
of statistical models for indirect determination of
423
permeability functions from soilwater characteristic

curves. Geotehnique, 53:279282.
Agus, S.S., Leong, E.C., Rahardjo, H. 2005. Estimating
permeability functions of Singapore residual soils. Engineering Geology, 78:119133.
Burdine, N.T. 1953. Relative permeability calculation
from poresize distribution data. Petroleum Transactions
American Institute of Mining. Metallurgical and Petroleum
Engineers, 198: 7178.
Chatterjea, K. 1989. Observations on the fluvial and slope
processes in Singapore and their impact on the urban environment. PhD Thesis, National University of Singapore.
Childs, E.C. & Collis-George, G.N.1950. The permeability
of porous materials. Proceedings of the Royal Society,
210A:392405.
Clennell, M.B. 1997. Tortuosity: a guide through the maze.
Special Publications on Geological Society, London, 122:
299-344.
Evangelista, A., Nicotera, M.V., Papa, R., Urciuoli, G.
2008. Field investigation on triggering mechanisms of
fast landslides in unsaturated pyroclastic soils. 1st European Conference on Unsaturated Soils, Durham, UK,
pp. 909915.
Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for
Unsaturated Soils. New York: John Wiley & Sons Inc.
Fredlund, D.G. & Xing, A. 1994. Equations for the soilwater
characteristic curve. Canadian Geotechnical Journal,
31:533546.
Gardner, W. R. 1958. Some steady-state solutions of the
moisture flow equations with application to evaporation from a water table. Soil Science Society of America
Journal, 85:228232.
Goh, S.G., Rahardjo, H. and Leong, E.C. 2010. Shear
strength equations for unsaturated soil under drying and
wetting. Journal of Geotechnical and Geoenvironmental
Engineering, 136(4):594606.
Knackstedt, M.A. & Zhang, X. 1994. Direct evaluation of
length scales and structural parameters associated with
flow in porous media. Phys. Rev. E., 50(3):21342138.
Krisdani, H., Rahardjo, H., Leong, E. C. 2006. Experimental
study of 1-D capillary barrier model using geosynthetic
material as the coarse-grained layer. Unsaturated Soils
2006 (Geotechnical Special Publications 147), ASCE,
Reston, VA, 16831694.
Kunze, R.J., Uehara, G., Graham, K. 1968. Factors important
in the calculation of hydraulic conductivity. Proceedings
Soil Science Society of America 32:760765.
Leij, F.J., Alves, W.J., and van Genuchten, M. Th. 1996.
The UNSODA unsaturated soil hydraulic database.
EPA/600/R-96/095, 103p.
Leong, E.C. & Rahardjo, H. 1997a. Review of soilwater
characteristic curve equations. Journal of Geotechical and
Geoenvironmental Engineering, 123(12):11061117.
Leong, E.C. & Rahardjo, H. 1997b. Permeability functions for unsaturated soils. Journal of Geotechical and
Geoenvironmental Engineering, 123(12):1118 1126.
Mbonimpa, M., Bedard, C., Aubertin, M., Bussiere, B. 2004.
A model to predict the unsaturated hydraulic conductivity
from basic soil properties. 57th Canadian Geotechnical
Conference.
Mendes, J. 2011. Assessment of the impact of climate change
on an instrumented embankment: an unsaturated soil
mechanics approach. PhD thesis, Durham University.
Mualem, Y. 1976. A new model for predicting the

hydraulic conductivity of unsaturated porous media.
Water Resources Research, 12:513522.
Mualem, Y. 1986. Hydraulic conductivity of unsaturated
soils: prediction and formulas. In A. Klute (ed.), Methods of Soil Analysis. Part 1: Physical and Mineralogical
Methods. 2nd Edition, Agronomy. American Society of
Agronomy, Inc., and Soil Science Society of America,
Madison, 799823.
National Environment Agency. 2010. Meteorological Services Data, Singapore.
Nemes,A., Schaap, M.G., Leij, F.J., and Wsten, J.H.M. 2001.
Description of the unsaturated soil hydraulic database
UNSODA version 2.0. Journal of Hydrology, 251:
151162.
Papa, R., Evangelista, A., Nicotera, M.V., Urciuoli, G. 2008.
Mechanical properties of unsaturated pyroclastic soils
affected by fast landslide phenomena. 1st European Conference on Unsaturated Soils, Durham, UK, pp. 917923.
Parametric Technology Corporation. 2010. Mathcad 15.
Needham, MA, USA.
Rahardjo, H., Li, X.W., Toll, D.G., Leong, E.C. 2001. The
effect of antecedent rainfall on slope stability. Journal of
Geotechnical and Geological Engineering. Special Issue
on Unsaturated and Collapsible Soils, 19:371399.
Rahardjo, H., Satyanaga, A., Leong, E.C. 2012a. Effects of
flux boundary conditions on pore-water pressure distribution in slope. Journal of Engineering Geology, Special
Issue on Unsaturated Soils: Theory and Applications
(available on line April 2012).
Rahardjo, H., Santoso, V.A., Leong, E.C., Ng, Y.S., Hua, C.J.
2012b. Performance of an Instrumented Slope covered by
a capillary barrier system. Journal of Geotechnical and
Geoenvironmental Engineering, 138(4), April 1.
Scholes, O.N., Clayton, S.A., Hoadley, A.F.A. 2007. Permeability anisotropy due to consolidation of compressible
porous media. Transport Porous Media 68:365387.
Soeiro, F. 1964. Contribution ltude du mouvement de
lhumidit dans les milieux poreux isothermes. Cahier de
la recherche, Eyrolles, Paris, 18.
Springman, S.M., Jommi, C., Teysserie, P. 2003. Instabilities on moraine slopes induced by loss of suction: a case
history. Geotechnique, 53(1):310.
Tami, D., Rahardjo, H., Leong, E-C., and Fredlund, D. G.
2004. Design and laboratory verification of a physical
model of sloping capillary barrier. Canadian Geotechnical Journal, 41(5):814830.
Toll, D.G., Rahardjo, H., Leong, E.C. 1999. Landslides in
Singapore. Proc. 2nd International Conference on Landslides, Slope Stability and the Safety of Infra-Structures,
Singapore, July 2728, pp. 269276.
Wyllie, M.R.J. & Gregory, A.R. 1955. Fluid flow through
unconsolidated porous aggregates. Indust. Eng. Chem.,
47(7):13791388.
Wyllie, M.R.J. & Rose, W.D. 1950. Application of the Kozeny
equation to consolidated porous media. Nature, 165:972.
van Genuchten, M.Th. 1980. A closed-form equation for
predicting the hydraulic conductivity of unsaturated
soils. Soil Science Society of America Journal, 44:
892898.
424

Capillary barrier system for landfill capping

H. Rahardjo, A. Satyanaga, F.R. Harnas, J.Y. Wang & E.C. Leong
School of Civil and Environmental Engineering, Nanyang Technological University, Singapore
ABSTRACT: Substantial economic and population growth generate a large amount of solid waste in the world.
It is common to dispose solid waste by dumping them in sanitary landfills. The most important requirement of
a landfill is that it does not pollute or degrade its environment. A method to reduce leachate from a landfill is to
reduce rainwater infiltration into a landfill. The use of a cover system for the landfill such as a capillary barrier
system (CBS) can reduce rainwater infiltration. A CBS is a man-made two-layer cover system designed as an
unsaturated system, harnessing the distinctly different hydraulic properties between a fine-grained layer (sand)
and a coarse-grained layer (gravel) of soils. Previous research works indicated that a single layer CBS could be
used as slope stabilization purposes against rainfall-induced slope failure in Singapore. In this study, a double
layer CBS is proposed to minimize the infiltration of rainwater into the landfill. The recycled materials were
used in the double layer CBS to reduce the cost associated with the construction of a cover system in the field
and to maintain environmental sustainability. One dimensional seepage analyses were performed to determine
the optimum thickness of fine-grained and coarse-grained layers within a double layer CBS and to determine
the suitable material combination within the double layer CBS. One dimensional infiltration tests were carried
out in the laboratory to verify the results of the numerical analyses.The results from the numerical analyses and
infiltration tests indicated that double layer CBS performed better than single layer CBS in minimizing rainwater
infiltration.
INTRODUCTION
Substantial amount of waste from household, industry, and highway construction are produced everyday
around the world including Singapore. The safe and
reliable long-term disposal of waste residues is an
important component of integrated waste management. Waste residues that are not incinerated and
recycled need to be stored in waste containment systems or landfills. Controlled placement of waste in
sanitary landfill greatly reduced the number of rodents
and insects, dramatically reduced public health risks,
prevent the contamination of waste to the surrounding groundwater and generally contributed to major
aesthetic improvements in waste disposal (Hatheway
and McAney, 1987; Tchobanologlous et al., 1993).
There are various philosophies to approach the design
and management of a landfill as reported by Rowe
et al. (1995), among which the role of cover system
should be noted. A cover system must be constructed
as impermeable as possible to minimize the generation
of leachate or contamination of waste into surrounding
groundwater.
Previous studies indicated that compacted clay liner
(CCL), geosynthetic clay liner (GCL) and evapotranspiration cover (ET) system are commonly used as
cover system. Each cover system has advantages and
disadvantages. Research works by Montgomery &
Parsons (1989), Suter et al. (1993), Albright et al.
(2004) and Benson (2007) showed that CCL system meet the permeability criteria for cover system.
However, the permeability of CCL material increases
with time due to desiccation cracking, wet-dry and
freeze-thaw effects, root penetration, and differential
settlement. Daniel and Koerner (1995), Daniel (1995)
and USEPA (1999) recommended the use of GCL system instead of CCL system since the permeability of
GCL material is much lower than that of CCL system
and the permeability of GCL material will not change
with time. However, they also observed the disadvantages of GCL system, such as: thin barrier layers within
GCL system are more vulnerable to construction damage or post construction puncture, material of GCL
is expensive and geosynthetic material within GCL
system has low hydrated shear strength. Nyhan et al.
(1990) and Anderson et al. (1999) studied ET system
that can also be used as an alternative cover system for
landfill since this system is cheap and the permeability
of material for this system will not change with time.
However, this system depends on the quality life of the
plant and is only applicable for area with dry climate.
Singapore is located in a tropical climate that is
characterized by uniform temperature, high humidity
and abundant rainfall throughout the year. The average daily temperature ranges between 24 C to 31 C
and the humidity ranges from 80% to 98%. A typical rainfall amount of 200 mm/month is observed in
Singapore with the highest monthly rainfall can be
425
observed during the rainy period in December and

January (Rahardjo et al., 2012a). A typical potential
evaporation of 5 to 7 mm/day is observed in Singapore (Rahardjo et al., 2012a). Rainfall and evaporation
affect the design of landfill cover. The rainfall intensity is much higher than the potential evaporation
rate in Singapore. Therefore, the current landfill cover
system which was developed in arid or semi arid
regions may not be applicable to Singapore.
In this study, a new cover system was developed
to minimize generation of leachate in Lorong Halus
sanitary landfill, the largest closed dumping ground
in Singapore. This landfill was used for dumping a
wide range of material including construction debris,
incineration ash and inert stabilized hazardous waste.
It is bounded by Buangkok East Drive on its western edge, Sungei Serangoon River running along its
northern and western perimeter and has an extensive
network of landfill gas collection pipes connected to
a flare system. Lorong Halus landfill was targeted for
potential redevelopment into commercial, industrial,
or even residential use in the future. Therefore, the
development of good cover system becomes necessary
to avoid contamination or pollution from leachate to
the surrounding environment in the future.
The new cover system utilizes capillary barrier
concept that has been proved to work effectively to
reduce the rainwater infiltration into soil layers and
to maintain the stability of slope (Rahardjo et al.
2012b, 2007). The new cover system utilizes recycled
materials (recycled concrete aggregate and recycled
asphalt pavement) to reduce the cost of the cover
system construction and to maintain environmental
sustainability.
2
SINGLE CAPILLARY BARRIER SYSTEM
A single capillary barrier (SCB) system is a two-layer

cover system designed as an unsaturated system by
using distinctly different hydraulic properties between
a fine-grained layer and a coarse-grained layer of
soil (Figure 1). The contrast in unsaturated hydraulic
properties, which are soil-water characteristic curves
and permeability functions, serves to minimize water
infiltration into the underlying soil. The infiltrated
water is then stored in the fine-grained layer and
ultimately removed by evaporation and transpiration,
lateral drainage through the slope and some time
percolation into the underlying layer.
The SCB materials must be selected properly with
careful consideration on the controlling parameters
associated with material properties in order to design
an effective SCB system. Previous research works by
Rahardjo et al. (2006) showed that there are three controlling parameters that must be considered in selecting the SCB materials, which are: the ratio between
the water-entry value of the fine-grained layers and the
coarse-grained layers (w -ratio), the water-entry value
of the coarse-grained layer and the saturated coefficient permeability of the fine-grained layer. It was
Figure 1. Soil-water characteristic curves and permeability

function of capillary barrier system.
also concluded by Rahardjo et al. (2006) that the minimum w -ratio should be 10 to create the barrier effect
between the fine-grained layer and the coarse-grained
layer and to minimize the infiltration of water into the
coarse-grained layer. The coarse-grained layer must
have a low water-entry value (preferably less than
1 kPa) in order to maintain the effectiveness of the SCB
system for a longer period of time. The saturated coefficient permeability of the fine-grained layer should
not be too low (preferably higher than 105 m/s) to
allow water to flow out from the fine-grained layer
by lateral diversion and as a result, the SCB system
remains effective. Previous study by Tami et al. (2007)
indicated that the fine-grained layer should have low
fines content so that the SWCC of the soil for the finegrained layer will be steep and the soil is able to drain
a large amount of water during a rainfall. In addition,
the use of soils with low fines content as the finegrained layer prevents the development of cracks in
the upper layer of SCB system, especially during dry
period when matric suctions are high.
Several laboratory tests have been carried out to
study the water flow through soil layers in SCB system. Yang et al. (2004) studied the performance of
SCB in minimizing rain water infiltration using 1-D
soil columns. It was concluded that a saturated coefficient permeability ratio of 2 to 3 orders of magnitude
between the fine-grained and the coarse-grained layers
was generally effective to produce a capillary barrier
effect. Tami et al. (2004) investigated the potential
use of SCB system as a slope stabilization technique
by limiting infiltration into a soil slope. An infiltration box of 2.45 m in length, 2 m in height and 0.4 m
in width was used to construct the laboratory capillary barrier model. Three combinations of materials
consisted of 0.2 m thick fine sand overlying 0.2 m
thick gravelly sand, 0.4 m thick fine sand overlying
0.2 m thick gravelly sand and 0.2 m thick silty sand
overlying 0.2 m thick gravelly sand were investigated.
426
The inclination angle of the SCB system model was

30 . The experimental results showed that SCB has
high potential application as a slope stabilization measure against rainfall-induced slope failures. Indrawan
et al. (2005) investigated the use of soil mixtures as
fine-grained layers in SCB system. The soil mixtures
used as the fine-grained layers were a soil mixture of
50% residual soil and 50% gravelly sand (RG-50), a
soil mixture of 75% residual soil and 25% medium
sand (RM-75) and a soil mixture of 95% residual soil
and 5% lime (RL-95). The experimental results indicated that the RG-50 and RM-75 were more effective
than the RL-95 in maintaining matric suctions during
wetting process. However, the RL-95 was found to be
more effective than the RG-50 and RM-75 in draining
water.
Krisdani et al. (2006) conducted a 1-D laboratory test to investigate the infiltration characteristics
through a capillary barrier system and the storage of
the fine-grained layer. The performance of capillary
barrier models constructed using different materials (i.e. geosynthetic material and gravelly sand) as
coarse-grained layer was also studied. The experimental results showed that capillary barrier effect existed
in both capillary barrier columns with geosynthetic
material and gravelly sand as the coarse-grained layer.
During the rainfall tests, the geosynthetic material was
found to be more effective than if gravelly sand was
used as the coarse-grained layer in a capillary barrier system. Krisdani et al. (2006) concluded that the
recovery of water storage of the SCB system is mainly
controlled by the lateral diversion flow and the evapotranspiration process. The evaporation process alone
is not sufficient to remove the infiltrated water in the
SCB system constructed in the tropical region.
Rahardjo et al. (2010) constructed an SCB system in the Bukit Timah Granite residual soil slope at
Ang Mo Kio and instrumented it using tensiometers
and piezometers. The slope experienced failure several times and had been repaired using SCB system.
The monitoring results showed that pore-water pressures under the slope with the capillary barrier system
were able to maintain negative pore-water pressures
(matric suction) under rainfall conditions, indicating
the effectiveness of SCB system in minimizing rainwater infiltration into soil layer. SCB system was
also constructed at a residual soil slope at Tampines
(Rahardjo et al., 2012b). The SCB system was constructed using fine sands as the fine-grained layer
and recycled concrete aggregates as the coarse-grained
layer. The monitoring results of tensiometers installed
below the SCB system area at Tampines showed that
the negative pore-water pressures during rainfall were
relatively constant at the slope with SCB system as
compared with those in the original slope. This indicated that the SCB system using recycled material was
also effective in minimizing rainwater infiltration into
soil layer.
Previous studies indicated that SCB system was
effective in minimizing rainwater infiltration into soil
layers of steep slopes. However, the SCB system needs
Figure 2. Cross section of SCB and DCB.
Figure 3. Typical pore-water pressure profiles for SCB and

DCB at initial and breakthrough conditions.
to be modified for its use on landfill with a gentle

slope (inclination angle of less than 10 ) in order to
expedite the removal of water from this cover system.
DUAL CAPILLARY BARRIER CONCEPT
In this study, it is proposed the use of dual capillary barrier (DCB) system. Dual capillary barrier
is two stacks of single capillary barrier (SCB). The
advantage of dual capillary barrier is an increase in
water storage as compared to that of SCB with similar total thickness. The schematic diagram of DCB
and SCB is shown in (Figure 2). The increase in water
storage can be illustrated using a typical example as
illustrated in Figure 3 and Figure 4. Figure 3 shows
the typical pore-water pressure profile at initial and
breakthrough conditions for SCB and DCB. The initial condition for both SCB and DCB is a hydrostatic
condition with an assumed water table at the bottom
of the cover (z = 0 m). The idealized pore-water pressure profile at breakthrough condition for a capillary
barrier has been reported by Stormont and Morris
(1998). For SCB, at breakthrough condition, the porewater pressure at the bottom of fine grained layer is
similar to the water-entry value of the coarse-grained
layer. The pore-water pressure decreases linearly (45 )
427
Figure 5. Grain size distributions of FRAP and CRAP.

Figure 4. Typical volumetric water content profiles for
SCB and DCB at initial and breakthrough conditions.
Table 1.
with increasing height of the fine-grained layer. For

DCB, similar principle applies to the pore-water pressure at the bottom and throughout the height of the
fine-grained layers at breakthrough condition. The
pore-water pressure at the upper coarse-grained layer
is assumed to be constant as represented by a vertical
line with a value similar to the water-entry value of the
coarse-grained layer.
In order to obtain the highest storage capacity, the
PWP at breakthrough should correspond to the highest
volumetric water content as possible. Figure 3 shows
that at breakthrough condition the range of PWP that
affects the water storage of the SCB is between the
water-entry value of the coarse-grained layer (assumed
to be 0.5 kPa) and the limiting PWP that corresponds
to the thickness of the fine-grained layer (50 cm or
5 kPa). Similarly, the range of PWP that affects the
water storage of DCB is from 0.5 kPa to 3 kPa. These
PWP ranges can be defined as the important PWP
ranges that affect the water storage of SCB and DCB,
respectively.
From the pore-water pressure profiles and the
water-characteristic curve (WCC) of the fine- and
coarse-grained materials, the volumetric water content
profiles of SCB and DCB can be drawn. The storage
capacity of SCB and DCB is calculated by measuring
the area bounded by the volumetric water content profile at the initial condition and the volumetric water
content profile at breakthrough condition. The typical
storage capacity of DCB can be seen to be higher than
the storage capacity of SCB as shown in Figure 4.
The mass of water stored in the fine-grained layer
of SCB or DCB can be calculated by integrating the
WCC of the fine-grained layer material with respect to
suction. The bottom integration limit is the water-entry
value of coarse-grained layer and the top integration
limit is the respective limiting suction.The integrations
show that the storage for a SCB is 0.035 m3 /m2 while
for a DCB is 0.038 m3 /m2 .An increase in water storage
of around 9% is achieved. The percentage of increase
in water storage depends on the material used. However, in general the water storage of DCB is higher as
compared to that of SCB. The increase in water storage
Basic properties of FRAP and CRAP.
Properties
FRAP
CRAP
Group symbol (USCS)

Specific gravity, Gs
Max. dry density, d max (Mg/m3 )
Min. dry density, d min (Mg/m3 )
Saturated permeability, ks (m/s)
SW
2.43
1.92
1.3
5 103
GP
2.41
1.84
1.46
5 101
of DCB can be attributed to the fact that the smaller

thickness of each fine-grained layer of DCB provides
a range of suction where volumetric water content at
breakthrough of the material is higher compared to the
volumetric water content at breakthrough of SCB.
LABORATORY COLUMN TEST
The performance of a dual capillary barrier was

assessed using one-dimensional column experiments.
The setup of the column is explained in detail in Yang
et al. (2004). In this study, two gradations of recycled asphalt pavement (RAP) were used, fine recycled
asphalt pavement (FRAP) and coarse recycled asphalt
pavement (CRAP). The use of RAP as capillary barrier
has never been investigated and it is a novel initiative
to reduce construction and demolition waste thrown
into a landfill. The grain size distributions and basic
properties of FRAP and CRAP are shown Figure 5 and
Table 1, respectively.
Water characteristic curve (WCC) is a relationship between water content and suction of a material
(Fredlund and Rahardjo, 1993). WCC is an important
relationship in unsaturated soil mechanics which governs the water storage of a material in unsaturated
condition. The WCC could be obtained by carrying
out an experiment in its wetting and drying paths.
In this study, the WCC data is obtained in its drying
path by using Tempe cell (Soil Moisture Corp). Subsequently, the WCC data are fitted using Fredlund and
Xing (1994) equation. The WCC data and the fitted
curve are shown in Figure 6. Best fitting parameters
428
Figure 6. Drying WCC of FRAP and CRAP fitted with

Fredlund and Xing (1994) equation
Table 2.
Fredlund and Xing (1994) WCC fitting parameters.
No
Description
Symbol
FRAP
CRAP
Saturated volumetric
water content
Residual volumetric
water content
Air entry value
Fitting parameter
0.34
0.38
0.106
0.101
a (kPa)
a (kPa)
n
m
r (kPa)
0.79
2.62
1.01
0.95
55
0.02
0.04
3.16
0.82
0.19
2
3
4
5
6
7
Figure 7. Permeability function of FRAP and CRAP.
following the Fredlund and Xing (1994) best-fitting

equation (Equation 1) are presented in Table 2.
where: e = base of natural logarithm; s = saturated

volumetric water content; = soil suction; a, n, m,
r = fitting parameters
Water flow in unsaturated soil follows a similar law as in saturated soil, Darcys law, except that
the permeability for an unsaturated soil is not constant. It depends on the soil suction. The relationship
between the coefficient of permeability and soil suction is called permeability function. In this study,
the permeability function is predicted from the WCC
of the materials using the statistical method originally proposed by Childs and Collis-George (1950) as
explained in Fredlund and Rahardjo (1993). The permeability functions for the materials used in the study
are shown in Figure 7.
Two column experiments are presented in this study.
The first column (SCB) comprised of 50 cm thick
FRAP and 50 cm thick CRAP. The second column
comprised two stacks of 25 cm thick FRAP over 25 cm
Figure 8. Schematic diagram of SCB and DCB columns.
thick CRAP (DCB). The schematic diagram of the two

columns and the location of the instrumentations are
shown in Figure 8. There were three tests conducted in
this column study. Firstly, an upward infiltration test
was conducted in order to saturate the column and to
remove the trapped air in the column. Subsequently, a
drawdown test was conducted to create a hydrostatic
condition by altering the position of water table from
the top of column (z = 1 m) to the bottom of column
(z = 0 m). The result of the drawdown test then became
an initial condition for the third test, rainfall test.
There were three rainfall rates applied to the column.
The results of one rainfall test of 50 mm/hr for 1 hour
are presented in this paper.
The pore-water pressure (PWP) head profiles for
SCB and DCB at various times are shown in Figure 9. The ideal initial condition for the experiment
was a hydrostatic condition. The hydrostatic condition means that the PWP head at the bottom of the
column is 0 m and the PWP head at the top of the column is 1.0 m. The ideal hydrostatic condition in both
SCB and DCB could not be reached. The PWP head
in the coarse-grained layer of the SCB only decreased
to 0.25 m (at the top of the coarse-grained layer) and
then in the fine-grained layer of SCB it decreased to
0.8 m (at the top of the fine-grained layer). This was
due to the fact that when the PWP head reached a
429
Figure 9. Pore-water pressure head profile during test for

SCB and DCB columns.
pressure around its air-entry value, the permeability

of the soil became very low and it would take a very
long time to reach an equilibrium condition. Similarly
for DCB column, the PWP head in the coarse grained
layer only decreased to around 0.2 m (at the top of
the coarse-grained layer) and the PWP head in the
fine-grained layer decreased linearly to around 0.5 m
(at the top of fine-grained layer). A closer inspection
showed that the PWP head at the interface of fine-and
coarse-grained materials for both SCB and DCB was
almost similar of around 0.2 m. The PWP head at
the interface may be limited by the residual PWP head
of the coarse-grained layer of around 0.2 m. In other
words, a thickness of more than 0.2 m for the coarsegrained layer may not increase the PWP head at the
interface.
When rainfall started, the PWP head increased to a
less negative value. In SCB, the PWP head increased
from 0.8 m to around 0.35 m on the top of the
fine grained layer during rainfall (up to 1hr). Subsequently, the PWP returned to a more negative value
(from 0.35 m to 0.6 m) 6 hrs after rainfall. During
the drying process, the PWP head in the coarse-grained
layer of SCB was still increasing. The PWP head profile of DCB moved to a less negative value during
rainfall especially in the upper part of the fine- and
coarse-grained layers and 6 hours after rainfall the
PWP head profile of DCB almost returned to its initial
condition. Because of the positions of instrumentation in the DCB, the pore-water pressure profiles of
DCB may not represent the exact pore-water pressure
profile. Hence, the comparison of pore-water pressure
profiles between SCB and DCB may not show clearly
the increase in water storage of DCB.The water storage
can be more accurately represented by the volumetric
water content profiles and the water balance graph.
Volumetric water content profiles for SCB and DCB
at various times are shown in Figure 10. Water storage
capacities of SCB and DCB are highlighted in the figure. It can be seen than the water storage of DCB is
slightly higher as compared to the water storage of
SCB. The area of that is bounded by volumetric water
content profiles at initial and breakthrough conditions
Figure 10. Volumetric water content profiles during test for

SCB and DCB columns.
Figure 11. Water balance during test for SCB and DCB
columns.
can also be calculated. The calculated water storage

capacity for SCB is 0.0175 m3 while the water storage
capacity for DCB is 0.02 m3 . An increase in water storage of 12% is observed. The explanation for the higher
water storage is presented in Section 3. The increase in
water storage is also shown by the water balance curves
for SCB and DCB in Figure 11. The figure shows that
drainage in SCB occurred earlier as compared to DCB.
This indicates that breakthrough occurred earlier in
SCB column as compared to DCB column. In addition,
the maximum water storage of DCB was higher than
the maximum water storage of SCB before the occurrence of breakthrough. This means that DCB can store
more water than SCB before breakthrough occurred.
5
NUMERICAL SIMULATION OF THE

INFILTRATION TEST
The infiltration processes from the laboratory infiltration column were simulated numerically using SVFlux
(Soilvision Systems Ltd). The model geometry was
one-dimensional with the same geometry as the actual
laboratory condition. Similar boundary conditions as
those applied in the actual laboratory column tests
430
Figure 12. Experimental and numerical pore-water pressure

profiles of SCB during wetting process of rainfall test.

profiles of SCB during drying process after rainfall test.
were applied. During the drawdown test, there is no

boundary condition assigned at the top boundary since
the water was expected to drain from the bottom of the
column while pressure head equal to zero was assigned
at the bottom. During the rainfall test, the top boundary
condition was set as a unit flux equals to the specified
rainfall, and the bottom boundary condition was set
as pressure head equal to zero. The drying WCC and
permeability functions of the materials were used in
the simulation. The numerical simulation results of the
infiltration tests were verified by comparing them with
the laboratory experimental data.
The pore-water pressure head profiles computed
from the numerical modeling and the laboratory experimental data for SCB are shown in Figure 12 and
Figure 13 for wetting and drying processes, respectively. The numerical modeling results show a good
Figure 14. Experimental and numerical volumetric water

content profiles of SCB during wetting process of rainfall
test.

content profiles of SCB during drying process of rainfall
test.
agreement with those observed in the laboratory experiments. During wetting process, the PWP head profile
moved towards less negative value and during drying
process the PWP head profile moved towards to a more
negative value. Comparisons of the volumetric water
content profiles obtained from the numerical simulations and the laboratory experiments during the rainfall
test are shown in Figure 14 and Figure 15 for wetting
and drying processes, respectively. It can be seen that
the trends of the profile are consistent with the laboratory experiments. However, a closer look shows a little
discrepancy especially in volumetric water content of
the coarse-grained soil that could be caused by poor
contact of the moisture sensor with the soils as also
reported by Indrawan et al. (2007).
431

profiles of DCB during rainfall test.

content profiles of DCB during rainfall test.
The pore-water pressure head profiles from the

numerical modeling and the laboratory experiment for
DCB are shown in Figure 16. The numerical modeling
results do not match the experimental data perfectly;
however, the trend of movement was similar. The volumetric water content profiles of the DCB calculated
from the numerical modeling and laboratory experiment are shown in Figure 17. Similar to the pore-water
pressure head, the trends of movement was similar but
small discrepancy occurs in the value of the volumetric water content especially near the surface of the
column.
DISCUSSION
The PWP and volumetric water content profiles from

laboratory column test showed agreement with the
results of numerical model. There is a slight discrepancy on values of volumetric water content which may
be caused by a non-uniform compaction process. The
non-uniform compaction process may cause differences in the hydraulic properties of the materials in
the column with the hydraulic properties used in the
numerical model.
Dual capillary barrier system is a novel method to
increase the water storage of a single capillary barrier
system. Current methods to increase the water storage
include increasing the thickness of the fine-grained
layer or substituting the fine-grained layer with soil
having a higher water storage. The advantage of DCB
as compared to other methods is that the total thickness
of the fine-grained layer of capillary barrier needs not
to be changed. As a result, DCB does not increase the
amount of materials required.
The thickness of layers in DCB can be optimized in
order to increase the performance of DCB. The thickness of the coarse-grained soil can be limited to the
water-entry value head of the coarse-grained soil. The
thicknesses of the upper fine-grained layer also can be
optimized to take into account the effect of evaporation. As reported by McCartney and Zornberg (2010),
evaporation is only effective to a certain depth of soil.
By having two fine-grained layers, the upper finegrained layer can be optimized to take advantage of
evaporation.
The effectiveness of a capillary barrier system does
not only depend on the water storage capacity, but
also on the lateral diversion capacity and evaporation capacity. Therefore, the investigation on lateral
diversion and evaporation capacity of DCB should be
conducted in future studies.
The use of numerical modeling as a tool to analyze
cover system problem has been practiced by engineers and researchers (OKane, 2002). Therefore, the
positive correlation between the numerical modeling
results and the laboratory experimental data for the
SCB column supported the use of numerical modeling
as one of the tools to analyze cover system.
Recycled aggregates have been used in civil engineering applications (Loehr et al., 2000; Consoli et al.,
2002). The use of RAP as material for capillary barrier
construction has not been tested before. Previously,
recycled concrete aggregate (RCA) has been successfully used as the coarse-grained layer material for a
capillary barrier that was used as slope stabilization
in Singapore (Rahardjo et al, 2012b). The increasing
trend in the use of recycled aggregates in civil engineering application is beneficial for environmental
sustainability.
CONCLUSIONS
The new cover system, dual capillary barrier system,

has been developed to minimize rainwater infiltration
into landfill in Singapore. The numerical analyses and
infiltration tests showed that fine and coarse recycled
asphalt pavement could be used as fine and coarsegrained materials, respectively, to create a capillary
barrier effect within a dual capillary barrier system.
The results of the numerical analyses and infiltration
432
tests using 1-D column also showed that dual capillary

barrier system was more effective than single capillary
barrier system in minimizing rainwater percolation
into the soil beneath the capillary barrier system. An
increase in water storage of 12% was observed from
the laboratory column test.
ACKNOWLEDGEMENT
This work was supported by a research grant from the
competitive research project (CRP) on Developing
Low-Cost Landfill Capping Technologies, National
Research Funding (NRF), Singapore.
REFERENCES
Albright, W.H., Benson, C.H. and Gee, G.W. 2004. Field
water balance of landfill covers. Journal of Environmental
Quality, 33:23172332.
Anderson, J.E., Nowak, R.S., Ratzlaff, T.D., and Markham,
O.D. 1993. Managing soil moisture on waste burial sites in
arid regions. Journal of Environmental Quality. 22:6269.
Benson, C.H., Sawangsuriya, A., Trzebiatowski, B. and
Albright, W.H. 2007. Post-construction changes in the
hydraulic properties of water balance cover soils. Journal of Geotechnical and Geoenvironmental Engineering.
133(4):349359.
Childs, E.C., and Collis-George, G.N. 1950.The permeability
of porous materials. Proc. of the Royal Society of London.
Vol. 210A, pp. 392405.
Consoli, N.C., Montardo, J.P., Prietto, P.D.M., and Pasa, G.S.
2002. Engineering behavior of a sand reinforced with plastic waste. Journal of Geotechnical and Geoenvironmental
Engineering 128(6):462472.
Daniel, D.E. 1995. Soil barrier layer versus geosynthetic
barrier in landfill cover systems. Landfill ClosuresEnvironmental Protection and Land Recovery, ASCE,
Geotechnical Special Publication, No. 53, R.Jeffrey Dunn
and Uday P. Singh, Eds., New York, 118.
Daniel, D. E., and Koerner, R. M. 1995. Waste containment
facilities: guidance for construction, quality assurance and
quality control of liner and cover system, ASCE.
Fredlund, D.G. and Rahardjo, H. 1993. Soil Mechanics for
Unsaturated Soils. John Wiley & Sons Inc., New York
Fredlund, D.G. and Xing, A. 1994. Equations for the
soil-water characteristic curve. Canadian Geotechnical
Journal. 31: 521532.
Hatheway, A. W., and McAney, C. C. 1987. An in-depth look
at landfill covers, Waste Age, 1(8).
Indrawan, I.G.B., H. Rahardjo and E.C. Leong. 2007. Drying
and Wetting Characteristics of a Two-Layer Soil Column.
Canadian Geotechnical Journal 44: 2032.
Indrawan, I.G.B., Rahardjo, H. and Leong, E.C. 2006. Effects
of Coarse-Grained Materials on Properties of Residuals
Soil. Journal of Engineering Geology, 82(3):154164.
Loehr, J.E., Bowders, J.J. and Salim, H. 2000. Slope Stabilization Using Recycled Plastic Pins-Constructability. Report
RDT 00-007, RI 98-007, Missouri Dept of Transportation,
Jefferson City, Missouri.
McCartney, J.S and Zornberg, J.G. 2010. Effects of infiltration and evaporation on geosynthetic capillary barrier performance. Canadian Geotechnical Journal 47:
12011213.
Montgomery, J. H., and Parsons, L. J. 1989. The Omega

Hills final cover test plot study: three-year data summary,
Annual Meeting of the National Solid Waste Management
Association, Washington, DC.
Nyhan, J.W., Hakonson,T.E., and Drennon, B.J. 1990.A water
balance study of two landfill cover designs for semiarid
regions. Journal of Environmental Quality. 19:281288.
OKane, M.A., Ayres, B., Christensen, D., Meiers, G., 2002.
CANMET-CETEM Manual on cover system design for
reactive mine waste, Report 689-01, CANMET-CETEM.
Krisdani, H., Rahardjo, H. and Leong, E.C. 2006. Experimental Study of 1-D Capillary Barrier Model using
Geosynthetic Material as the Coarse-Grained Layer.
Proc. 4th International Conference on Unsaturated Soils.
Geo Institute. Phoenix, Arizona, U.S.A., 25 April,
pp. 16831694.
Rahardjo, H., Tami, D. and Leong, E.C. 2006. Effectiveness
of Sloping Capillary Barriers Under High Precipitation
Rates. Proc. 2nd International Conference on Problematic
Soils, Petaling Jaya, Selangor, Malaysia, 35 December,
pp. 3954.
Rahardjo, H., Krisdani, H. and Leong, E.C. 2007. Application of Unsaturated Soil Mechanics in Capillary Barrier
System. Proc. 3rd Asian Conference on Unsaturated Soils.
Nanjing, China, 2123 April, pp. 127137.
Rahardjo, H., Hua, C.J., Leong, E.C. and Santoso, V.A. 2010.
Performance of an Instrumented Slope under a Capillary
Barrier System. Proc. 5th International Conference on
Unsaturated Soils, Barcelona, Spain, 68 September, 2,
pp. 12791284.
Rahardjo, H., Satyanaga, A. and Leong, E.C. 2012a. Effects
of Flux Boundary Conditions on Pore-water Pressure
Distribution in Slope. Journal of Engineering Geology,
Special Issue on Unsaturated Soils: Theory and Applications (available on line April 2012).
Rahardjo, H., Satyanaga, A. and Leong, E.C. 2012b. Unsaturated Soil Mechanics for Slope Stabilization. Southeast
Asian Geotechnical Journal, 43(1): 4858.
Rahardjo, H., V.A. Santoso, E.C. Leong, Y.S. Ng and C.P.H.
Tam. 2012c. Use of Recycled Concrete Aggregates in
a Capillary Barrier for Slope Stability. Proceedings of
11th International and 2nd North American Symposium
on Landslides ISL NASL 2012, Banff, Canada.
Rowe, R. K., Quigley, R. M., and Booker, J. R. 1995. Clayey
barrier systems for waste disposal facilities, E&FN Spon,
London.
Stormont, J.C and Morris, C.E. 1998. Method to estimate
water storage capacity of capillary barriers. Journal of
Geotechnical and Geonvironmental Engineering 124(4):
297302.
Suter, G. W., Luxmoore, R. J. And Smith, E. D. 1993. Compacted soil barriers at abandoned landfill sites are likely
to fall in the long term. Journal of Environmental Quality,
22(2), 217226.
Tami, D., Rahardjo, H. and Leong, E.C. 2007. Characteristics
of Scanning Curves of Two Soils. Soils and Foundations,
Japanese Geotechnical Society, 47(1): 97109.
Tchobanoglous, G., Theisen, H., and Vigil, S. 1993. Integrated solid waste management: engineering principles
and management issues. Irwin/McGraw-Hill series in
Water Resources and Environmental Engineering.
Yang, H., Rahardjo, H., Wibawa, B., and Leong, E.-C. (2004)
A soil column apparatus for laboratory infiltration study.
Geotechnical Testing Journal. 27(4): 347355.
USEPA 1999. Evaluation of subsurface engineered barrier at
waste site. EPA 542-R-98-005, Office of Solid Waste &
Emergency Response (5102G).
433

Geosynthetic clay liner gas permeability relationship with moisture

content and suction under pre-conditioning stresses
M.A. Rouf, R.M. Singh & A. Bouazza
R.K. Rowe
Department of Civil Engineering, Queens University, Kingston, Canada
ABSTRACT: Series of gas permeability tests, with monitoring of gravimetric moisture content and total
suction, were conducted on a commercially available needle punched geosynthetic clay liner (GCL). The GCL
samples were partially hydrated with de-ionized water under 2 kPa and 20 kPa confinement respectively, prior
to testing. Gas permeability measurements were conducted at differential pressures ranging from 1 to 10 kPa
and at different gravimetric moisture contents and suctions. Gas permeability was found to decrease with the
increase of moisture content and decrease of suction and vice versa. It was also found that the effect of preconditioning stresses was more pronounced at a gravimetric moisture content greater than 50% and at suction
less than 1.6 MPa.
INTRODUCTION
Geosynthetic clay liners (GCLs) are most typically

comprised of a thin layer of bentonite contained
between two layers of geotextile with the components
being held together by needle-punching or stitch bonding. They are widely used in final cover systems of
modern landfills to minimise migration of gases and
infiltration of meteoric water. In this respect, there is a
wide body of work available on their hydraulic performance. However, experimental measurements of their
permeability to gases are less widely available, and
only recently have information on their gas advective flow performance became available in the context
of landfill capping (Didier et al. 2000, Bouazza &
Vangpaisal 2003, Vangpaisal & Bouazza 2004,
Mendes et al. 2010, Pitanga et al. 2011).
For the GCL to efficiently minimize gas flow, the
bentonite component of GCL must be sufficiently
hydrated. Earlier work has shown that the gas permeability of GCLs varied largely with changes in
gravimetric moisture content as well as the form of
bentonite (powdered or granular) used and indicated
that very large attenuation of gas migration could
occur if the GCL was sufficiently hydrated (Didier
et al. 2000, Bouazza & Vangpaisal 2003, Vangpaisal &
Bouazza 2004, Pitanga et al. 2011). More importantly,
it was also reported that when a GCL lost the absorbed
water, its gas permeability significantly increased due
to the shrinkage of the bentonite component and, in
extreme cases, due to the formation of desiccation
cracks which provided preferential gas flow paths
(Bouazza et al. 2006, Mendes et al. 2010).
The initial state of the bentonite within the GCL, as

installed, is at an as-manufactured gravimetric moisture content (10%) that is insufficient for the GCL
to minimize gas flow. As a result, it is expected that
passive hydration ought to begin at the time of installation and should be complete prior to significant contact
with gas. However, this potentially introduces a high
degree of uncertainty since there is no guarantee that
the GCL will reach full hydration and will more likely
to be in an unsaturated condition prior to the occurrence of gas migration. This latter aspect highlights
also the need to quantify the water retention of GCLs
and its effect on gas permeability.
This paper presents the results from a series of
gas permeability tests conducted on a needle-punched
GCL which was partially hydrated to cover a range
of gravimetric moisture content and suction, under
two different curing conditions. The differences in
gas flow parameters are identified in relation to
the GCL gravimetric moisture content, suction, preconditioning stress and applied differential pressure
during measurements.
MECHANISM OF GAS TRANSPORT
The movement of gas in porous media such as soil

or GCLs occurs by two major transport mechanisms:
advective flow and diffusive flow. In advective flow,
the gas moves in response to a gradient in total pressure. To equalize pressure, a mass of gas travels from a
region of higher pressure to a lower one. In the context
435
of landfills, the primary driving force for gas migration, especially through cover systems, is a pressure
differential due to natural fluctuations in atmospheric
pressure and/or change in the leachate/water table or
temperature. Gas movement by diffusion occurs due
to molecular interactions. When a gas is more concentrated in one region of a mixture than another, it is
likely that this gas diffuses into the less concentrated
region. Thus the molecules move in response to a partial pressure gradient or concentration gradient of the
gas. The present paper will focus only on gas transport
due to advective flow caused by pressure gradient.
Earlier work by Alzaydi et al. (1978), Bouazza &
Vangpaisal (2003), Vangpaisal & Bouazza (2004)
showed that Darcys law could provide a fair approximation of gas flow in a low permeability material. Furthermore, Massmann (1989) indicated that a
groundwater flow model provided a good approximation for gas transport up to a differential pressure of
50 kPa.
Based on Darcys law, the one-dimensional volumetric flow rate Q (m3 s1 ) of gas in porous media is
given as:
3.1 Geosynthetic clay liner

A commercial GCL regularly marketed in Australia
(Elcoseal X2000), was used in the present investigation. It is composed of powdered sodium bentonite
(commercially known as Trugel) needle-punched and
thermally locked between a nonwoven geotextile cover
layer and a nonwoven geotextile reinforced by a slit
film woven geotextile carrier. The mass of bentonite
(Mbent ) was calculated from the difference between the
mass per unit area of the GCL and the mass per unit
area of the geotextiles (Mbent = MGCL Mgeo ). MGCL
was measured following the procedure outlined in
ASTM D5993; whereas the value of the mass per unit
area of the geotextiles (Mgeo ) was based on the information supplied by the manufacturers. The mass per
unit area of GCL and dry bentonite was found to vary
from 4.5 to 5.8 kg m2 and 3.3 to 4.7 kg m2 , respectively. The thickness of the GCL at a dry state or as
received conditions was found to vary between 7.71 to
8.91 mm.
3.2 Sample preparation
noindent where k = intrinsic permeability of the

porous material (m2 ); A = cross section of the porous
material (m2 ); dP/dx = pressure gradient; and =
dynamic viscosity of gas (N s m2 ). Intrinsic permeability is considered a property of porous material
which is independent of permeating liquid or gas.
The gas permeability, K (m s1 ), can be calculated
from the equation as shown below:
where = density of the gas (kg m3 ) and g =

gravitational acceleration (m s2 ).
Assuming that landfill gas behaves as an ideal gas
and that the continuity equation for a gas applies, then
equation (1) becomes:
For a sample of length L (m), the solution of the

above equation is subjected to boundary conditions,
P = P1 at x = 0 and P = P2 at x = L, where P2 is the
atmospheric pressure. The mass flow rate entering the
atmosphere (Q2 ), in terms of boundary pressure can
be obtained by integrating equation (3) and is given
as follows:
GCL samples were prepared to cover various gravimetric moisture contents and suction values. Firstly,
125 mm diameter GCL samples were cut using a sharp
knife and a plastic disc as cutting base. One sample
from each sheet was used to measure initial thickness,
mass, and gravimetric moisture content, respectively.
Then the samples were placed on a saturated porous
sponge material to hydrate for different specified time
periods (bottom-up hydration process). Once the target time was reached the samples were removed from
the hydration set up and stored in a double re-sealable
plastic bag for curing.The hydration of the sample took
place gradually until all available water was absorbed
by the bentonite as dry bentonite has a very high negative water potential. For this reason, the curing process
of GCL after hydration was essential. The specimens
were cured following two different methods. In the
first method, the GCL was kept under a normal stress
of 20 kPa by direct loading, to simulate the weight
of 1 m thick soil cover in a landfill cover system. In
the second method, the GCL was placed under 2 kPa
confinement to provide a minimum stress to ensure
uniform distribution of moisture content in the sample.
A curing period of 710 days was deemed sufficient
to homogenize the gravimetric moisture content of the
bentonite component (Vangpaisal & Bouazza 2004).
However, a curing period of at least10 days was used to
ensure a uniform distribution of moisture in the GCL
specimen prior to the gas permeability test.
3.3 Gas permeability cell, test procedures and
water retention
The GCL gas permeability cell was designed and
developed by Bouazza & Vangpaisal (2003). The cell
consisted of two distinct parts: a base cylinder, and an
436
Figure 1. Cross section of gas permeability cell (modified

from Bouazza and Vangpaisal 2003).
upper cylinder with piston. The two parts were held

together with retaining threaded rods (Fig. 1).
The purpose of the piston situated in the upper cylinder was to transmit the applied confining stress to
the GCL sample. The connections of the upper and
the base cylinders, and the piston were sealed using
O-rings. The base cylinder had two different inside
diameters. The upper part had a diameter of 130 mm
and the lower part had a diameter of 100 mm, creating a shoulder on its wall. This shoulder was used to
accommodate the GCL sample and the upper cylinder.
The effective gas flow area of the gas permeability cell was 7.85 103 m2 .Since nitrogen gas ( =
1.76 105 N s m2 , = 1.165 kg m3 at 20 C and
atmospheric pressure) is relatively inert and has very
low water solubility, pressurized nitrogen gas was used
in this study as permeating gas.
To measure gas permeability, nitrogen gas was supplied to the top of the cell, permeated through the GCL
sample and flowed out from the base of the cell where
five (one at a time) gas flow meters (GFM17 Mass
Flow Meters, Aalborg, Denmark; accuracy: 1.5% of
full range at 20 C and atmospheric pressure), ranging
from 010 mL/min up to 015 L/min, were connected
to cover the different gas flow rates. The outflow port
was connected to atmospheric pressure. The differential gas pressure was estimated from the difference
between the pressure supply and atmospheric pressure.
The setting of the GCL sample in the cell followed
several steps. A short description of the process is
presented in the following: firstly, a filter paper was
placed on the bottom part of the lower chamber to
prevent migration of small particles of gravel into the
drainage port. Next, gravel was added to reach a height
equal to that of the inner lower chamber and carefully
leveled to provide a smooth surface. Then, a hydrated
GCL sample of 125 mm diameter was seated on top of
the gravel layer and on the shoulder of the base cylinder. The gap between the edge of the GCL and the
cylinder wall was filled with a bentonite paste similar
to the bentonite used in the GCL. Then the upper chamber was gently fitted and secured to the base chamber
by tightening the retaining rods. The same bentonite

paste was also smeared on the perimeter of the geotextiles, and on the upper and lower grooves to complete
the lateral sealing at the interface between the two
chambers of the cell. Again, gravel was added on top of
the GCL sample until the required height (20 mm) was
reached. Then the piston was placed above the gravel
layer.After completing the assembling process, the cell
was placed on a stand equipped with a load hanger system and a normal stress of either 2 kPa or 20 kPa was
applied on the sample. Finally, pressurized nitrogen
gas was supplied to the system from the top port of
the piston; the applied gas passed through the GCL
sample and vented out through the flow meter to the
atmosphere. Gas flow rate, differential pressure, room
temperature and atmospheric pressure were recorded
periodically. After completion of the test, the GCL
sample was taken out of the cell and visually inspected
for the presence of breakthrough or leakage spots, if
any. Final GCL height was then determined for posttest analyses. Three small pieces of 30 mm diameter
specimen were cut from the tested 125 mm diameter sample to measure total suction by WP4C dew
point potentiometer (Decagon Devices, USA) and subsequent measurement of moisture content was done.
For this study, moisture content of the top and bottom gravel layers was assumed to be constant before
and after gas permeability test. It was also assumed
that the pressure drop in gravel layers was negligible
(Bouazza & Vangpaisal 2003, Vangpaisal & Bouazza
2004) compared to differential pressure used in these
experimentations. It is also noteworthy that gas permeability measurement under steady state condition
at one single differential pressure takes 1030 minutes. Therefore, the limitation of this apparatus is that
it takes long time to perform the test. A full description of the cell, testing procedures, limitation of set-up
and duration of test are given in Bouazza & Vangpaisal
(2003), Vangpaisal & Bouazza (2004).
GCL samples were hydrated to cover a wide range

of gravimetric moisture contents and suctions. All the
tests were conducted in a temperature controlled room
(set at 20 1 C). Therefore, the viscosity and density
of the nitrogen gas used in this study were considered
to be constant.
In case of gas flow through porous media, application of Darcys law (equation 4) shows that flow rate
is proportional to P21 P22 instead of P as discussed in
Bouazza & Vangpaisal (2003), Vangpaisal & Bouazza
(2004).
Figure 2 shows typical variation of gas flow rate
against difference of inflow and outflow gas pressure
squared. The gas flow rate at high gravimetric moisture content with corresponding low suction and low
gravimetric moisture content with corresponding high
suction were plotted against P21 P22 in Figure 2a and
2b, respectively, both under 2 kPa and 20 kPa curing
437
Figure 3. Intrinsic gas permeability against differential

pressure for both 2 kPa and 20 kPa curing.
Figure 2. Gas flow rate with P21 P22 (a) at high moisture
content and low total suction (b) at low moisture content and
high total suction under 2 kPa and 20 kPa curing.
stress conditions. In Figure 2a, it is observed that the

gas flow rate varied linearly with P21 P22 for GCL
samples prepared at high gravimetric moisture content
and low suction under both curing stress conditions
used in this study. Similarly, gas flow rate was found
to vary linearly with P21 P22 at low gravimetric moisture content and high suction condition (Fig. 2b) for
the two pre-conditioning stressed investigated in this
study. This linear behaviour indicated that gas flow
was laminar during the experiments. Additionally, the
Reynolds number estimated by Bouazza & Vangpaisal
(2003), Vangpaisal & Bouazza (2004) for similar GCL
material showed that the value was within the range
given by Bear (1972) for both dry and wet conditions
used in their study. Furthermore, McBean et al. (1995)
reported that for landfill gas migration and recovery
Darcys law will be applicable if the characteristics
grain sizes of porous media are smaller than 2 mm.
The GCL used in this study has bentonite clay that has
grain size much less than 2 mm. Therefore, it can be
assumed that Darcys law is applicable in this study.
It is also shown in Figure 2b that gas flow rate
increased with the decrease of gravimetric moisture content and increase of suction for both preconditioning stresses, indicating that air filled pores
increased with the decrease of gravimetric moisture

content irrespective of curing stress difference. However, Figure 2a shows that at high moisture content the
curing stress governs the gas flow rate. The higher the
curing stress, the lower is the gas flow rate.
Intrinsic gas permeability calculated using equation (4) is plotted in Figure 3 with respect to increase
and decrease of differential pressures for both 2 kPa
and 20 kPa cured samples at two different gravimetric
moisture contents with their corresponding suctions.
The differential gas pressure, for most of the tests, was
restricted to the range of 3 to 10 kPa. This is because
the build up of gas pressure under the cover system
in landfill sites is unlikely to be higher than 10 kPa
(McBean et al. 1995). Figure 3 shows that the intrinsic
gas permeability remained constant over the range of
differential pressure used in this investigation. It did
not change during increase of pressure from 3 kPa to
10 kPa or decrease of pressure from 10 kPa to 3 kPa
suggesting a good repeatability in the measurements.
It is also observed that the intrinsic gas permeability increased with the increase of suction (2 MPa to
44 MPa) and decrease of gravimetric moisture content
(54% to 13%).
Figure 4 shows the GCL water retention curve on
the wetting path. The measured GCL total suction varied approximately from 0.5 MPa to 140 MPa and the
corresponding gravimetric moisture content was in the
range of 5% to 90%. The measured GCL water retention curve exhibited typical trend for fine grained soil
both under 2 kPa and 20 kPa curing stress. Furthermore, it can be noticed that only a small change in the
total suction (0.85 MPa to 3 MPa) caused large and
rapid decrease in moisture content due to entering air
in GCL sample. Beyond 3 MPa suction, there was a
large increase in suction due to only a small decrease
in gravimetric moisture content.
The estimated gas permeability (Fig. 5a) and gas
permittivity (Fig. 5b) were plotted against gravimetric
moisture content for both 2 kPa and 20 kPa stress curing conditions. Additionally, data from Vangpaisal &
Bouazza (2004) were included in Figures 5a and 5b,
438
Figure 4. GCL water retention curve on the wetting path.
Figure 5. (a) Gas permeability and (b) gas permittivity

with respect to gravimetric moisture content under 2 kPa and
20 kPa curing.
respectively, for comparative purposes. Vangpaisal &

Bouazza (2004) used similar type of material as
the present study to measure gas permeability under
20 kPa confinement conditions.
Figure 5a shows that gas permeability decreased
with the increase of gravimetric moisture content for
both pre-conditioning stresses. It can also be observed
that gas permeability decreased one order of magnitude due to gravimetric moisture content increasing
from 7% to 85% at 2 kPa pre-conditioning; while
at 20 kPa stress curing condition, gas permeability
decreased a little bit more than two orders of magnitude due to gravimetric moisture content increase
of around 50%. It can also be observed that Vangpaisal & Bouazza (2004) data agreed well with the
current measured gas permeability for samples tested
under the same conditions. It is to be noted that up
to 50% gravimetric moisture content no marked difference in gas permeability values was observed for
the conditions considered in the present investigation.
This might be due to the fact that in this range (i.e.,
10 to 50% moisture) there was insufficient moisture
to hydrate the samples. However, at high gravimetric
moisture content, a noticeable difference in gas permeability values was observed. The gas permeability
reduced by one to two orders of magnitude for samples
cured under 20 kPa compared to samples cured under
2 kPa. This reduction is considered to be due to a much
greater reduction of pore space and disruption of the
air filled pore networks at high stress compared to
low stress pre-conditioning. These results imply that,
for this GCL, the effect of pre-conditioning stress is
insignificant for up to 50% gravimetric moisture content; beyond this range gas permeability was greatly
affected by preconditioning stresses.
The ability of GCLs to allow the flow of gas can
also be expressed in terms of gas permittivity (). It
is defined as the cross plane permeability (K) divided
by the GCL thickness (L), = K/L.
The variation of the permittivity against gravimetric
moisture content is plotted in Figure 5b. The variation of permittivity followed the same trend as the gas
permeability results shown in Figure 5a. A relatively
low reduction (less than one order of magnitude) in
gas permittivity was observed over the range of gravimetric moisture content investigated for samples cured
under 2kPa whereas a large reduction (up to two orders
of magnitude) was observed for samples cured under
20 kPa. Similar to the observations made for the gas
permeability variation, the effect of pre-conditioning
stresses is insignificant for up to 50% moisture content; beyond this range gas permittivity was found to
be affected by preconditioning stresses.
The calculated gas permeability (Fig. 6a) and gas
permittivity (Fig. 6b) were plotted against the measured total suction for the pre-conditioned samples.
Figure 6a shows that gas permeability increased with
the increase of total suction. Under the 2 kPa curing condition, gas permeability increased by one
order of magnitude when total suction increased from
0.85 MPa to 140 MPa. Under 20 kPa curing condition,
gas permeability increased by a little bit more than
two orders of magnitude when total suction increased
from 1.2 MPa to 88 MPa. It is also observed from Figure 6a that at low suction (suction less than 1.6 MPa)
gas permeability increased at low curing stress compared to high curing stress. However, at suction higher
than 1.6 MPa, gas permeability for both 2 kPa curing
439
Figure 7. Gas permeability against degree of saturation

under 2 kPa and 20 kPa curing.
Figure 6. (a) Gas permeability and (b) gas permittivity

against total suction under 2 kPa and 20 kPa curing.
and 20 kPa curing showed no noticeable difference.

This result indicates that the pore structure of the GCL
samples changed at low suction conditions under the
two different pre-conditioning stresses.
Figure 6b shows that gas permittivity increased with
the increase of total suction for both pre-conditioned
samples. Similar to Figure 6a, it is observed that at low
suction (suction less than 1.6 MPa) gas permittivity
increased at low curing stress compared to high curing stress. This might be due to reduction of air filled
pore spaces and pore network connectivity in higher
stressed samples compared to lower stressed samples.
At suctions higher than 1.6 MPa, gas permittivity for
both pre-conditioned samples remained the same.
The variation of gas permeability is plotted against
degree of saturation in Figure 7. The gas permeability
results with respect to degree of saturation shows the
similar trend of gas permeability results against moisture content in Figure 5a for both pre-conditioning
stresses. It is shown from Figure 7 that up to 30%
degree of saturation no marked difference in gas
permeability values was observed for the conditions
considered in the present investigation. This might be
due to the fact that in this range (i.e., 3 to 30% degree of
saturation) there was insufficient moisture to hydrate
the samples. However, at high degree of saturation,
a noticeable difference in gas permeability values was

observed. The gas permeability reduced by one to two
orders of magnitude for samples cured under 20 kPa
compared to samples cured under 2 kPa. This reduction is considered to be due to a much greater reduction
of pore space and disruption of the air filled pore
networks at high stress compared to low stress preconditioning. These results imply that, for this GCL,
the effect of pre-conditioning stress is insignificant for
up to 30% degree of saturation; beyond this range gas
permeability was greatly affected by preconditioning
stresses.
CONCLUSIONS
The gas permeability of a needle punched GCL was

found to be influenced by both gravimetric moisture
content and total suction. Furthermore, the effect of
preconditioning stresses was found to be insignificant
at low moisture content (<50%) and corresponding
high suction (>1.6 MPa) while the pre-conditioning
stress effect became more significant when gravimetric moisture content was greater than 50% and
corresponding total suction was less than 1.6 MPa.
These initial findings suggest that GCLs should be
subjected to more than a nominal overburden pressure
of 2 kPa and hydrated to more than 50% gravimetric
moisture content to minimise possible gas migration
by advection. More research is needed to evaluate the
level of overburden pressure required for a given moisture content, but certainly the 20 kPa used in these
experiments resulted in low gas permeability when the
GCL was hydrated to a gravimetric moisture content
of 68%. The relationship between overburden stress,
moisture content and gas permeability may depend
on the particle size of the bentonite and other GCL
characteristics, and these results cannot be assumed to
be applicable to GCLs other than the one examined
without independent verification.
440
ACKNOWLEDGEMENT
The present study was financially supported by a discovery grant from the Australian Research Council.
Our sincere appreciation is extended to the council.
REFERENCES
Alzaydi, A. A., Moore, C. A. A. & Rai, I. S. 1978. Combined
pressure and diffusional transition region fow of gases in
porous media. AICheE, 24: 3543.
Bear, J. 1972. Dynamics of fluid in porous media, Dover,
New York: American Elsevier Publishing Company, Inc.
Bouazza, A. & Vangpaisal, T. 2003. An apparatus to measure
gas permeability of geosynthetic clay liners. Geotextiles
and Geomembranes, 21: 85101.
Bouazza, A., Vangpaisal, T. & Jefferis, S. 2006. Effect of
wet-dry cycles and cation exchange on gas permeability
of geosynthetic clay liners. Journal of Geotechnical and
Geoenvironmental Engineering, 132: 10111018.
Didier, G., Bouazza, A. & Cazaux, D. 2000. Gas permeability of geosynthetic clay liners. Geotextiles and
Geomembranes, 18: 235250.
Massmann, J. W. 1989. Applying groundwater flow models
in vapor extraction system design. Journal of Engvironmental Engineering, ASCE, 115: 129149.
McBean, E. E., Rovers, F. A. & Farquhar, G. J. 1995. Solid
waste landfill; Engineering and design, Englewood Cliffs,
N.J.: Prentice-Hall PTR.
Mendes, M. J. A., Pierson, P., Touze-Foltz, N., Mora, H. &
Palmeira, E. M. 2010. Characterisation of permeability to
gas of geosynthetic clay liners in unsaturated conditions.
Geosynthetics International, 17: 344354.
Pitanga, H. N., Pierson, P. & Vilar, O. M. 2011. Measurement
of gas permeability in geosynthetic clay liners in transient
flow mode. Geotechnical Testing Journal, 34.
Vangpaisal, T. & Bouazza, A. 2004. Gas permeability of
partially hydrated geosynthetic clay liners. Journal of
Geotechnical and Geoenvironmental Engineering, 130:
93102.
441

Modeling of strain localization around the radioactive waste

disposal galleries
F. Salehnia, R. Charlier & S. Levasseur
Departement ArGenCo, Universit de Lige, Lige, Belgium
ABSTRACT: The excavation process of the underground repository galleries creates an Excavation Damaged
Zone (EDZ) in which the properties of the host media are locally altered. Since the development of shear
bands has been observed during the excavation of underground research facilities in potential clay host rocks,
it is proposed here to model the extension of damaged zone by the strain localization approach. Numerical
modeling of strain localization needs a specific approach to overcome the practical problem of mesh size
dependency within the framework of classical finite elements. To overcome this problem, the second gradient
method which has been incorporated into the finite element code Lagamine is proposed in this paper. The
objective of this study is to perform the hydro-mechanical modeling of a gallery excavation in Boom Clay
(Belgium) accomplished by the analysis of strain localization in shear bands mode: the thickness and evolution
of EDZ during the gallery excavation are exhibited through the analysis of plastic zone and increment of deviatoric
strain.
INTRODUCTION
Radioactive waste is an inevitable outcome of various

industrial and medical activities, such as material testing, manufacturing of some semi-conductors, radiotherapy, food sterilization and electricity production.
Nuclear power can be quantitatively an important contributor to the global electricity production capacity.
Based on the World Nuclear Associations announcement, nuclear power supplies about 1314% of the
worlds electricity requirements. However, it produces
some amount of potentially highly hazardous waste
which needs to be carefully managed. In fact, some
of the nuclear wastes will remain hazardous to man
and the environment for hundreds of thousands of
years and should thus be isolated for a long period
of time.
There is an international consensus on the advisability of storing high-level nuclear waste in deep
geological repository. The depth of a repository (several hundreds of meters) should ensure isolation from
the biosphere, while host rock characteristics such as
a low permeability should ensure that in the event of a
loss of containment due to the degradation of the engineered barrier system, radionuclide releases rates will
remain limited.
The north-east of Belgium is characterized by a
thick deposit of over-consolidated marine clay which
is known as Boom Clay. Clays are considered as possible hosts for geological disposal systems because
of their low permeability, large retention (sorption)
capacity for many radionuclides and their self-sealing
capacity. Boom Clay is present in Belgium in a

region with no volcanic activity and only low seismic
activity.
Underground excavation in deep geological host
rocks inevitably induces a stress redistribution which
results in the creation of a zone with significant
irreversible deformations and changes of hydromechanical properties; it is named as Excavation
Damaged Zone (EDZ). As the rock is damaged, crack
networks are created, which could constitute preferential flow paths, depending on the network connectivity.
Consequently, the rocks hydraulic conductivity can be
increased significantly and it may become heterogeneous and anisotropic (Shao et al. 2005, Levasseur
et al. 2010), although depending on the host rock
material, some self-sealing may develop after the
excavation. Therefore, the numerical modeling of the
extension of the damaged zone and the evolution of
its permeability remains important for building confidence in the long-term safety of high-level radioactive
waste repositories.
Based on the experimental observations in geomaterials, there is either the localized rupture (Lenoir
et al. 2007) due to the localized micro-crack propagation or the diffuse mode of failure (Levasseur et al.
2013). Considering the first phenomenon, strain localization can be mostly observed in shear band mode
leading to rupture in geomaterials. Thus the development of strain localization as a classical mode of
failure of geomaterials can be studied in order to
modelize the extension of EDZ and the fracturing
structure.
443
2
2.1
NUMERICAL MODELING OF STRAIN

LOCALIZATION
First gradient law
The Drucker-Prager yield surface (Equation 1) in the

framework of a frictional elasto-plastic model is used
as the constitutive mechanical law:
where c = cohesion; c = compression friction angle;

the first stress invariant, I = ij ij ; the second deviatoric stress invariant, II , is defined by Equation 3 in
which ij = deviatoric stress tensor; and m is given as
the following (Equation 2):
Figure 1. Hardening hyperbolic relation for two values of

cofficient B .
where hyperbolic functions are used as Equations 58:
size and orientation (Collin et al. 2009), a specific

approach is needed to overcome this problem in order
to properly model the localization phenomenon and
post-peak behavior. Among the different technics, the
second gradient model (Chambon et al. 1998 & 2001)
which is a model with internal length is used as the
regularization method. The continuum is enriched by
microstructure effects in the second gradient method;
so the kinematics includes macrokinematics as well
as microkinematics (Mindlin 1964, Germain 1973,
Collin et al., 2006).
Equations 910 are two balance equations (in the
weak form) which should be solved in the coupled
second gradient model for every kinematically admissible virtual displacement field ui and virtual pore
water pressure field pw :
where c0 = initial compression friction angle;

cf = final compression friction angle; c0 = initial
cohesion; cf = final cohesion; dec /decc = values of
equivalent plastic strain from which hardening/
softening starts; and B /Bc = values of equivalent
plastic strain for which half of hardening/softening
on friction angle and cohesion is achieved (Fig. 1)
(Barnichon, 1998).
where = current solid configuration (volume);

ijk = double stress, dual of the (micro) second gradient, which needs an additional constitutive law, and
it has no link with the pore water pressure; Gi = body
force per unit volume; ti = external traction (classical)
forces per unit area; T i = an additional external (double) force per unit area that both ti and Ti applied on a
part of the boundary of ; Dui is the normal deriva = time derivative of the water mass inside
tive of ui ; M
; mi = mass flow; Q = a sink term; q = part of the
boundary where the input water mass per unit area q
is prescribed; and ij = the total stress field which is
defined according to Bishop & Terzaghis postulate
as Equation 11:
The friction angle and/or cohesion hardening and/or

softening can be occurred as a function of Von Mises
equivalent plastic strain (Equation 4):
2.2
Second gradient law
Since numerical modeling of strain localization using

the classical finite element is depending on the mesh
where ij = effective stress; Sr,w = water saturation

degree; and ij = the Kronecker symbol.
444
Table 1.
Mechanical parameters of model.
Parameter
Symbol
Value
Unit
Young modulus
Poisson ratio
Specific mass
Initial friction angle
Final friction angle
Hardening/softening
coefficient
Initial cohesion
Final cohesion
Hardening/softening
coefficient
Dilatancy angle
0
f
B
300
0.125
2700
8
18
0.001
MPa
Kg/m3
c0
cf
Bc
300
30
0.01
kPa
kPa
Table 2.
Hydraulic parameters of model.
Figure 2. Geometry of the model.

Parameter
Symbol
Value
Unit
ijk in the second gradient law is defined with an

elastic law (Mindlin, 1964) as a function of (micro)
second gradient of the virtual displacement. It depends
on one elastic parameter D that the shear band width is
proportional to this elastic parameter (Chambon et al.
1998 & Kotronis et al. 2007).
Water permeability
Specific mass of water
Porosity
Water compressibility
Van Genuchten parameter*
kw
w
1/w
m
n
P
3 1019
1 103
0.39
5 1010
0.32
1.47
3.44
m2
Kg/m3
Pa1
MPa
*Water retention curve and water permeability based on the

Van Genuchten model (Van Genuchten, 1980).
RESULTS OF NUMERICAL SIMULATIONS

OF A GALLERY EXCAVATION
A two-dimensional hydro-mechanical modeling of

a gallery excavation has been performed in the
plain strain condition using the Finite Element code,
Lagamine, from the Universit de Lige.
The studied gallery is a synthetic case defined by
analogy with one of the principle galleries excavated
in the underground research laboratory (URL) close to
the city of Mol, Belgium, in order to study the feasibility of high-level nuclear waste disposal in the Boom
Clay layer. In fact, the following computed results estimate the evolution of EDZ around this gallery through
the shear band analysis. However, this study is now
limited to the numerical modeling with no comparison
with the field tests results.
3.1 Geometry and initial conditions
Figure 2 shows the two-dimensional mesh geometry
of the model. Due to the symmetry of gallery, only a
quarter of it has been modeled with a radius of 1.25 m.
The blocked displacements can be observed in Figure 2. The anisotropic initial stresses were applied as
yy = 4.5 MPa and xx = zz = 3.825 MPa which were
decreased to 100 KPa on the gallery wall during the
excavation phase (24 hours) and remained constant
until the end of simulation (3.5 years). Moreover, The
initial pore water pressure was equal to 2.25 MPa.
The other boundries were assumed to be imprevious.
3.2 Hydro-mechanical properties

Considering the initial conditions brought in the last
section, the model has been analyzed using a frictional elasto-plastic model with the Drucker-Prager
yield surface (first gradient law) and the second
gradient method; the cohesion softening and friction angle hardening were applied. The applied flow
model lets reproducing the existed transfers in porous
media.
Tables 12 show the hydraulic and mechanical
parameters for Boom Clay which were used in the
gallery excavation modeling. However, several calculations were performed in the framework of parametric
study in which the influences of some parameters of
Table 1 were carefully studied; so the brought value
of final cohesion will be more explained in the next
section based on the fact of considering the cohesion
softening.
3.3 Results
3.3.1 Parametric study
In order to understand precisely about the influences of the main mechanical parameters on strain
localization phenomenon, several calculations have
been performed including the influence of cohesion
softening (parameters cf and Bc ) and secondgradient
445
Figure 3. Increment of deviatoric strain as long as the evolution of strain localization after a) 1 day of excavation phase; b)
10 days and; c) 1277.5 days = 3.5 years, influence of second
gradient elastic modulus, D = 0.2 N.
Figure 4. Increment of deviatoric strain as long as the evolution of strain localization after a) 1 day of excavation phase;
b) 10 days and; c) 1277.5 days = 3.5 years, influence of
second gradient elastic modulus, D = 2000 N.
446
elastic modulus D. In this paper, it is presented only

the influences of second gradient elastic modulus (D)
which needs to be obtained numerically since the
experimental works cannot give it.
As the consequence, the second gradient parameter affects significantly on the width of the shear
strain localization bands. Using the parameters of
Tables 12 (except of final cohesion which was taken
as cf = 200 kPa), Figures 34 show the Increment of
deviatoric strain in three different time steps during
the calculation (after one day of excavation phase, 10
days, and 3.5 years that is the end of simulation) for
two values of D. The smaller second gradient modulus
(and implicitly the internal length) as it is observed
in Figures 3(b)4(b) makes larger the apparent softening effect on launching the localization. Moreover
Figures 3(c)4(c) show that increasing the second
gradient parameter induces the wider bands of localization. Therefore, the value of D should be carefully
chosen considering the mesh element size and it is

considered here as D = 2000 N which seems suitable.
Similarly, the numerical simulations have shown
the role of the cohesion softening in order to initiate
the strain localization around the gallery. So based
on these results, the initial cohesion was divided by
factor 10 to give the final cohesion as it was brought
in Table 1.
3.3.2 Study of strain localization pattern
Taking into account the results of the parametric study,
the calculation has been done with the second gradient
elastic modulus D = 2000 N as well as the parameters
of Table 12.
The results of increment of the deviatoric strain
and plasticity of the elements in four different time
steps during the calculation (after one day of excavation phase, 100 days, 500 days and 3.5 years that
is the end of simulation) are brought in Figures 56.
Figure 5. Increment of deviatoric strain as long as the evolution of strain localization after a) 1 day of excavation phase;
b) 100 days; c) 500 days and; d) 1277.5 days = 3.5 years.
447
Figure 6. Plasticity as long as the evolution of strain localization after a) 1 day of excavation phase; b) 100 days; c) 500 days
and; d) 1277.5 days = 3.5 years.
In term of the excavation damaged zone, finally the

extension of the damaged zone is as 10.25 m horizontally and 3.5 m vertically (the coordinate of 11.5 m
horizontally and 4. 75 m vertically) for which the effect
of initial anisotropic stress is obvious. The modeling
provides some information about the fracturing structure and its evolution, although it still should go beyond
some limitations in order to consider the rock properties changes and anisotropic permeability during the
localization.
Figure 7 shows three studied cross-sections (horizontal, 45 degree and vertical) around the gallery for
which the evolutions of displacements and pore water
pressure have been studied.
Evolution of pore water pressure and displacements for the mentioned cross-sections at different
times during the simulation (from 1 day which is end
of excavation until 1277.5 days = 3.5 years which is
end of calculation) are observed in Figures 89. As
it was already expected, appearance of the bands of
Figure 7. The studied cross-sections.
448
Figure 8. Evolution of pore water pressure for a) horizontal

cross-section A-A; b) 45? inclined cross-section B-B and;
c) vertical cross-section C-C.
Figure 9. Evolution of displacement of a) dx for the horizontal cross-section A-A; b) dx for the 45 inclined cross-section
B-B; c) dy for the 45 inclined cross-section B-B and; d) dy
for the vertical cross-section C-C.
localization and its development influence clearly in

the evolution of pore water pressure (Figure 8) and
displacements (Figure 9) with some falls and rises in
the pattern of their evolution.
CONCLUSIONS
Boom clay is considered as a potential host rock formation for the deep geological disposal of high-level
449
REFERENCES
Figure 9. Continued
radioactive waste in Belgium due to its favorable properties, among which its very low hydraulic conductivity. As the creation of an Excavated Damaged Zone
(EDZ) around the galleries is inevitable, this permeability will be locally increased during the excavation,
while some self-sealing occurs afterwards. Modeling
the development of strain localization bands may allow
a better estimation of evolution of the hydraulic conductivity in the EDZ. In other words, studying the
evolution of fracture network in the framework of
strain localization subsequent to rocks damage permits a better understanding of the rocks hydraulic
conductivity changes.
The excavation damaged zone and strain localization bands during the excavation phase of a gallery at
Mol (Belgium) have been modeled realistically using
the second gradient model (Chambon et al. 1998 &
2001). The results simulated the evolution of the localization bands and the extension of EDZ as 10.25 m
horizontally and 3.5 m vertically. As it was observed,
the effect of strain localization is obvious in terms of
evolution of pore water pressure and displacements.
Moreover, modeling confirms that permeability is not
anymore homogeneous in the damaged zone and it is
changed within the shear bands; this phenomenon will
be focused in a future study.
Barnichon, J.D. 1998. Finite element modeling in structural

and petroleum geology: PhD thesis. Lige: Universit de
Lige.
Chambon, R., Caillerie, D. & El Hassan, N. 1998. Onedimensional localisation studied with a second grade
model. Eur. J. Mech. A/Solids 17(4): 637656.
Chambon, R., Caillerie, D. & Matsushima, T. 2001. Plastic
continuum with microstructure, local second gradient theories for geomaterials: localization studies. International
Journal of Solids and Structures 38: 85038527.
Charlier, R. et al. 2010. Thermal impact on the damaged zone
around a radioactive waste disposal in clay host rocks.
European commission.
Collin, F., Chambon, R. & Charlier R. 2006. A finite element
method for poro mechanical modelling of geotechnical problems using local second gradient models. Int. J.
Numer. Meth. Eng. 65 (11): 17491772.
Collin, F., Levasseur, S. & Chambon, R. 2009. Numerical
post failure methods in multiphysical problems. Eur. J.
Environ. Civ. Eng 13(78): 9831004.
Germain, P. 1973. The method of virtual power in continuum
mechanics. Part 2: Microstructure. SIAM J. Appl. Math
25: 556575.
Kotronis, P., Collin, F., Bsuelle, P., Chambon, R. & Mazars,
J. 2007. Local Second Gradient Models and Damage
Mechanics: 1D Post Localization Studies in Concrete
Specimens. 7th International Workshop on Bifurcation,
Instabilities and Degradation in Geomechanics: 127142.
Lenoir, N., Bornert, M., Desrues, J., Bsuelle, P. & Viggiani,
G. 2007. Volumetric digital image correlation applied to
X-ray microtomography images for triaxial compression
tests on argillaceous rock. Strain 43(3): 193205.
Levasseur, S., Charlier, R., Frieg, B. & Collin, F. 2010. Hydromechanical modelling of the excavation damaged zone
around an underground excavation at Mont Terri Rock
Laboratory. Int. J. Rock Mech. Min. 47: 414425.
Levasseur, S., Collin, F., Charlier, R. & Kondo, D. 2013.
A micro-macro approach of permeability evolution in
rocks excavation damaged zones. Comput. Geotech. 49:
245252.
Mindlin, R.D. 1964. Micro-structure in linear elasticity. Arch.
Ration. Mech. Anal. 16: 5178.
Shao, J.F., Zhou, H. & Chau, K.T. 2005. Coupling between
anisotropic damage and permeability variation in brittle
rocks. Int. J. Numer. Anal. Methods Geomech 29:1231
1247.
Van Genuchten, M. TH. 1980. A closed-form equation for
Soil Sci. Soc. Am. J. 44:892898.
World Nuclear Association. www.world-nuclear.org.
ACKNOWLEDGEMENT
The authors would like to thank the Belgian Agency
for Radioactive Waste ONDRAF/NIRAS for their
financial support in this project.
450

Constitutive model for unsaturated soils based on the effective stress

H. Shin
University of Ulsan, Ulsan, Republic of Korea
S.R. Lee
Korea Advanced Institute of Science and Technology, Daejeon, Republic of Korea
ABSTRACT: The importance of unsaturated state in various geo-engineering problems has led to advance
mechanical constitutive model emulating behavior of unsaturated soils in response to thermo-hydro-mechanical
loading. Elasto-plastic mechanical constitutive model for unsaturated soil is formulated based on Bishops
effective stress. Effective stress and temperature are main variables in constitutive equation, and incremental
formulation of constitutive relationship is derived to compute stress update and stiffness tensor. Numerical
simulations involving coupled THM (Thermo-Hydro-Mechanical) processes are conducted to discuss numerical
stability and applicability of developed constitutive model: one-dimensional test, and clay-buffering at high
level radioactive waste disposal. Numerical results demonstrated that developed model can predict very complex
behavior of coupledTHM phenomena and is applicable to geo-engineering problems under various environmental
conditions, as well as interpret typical behavior of unsaturated soils.
INTRODUCTION
Recently diverse engineering problems identify the

important of unsaturated state of soils, and there is also
a growing interest and research to consider the effect
of temperature on the deformation of unsaturated soils
(Gens, 2010; Mitchell, et al., 2005). The behavior of
unsaturated soil is governed by thermo (T), hydraulic
(H), and mechanical (M) factors, and the interactions
among these factors affect each other.
Application areas of the research on the unsaturated soils can be general unsaturated soil behavior (impact of climate change on soil structure,
underground energy storage), heat-related fields
(geo-thermal energy, nuclear waste disposal, repeated
temperature load on the pavement, permafrost, powerlines buried ground), and soil environmental and
energy sectors (landfill leakage, CO2 sequestration,
CH4 hydrate), to understand short and long term unsaturated soil behaviors of coupled THM phenomena.
One of the most important factors to interpreting
fully coupled THM phenomena is mechanical constitutive model for unsaturated soil. Alonso et al. (1990)
first presented a complete formulation of elasto-plastic
constitutive law for HM phenomena in unsaturated
soil, and the proposed model has been called BBM
(Barcelona Basic Model) and considered as the fundamental model for unsaturated soil.After then numerous models have been proposed (refer to Sheng, 2011),
but basically they belong to the BBM except using different deformation characteristics or using different
stress variables.
BBM used two stress variables; net stress (=total

stress-air pressure) and suction (=air pressure-fluid
pressure), and LC (loading-collapse) yield function
is introduced to define variation of pre-consolidation
pressure with respect to suction increase. This framework of the model can predict volumetric change due
to wetting (which can be collapse or swelling depending on the magnitude of applied stress) or change in
shear strength by suction increase. However using net
stress and suction as stress variables can cause discontinuities in the stress field at the transition zone
of saturated and unsaturated conditions (Sheng, et al.,
2004). Therefore, Bishops effective stress (or average
stress) and suction have been used in the mechanical
model formulation in unsaturated soils (Nuth et al.,
2008; Wheeler et al., 2003; Zhao et al., 2010).
The effect of temperature on mechanical behavior of
soils has been studied for geotechnical structures with
the purpose of facilitating or inhibiting of heat spread
(Vulliet, et al., 2002). Laboratory experiments on saturated soils have conducted to understand the effect
of temperature on pre-consolidation stress (Eriksson,
1989), compressibility (Campanella, et al., 1968) and
shear strength (Cekerevac, et al., 2004; Hueckel, et al.,
1990; Santamarina, et al., 2009), but experimental
results for the unsaturated soil are very limited. In
tri-axial compression tests, pre-consolidation pressure
decreased with temperature increasing under constant
suction condition (Romero, 1999; Wiebe, et al., 1998).
In this paper, we derived elasto-plastic mechanical constitutive model for unsaturated soil based on
the effective stress. The developed model used two
451
stress variables: effective stress and temperature, and

formulated as an incremental form.
2
MECHANICAL CONSTITUTIVE MODEL
The proposed model is based on the BBM, and

Bishops effective stress is used as a stress variable to
maintain continuity of stress variable in the transition
between unsaturated and saturated regions. The effect
of temperature on deformation behavior of unsaturated soil is considered from the experimental results
(Laloui, etc., 2003). Stress and strain are positive value
for compression.
2.1
Stress variables
Constitutive law to describe mechanical behavior can

be expressed by various stress variables. In saturated soils, only effective stress is a stress variable
to determine soil behavior (Terzaghi, 1936). However
selection of stress variables in unsaturated soils is still
critical issue.
Basis constitutive model in unsaturated soil, BBM
uses net stress and suction as independent stress variables, but net stress causes discontinuity across the
transition of unsaturated and saturated zones (Loret
et al., 2000; Sheng et al., 2004). Since prediction
capability of effective stress to shear strength and
volumetric deformation was proved to be appropriate (Khalili, et al., 2004; Sheng, et al., 2003), model
formulation based on the effective stress has been
increasing. The use of suction as a stress variable is
still matter of debate (Laloui, etc., 2009; Loret, etc.,
2002), because it already affects effective stress. In the
developed model, suction is used as an internal variable, only affect the pre-consolidation stress (Bolzon
etc., 1996; Loret, etc., 2002). Also the commonly used
effective stress, Bishops stress (Bishop, 1959; Sheng,
et al., 2004) was used in the model.
Figure 1. Three-dimensional configuration of THM

mechanical constitutive model based on effective stress
(T2 > T1 ).
into MCC (Modified Cam Clay) model which has

a oval shape of yield surface (Alonso, et al., 1990;
Jommi, 2000; Laloui, et al., 2009; Sheng, et al., 2004).
The pre-consolidation pressure pc has the following characteristics: suction increases pc non-linearly
(Alonso, etc., 1990), and temperature tends to decrease
pc (Francois, et al., 2008; Laloui, et al., 2003).
where pre-consolidation pressure can be expressed

as pc = pr (p0 /pr )(0)/(s) + Sr s exp (4 T ), and
p0 = p [1 log (T /T0 )] = p Rp . Compression index
considering internal variable suction is (s) = (0)
[(1 r) exp (s) + r], and Cc (s) = 2.3(s).
In the equation (2), the transition from unsaturated
to saturated regions has suction to disappear, and finial
yield function shows smooth yield function of MCC
model considering temperature. Figure 1 show a threedimensional LC with respect to effective stress (p, q),
internal variable suction, and temperature.
Mechanical plastic potential function determining
incremental direction of plastic deformation should
avoid over-estimation of at rest earth pressure k0 from
using associate flow rule and meet Jakys law (Alonso,
etc., 1990), so it was defined as follows.
where is total stress, is Bishops effective stress, Pg
and Pl are gas and fluid pressure 1 = ij is Kroneckers
where
delta tensor, and is Bishops coefficient of effective

stress which can be simplified to degree of saturation
Sr (Jommi, 2000; Schrefler et al., 2001; Sheng et al.,
2004). Suction s is the difference between gas and fluid
pressure, is net stress. Later, expresses effective
stress, and is for net stress.
2.3 Incremental relationship between

stress and strain
2.2 Yield surface and plastic potential function

Effective stress and temperature are used to define
yield surface and plastic potential function which
describe mechanical behavior of unsaturated soils. The
proposed yield function takes suction and temperature
Total strain increment can be divided into two mechanical strain components from effective stress and temperature. In addition, the elastic and plastic mechanical
strain increment can be divided into parts.
452
where elastic mechanical strain increment is d e<m> =
De1 : d (Sheng et al., 2008), and De = (K 2G/3)
1 1 + 2G I is (elastic stiffness tensor). K and G are
elastic bulk and shear modulus, and 1 1 = ij kl ,

I = (ik jl + il jk )/2.
Incremental relations of the effective stress and

strain can be expressed as follows:
Plastic flow rule determines increment of mechanical plastic deformation in a direction perpendicular
to plastic potential function with a magnitude of
non-negative scalar multiplier d.
Figure 2. Comparison of swelling pressure due to water wetting in laboratory tests with FEBEX bentonite (Villar et al.,
2008) and the numerical results (solid line).
Typical relationship between stress and strain for

net stress, suction and temperature is defined by
d = Dep : d W ds RdT , but suction is already
2.4 Consistency condition

Even if the yield surface can move itself in space
or change its configuration by the hardening, stress
state must be in or on the yield surface. This consistency condition and its differential are represented as
follows:
Unknown multiplier d can be evaluated by consistency condition and stress-strain relationship.
2.5 Incremental form of mechanical constitutive

model
NUMERICAL ANALYSES
Numerical simulations are performed to solve fully

coupled THM phenomena using the developed
mechanical model. Since the solution of THM behavior of the porous material does not exist yet (Schrefler et al., 2001) and required material properties
in the constitutive model for unsaturated soils is
very limited, soil properties are calculated from
previous experimental results (Alonso et al., 1990;
Gens, 2010). Numerical simulations are conducted
for one-dimensional wetting process in unsaturated
soil under and long-term behavior of buffering material in high-level radioactive waste repository, and
they assured convergence and applicability of the
developed mechanical constitutive model.
3.1
Incremental stress update equation and the stiffness

tensor on the effective stress and temperature stress
can be obtained as follows.
relation.
Developed model is implemented to fully coupled
THM finite element code (element type: 8 nodes
for displacements, 4 nodes for fluid pressure and
temperature) (Shin, 2011).
3
(7)
engaged in the effective stress in the effective stress

based formulation and W disappear in the incremental
One-dimensional swelling test
Backfill material between in-situ soil and a structure

plays an important role to protect an earth-structure
against external loads, and expansion pressure of
filling material due to water wetting should not damage the structure (Villar et al., 2008). Wetting of
unsaturated soil produces expansion pressure which
depends on initial suction and dry unit weight of the
soil. Numerical analyses are carried out to estimate
swelling pressure during wetting process in unsaturated soil at various initial suctions, and numerical
results were compared with those of the laboratory
experiments with FEBEX bentonite (Villar et al.,
2008).
453
Figure 3. Geotechnical behavior of the engineered barrier system over time (without radioactive decay heat). (a) cross-section
of Korea barrier system (Cho et al, 2010), (b) 2-dimensional axis-symmetric analysis domain, (c) change in pore water pressure
Pl with time at various locations, (d) change of degree of saturation Sr , (e) void ratio change, (f) horizontal displacement of
each location, (g) vertical net stress v , (h) net horizontal stress h .
Numerical results tend to very similar to experiments results, and expansion pressure is rapidly
increased above dry unit weight 1.6 tonf/m3 (Figure 2).
Additional analyses showed that swelling index
(swelling index, ) has little effect on the swelling
pressure. On the other hand, compression index
(compression index, ) induced major change in
swelling pressure with the same dry unit weight of

the soil.
3.2 Behavior of buffering clay in high-level
radioactive waste repository
Understanding of THM behavior of high level radioactive waste disposal system built deep underground
454
Figure 4. Geotechnical behavior of the engineered barrier system over time (with radioactive decay heat). Variation (a)
radioactive decay heat release with time (P[W /m] = 350 e0.0219Year , Millard, etc., 2005), (b) temperature variations with
various locations, (c) pore-water pressure with time Pl , (d) change of degree of saturation Sr , (e) void ratio change, (f)
horizontal displacement of each location, (g) vertical net stress v , (h) net horizontal stress h .
plays an important role in optimal design of clay

buffering and securing long-term stability of storage
facilities. Clay-buffering (typically, compacted expansive clays) is under complex THM interaction between
canister and surrounding rock, which determines longterm behavior of the whole repsotiry system.
Material properties of the buffering material are
based on the laboratory test on FEBEX benonite
(Gens, 2010), and the canister was assumed to be

a rigid body. Repository system was located at initially fully saturated rock in 500m depth, and air
pressure is always constant zero. The initial stress
conditions are v = v + Pl = 500 t/m2 + 500 t/m2 ,
h = h + Pl = k0 v + Pl = 250 t/m2 + 500 t/m2 . Initial temperature is assumed 30 C (15 C on the ground
surface, geothermal gradient = 2530 C/km).
455
Engineered barrier system chosen in this analysis is

a Korean disposal system (Cho et al, 2010). After tunnel excavation, exposed surface has zero pore-water
pressure, and clay-buffering between the canister and
surrounding rock is assumed to have initial suction
s = 50 tonf/m2 (initial degree of saturation = 99.5%).
The effect of radioactive decay heat from high-level
radioactive waste on the behavior of clay buffering
and surround rock was investigated by two separate
numerical simulations; with and without considering radioactive decay heat. The two numerical results
showed that clay buffering installed engineering system promotes stress and displacement in the barrier
system through thermal, hydraulic, and mechanical
interactions (Figure 3 and 4).
THM behavior of the disposal system with time
could be summarized as follows.
1) Point a and b showed almost the same mechanical and hydraulic behavior due to low permeability
of clay buffering in both simulations, with and
without radioactive decay heat.
2) In the case considering thermal diffusivity from
the canister, temperature in clay buffering and
surrounding rock reached peak value and then
decrease, after heat generation from canister weakens (Figure 4a, 4b). Predicted maximum temperature (78 C) is lower than critical temperature
causing physical and chemical degradation of clay
buffering against the appropriate performance of
the barrier system (Korea Atomic Energy Research
Institute, 2003).
3) Inflow of groundwater due to hydraulic gradient
between surrounding rock and buffering material
starts to fully saturate clay buffering around 16
years later, causing dramatic hydraulic and mechanical change of the system (Figure 3c, 4c). Hydraulic
conductivity of the surround rock is the critical
property to determine the time for this full saturation of clay buffering; lower hydraulic conductivity
of the rock linearly delays the saturation time of the
buffering material. While, the second water pressure rise in the clay buffering is 1,500 years that
is determined by the hydraulic conductivity of clay
buffering.
Radioactive decay heat generates additional thermal pressurization (Rutqvist, etc., 2011) due to
the low hydraulic conductivity of the surrounding
rock (Figure 4c). Also, the degree of saturation of
the rock is sharply reduced and then re-saturated
(Figure 4d; Gens, 2002).
4) Volume expansion of clay buffering due to groundwater influx (16 year) rapidly increase the void
ratio (Figure 3e), and causes horizontal displacement of point b toward the canister (Figure 3f).
After full saturation of clay buffering, buffering
material adjacent to the canister is swelled and
void ratio at point c decreases and horizontal displacement of the point b turns the direction to
the surrounding rock. Swelling of the clay buffering produced overall expansion of the engineering
barrier system (Figure 3f, point c).
Effect of radioactive decay heat produces thermal expansion of clay buffering at point b. But
point c tends to shrink (Figure 4f), when heat
release from the canister reaches maximum around
1 year later.
5) Swelling pressure of the clay buffering acting on the canister increases over time up to
500 tonf/m2 (5 MPa) in both cases (Figure
3g h, 4gh). Canister design shall consider this
expansion pressure, as well as other external loads.
CONCLUSION
Fully coupled THM soil behaviors in various applications of geotechnical engineering increases the need of
numerical tools to analyze these complex soil phenomena. Mechanical constitutive model should consider
deformation due to stress, fluid pressure, and heat.
THM mechanical elasto-plastic model for unsaturated soil based on the effective stress is presented in
this study. Developed model uses effective stress and
temperature as the main variables, and formulated in
incremental form. THM finite element code executed
the developed model to interpret THM phenomena of
unsaturated soils, and assured the stability and applicability of the model through numerical examples.
REFERENCES
Alonso, E.E., Alcoverro, J., Coste, F., Malinsky, L., MerrienSoukatchoff, VOL., Kadiri, I., Nowak, T., Shao, H.,
Nguyen, T.S., Selvadurai, A.P.S., Armand, G., Sobolik,
S.R., Itamura, M., Stone, C.M., Webb, S.W., Rejeb, A.,
Tijani, M., Maouche, Z., Kobayashi, A., Kurikami, H., Ito,
A., Sugita, Y., Chijimatsu, M., Borgesson, L., Hernelind,
J., Rutqvist, J., Tsang, C.F., and Jussila, P. (2005), The
FEBEX benchmark test: Case definition and comparison
of modelling approaches, International Journal of Rock
Mechanics and Mining Sciences, Vol. 42, pp. 611638.
Alonso, E.E., Gens, A., and Josa, A. (1990), A constitutive
model for partially saturated soils, Geotechnique, Vol.
40, pp. 405430.
Campanella, R.G., and Mitchell, J.K. (1968), Influence
of temperature variations on soil behavior, Journal of
the Soil Mechanics and Foundations Division, Vol. 94,
pp. 709734.
Cekerevac, C., and Laloui, L. (2004), Experimental study
of thermal effects on the mechanical behaviour of a
clay, International Journal for Numerical and Analytical
Methods in Geomechanics, Vol. 28, pp. 209228.
ENRESA (1998), FEBEX: Full-scale engineered barriers
experiment in crytalline host rock: Preoperational thermohydro-mechanical (THM) modelling of the in-situ test,
Publicacin tecnica num.
Eriksson, L.G. (1989), Temperature effects on consolidation properties of sulphide clays, Proceedings of the12th
International Conference on Soil Mechanics and Foundation Engineering, Rio de Jeneiro, pp. 20872090.
Francois, B., and Laloui, L. (2008), ACMEG-TS: A constitutive model for unsaturated soils under non-isothermal
conditions, International Journal for Numerical and
Analytical Methods in Geomechanics, Vol. 32, pp. 1955
1988.
456
Hueckel, T., and Baldi, G. (1990), Thermoplasticity of saturated clays. Experimental constitutive study, Journal of
geotechnical engineering, Vol. 116, pp. 17781796.
Gens, A., Guimaraes, L.D., Garcia-Molina, A., and Alonso,
E.E. (2002), Factors controlling rock-clay buffer interaction in a radioactive waste repository, Engineering
Geology, Vol. 64, pp. 297308.
Gens, A. (2010), Soil-environment interactions in geotechnical engineering, Geotechnique, Vol. 60, pp. 374.
Laloui, L., and Cekerevac, C. (2003), Thermo-plasticity of
clays: An isotropic yield mechanism, Computers and
Geotechnics, Vol. 30, pp. 649660.
Lewis, R.W., and Schrefler, B.A. (1998), The finite element method in the static and dynamic deformation and
consolidation of porous media, New York, John Wiley.
Mitchell, J.K., and Soga, K. (2005), Fundamentals of soil
behavior, New York Wiley, 577 p.
Rutqvist, J., Barr, D., Birkholzer, J.T., Fujisaki, K., Kolditz,
O., Liu, Q.S., Fujita, T., Wang, W.Q., and Zhang, C.Y.
(2009), A comparative simulation study of coupled THM
processes and their effect on fractured rock permeability around nuclear waste repositories, Environmental
Geology, Vol. 57, pp. 13471360.
Santamarina, J.C., and Shin, H. (2009), Friction in granular media, in Hatzor, Y.H., Sulem, J., and Vardoulakis,
I., eds., Meso-scale Shear Physics in Earthquake and
Landslide Mechanics, CRC Press, pp. 157188.
Schrefler, B.A., and Scotta, R. (2001), A fully coupled
dynamic model for two-phase fluid flow in deformable
porous media, Computer Methods in Applied Mechanics
and Engineering, Vol. 190, pp. 32233246.
Sheng, D., Sloan, S.W., and Gens, A. (2004), A constitutive model for unsaturated soils: thermomechanical and
computational aspects, Computational Mechanics, Vol.
33, pp. 453465.
Sheng, D. (2011), Review of fundamental principles in
modelling unsaturated soil behavior, Computers and
Geotechnics, Vol. 38, pp. 757776.
Shin, H. (2011), Formulation of Fully Coupled THM Behavior in Unsaturated soil, Journal of Korean Geotechnical
Society, 27(3), pp. 7583
Villar, M.VOL., and Lloret, A. (2008), Influence of dry density and water content on the swelling of a compacted
bentonite, Applied Clay Science, Vol. 39, pp. 3849.
Vulliet, L., Laloui, L., and Harding, R. (2002), Environmental geomechanics:An introduction, in Environmental
Geomechanics, Lausanne, EPFL-Press, pp. 312.
Wheeler, S.J., Sharma, R.S., and Buisson, M.S.R. (2003),
Coupling of hydraulic hysteresis and stress-strain
behaviour in unsaturated soils, Geotechnique, Vol. 53,
pp. 4154.
Wiebe, B., Graham, J., Tang, G.X.M., and Dixon, D. (1998),
Influence of pressure, saturation, and temperature on
the behaviour of unsaturated sand-bentonite, Canadian
Geotechnical Journal, Vol. 35, pp. 194205.
Zhao, C.G., Liu, Y., and Gao, F.P. (2010), Work and energy
equations and the principle of generalized effective stress
for unsaturated soils, International Journal for Numerical and Analytical Methods in Geomechanics, Vol. 34,
pp. 92036.
457

Crack control of landfill liner and cap materials using nano-alumina powder
M.R. Taha & O.M. Taha
Dept. of Civil and Structural Engineering, Universiti Kebangsaan Malaysia, Selangor, Malaysia
ABSTRACT: Desiccation cracking is a widespread phenomenon in clay soils which may significantly increase
the hydraulic conductivity of soil. This problem is one of the main concerns in the design and construction of
landfill covers, particularly in barren regions. Tests were conducted to investigate the effects of nano-alumina
on the desiccation cracking of four soil samples of different plasticity index. Surficial dimensions of cracks
in the soils were monitored and quantified using the crack intensity factor (CIF). The results showed that the
required amounts of nano-alumina mixed with soil were very low, which give a little increase in maximum dry
density. Furthermore, it was observed that the optimum water content, volume change and crack intensity factor
of soil specimens decreased measurably with addition of these low amounts of nano-alumina particles. This will
generally decrease the development of cracks in the soil and thus reducing its hydraulic conductivity.
INTRODUCTION
Desiccation tends to cause clayey soils to shrink due

to the loss of soil water content. The reduction of
moisture developed the matric suction in soil leading
to an increase in shrinkage strain eventually followed by crack (Nahlawi and Kodikara, 2006, Omidi,
1993). According to Albrecht and Benson (2001), soil
should be compacted at the maximum dry density
and optimum water content, which should result in
low void ratio to achieve low shrinkage strain after
saturation.
In dealing with shrinkage strain and desiccation
in soil, another important parameter accompanying
them is the hydraulic conductivity. In general, the soil
should be compacted to achieve a hydraulic conductivity of 1 109 m/s or less. Thus, the acceptable zone
of water content and dry unit weight that yielded lower
hydraulic conductivity is usually the line of optimums
and some part of wet side of line of optimum. The volume change increase with the higher molding water
content varied greatly between the dry and wet sides
of the line of optimum, where increase towards the wet
side and decreased towards the dry side (Daniel and
Wu, 1993, Fu et al., 2011).
One of the methods to reduce the shrinkage strain
and hydraulic conductivity is to increase compaction
effort or energy. This method sometimes may cause an
increase in construction cost. The increase in the compaction effort increases the dry density and decreases
the water content but for higher water content soils, the
increase in the compaction effort does not reduce the
shrinkage strain significantly (Osinubi and Nwaiwu,
2008). Many materials have been studied and used to
reduce shrinkage in soils. For example, mixing the soil
with fiber can reduce the shrinkage strain up to 90%
but the fiber increases the hydraulic conductivity up to

50 times the hydraulic conductivity of the soil (Rifai,
2000, Harianto et al., 2009).
This paper presents the effect of addition of nanoalumina on soil compaction condition, soil volume
change and crack intensity factor (CIF) of the soil
samples.
2
MATERIALS
2.1 Soil
Four types of soils are used in this study, i.e. two low
plasticity and two high plasticity clays (i.e CL and
CH soils, respectively). M1 is a local soil obtained
from within the campus of the Universiti Kebangsaan
Malaysia. This is a residual soil of sedimentary origin.
Soils M2, M3, and M4 are soil M1 with 5%, 10% and
20% bentonite for creating soils with higher plasticity
index.
The specific gravity of all the soils were almost
the same but their plasticity index significantly differ
from each other. The properties of soil are shown in
the Table 1.
2.2 Nano-alumina
The nano-alumina used in this study was Ultrapure
Gamma-Alumina (-Al2 O3 ) powder with purity equal
to 99.99% supplied by Inframat Advanced Materials, Manchester, Connecticut, USA. The specification
and other detailed information are shown in Table 2.
Pictures of nano-alumina taken by scanning electronic microscope (SEM) and Transmission electron
microscope (TEM) are shown in Figures 1 and 2.
459
Table 1.
Basic properties of the soil used in this study.

Values and descriptions
Characteristics
M1
Specific gravity
Plasticity index (%)
Linear shrinkage (%)
Passing No. 200 sieve (%)
Clay content (<2 m) (%)
Unified Soil Classification
System (USCS)
Compaction properties
Optimum water content (%)
Maximum dry unit
weight (kN/m3 )
2.61
2.61
2.61
2.62
16.96 28.26 36.18 69.01
8.2
12.86 15.9
23.57
47.16 49.80 52.44 57.73
18
22.9
28.5
38.5
CL
CL
CH
CH
Table 2.
M2
M3
M4
14.29 16.01 18.63 20.3

18.05 17.70 16.93 16.12
Figure 2. Nano-alumina material as shown in the transmission electron microscope (TEM).
Properties of nano-alumina powder.
Property
Nano-alumina
Particle density (g/cm3 )

Surface area (m2 /g)
Formula
Average particle size
Solubility in water (%)
Color
3.6
>150
Al2 O3
2050
Insoluble
Off white
4 TEST PROCEDURES
Figure 1. Nano-alumina material as shown in the scanning

microscope electron (SEM).
PREPARATION OF SOILNANOMATERIAL
MIXTURES
The soil sample was dried in an oven before grinding. Then, the soil and nano-alumina were mixed by
spraying the nano-alumina powder over the soil in
six layers to prepare the mixtures. The soil and nanoalumina were blended together under dry conditions.
The dry mixtures were mixed with the required amount
of water established to give the optimum Proctor moisture content.All mixing was done manually, and proper
concern was taken to prepare homogeneous mixtures
at each stage.
Theexperimental procedures involved determining

compaction behavior of unamended soils and soils
mixed with various amounts of nano-alumina: 0.0%,
0.05%, 0.075%, 0.1%, 0.15%, and 0.3% the percentage expressed by weight of dry soil. The compaction
curves were determined according to standard test
method ASTM D698 for both unamended soil and soil
with nano-alumina. Compacted samples in Standard
Proctor mold were used for measuring the shrinkage
and expansive strains in addition to running hydraulic
conductivity tests. Volume change tests were conducted on compacted samples of unamended soil
and soilnano-clay mixtures.After the compaction was
completed some of the soil sampleswere saturated
with water and the other left to dry directly under
oven temperature approximately 34 2 C. The saturation was conducted by permeating the specimens in
flexible-wall permeameters following ASTM D5084.
Each specimen was dried after having been saturated
by permeation.
When a specimen was removed from the permeameter, measurements were made of height, diameter,
and weight. The specimen was then placed in a pan in
the oven at temperature (approximately 34 C 2 C)
and allowed to dry slowly.
After the specimens were fully desiccated, measurements were made of height, diameter, and weight.Then
the total volume change (total strain) and crack intensity factor were measured. The total strain is the total
volume change in soil and approximately the sum of
shrinkage and expansive volume changes (Albrecht
and Benson, 2001, Rifai, 2000). The crack intinsity
factor (CIF) measured by dividethe surficial dimensions of cracks (cracked area) over the total area of the
soil samples, as expressed in equation (1) (Harianto
et al., 2008):
460
Figure 3. Maximum dry density and optimum moisture

content relationship of soil type M1 mixed with different
nano-alumina content.

content relationship of soil sample M3 mixed with different

where Ac is the desiccation crack area and At is the

total surface area of the sample.
5
5.1

Effect of nano-alumina on the compaction
conditions and plasticity index
The compaction curves showing the influence of nanoalumina on the optimum water content and maximum
dry density for soil samples M1, M2, M3, and M4
are shown in Figures 3, 4, 5, and 6. There is a slight
decrease in the optimum water content and slight
increase in maximum dry density due to addition
of nano-alumina. From Figure 3, the maximum dry
density in the soil sample M1 was achieved by the
addition of 0.05% nano-alumina and for soil sample
M2, the maximum dry density was obtained at 0.075%

nano-alumina content (Fig. 4).
For soil samples M3 and M4 the higher dry density
was achieved at 0.1% nano-alumina content (Figs 5 &
6). Due to the low nano-alumina content mixed with
soil samples, there is no marked increase in dry density.
From Figure 7, it is shown that theincrease in nanoalumina causes some decrease in plasticity index. This
happened possibly because density of nano-alumina
particles is higher than density of soil particles. Moreover, the addition of nano-alumina powder decreases
the liquid limit in soil due to the high surface area of
nano-alumina particles. However, the decrease in the
plasticity index was not very significant, due to the
low percentages used in this study.
5.2 Effect of nano-alumina on the soil volume
change and crack intensity factor
The total volume change (total strain) decreases
slightly at 0.05% and 0.075% nano-alumina then
461
Figure 7. Plasticity index vs. nano-alumina content.
Figure 9. Crack intensity factor vs. nano-alumina content.
Figure 8. Total volume change vs. nano-alumina content.
increases for soil samples M1 and M2. However, for

soil samples M3 and M4, the total volume change (total
strain) decreases at 0.05%, 0.075%, and 0.10%, and
then increase at 0.15% and 0.3% nano-alumina contents as shown in Figure 8. The optimum nano-alumina
contents were 0.05%, 0.075%, 0.1%, and 0.1% for soil
type M1, M2, M3, and M4, respectively. In this study,
only soil samples M3 and M4 were selected for crack
intensity factor (CIF) test due to the cracks in soil samples M1 and M2 were too small and were unable to be
measured.
From Figure 9, there is a measureable reduction in
crack intensity factor for both soil samples M3 and M4.
The maximum reduction in CIF was noted at 0.1%
nano-alumina content. The reduction in crack intensity factor (CIF) is due to the decrease in total strain
by the reduction of volume changes in soil samples
after adding nano-alumina material. The reduction in
total volume change depends on the increases amount
in dry density and the reduction amount in plasticity
index results in soil samples byadding nano-alumina
material. In general, the reduction in CIF means less
cracks (also reduction in crack size) should appear in
the soil thus reducing hydraulic conductivity.
Comparing to our previous studies, a small amount
of nano-alumina material or nano-clay can reduce
the soil hydraulic conductivity without increase the
swelling potential in soil samples (Taha and Taha,
2012a, Taha and Taha, 2012b).
CONCLUSION
This study attempts to improve desiccation crack properties of soils using nano-alumina. From the above
results, the required amount of nano-alumina mixed
with soil should be equal or less than 0.1% of dry soil
weight to achieve an improvement in soil without any
or little side effects (such as change in optimum dry
density, plasticity index).The small amounts of nanoalumina can reduce the total volume change of soil
and the crack intensity factor (CIF). The reduction
in soil volume change and crack intensity factor is
due to the increase in soil density. The increase in soil
density by adding nano-alumina material was possibly
due to the high specific gravity of nano-alumina and
the friction between soil particles and nano-alumina
particles.Moreover, particle sizes of nano-alumina are
less than the size of clay particles, which possibly
allow nano-alumina particles to enter in-between the
clay particles and reduce soil void ratio eventually
increasing the density and reducing the total strain.
REFERENCES
ALBRECHT, B. A. & BENSON, C. H. (2001) Effect
of desiccation on compacted natural clays. Journal of
6775.
DANIEL, D. E. & WU, Y.-K. (1993) Compacted Clay Liners and Covers for Arid Sites. Journal of Geotechnical
Engineering, 119, 223237.
FU, X., SHAO, M., LU, D. & WANG, H. (2011) Soil water
characteristic curve measurement without bulk density
changes and its implications in the estimation of soil
hydraulic properties. Geoderma, 167168, 18.
HARIANTO, T., HAYASHI, S., DU, Y.-J. & SUETSUGU, D.
(2008) Effects of FiberAdditives on the Desiccation Crack
Behavior of the Compacted Akaboku Soil as A Material
for Landfill Cover Barrier. Water, Air, and Soil Pollution,
194, 141149.
HARIANTO, T., HAYASHI, S., DU, Y. J. & SUETSUGU,
D. (2009) Experimental Investigation on Strength and
Mechanical Behavior of Compacted Soil-fiber Mixtures.
IN LI, G., CHEN, Y. & TANG, X. (Eds.) Geosynthetics
in Civil and Environmental Engineering. Springer Berlin
Heidelberg.
NAHLAWI, H. & KODIKARA, J. K. (2006) Laboratory experiments on desiccation cracking of thin soil
462
layers. Geotechnical and Geological Engineering, 24,

16411664.
OMIDI, G. (1993) Desiccation shrinkage and cracking in soil
liner. Texas A&M University.
OSINUBI, K. J. & NWAIWU, C. M. O. (2008) Desiccationinduced Shrinkage in Compacted Lateritic Soils. Geotechnical and Geological Engineering, 26, 603611.
RIFAI, S. M. (2000) Impact of Polypropylene Fibers on
Desiccation Cracking and Hydraulic Conductivity of
Compacted Clay Soils, Wayne State University.
TAHA, M. R. & TAHA, O. M. (2012a) Investigation on

Hydraulic Conductivity and Desiccation Crack Behavior
of Compacted Soil-Nano Alumina Mixture. 4th International Symposium on Nanotechnology in Construction.
Agios Nikolaos, Crete, Greece.
TAHA, O. M. E. & TAHA, M. R. (2012b) Experimental study
on the expansive-shrinkage of residual soil mixed with
nanoclay. 5th Jordanian International Civil Engineering
Conference (JICEC5). Jordan.
463
Underground energy issues

Coupled phenomena induced by freezing in a granular material

F. Casini, A. Gens & S. Olivella
ETCG, Universitat Politecnica de Catalunya, Barcelona, Spain
G.M.B. Viggiani
DICII, UNiversit di Roma Tor Vergata, Roma, Italy
ABSTRACT: Artificial Ground Freezing (AGF) is a controllable process that can be used by engineers to
stabilise temporarily the ground, provide structural support and/or exclude groundwater from an excavation until
construction of the final lining provides permanent stability and water tightness. AGF is often carried out based
on an observational approach, while a full understanding of the freezing process and of the generation of freezinginduced heave and settlements is still far from being achieved. In this work, the process of ground freezing is
studied using a constitutive model that encompasses frozen and unfrozen behaviour within a unified effectivestress-based framework and employs a combination of ice pressure, liquid water pressure and total stress as
state variables. The parameters of the constitutive model are calibrated against experimental data obtained from
samples retrieved during construction of Napoli underground, in which AGF was extensively used to excavate
in granular soils and weak fractured rock below the ground water table.
INTRODUCTION
Frozen ground is soil or rock with a temperature below

0 C. The thermo-hydro-mechanical processes induced
by freezing and thawing of pore fluid within soils are
complex and can have significant mutual geotechnical
interaction (Nishimura et al., 2009). As the temperature decreases the ice content increases and the ice
becomes a bonding agent between soil particles or
blocks of rock increasing the strength of the soil/rock
mass and modifying the pore water pressures and the
effective stress on the soil skeleton, which, in turn,
induces mechanical deformation. At the same time,
any changes in the hydraulic and mechanical boundary
conditions can affect the thermal processes by advection and changes of ice and water contents (Gens,
2010).
Natural freezing occurs seasonally in many areas
of the globe, and can adversely affect the engineering
performance of roads and pavements as ice lens form
and grow. In cold regions, perennially frozen ground
(or permafrost) remains at a temperature below 0 C
continuously from year to year. Several distinct features are associated with perennially frozen ground
such as ice-wedge and thermokarst topography. Engineering considerations require an understanding of the
freezing process, the effects of thawing frozen ground,
seasonal frost heave and settlement and how useful
aspects of frozen ground can be utilized. These include
stability, high strength, and water tightness that can
be utilized in frozen earth structures for construction
purposes (Andersland & Landany, 2004).
On the other hand, man-made or Artificial Ground

Freezing (AGF) is a controllable process and can be
used profitably by civil and mining engineers to temporarily provide structural support and/or to exclude
groundwater from an excavation until construction of
the final lining provides permanent safety. The process
was originally applied mainly to vertical openings,
such as shafts or pits, but, with the increasing ability
to drill and install freezing tubes horizontally, other
excavation works, such as tunnels, were considered.
Besides protecting excavations, AGF has also been
used to stabilize slopes, to sample coarse grained soil,
to construct temporary access roads, and to maintain
permafrost below overhead pipeline foundations and
below heated buildings (Harris, 1995).
AGF is one of the construction techniques that were
extensively adopted during construction of Line 1 of
Napoli Underground, to ensure stability and waterproofing of the platform tunnels and inclined passageways during excavation below the ground water
table through loose granular soils of pyroclastic origin
(Pozzolana) and a fractured soft rock (Neapolitan Yellow Tuff) (Viggiani & de Sanctis, 2009; Cavuoto et al.,
2011; Russo et al., 2012).
AGF was carried out by driving freeze tubes into
the ground parallel to the tunnel length around the
future excavation section, and then circulating a refrigerating fluid into the tubes until the temperature of
the ground around the tubes was below the freezing
point of the groundwater. Freezing was activated with
nitrogen and maintained with brine. The contractor
specified that excavation should be undertaken within
467
1 m thick frozen collars with external surfaces at a

temperature of 10 C. The growth of the frozen body
was monitored with temperature sensors located along
chains parallel to the freeze pipes. Due to the complexity of the works, the construction of the line was
accompanied by an intense programme of monitoring designed to measure and/or control the effects
of construction on adjacent structures, and, for the
extension and completeness of the monitoring, represented a unique opportunity to collect field data on
the performance of AGF.
Several Authors have attempted to back analyse
and interpret different aspects of the freezing process. Viggiani & de Sanctis (2009) analysed transient
heat propagation numerically using the finite element
code ABAQUS. The thermal properties of the soil
were obtained by back analysis of an instrumented
trial field in which the ground temperature around the
freezing holes was measured during cycles of freezing
and thawing. An attempt to predict ground heave on
freezing and subsequent settlements on thawing with
a decoupled approach, was also carried out by finite
element analyses imposing freezing-induced volume
strains to the ground (De Santis, 2006).
Colombo (2010) also tackled the problem of heat
propagation, comparing the results of a theoretical
analyses due to Sanger & Sayles (1979) and those
obtained by finite element analyses. In this case
the thermal properties of the ground were assigned
based on literature data. Both the analytical and FEM
approach were adopted to analyse realistic layout of
freezing tubes similar to those used during construction of Napoli Underground and the results compared
with experimental data.
Papakonstantinou et al. (2012) examined the temperature histories monitored within the ground during
the freezing process numerically, in this case taking
into account the thermo-hydraulic coupling. The influence of a number of parameters, including the spacing
between the freeze pipes and the thermal conductivity
of the ground was investigated to gain an understanding of the thermal behaviour of the ground during
activation of artificial freezing with nitrogen.
The goal of this work is that of studying the freezing process considering the full thermo-hydraulicmechanic coupling; to this purpose a research project
bringing together constitutive modelling, laboratory
tests and field data has been started as an international
co-operation involving UPC Barcelona, Universit di
Roma Tor Vergata and technical personnel, and engineers involved in the design and construction of Napoli
underground. The final goal of the work is that to
be able to model the construction process, thus gaining confidence in the design of other works from the
engineering point of view.
This paper presents the first results obtained
using a fully thermo-hydro-mechanical (THM) model
(Nishimura et al. 2009), calibrated against experimental data obtained under temperature controlled
conditions on Pozzolanas retrieved from the subsoil in
Municipio Station. The testing programme was carried
Figure 1. Suction development at ice water interface.
out by Tecno-in SpA (www.tecnoin.it/en) as part of

the geotechnical investigation for the works of Napoli
underground.
2
CONSTITUTIVE MODEL ADOPTED
The THM model developed by Olivella et al. (1994,

1996) and Gens et al. (1998) for high temperature
problems involving gas phase has been extended to
low temperature problems by Nishimura et al. (2009).
In this case, the gas phase is replaced by a second solid phase representing ice. The formulation
was implemented in CODE_BRIGHT (Olivella et al.
1996).
2.1 Freezing retention model
The freezing storage mechanism, which links the saturation degree of the liquid phase Sl (unfrozen water)
to the thermodynamic properties of the soil, must be
defined.As the temperature decreases, a liquid-ice surface tension, li , develops at the interface of the two
phases, which must be balanced by the difference of
the pressures in frozen and liquid water, Pi and Pl as
schematically shown in Figure 1.
The equilibrium between the liquid water and ice
phases is described by the Clausius-Clapeyron equation, derived from the equilibrium of the chemical
potential between two phases. The integrated form
of the equation using the atmospheric pressure and
a temperature T = 273.15 K as references is given by:
where l is the specific latent heat of fusion, and i and

l are the mass densities of the frozen and liquid water
respectively.
The van Genuchten (1980) equation is used to
represent the freezing retention model:
468
where m is a material constant and P is the ice entry

value. This is related to the air entry value in unsaturated condition as P = Pa li /la where la and li are
the liquid-air and liquid-ice surface tensions, and Pa
the air entry value.
The suction, defined as s = Pi Pl , depends on
temperature, T , and liquid pressure, Pl :
Substituting equation (3) into equation (2) it is

possible to obtain the relation between Sl and T .
Finally, the relative permeability is obtained from
the van Genuchten equation together with the following relation linking the relative permeability to the
saturation degree of the liquid phase:
2.2 Mechanical model

The stress variables adopted in the formulation are the
net stress, ij,n :
and the suction s, see equation (3).

The Barcelona Basic Model (BBM, Alonso et al,
1990) has been extended to describe the essential features of frozen-unfrozen soils where ij,n represent the
net stress and s the suction, to obtain the Barcelona
Freezing Model (BFM)
The yield function is given by:
in which q is the deviatoric stress, pn is the mean net

stress, pn0 is the net pre-consolidation pressure, and
ps0 , k, and M are material constants.
The loading-collapse curve (LC), describing the
evolution of pre-consolidation pressure with suction,
is given by:
Figure 2. Experimental set-up (after Cantone et al., 2006).
LABORATORY TESTS
The tests were performed using a double walled triaxial cell originally developed by Tecno-in SpA which
works under temperature controlled conditions, see
Figure 2.
The cell fluid (glycol), acts also as the refrigerating
medium and is circulated in an inner cylinder containing the sample, while the outer cylinder is kept under
vacuum to limit condensation. The loading ram moves
vertically in a sleeve seal, which is connected to an
LVDT for the measurement of vertical displacements.
The axial load is measured using an external load cell
fixed under the cross bar of the loading frame. Tests are
generally performed at controlled displacement rate.
There are three independent pressure circuits, one controlling the cell pressure and two controlling the pore
water pressure in the sample. A probe placed in the
middle of the samples monitors its temperature. Further details on the experimental set-up are given by de
Sanctis (2007).
The laboratory tests were performed to characterise the mechanical behaviour of the natural soils at
low temperatures. Undisturbed samples were retrieved
from the site of Muncipio station at different depths,
for a total of 10 samples in the granular deposits
(Pozzolanas) and 9 samples from the underlying soft rock (Neapolitan Yellow Tuff). The tests
included unconfined and triaxial compression tests
and indirect extension tests on unfrozen, frozen and
frozen/unfrozen samples.
3.1 Experimental results

Figure 3 shows the stress paths followed by three
samples of Pozzolana in the p:T and p:q planes.
The main phases of the tests are:
where:
and (0), , pc , , and r are all material constants. For

further details the reader is referred to Alonso et al.
(1990), Nishimura et al. (2009) and Gens (2010).
469
drained isotropic compression to mean total stress

p = 200 350 kPa;
freezing to temperatures T = 5 10 C over a
time of about five hours followed by an equalization
stage at constant temperature;
Figure 3. Stress path followed in test in the p T and p q

planes.
axial loading at controlled displacement rate of

106 mm/s for tests TX1 and TX2 and 107 mm/s
for test TX3;
thawing after axial loading under maintained displacement.
Figure 4 reports the results of the three tests in
terms of deviatoric stress, q, volumetric strain, v ,
temperature, T , and axial displacement, , versus
time, t.
The volume strain recorded during all tests indicates a tendency for the sample to increase its volume
during the initial part of the freezing stage. This is followed by an apparent decrease of the volume strain
before reaching an approximately constant value during equalisation. It must be noted, however, that the
water in the drainage circuit freezes and so the external
measurements of volume strain are not reliable below
0 C. Different methods to measure the volume strain
of the sample are being developed to overcome this
problem.
The peak deviatoric stress, qp , obtained from the
shearing stage of the two tests at T = 10 C are
almost the same (qp = 3565 kPa in test TX1, at a cell
pressure pc = 200 kPa, and qp = 3562 kPa for test TX2,
at pc = 350 kPa) indicating that, in this range of stress,
the effect of confining pressure is negligible compared
to the bonding effect of the ice.
The shearing stage of test TX3 was carried out
at a temperature ranging between 6 and 4 C,
see Figure 4(c). In this case, the peak deviatoric
stress, obtained for = 2 mm and at a temperature
T = 6 C, was qp = 1740 kPa, after which the temperature increased up to 4 C and the deviatoric stress
decreased to 1200 kPa.
During the shearing stage the external system for
volume strain measurements records a small contraction of the sample followed by dilation as the peak
deviatoric stress is attained. For the reasons outlined
above, however, the reliability of these measurements
of volume strain is questionable.
4
MODELLING
Table 1 reports the values of the parameters adopted to

define the freezing retention model, together with the
residual degree of saturation of liquid water, Slres and
Figure 4. Experimental results: (a) TX1, (b) TX2, and (c)

TX3.
470
Table 1.
Parameters of Freezing Retention Model (FRM).
P(kPa)
Slres
Kintr (m2 )
5.2
0.366
0.10
1013
0.5
Table 2.
Parameters of Barcelona Freezing Model (BFM).
(0)
pc
MPa
G
MPa
0.02
0.002
0.13
1.3
58
10
1.2
1.3
40
the initial porosity of the soil, n. Table 2 summarises

the values of the parameters of the Barcelona Freezing
Model adopted in the numerical simulations.
The parameters defining the freezing retention
model and the loading collapse curve were calibrated
using literature data obtained by Nicotera (2000) and
Picarelli et al. (2007) on Pozzolanas similar to those
under examination although retrieved from different
sites. Parameters M and G were obtained by backanalysis of the behaviour of unfrozen samples reported
in de Sanctis (2007), while the remaining parameters were calibrated against the experimental results
reported herein.
Axis-symmetric coupled THM analyses were performed to validate the BFM under different temperatures and mean confinement stress, using the mesh
shown in Figure 5. A step of 24 hour of linearly
decreasing temperature to T = 5 or 10 C was
applied to the top and right boundary of the mesh,
both allowed to move during freezing. The initial pore
water pressure in the sample at the beginning of freezing was set to zero, and the top and lateral boundaries
of the mesh were drained. Axial loading was carried
out imposing a constant rate of vertical displacement,
vp , to the top boundary of the mesh.
Figure 6 shows the contours of temperature, T ( C),
liquid water pressure, Pl (MPa), porosity, n, and equivalent degree of saturation Seq (= Sl + (iw /lw )Si ) at
a specific time (t = 14 h) for the simulation of the
freezing stage of test TX3.
The model reproduces correctly the advancement of the freezing front from the boundary of
the sample towards its centre, with a gradient
T /r 1.2/0.019 ( /m), see Figure 6(a). Due to
the decreasing temperature, the liquid water pressure
becomes negative where the freezing front advances,
see Figure 6(b). Also, in the frozen area, there is a
marked increase of porosity induced by phase transformation (from water to ice) coupled with the changes
of liquid water pressure, see Figure 6(c), and a corresponding decrease of liquid water saturation, see
Figure 6(d).
Figure 7 shows the deviatoric stress versus axial
strain curves obtained from the numerical simulation
of the shearing stage of all three tests, together with
Figure 5. Geometry of the finite element mesh used in the

analysis and boundary conditions applied.
Figure 6. Test TX3 Results of numerical simulations:

freezing stage.
the experimental results. Consistently with the experimental conditions, the imposed displacement rate in
the simulation of test TX3 was one order of magnitude less than that adopted in the simulation of tests
TX1 and TX2. Also, the values of the initial unfrozen
over consolidation mean effective stress in the numerical simulation of the tests on different samples were
not the same (pc = 400 kPa for samples TX1 and TX3,
and pc = 330 kPa for sample TX2). This is due to the
different depth of retrieval of the different samples.
The agreement between model predictions and
experimental data is very satisfactory both for the samples tested at the same temperature with two different
confining stress (TX1 and TX2) and for those tested at
the same confining stress at two different temperatures
(TX2 and TX3).
The final thawing stage under maintained displacements was also simulated numerically for all tests.
Figure 8 shows the computed contours of temperature,
T ( C), liquid water pressure, Pl (MPa), and equivalent
degree of saturation, Seq , at a specific time in the simulation of the thawing stage for test TX3. The model
correctly reproduces the advancement of the thawing
471
compression at different temperatures and confinement pressures. The performance of the model is satisfactory during all stages of the test, including drained
compression, freezing, equalisation, axial loading in
frozen conditions, and thawing.
Further work, currently under way, includes modifications to the temperature controlled triaxial equipment to measure volume strains of frozen soil and
change the freezing mechanism such that the freezing front will proceed from the centre of the sample
towards its boundaries, with an effect on the hydraulic
boundary conditions. From the point of view of constitutive modelling modifications to the present formulation are being examined to include the viscous
behaviour of the ice phase, mechanical degradation
on cycles of freezing and thawing, and the adoption of
the Bishop stress as a constitutive variable.
Figure 7. Measured and predicted stress strain behaviour in

triaxial compression.
ACKNOWLEDGMENTS
The financial support of the European Commission for the first author through the Marie Curie
Intra European Fellowship (EU FP7-NuMAGF, grant
agreement 272073) is acknowledged. The authors are
grateful to Studio Cavuoto, Tecno-in SpA and professor Alessandro Mandolini for their technical support
as well as their permission to publish research results.
REFERENCES
Figure 8. Test TX3 - Results of numerical simulations:

thawing stage.
front from the boundary of the sample to its centre

with a gradient T /r 0.01/0.19 ( /m). Due to the
increasing temperature, the liquid pressure increases
as the thawing front advances. The liquid water pressure is higher in the upper part of the sample due to
the axial load transmitted by the top platen.
In this step of increasing temperature (a > 0.03) the
model reproduces quite well the observed behaviour
during axial loading.
CONCLUSIONS AND PERSPECTIVE
This is paper has illustrated the first results of

a research into the fundamentals of frozen soil
behaviour, bringing together constitutive modelling,
laboratory tests and field data. A fully coupled thermohydraulic-mechanical model extended to low temperature problems has been validated in freezing and
thawing against experimental data obtained in triaxial
Alonso, E. E., Gens, A. & Josa, A. (1990). A constitutive

model for partially saturated soils. Gotechnique 40(3):
405430.
Andersland, O.B. & Ladanyi, B. (2004). Frozen ground
engineering. Wiley and Sons Inc., New Jersey.
Cantone, A., De Sanctis, L. & Mandolini, A. (2006). Interventi di protezione degli scavi di gallerie mediante congelamento. Attivit sperimentali nella Stazione Muncipio della Metropolitana di Napoli. Incontro Annuale dei
Ricercatori di Geotecnica, Pisa, Italy.
Cavuoto, F., Corbo, A., Fico, R., De Risi, A. & Giannelli, F.
(2011). La metropolitana di Napoli: la galleria di scavalco
della stazione Toledo. Confronto tra misure dei cedimenti
e analisi numeriche. In: XXIV Convegno Nazionale di
Geotecnica, Napoli:19.
Colombo, G. (2010). Il congelamento artificiale del terreno
negli scavi della metropolitana di Napoli. Rivista Italiana
di Geotecnica, XLIV(4), 4262.
de Sanctis, L. (2007). Protezione degli scavi di gallerie
mediante congelamento. Rapporto tecnico CONSORZIO
IKOTEKNE: 1119.
De Santis, D. (2006). Osservazione sperimentale e analisi
numerica degli interventi di protezione dello scavo con
AGF. Master Thesis, Universit di Roma Tor Vergata (in
Italian)
Gens,A. (2010). Soil-enviroment interactions in geotechnical
engineering. Gotechnique 60(1): 374.
Gens, A., Garcia-Molina, A. J., Olivella, S., Alonso, E. E. &
Huertas, F. (1998). Analysis of a full scale in situ test simulating repository conditions. Int. J. Numer. Anal. Methods
Geomech. 22(7): 515548.
Harris, J.S. (1995). Ground freezing in practice. Thomas
Teldford Ltd, London
472
Nicotera, M.V. (1998). Effetti del grado di saturazione sul

comportamento di una pozzolana del napoletano. PhD
Thesis, Universit di Napoli Federico II (in Italian).
Nishimura, S., Gens, A., Olivella, S. & Jardine, R.J. (2009).
THM-coupled finite element analysis of frozen soil: formulation and application. Gotechnique 59(3): 159171.
Olivella, S., Carrera, J., Gens, A. & Alonso, E. E. (1994).
Nonisothermal multiphase flow of brine and gas through
saline media. Transp. Porous Media 15(3): 271293.
Olivella, S., Gens, A., Carrera, J. & Alonso, E. E. (1996).
Numerical formulation for a simulator CODE_BRIGHT
for thecoupled analysis of saline media. Engng Comput.
13(7): 87112.
Papakonstantinou, S., Anagnostou, G. & Pimentel, E. (2012).
Evaluation of ground freezing data from the Naples subway. Proceedings of the ICE: Geotechnical Engineering.
DOI: 10.1680/geng.10.00099
Picarelli, L., Evangelista, A., Rolandi, G., Paone, A.,
Nicotera, M.V., Olivares, L., Scotto Di Santolo, A.,
Lampitiello, S. & Rolandi, M. (2007). Mechanical
properties of pyroclastic soils in Campania Region.

1st International Workshop on Characterisation and
Engineering Properties of Natural Soils;Singapore 34:
23312383.
Russo, G., Viggiani, C. &Viggiani, G.M.B. (2012). Geotechnical design and construction issues for lines 1 and 6 of the
Naples underground. Geomechanik und Tunnelbau 5(3):
300311.
Sanger, F.J. & Sayles, F.H. (1979). Thermal and rheological
computations for artificially frozen ground construction.
Engineering Geology, 13: 311337
van Genuchten, M. Th. (1980). A closed-form equation for
Soil Sci. Soc. Am. J. 44: 892898.
Viggiani, G.M.B. & de Sanctis, L. 2009. Geotechnical aspects
of underground railway construction in the urban environment: The examples of Rome and Naples. Geological
Society Engineering Geology Special Publication 22(1):
215240. www.tecnoin.it/en/www.studiocavuoto.com/
main/intro/intro.php
473

Numerical study of the heating-cooling effects on the geotechnical

behaviour of energy piles
A. Di Donna, F. Dupray & L. Laloui
Swiss Federal Institute of Technology, EPFL, Lausanne, Switzerland
ABSTRACT: Energy geostructures represent a renewable and clean source of energy which can be used for
heating and cooling of buildings and de-icing of infrastructures. This technology couples the structural role of
geostructures with the energy supply, using the principle of shallow geothermal energy. Heat is extracted from
the ground during winter and injected into the ground during summer. This represents an additional thermal
loading, seasonally cyclic, which is imposed to the soil and the structure itself. In this paper, the influence of
temperature on the different aspects involved, accordingly to the Eurocode 7, in the geotechnical design of deep
foundations is analysed. This includes mainly the admissible displacements, the structural failure (concrete)
and the geotechnical capacity. The numerical results show that the thermal load has an effect both in terms of
piles displacements and bearing capacity which are acceptable in normal working conditions but deserve to be
considered in the design practice.
INTRODUCTION
Energy geostructures are spreading rapidly in Europe

and all around the world. They represent a renewable and clean source of energy which can be used
for heating and cooling of buildings, infrastructures,
stations, etc. This technology couples the structural
role of geostructures with the energy supply, using
the principle of shallow geothermal energy. Practically, a certain number of pipes are embedded into
the concrete structures with a heat carrying fluid
which circulates through them and exchanges heat
with the ground. The pipes are then connected to a
heat pump system, which circulates the fluid in the
heating-cooling plant of the building. This system
allows extracting heat from the ground during winter to satisfy the heating needs of the buildings and
injecting heat in the ground during summer, to satisfy
the conditioning needs. Ideally, the heat injected into
and extracted from the ground should balance over an
annual cycle, thus maintaining the average soil temperature constant at the long term. The efficiency of the
system depends on the soil properties, but the ground
can generally be used as a heat storage medium. In
this sense, the heat injected during summer can be
re-used during winter. The advantage of this technology is that it incorporates the geothermal equipment
inside the geostructures which are already required
for stability reasons, thus reducing the initial costs of
construction with respect to other geothermal installations. Since the beginning of 1980s geothermal energy
has been increasingly obtained from foundation elements (Brandl, 2006). As the energy geostructures
represent a new engineering technology and they are
becoming more and more common, an improved scientific knowledge of their behaviour is necessary. In
the recent years, several efforts have been devoted
to investigate and optimize the energy performance
of such structures (Pahud, 2002). With regards to
their mechanical behaviour and geotechnical design,
efforts have been devoted to the in-situ characterization (Laloui et al., 2003; Bourne-Webb et al., 2009;
Amatya et al., 2012), the numerical analysis (Laloui
et al., 2006, Dupray et al., 2013) and the development of design tools (Knellwolf et al., 2011). An up
to date state of the art on the subject can be found
in Laloui and Di Donna (2011). The goal of this
paper is to investigate numerically the behaviour of
energy pile foundations, focusing on their long term
response to a seasonally cyclic thermal loading. For
this purpose, a finite element modelling of an energy
pile foundation is presented and the results obtained
after the application of a seasonally cyclic thermal
loading, are discussed. In the first part of the paper,
the considered mathematical formulation to solve the
problem is introduced. Then, the constitutive model
used to reproduce the thermo-mechanical behaviour
of the soil is described, with particular attention to
the thermal cyclic aspect. In the following sections, the
geometry and the materials properties chosen for the
model are presented, as well as the applied mechanical and thermal loading. Finally the results are discussed, focusing mainly on the effects of temperature
changes on different aspects involved in the geotechnical design of energy piles.Accordingly to the Eurocode
7 (BS EN 1997), both the design at the Ultimate
Limit State and at the Serviceability Limit State are
studied.
475
MATHEMATICAL FORMULATION
where k is the permeability tensor. The energy conservation equation reads:
The considered problem is represented by a concrete

pile foundation subjected to a mechanical (vertical)
load and able to exchange heat with the surrounding
ground. Both the concrete and the soil are considered as porous materials composed by a solid and
a liquid phase. The whole medium is considered to
be fully saturated of water. The concrete behaviour
is modelled with a thermo-elastic response, while the
soil behaves accordingly to a cyclic thermoelasticthermoplastic constitutive model. Hence, three main
aspects are involved in the problem, which are the
mechanical, the thermal and the hydraulic ones. The
mechanical aspect is represented by the stress-strain
response of the soil and the piles, the thermal one by
the energy exchanged between the foundation and the
ground, and the hydraulic one by the presence of water
in the soil. These three aspects are influenced by each
other as, for instance, the solid volume variations are
affected by temperature, the heat exchanged depends
on the presence of water flow, the water density varies
with thermal loading and the mechanical response of
a porous material depends on the pore water pressure
(effective stress concept). A coupled Thermo-HydroMechanical (THM) formulation for porous media is
needed to correctly analyse such problem. The software used for the work presented in this paper is
called LAGAMINE and the coupled THM formulation
implemented in it was developed by Charlier (1987)
and Collin (2003). It is summarised in the following.
2.1
Governing equations
The equilibrium and balance equations, as well as the

water and heat diffusions, are expressed in the moving
current configuration through a Lagrangian-updated
formulation. The equilibrium equation reads:
where div denotes the divergence, : the tensorial product, D the constitutive tensor, the strain
tensor which depends on the solid displacements us ,
grad the gradient, pw the pore water pressure, the
total density (combining the water and solid densities,
respectively w and s ) and g the gravity. The mass
conservation equation reads:
where t denotes the time derivative, n the porosity,

w the compressibility of water, T the temperature
(supposed to be the same in each phase), w and s
the volumetric thermal expansion coefficients respectively of the water and of the solid and urw the relative
velocity of water with respect to the solid. This latter
is expressed by the Darcys law, as:
where c is the soil specific heat (combining the solid

and water components, cps and cpw ), the soil total
thermal conductivity (combining the solid and water
components, s and w ).
2.2 Mechanical constitutive models
From the constitutive point of view, the piles and
the slab are made with concrete and are supposed to
behave thermo-elastically. The soil behaviour is simulated by a cyclic thermoelastic-thermoplastic constitutive model called ACMEG-T (Laloui & Franois,
2009). This model allows considering also the thermal cyclic effects on the response of the material.
It belongs to the Cam-Clay family and is based on
the critical state theory. The isothermal part is based
on the works of Hujeux (1979). Various successive
improvements were made to extend it to the nonisothermal conditions: Laloui (1993), Modaressi and
Laloui (1997), Laloui and Cekerevac (2008). Accordingly to the elasto-plasticity theory, the tensor of the
total strain increment d is decomposed into elastic
de and plastic dp components, so that:
In the elastic domain and in non-isothermal conditions, the increment of volumetric deformation dev is
equal to:
where p is the mean effective stress and K the bulk

modulus. The deviatoric component of the elastic
deformation ded is not affected by temperature, thus it
reads:
where q is the deviatoric stress and G the shear modulus. The two elastic modula are non linear functions
of the mean effective stress, as:
where Kref and Gref are the two modula at the reference mean effective stress pref and ne is a material
parameter. The plastic response is described by twomechanisms, one isotropic and one deviatoric, which
are coupled together. A pure isotropic loading causes
only volumetric plastic deformation, while a pure deviatoric loading causes both deviatoric and volumetric
476
plastic deformation. Both of them are temperature

dependent, so that when a thermo-mechanical load
leads the stress point on one of these two surfaces thermal plastic deformation (contraction) is developed.
The isotropic yield limit fiso reads:
where pc0 is the pre-consolidation pressure at initial

temperature T0 and before plastification, the plastic
coefficient, T a material parameter defining the shape
of the thermal yield limit and riso the degree of mobilization of plasticity of the isotropic mechanism. This
latter evolves with the volumetric plastic deformation,
as:
Figure 1. Geometry and boundary conditions of the model.
where reiso is its initial value, c is a material parameter

p,iso
and v the volumetric plastic deformation induced
by the isotropic mechanism. The deviatoric yield limit
fdev reads:
where M is the slope of the critical state line (CSL)

in the p -q plane, b and d two material parameters and
rdev the degree of mobilization of plasticity for the
deviatoric mechanism. This latter parameter has the
same role than riso and reads:
where redev is its initial value, a is a material paramp

eter and d the deviatoric plastic deformation. The
flow rule is associated for the isotropic mechanism,
but not associated for the deviatoric one. This means
that, calling the isotropic and deviatoric plastic potentials respectively giso and gdev , giso = fiso but gdev = fdev
and:
where is a material parameter which expresses the

dilatancy rule, as
In order to take into account the cyclic effects, the

degree of plastification for the isotropic yield mechanism riso evolves during both mechanical and thermal
cycles. In particular, during thermal unloading (cooling) it decreases, even if the response is perfectly
elastic. Therefore, during each re-heating phase, the
degree of plastification for the isotropic mechanism
reads:
where pcyc is the pre-consolidation pressure at the end

p,iso,cyc
of the previous cooling phase and v

the volumetric plastic deformation induced by the isotropic
mechanism starting from the last re-heating.
NUMERICAL MODEL
The modelled foundation is composed of a slab with

7 rows of 15 energy piles each. The piles have a diameter of 80.0 cm, a length of 20.0 m and spacing of
7.0 m in both directions. The slab is 0.5 m thick and
110 m long. For the sake of simplicity and thanks to
the symmetry of the studied case, the numerical model
is limited to only 4 piles in 2D plane strain conditions
(Figure 1). This latter assumption involves the fundamental simplification of considering a circular pile as
an infinite wall, in the plane perpendicular to the one of
the simulation. From a mechanical point of view, as the
response of vertically loaded piles is controlled mainly
by their axial stiffness, this transition from 3D to 2D
plane strain conditions is taken into account by considering, for the piles, an equivalent Youngs modulus
477
Table 1. Water and heat diffusion parameters.
k[m/s]
n []
w [ C1 ]
w [W/mK]
cpw [J/KgK]
s [W/mK]
cps [J/KgK]
Table 2.
Table 3.
Parameters of the ACMEG-T model.
Soil
Concrete
Elastic Parameters
Plastic Parameters
108
0.39
2 104
0.6
4186
2.5
930
109
0.12
2 104
0.6
4186
2.2
930
Kref [MPa]
Gref [MPa]
pref [MPa]
ne [-]
s [K1 ]
M [-]
[-]
[-]
a [-]
b [-]
c [-]
d [-]
riso [-]
rdev [-]
T [-]
Properties of concrete.
83.3
38.5
1.0
1.0
1 105
1.42
10.0
1.0
0.001
1.0
0.02
2.0
0.3
0.3
0.35
Concrete
K [GPa]
G [GPa]
s [kN/m3 ]
s [ C1 ]
cyclic thermal load is applied in terms of injected and

extracted power from each pile.The maximum injected
and extracted power is 150 W per m length of pile.
23.3
14.0
21.8
3.6 105
4
Eeq , computed, as suggested by Prakoso and Kulhawi
(2002), as:
where np,row is the number of piles in a row in the third

direction (15 in this case), Ap the pile base area, Ep
the pile Young modulus, Lr the length of the slab in
the third direction (110 m in this case) and D the pile
diameter. For the other parameters, such as the porosity, the permeability and the thermal conductivity and
capacity, a weighted average over the soil and concrete
properties is assumed for the piles. Also, in order to
take into account the fact that the yield and plastic
potential surfaces of the ACMEG-T model are circular in the deviatoric plane (i.e. M does not depend on
the Lode angle), an equivalent friction angle (M) for
plane strain conditions is used in the simulations (Potts
and Zdravkovic, 1999).The boundary conditions of the
model are represented in Figure 1. The initial temperature of the soil and the piles is 11 C and the initial
pore water pressure is assumed to be hydrostatic with
the water table at the surface level. The temperature of
the slab is fixed at 15 C throughout the computation,
in order to take into account the regulated temperature
of the over building. The parameters used for the heat
and water flow problem are collected in Table 1. The
mechanical properties used for the piles and the slab
are summarized in Table 2. The ACMEG-T parameters used to simulate the soil are collected in Table 3.
The soil mass is divided into five layers with the same
mechanical properties but different preconsolidation
pressures to approach the normally consolidated conditions. The interface between the pile and the soil is
modelled with a thin layer of elements that behaves
accordingly to the ACMEG-T model. A mechanical
load of 1700 kN is applied on each pile and kept
constant thought out the computation. A seasonally
The first step of the calculation is represented by the

application of the mechanical load on the piles. The
maximum displacement induced in this phase is about
2.5 cm at the head of the first pile on the left (centre
of the foundation) and almost zero far from the foundation. Besides, a differential displacement between
the centre of the foundation and the external pile
(the first on the right) is shown. However, it is lower
than 0.5 cm and this is in accordance with the limits
imposed by the Eurocode 7 (BS EN 1997). The second and main part of the simulation is represented by
the application of a seasonally cyclic thermal loading.
The main purpose is to investigate the effects of temperature variations on the different aspects involved in
the geotechnical design of such a foundation. Accordingly to the Eurocode 7, a pile foundation subjected to
an axial (compressive or tensile) load must be designed
accounting for its bearing resistance (geotechnical stability) and structural failure (compression and tension
of concrete) at the Ultimate Limit State (U.L.S.) and
excessive displacements at the Serviceability Limit
State (S.L.S.). All these aspects are analysed in the
following, considering the additional imposed thermal
loading. First, the discussion focuses on the thermal
field applied by the piles on the surrounding soil and
the long term thermal equilibrium of the ground. This
allows to quantify the thermal load applied and to link
it with the induced effects described in the following
part of the paper. Later, the deformation of the soil
under cyclic thermal loading and the consequent displacements of the foundation are discussed (S.L.S.).
This aspect represents the first key point of the paper,
as in the design practice it is often one of the most
constraining limitation when dealing with deep foundations. In the last part, the state of stress in the soil,
in the piles and at the interface between them is investigated. Particular attention is paid to the long term
distribution of the load transmitted between the lateral surface and the pile tip during the cyclic thermal
478
Figure 2. Evolution of temperature in the soil.
Figure 3. Displacements of the piles during thermal cycles.
loading. This represents the second key point of the

paper (U.L.S.).
4.2 Soil strain and foundation displacements
4.1 Temperature field

The power injected and extracted from the piles generates a heat storage volume all around them, whose
dimension depends on the injected-extracted power,
the soil permeability, the presence of water flow and
the ground thermal properties. In this case study, there
is no imposed water flow, the soil has a relatively
low permeability of 108 m/s (silty clay) and a good
thermal capacity, so that the ground represents a heat
storage medium: the heat is stored during summer and
extracted during winter preserving the thermal equilibrium of the ground. The evolution of temperature
in the soil between the second and the third pile (at
3.5 m from both) at different depths is presented in
Figure 2. The temperature variations become negligible starting from about 40 m depth. The imposed
injection-extraction ratio equal to one allows maintaining the mean soil temperature constant at the long
term and almost equal to the natural one (i.e. the one
initially imposed of 11 C). After the first 5 years,
the temperature in the most thermally solicited zone
(see the point at 10 m depth) oscillates among a maximum value of 17 C during summer and a minimum
value of 10 C during winter. The points at 1 and 20 m
depth are influenced respectively by the slab temperature (which is kept constant at 15 C) and thermal
equilibrium with the soil below. Also, the temperature is almost uniform along the piles, except at the
extremes, where it must satisfy the equilibrium with
the soil below and the slab above. After the first heating and cooling phases, the maximum and minimum
temperatures reached inside the piles are 27 and 4 C
respectively. However, at the long term the pile temperature stabilizes between a maximum value of 25 C and
a minimum one of 2 C. This reduction with respect to
the first cycle is due to the thermal losses in the soil,
taken into account in the computation by the second
and third terms equation (4).
Accordingly to the ACMEG-T model, the response

of the soil is thermoelastic-thermoplastic. This means
that its thermal deformation has an elastic component (reversible), which is dilative during heating and
contracting during cooling, and a plastic component,
which is contracting during heating and irreversible.
The entity of each component depends on the initial
state of the soil, i.e. on its over-consolidation ratio,
and on the soil properties. The thermal deformation
of the soil results in additional displacements of the
foundation. From the design point of view, this has to
be included in the analysis at the Serviceability Limit
State (S.L.S). The long term thermal contraction of
the soil induces additional settlements of the piles and
consequently the over-structure. Besides, there is the
elastic component (reversible) of the thermal deformation of both the piles and the soil which makes the
foundation moving upward during heating and downward during cooling. Figure 3 shows the displacement
of the slab, at the head of the central and external piles.
The additional irreversible displacement induced by
the cyclic thermal loading is about 0.3 cm and almost
completely developed during the first five years. Its
magnitude depends on the initial state of stress and
the material parameters introduced for the soil. Consequently it may vary for each specific case. However,
the parameters used for this simulation are realistic and
in this specific case the thermal irreversible settlement
is about the 12% of the mechanical one. The elastic
component of the thermal displacement is represented
by the amplitude of the oscillations of the curve represented in Figure 3. It is about 0.7 cm for all the
piles. Also, it is worth mentioning that the differential
displacement of about 0.5 cm corresponds to the one
induced by the mechanical load. The thermal cyclic
loading does not induce any additional differential settlement because all the piles are heated and cooled
equally and together. A non-uniform thermal loading
would lead to differences, in terms of soil and piles
deformation, between the most and the less heated
479
Figure 4. State of stress in the soil between the piles at

17 m depth during thermal cycles.
zones. This would consequently result in differential

settlements for the over-structure.
4.3
Stress state in the soil
It is known that the higher thermal expansion coefficient of water with respect to the solid particles
induces an excess pore pressure during heating, whose
entity depends on the soil properties (i.e. permeability) and heating speed. If the soil is permeable and
the temperature is increased slowly enough, the heating phase approaches the drained conditions and the
pore pressures have time to dissipate. In the studied
case, the pore water pressure oscillates accordingly
to the thermal cycles, with negligible annual amplitude of less than 3 kPa, and remains constant cycle
after cycle. The low thermal-induced excess pore water
pressure is coherent with the temperature variation
imposed and the permeability of the soil (Table 1).
It is possible to state that the heating phase occurs
in almost drained conditions, which is likely the case
in all energy pile foundations. Coherently with the
evolution of pore water pressure, the vertical and the
horizontal effective stresses decrease during heating
(Figure 4). The excess pore pressure is then dissipated
during the cooling phase and accordingly, the effective stresses re-increase. Interestingly, the horizontal
stress (and also the vertical one, even if less significantly) decreases during the first 5 cycles. This results
from the thermal collapse of the soil (thermo plasticity). In other words, the irreversible soil contraction
induced by heating leads to a reduction of the confinement during the first cycles. This effect stops when no
more plasticity is generated, i.e. after about 5 years.
4.4
Structural and geotechnical capacity of the

piles
An energy pile behaves similarly to a concrete bar subjected to an axial load and a thermal solicitation. If it
is free to deform, it expands when heated and contracts when cooled. More likely the surrounding soil,
the structural rigidity at the head and the eventual stiff

layer at the toe (end bearing piles) represent a partial
constrain to its free thermal dilation. Thus, the portion of strain which is prevented generates additional
thermal stresses in the pile. The magnitude and distribution of the additional axial stresses depend on
the entity of restraints (Bourne-Webb et al., 2009,
Amatya et al. 2012). In an idealized framework, when
a heating phase is applied, the pile tends to expand,
and any restraint offered to this movement will lead
to additional compressive stresses. Conversely, during cooling, the pile tends to contract and any restrain
will lead to additional tensional stresses. In practice,
the additional stresses can be contractive or dilative
depending also on the global response of the foundation and the interaction with the other piles. Generally
speaking, it is worth mentioning that from a structural
point of view the stresses transmitted inside a concrete
pile do not represent a significant risk for the stability
of the foundation. In effect, deep foundations are able
to support high internal stresses, while the stability
problems might arise rather in terms of geotechnical
bearing capacity. Assuming a concrete 28 days compressive strength fcu of 25 MPa, the piles analysed in
this study have a structural capacity (in compression)
of 12 MN, which is largely sufficient to carry both the
mechanical and the thermal load applied.This is thanks
to the high resistance that concrete shows in compression. Conversely, the concrete is not able to support
tension, so a possible issue could be represented by
the eventual tensile stresses, which are more likely to
develop during the cooling phase. However, in normal conditions, the thermal induced tensile stresses are
unlikely to be high enough to overcome the compressive ones induced by the mechanical load. So, besides
the issues related to the displacements, the most constraining aspect which determines the maximum load
that a pile is able to carry is represented by its geotechnical capacity. The maximum load that a pile is able to
support can be split into a lateral Qs,max and a tip Qb,max
components, computed as (Lancellotta and Calavera,
1999):
where H is the pile height, h the horizontal effective stress normal to the pile-soil interface at depth z
and angular coordinate , the friction angle at the
interface, R the pile radius, Cu the undrained shear

resistance of the soil, r the radial coordinate and vb
the vertical effective stress at the pile base. In a general
way, the different parameters involved (i.e. and Cu )
might be affected by the changes in temperature and
this would have an effect on the final resistance. More
likely, coherently with Figure 4, the most significant
aspect is the effect of heating and cooling cycles on the
state of stress at the contact between the pile and the

soil (h and vb
in equations (17) and (18)). Finally,
480
Figure 5. Evolution of the load transmitted through the

lateral surface and the pile tip during thermal cycles.
it is generally recommended (SIA D0 190) that the

freezing of the interface is to be avoided to prevent the
consequent lose of lateral friction. In the analysed case,
the minimum temperature remains slightly higher than
zero so this problem does not arise. Running the same
calculation with a maximum injected-extracted power
of 175 W/m instead of 150 W/m induces a minimum
interface temperature of 0 C or less. For this reason it
is important to analyse each specific case and do not
exceed the maximum possible thermal exploitation.
Except the freezing problem and with the purpose to
investigate the effect of a thermal cyclic loading on the
geotechnical capacity of the energy piles modelled, the
load transmitted along their interface Qs and at their
base Qb during the heating and cooling cycles has been
computed starting from the numerical results as:
where is the shear stress at the interface between the

piles and the soil. The evolution of Qs and Qb during
the heating and cooling cycles is presented in Figure 5. Initially, before the application of the thermal
load, the 70% of the mechanical load is supported laterally (about 1200 kN) and the remaining 30% at the
tip (about 500 kN). Then, the portion of load transmitted along the shaft reduces cycle after cycle during
the first 5 years. This is mainly due to the reduction
of the normal stress to the pile-soil interface (h in
equation (17)). The portion of load that can no more
be supported laterally is thus transmitted though the
base and Qb increases during the first 5 years. After
five years (stabilization) the portion transmitted at the
interface is 66% and the remaining passes through the
base.
CONCLUSION
The main goal of this paper is to investigate the effects

of the thermal cyclic loading which is applied to energy
piles and to the surrounding soil, in addition to the
structural mechanical loading, in terms of stresses and
displacements at the long term. The imposed temperature variations can affect both the Ultimate and the
Serviceability Limit State. This is mainly due to the
influence that the temperature has on the soil properties and on the interface behaviour. Several years of
seasonally cyclic thermal loading on an energy pile
foundation were studied numerically and the results
discussed. The soil behaviour is reproduced with a
thermoelastic-thermoplastic model. As a consequence
during heating it might show elastic (dilative and
reversible during cooling) or elasto-plastic (contracting and partially irreversible during cooling) deformation depending on the specific conditions. The former
induces seasonally cyclic vertical displacements of the
foundation while the latter induces an irreversible settlement of the structure itself. Moreover, the model
is able to consider also the cyclic accommodative
thermal aspect on the soil behaviour, so that plastic
deformation is not only developed during the first thermal cycle but also during a certain number of cycles
after the first one. In the case considered the additional irreversible settlement induced by the thermal
loading is about 0.3 cm (12% of the one induced by
the mechanical load) while the amplitude of the elastic cyclic displacement of the piles is about 0.7 cm
(Serviceability Limit State). From the structural failure point of view, the heating and cooling phases result
respectively into additional compressive and tensile
stresses inside the piles, which are due to the partial
constrains represented by the surrounding soil and the
slab to the free thermal deformation of the foundation.
However, the thermal induced stresses are negligible
with respect to the concrete resistance (Structural Ultimate Limit State). Regarding the soil state of stress, the
pore water pressure increases slightly during heating
and decreases slightly during cooling, among constant bounds. Coherently, the vertical and horizontal
effective stresses decrease during heating and increase
during cooling. Moreover, the soil plastic contraction
developed during the first thermal cycles induces a
reduction of the confinement cycle after cycle. This
effect is shown not only in the soil mass but also at the
pile-soil interface and leads to a reduction of the normal load to the pile shaft, which results in a decrease of
the pile lateral bearing capacity.The portion of external
load which is initially transmitted thought the interface
reduces cycle after cycle during the first 5 years (from
70 to 66%) and it is transmitted thought the base of the
pile (Geotechnical Ultimate Limit State).
ACKNOWLEDGEMENTS
This research project was funded by the Swiss Federal
Office of Energy (contract Nb. 154426).
481
REFERENCES
Amatya, B.L., Soga, K., Bourne-Webb, P.J. and Laloui,
L. 2012. Thermo-mechanical behaviour of energy piles.
Bourne-Webb, P.J, Amatya, B., Soga, K., Amis, T., Davidson,
C., and Payne, P. 2009. Energy pile test at Lambeth College, London: geotechnical and thermodynamic aspects
of pile response to heat cycles. Gotechnique 59(3):
237248.
Brandl, H. 2006. Energy foundations and other thermo-active
ground structures. Gotechnique, 56(2): 81122.
BSI 2004. BS EN 1997-1. Eurocode 7: part 1, section 7.
British Standards Institution, London.
BSI 2007. BS EN 1997-2. Eurocode 7: part 2. British
Standards Institution, London.
Charlier, R. 1987. Approche unifie de quelques problmes
non linaires de mcanique des milieux continus par la
mthode des lments finis. PhD thesis, Universit de
Lige.
Collin, F. 2003. Couplages thermo-hydro-mcaniques dans
les sols et les roches tendres partiellement saturs. PhD
thesis, Universit de Lige.
Dupray, F., Laloui, L. and Kazabgba, A. Understanding
the thermo-hydro-mechanical behaviour of seasonal heat
storage in an energy pile foundation. Submitted.
Hujeux, J.C. 1979. Calcul numrique de problmes de consolidation lastoplastique. PhD Thesis, Ecole Centrale,
Paris.
Knellwolf, C., Peron, H. and Laloui, L. 2011 Geotechnical
analysis of heat exchanger piles. Journal of Geotechnical and Geoenvironmental Engineering, ASCE 137(10):
890902.
Laloui, L. 1993. Modlisation du comportement thermohydro-mcanique des milieux poreux anlastique. PhD
Thesis, Ecole Centrale de Paris.
Laloui, L. and Cekerevac, C. 2008. Non-isothermal plasticity
model for cyclic behaviour of soils. International Journal
for Numerical and Analytical Methods in Geomechanics,

32(5): 437460.
Laloui, L., and Di Donna, A. 2011. Understanding the
Thermo-Mechanical Behaviour of Energy Piles. Proceedings of ICE Civil Engineering 164, 184191.
Laloui, L., and Franois, B. 2009. ACMEG-T: soil thermoplasticity model. Journal of Engineering Mechanics,
ASCE, 135(9): 932944.
Laloui, L., Moreni, M. and Vulliet, L. 2003. Comportement
dun pieu bi-fonction, fondation et changeur de chaleur.
Canadian Geotechnical Journal, 40(2): 388402.
Laloui, L., Nuth, M. and Vulliet, L. 2006. Experimental
and numerical investigation of the behaviour of a heat
exchanger pile. International Journal for Numerical and
Analytical Methods in Geomechanics, 30(8): 763781.
Lancellotta, R. and Calavera, J. 1999. Fondazioni.
McGraw_hill eds.
Modaressi, H. and Laloui, L. 1997. A thermo-viscoplastic
constitutive model for clays. International Journal for
Numerical and Analytical Methods in Geomechanics,
21(5): 313315.
Pahud, D. 2002. Geothermal energy and heat storage.
SUPSI-DCT-LEEE, Laboratorio di Energia, Ecologia ad
Economia.
Potts, D. M and Zdravkovic, L. 1999. Finite element analysis in geotechnical engineering: theory. Thomas Telford
Limited eds.
Prakoso, W. A., and Kulhawy, F. H. 2002. Contribution to
piled raft foundation design. Journal of Geotechnical and
Geoenvironmental Engineering, ASCE 127(1): 1724.
SIA DO 190 2005. Utilisation de la chaleur du sol par
des ouvrages de fondation et de soutnement en bton.
Guide pour la conception, la ralisation et la maintenance. Socit Suisse des ingnieurs et des architects.
Documentation SIA DO 190.
482

Effect of groundwater flow on the THM behavior of an energy pile

F. Dupray, M. Baehler & L. Laloui
cole Polytechnique Fdrale De Lausanne (EPFL), School of Architecture, Civil and Environmental Engineering
(ENAC), Laboratory for Soil Mechanics (LMS)
ABSTRACT: Bridges are one of the road structures that are most exposed to icing, a problem that is widely
dealt with through salting, a technique that reduce the lifespan of the bridge. Seasonal storage of solar heat
energy captured directly through the asphalt layer is a promising solution that can be achieved cost-effectively
only if a necessary geostructure is used as heat-exchanger. Bridge piles and abutments are founded on piles that
can be used as heat exchangers. Seasonal storage and natural thermal reload are two possible solutions for the
operation of such a system, depending on local conditions. In particular, the presence of groundwater flow is
thought to be a significant factor in such a design and is taken into account. The study aims at determining the
geotechnical and energy design parameters through thermo-hydro-mechanical numerical simulations. A single
heat-exchanger pile is taken as an example in this study. Given the presence of water flow, a three-dimensional
FEM simulation needs to be carried out. The soil is described as a thermo-elastic porous medium, and various
groundwater flow scenarios are studied. The capture of energy and the needs for deicing are based on the few
existing structures that use other means of energy exchange with the ground. The article proposes an overview
of all coupled phenomena, between thermal, hydraulic and mechanical aspects. Results show that only specific
sites can be considered for using heat-exchanger piles for deicing bridges, but that the efficiency of the solution
in those sites is high. Possible stability problems, in terms of foundation and structure are also considered, such
as vertical displacements due to the dual use of the foundation piles.
1
1.1
INTRODUCTION
Principle and history
The traditional way of tackling the problem of snow

and ice on road infrastructures is to use snow plows and
salts. For the removal of snow, snow plows are still the
only widely usable technique, hot water melting being
limited to few potential sites close to geothermal wells
(Lund, 1999, Wang et al., 2008). But ice formation is
different in that it can be more dangerous but is more
prone to appear in specific sites instead of over the
whole road network. Bridges, which by definition do
not benefit from contact with the ground, are especially
sensitive to ice formation, a problem which is only
solved by heavy use of salts that reduce the freezing
point of water locally. This has been identified as causing two kinds of problems, one for the environment, the
second for the durability of the structure. One potential
technique for the reduction of the use of salts in those
situations is to use the resource of shallow geothermal
energy, i.e. energy that can be extracted from shallow ground, and which origin is mainly unrelated to
natural thermal gradient. Energy sources can be solar
energy gathered through asphalt or energy transported
by groundwater flow. Previous full-scale experiments,
such as the SERSO project in Switzerland (operational
since 1994) have proven that it is possible to store solar
energy in the ground during summer and to retrieve it
during winter, but at a high initial cost due to the need

for a specifically built storage volume (Hopkirk et al.,
1994). This article investigates the capability of a less
expensive device, energy pile, in its possible use as
the foundation for pillars and abutments of bridges.
Due to the combined use as structural foundation and
heat-exchanger, energy piles necessitate a lower initial cost for comparable running costs (Pahud and
Hubbuch, 2007).
1.2
Content of the study
The goal of the study is to evaluate the energy that

can be retrieved for bridge surface heating through
energy piles that are used in small groups, as would be
the case under the foundation of pillars or abutments.
Such energy geostructures are expected to behave
quite differently from large energy storage systems
in the ground, because of their geometry and reaction to heat exchange. Groundwater flow has been
identified as a major factor in the behavior of heat
stores. Therefore the study includes its effect on the
heat exchange. The existing ways of quantifying the
energy needs for deicing of bridges are compared with
the values of retrievable energy from the considered
energy foundation in order to evaluate the feasibility
of the system.
It is also interesting to analyze the trends that
can be expected from such a foundation in terms of
483
geotechnical response: stress changes in the pile and

the soil, and possible displacements of the foundation.
In order to analyse all these problems, a thermo-hydromechanical finite-element simulation is performed on
a case study.
A case study was devised for preliminary investigation on the convection effect due to underground
water flow in relationship to both heat storage and
mechanical response of the pile. This consists in a single pile that is used as heat exchanger in both heating
and cooling regimes. Four cases are studied: one without groundwater flow and with heat storage, one with
moderate flow and heat storage, one with moderate
flow and without heat storage and one with important
flow without heat storage. The thermal responses of
the cases are compared as well as the corresponding
mechanical behaviors.
2
value is in the order of 60 to 200 kWh/m2 /yr depending

on local conditions.
2.2 Possible solar energy retrieval

The other side of the problem is to evaluate the energy
that is available and renewable on a yearly basis. As
has been said, two mains sources can be identified:
groundwater flow, assumed at constant temperature,
and solar energy, ideally retrieved from the bridge itself
and stored seasonally. The retrievable solar energy
through piping located under asphalt is found to be
in the order of 20% of the total irradiated energy
(Bobes-Jesus et al., 2013), which in Switzerland is sufficient to cover winter needs, total irradiated energy
on horizontal surface being generally higher than
1000 kWh/m2 /yr.
EVALUATION OF THE DEMAND

2.3 Possible groundwater flow
2.1 Winter needs

Various papers (Lund, 1999, Liu et al., 2007) have
reviewed the only existing handbook for the evaluation
of heat requirements for the deicing and snow melting
of bridges, the ASHRAE Handbook for HVAC applications (ASHRAE, 1995), and came to the conclusion
that the proposed approach is correct on a seasonal
basis but presents limitations in the transient behavior. As the interest of this study is on the seasonal
behavior, the ASHRAE method is used for the evaluation of the heat needs. Based on meteorological data
for the period of interest (i.e. temperatures lower than
or equal to the set point temperature of the system,
generally 0.5 C), a total amount of energy can be calculated, taking into account the two aspects of snow
melting and temperature control. The general equation
describing surface heating power needed to melt snow
is as follows:
where q0 is the total surface heating power needed, qs

is the term related to snow temperature increase, qm
is the term related to latent heat of melting, Ar is the
free area ratio, and qe and qh losses terms related to
evaporation and convection/radiation respectively. qs
is proportional to snowfall rate and temperature while
qm is only proportional to snowfall rate. Evaporation
losses terms are more complex as they involve wind
speed and relative humidity of the atmosphere.
The heating power needed to maintain the temperature of the water film at a set value (generally 0.5 C)
in the absence of snow is identical to the term qh in
equation (1), and has been evaluated empirically as
follows:
where V is the wind speed in km/h, Tf the film temperature and Ta the air temperature in C. The resulting
It should be noted that groundwater flow is known to

be detrimental to the operation of solar energy storage systems (Pahud, 2007). The thermal efficiency of
a heat storage system is based on its ability to limit
thermal losses. These losses depend on ground parameters (thermal conductivity, volumetric heat capacity,
and permeability) and also on hydraulic ground conditions (degree of saturation and presence of a water
flux). Seasonal energy storage therefore requires particular ground conditions, such as the absence or small
magnitude of groundwater flow. A numerical study
by van Meurs (1985) concerning a porous medium
with homogeneous hydraulic properties indicates that
a heat storage unit requires a protecting hydraulic
screen if the groundwater flow exceeds 0.05 m per day
(Nordell and Hellstrm, 2000). While thermal conductivity is necessary to spread heat from the piles to
the ground, heat losses from the storage area increase
with increasing thermal conductivity (Schmidt and
Mangold, 2006).
MODELLING OF THE PILE
3.1 Geotechnical context

A context was devised in the study based on a possible
site for such thermo-active bridge in the Rhne valley, Switzerland. A complete geotechnical study was
performed on the real site, and offered precious information regarding the geotechnical characteristics of
the encountered layers, as well as on the nature of these
soils. Knowing the nature of the soils allows the evaluation of their hydraulic conductivity as well as their
thermal characteristics, in the absence of actual measurements. In order to properly identify and analyze the
effects of convection in the process, only the main layer
for the length of the piles was considered. This layer is
made of coarse glacial-alluvial deposits. Though they
are well-graded, their hydraulic conductivity is high at
484
and temperature T through the water bulk modulus w and the volumetric water thermal expansion
coefficient wT :
where w0, pw0 and T0 are the initial values of water

bulk density, pore water pressure, and temperature,
respectively. The liquid phase motion is governed by
Darcys law:
Figure 1. View of the model including dimensions.
3.5 104 m/s (at 20 C). Due to the hydraulic conditions, significant groundwater flow could take place
in this layer, which is between 7 and 20 m thick and
remains saturated.
3.2
Geometry of the problem
A bridge recently built in the area gave indications

regarding the piles that would be designed for purely
geotechnical reasons, and this served as a basis for the
study, in order to determine whether they could be sufficient for the thermal part of the problem as well. Pile
length varied between 20 and 38 m and the average
length was 23 m, which is chosen as the length of the
simulated pile. Chosen pile diameter is 1 m. A view of
the model, including dimensions is proposed in Figure 1. Due to the simulation of groundwater flow in
a non-isothermal model, care was taken in the dimensions of the model so that cooler water does not reach
the boundary during the simulation, which explains the
length of 500 m downwards the pile in the direction of
flow. This eliminates for a defined period the practical
problem encountered at boundaries with the classical
assumption of unique temperature for soil and water.
3.3
General THM formulation
The diffusive model that is used in this simulation was

implemented in the software Lagamine developed by
Collin (2003). A brief description of these diffusive
equations for a water-saturated porous medium is presented here. The equilibrium and balance equations, as
well as the water and heat flows, are expressed in the
moving current configuration through a Lagrangianupdated formulation (Charlier, 1987). The local mass
conservation equation for water is as follows:
where w is the bulk density of water, fl is the macroscopic velocity of water, Qw is the volume source of
water, and n is the porosity.
The conservation and balance equations of the system need to be expressed in terms of primary state
variables pw and T . The liquid water bulk density
is related as follows to the pore water pressure pw
where k is the isotropic intrinsic permeability, and w

is the dynamic viscosity of the water, which depends
linearly on temperature. The hydraulic conductivity is
therefore depending on temperature.
Collin (2003) calculated the mass variation from
eq. (4) when local volume variations are added. Such
a relation can be introduced in eq. (3) and used with
eq. (5) to obtain the conventional mass conservation
equation for a saturated porous medium:
where X is the time derivative of X and v is the

volumetric soil skeleton strain.
The energy balance equation of the saturated soil
has the following form:
where fT is the heat flow, QT is a volume heat source,

and and Cp are the density and specific heat of
the mixture (solid matrix with voids filled by liquid),
respectively. Those parameters are deduced from the
properties of each phase:
where s is the soil grain bulk density, and cp,w and

cp,s are the specific heat of water and solid medium,
respectively.
Heat transport is governed by conduction and
convection:
where is the thermal conductivity of the mixture, as

deduced from the thermal conductivity of each phase:
where s and w are the thermal conductivities of the

solid and liquid phases, respectively.
485
Finally, the mechanical balance equation is simply

written:
where = n w + (1 n) s is the density of the whole

medium and is the material effective stress.
3.4
Characteristics of the model
The numerical simulations are run with FEM code

Lagamine (Charlier et al., 2001). The finite element
mesh is composed of 8022 nodes and 6340 hexahedral linear elements. The fully coupled problem can
be written in terms of solid displacements, liquid pressure and temperature directly from equations (6), (7)
and (11). As can be seen, a plane of symmetry is used
to reduce the models size, and the shape of the pile is
hexagonal. The initial pore water pressure corresponds
to the hydrostatic profile with a water table located at
the surface. Temperature is imposed on all faces except
the surface, which is a no flow boundary. The initial
temperature is 12 C throughout the model, which is
a typical value for Switzerland. The heat exchange is
represented by a heat source/sink in the volume of
the pile.
3.5 Loading path
Three stages corresponding to the three aspects of the
problem are necessary to describe the load path. The
first aspect is linked to the hydraulic behaviour. In
the case of presence of a water flow, it is considered
as perpendicular to the bridge (parallel to the canal).
Such water flow is imposed through a constant water
pressure difference between the sides of the model
which is let equilibrate throughout the model before
the beginning of the mechanical and thermal parts.
Two waterflow values are evaluated, a high value of
0.3 m/d and a low value of 0.015 m/d.
The second aspect is mechanical. Initial stresses due
to gravity are introduced in the model assuming a coefficient of earth pressure at rest of K0 = 1. The loading
is then applied on top of the pile. The individual pile
was designed to withstand a load of 2750 kN, which
is applied as a stress of 3.9 MPa on the surface of
the pile. All subsequent relative values (displacements,
stresses) are based on the reference state at the end
of this stage.
The final aspect is the thermal loading. As has been
noted, two strategies are investigated regarding the
source of energy. In the first case, solar energy absorption is used as the heat renewal source. This means that
during summer months, heat is injected from the piles
to the surrounding soil. From June to August, the injection design value is constantly applied, while May and
September are used as ramp-up, resp. ramp-down periods. April and October are not used for heat exchange
in that case.
Then for both cases, a similar strategy is used for
heat extraction from the ground to the bridge. From
Figure 2. Description of the thermal extraction (positive) and injection (negative) expressed as the linear power
imposed along the pile, for the case with low water flow.
Table 1. Thermal and mechanical parameters of the
involved materials.
parameter
soil
concrete
porosity []
density [kg/m3 ]
hyd. cond. at 20 C [m/s]
therm. cond [W/m/K]
specific heat [J/kg/K]
Youngs modulus [MPa]
Poissons ratio []
0.34
2122
3.5 104
1.59
1426
20
0.3
0.12
2408
9 109
1.56
1045
3500
0.2
December to February, the extraction design value

is applied constantly, while November and March
are used as ramp-up, resp. ramp-down periods. With
underground water flow (or geothermal renewal), the
rest of the year does not see any heat exchange. The
procedure is repeated for four years.
This study aims at finding the potential of such a
solution for the deicing of bridges. This means that all
design values were tailored to extract a maximum of
energy while maintaining a sustainable level of temperature during the four years. Therefore the extraction
design value must respect two criterions: it must not
freeze the concrete or the soil, for geotechnical reasons, and it must not exceed the thermal reload (all
sources combined).
3.6 Parameters of the study
Though expressed for the components of the mixture, the thermal and mechanical parameters can more
readily be summed up in Table 1, for the saturated
materials.
4
RESULTS
4.1 Base case: low flow with solar energy storage

The groundwater flow in that case is set at 0.015 m/d,
which is lower than the value used as a reference to
486
Figure 3. Temperature evolution in the pile during the first

year.
determine the need for a screen in an underground thermal storage area (Nordell and Hellstrm, 2000). The
fact that a single pile, and therefore a smaller volume,
is considered, is taken into account into this choice.
The achieved highest extraction design value is found
to be 46 W/m, which lies in the observed range of 30
to 70 W/m (Pahud, 2007). The corresponding injection
design value is found to be 50 W/m in order to allow a
sustainable use of the heat-exchanger pile in the longterm. Figure 3 shows the temperature profile in the
pile at different times during the first year.
The effect of this range of temperatures on the
mechanical behavior of the pile can be evaluated
through the evolution of stresses with time during the
considered period. Three points are considered and the
average of thermally induced stresses at three integration points is shown in Figure 4. The evolution is rather
homogeneous throughout the pile, though maximum
thermal stresses are observed in the middle of the pile.
The magnitude of 0.2 MPa corresponds to 15 kPa/ C,
which is logical in the case of low-restraint end conditions seen in an isolated pile, whereas a totally
restrained pile would exhibit values around 300 kPa/ C
(Amatya et al., 2012). These values do not cause concern for the structural integrity of the pile, which is
loaded to 3.9 MPa and can be designed to handle the
difference. The other aspect of the structure behavior is linked with displacements, especially the ones
observed at the surface and which are transferred to the
superstructure. Figure 5 shows the thermally induced
displacements at the surface and at three other points
in the pile. An amplitude of 2.5 mm is seen, which
is again well within the acceptable displacements of a
bridge deck, with the possible exception of short spans
(less than 10 m) in cases with the strictest deflection
requirements, in the order of L/5000 where L is the
length of the span. With the usual
4.2
Case 2: low flow with natural reload
The achieved extraction design value is found to be

only slightly lower at 42.5 W/m. The reason for this
Figure 4. Evolution of thermally induced vertical stresses

(compression negative) in the pile with time.
Figure 5. Evolution of vertical displacements of some

points in the pile with time.
small difference can be found in the thermal behavior

of the system in a comparison between the base case
and this one, as is done in Figure 6. Just before the
winter season, the temperature difference between the
two cases is limited to less than 2 C, while it was more
than 10 C at the end of the solar reload period. This
demonstrates the inefficiency of solar energy storage
with such a small system when compared to a building foundation. The benefit of solar energy retrieval
would be limited to two aspects: ensuring a return to
the natural temperature even after a demanding winter, and controlling the asphalt temperature during hot
periods, which is a gain for its durability and the safety
of users.
From a mechanical point of view, the difference lies
in the yearly evolution of displacements is the major
one. Winter periods are comparable between the two
cases, as can be seen in Figure 7. The total amplitude
in the geothermal case is limited to 1.3 mm, which
is another positive aspect for the geothermal solution
rather than the solar energy use in small foundations.
487
Figure 6. Temperature profiles during the end of the thermal

reload period at a depth of 20 m, in the direction of the flow.
Figure 7. Evolution of vertical displacement on top of the

pile in the solar and geothermal scenarios.
4.3
Figure 9. Temperature evolution in the pile for two different

flow and extraction rates.
Case 3: no flow with solar energy storage
An additional case was tested to determine whether

the low efficiency of solar energy storage should be
attributed to the groundwater flow or purely to the
geometry of the system. In this case, no groundwater
flow is present. The extraction design value is found to
be equal to that obtained in the base case. Only minimal changes can be found in the temperature evolution
in the vicinity of the pile, which do affect neither the
thermal nor the mechanical behavior of the system.
These changes are shown in the temperature profiles
at a depth of 20 m, represented in Figure 8, and in
which the shift to the right of the full-line curve (low
flow) is visible. Consequently, the geometry of the system is the main reason for the observed inefficiency,
and groundwater flow should be seen as beneficial in
any case.
4.4
Figure 8. Temperature profiles during the end of the thermal

reload period at a depth of 20 m, in the direction of the flow.
Case 4: high flow with natural reload
The evaluation of the potential of isolated (or small

groups of) geothermal energy piles has highlighted
the differences between these and a building foundation. Groundwater flow being beneficial here, a
higher water flow of 0.3 m/d is also considered in what
constitutes the fourth scenario. The same extraction
rate profile shown in Figure 2 is used, but a different extraction design value has been evaluated. In the
present case, an extraction rate of 70 W/m, which is
the maximum of the range of values that are achieved
with conventional heat pumps and energy piles can be
extracted thanks to the natural thermal reload brought
by the groundwater flow. The compared evolution of
temperature in both geothermal cases (cases 2 and 4)
is shown in Figure 9. The convective effect of the high
water flow is clearly visible as temperature reaches a
plateau during the extraction phase, here at 3.5 C. The
thermal reload is also fast and complete, while in the
low flow case, a balanced state over a year is reached
only on fourth year, and at a lower temperature level
than the natural one.
The displacements are evolving similarly to the
other geothermal case, as can be seen in Figure 10.
Again, they do not represent a concern for a bridge.
488
ACKNOWLEDGEMENTS
The authors wish to thank the Swiss association of
road and transportation experts, VSS, for supporting
project VSS 2010-503.
REFERENCES
Figure 10. Displacement evolution on top of the pile for two

different flow and extraction rates.
4.5 Use for deicing

The presented study has allowed the identification of
the favorable uses of heat exchanger piles in the context of bridge deicing. The total amount of energy that
can be retrieved from one pile in each of the four cases
leads to a maximum of 5800 kWh per pile per year
in the case of high natural reload through groundwater flow, while less favorable case leads to 3480 kWh
per pile and per year. If we consider an average heat
demand of 100 kWh/m2 per year, this translates into a
range of 34 to 58 m2 of deck per pile, or into a more
general unit, 1.5 to 2.5 m2 of deck per m of pile.
5
CONCLUSIONS
The thermo-hydro-mechanical behavior of a heatexchanger pile has been studied with the addition of
convective heat transport by groundwater flow. The
study aimed at demonstrating that such a study could
indeed be performed in a fully coupled manner, simplifying the design process on both the energy and the
structural sides; and it also aimed at evaluating the
potential of such foundations for the case of bridge
deicing. The results are positively showing that the
two criteria can be met as long as some specifications
of the bridge and location are considered. The ratio of
deck surface to pile length should be lower than 1.5 to
2.5 m2 /m in typical temperate European climate, and
the bridge should be long enough to allow service displacements in the order of a few millimeters. It must be
noted that the economy and efficiency of the system is
higher when groundwater flow is present, and reaches
the limits in terms of power of ground heat-exchangers
that use water as a fluid.
AMATYA, B., SOGA, K., BOURNE-WEBB, P. J.,

AMIS, T. & LALOUI, L. 2012. Thermo-mechanical
Behaviour of Energy Piles. Gotechnique, 62, 503519.
ASHRAE 1995. Snow melting. In: ASHRAE (ed.) Heating,
Ventilating and Air-Conditioning Applications. Atlanta.
BOBES-JESUS, V., PASCUAL-MUOZ, P., CASTROFRESNO, D. & RODRIGUEZ-HERNANDEZ, J. 2013.
Asphalt solar collectors: A literature review. Applied
Energy, 102, 962970.
CHARLIER, R. 1987. Approche unifie de quelques problmes non linaires de mcanique des milieux continus par
la mthode des lments finis. PhD PhD Thesis, Universit
de Lige.
CHARLIER, R., RADU, J.-P. & COLLIN, F. 2001. Numerical modelling of coupled transient phenomena. Revue
Franaise de Gnie Civil, 5, 719743.
COLLIN, F. 2003. Couplages thermo-hydro-mcaniques
dans les sols et les roches tendres partiellement saturs.
PhD PhD Thesis, Universit de Lige.
HOPKIRK, R. J., HESS, K., EUGSTER, W. J. & KNOBEL,
P. 1994. Serso: Pilotprojekt zur Sonnenenergierckgewinnung aus Strassenoberflchen. In: BERN, B. F. S. T. D. K.
(ed.) Technischer Bericht. Bern.
LIU, X., REES, S. J. & SPITLER, J. D. 2007. Modeling snow melting on heated pavement surfaces. Part I:
Model development. Applied Thermal Engineering, 27,
11151124.
LUND, J. W. 1999. Geothermal snow melting. Transactions
of the Geothermal Research Council, 23, 127133.
NORDELL, B. & HELLSTRM, G. 2000. High temperature
solar heated seasonal storage system for low temperature
heating of buildings. Solar Energy, 69, 511523.
PAHUD, D. 2007. Serso, stockage saisonnier solaire pour
le dgivrage dun pont. Berne (CH): Office fdral de
lnergie.
PAHUD, D. & HUBBUCH, M. Year. Measured Thermal Performances of the Energy Pile System of the Dock Midfield
at Zrich Airport. In: Proceedings European Geothermal
Congress 2007, 2007 2007 Unterhaching, Germany.
SCHMIDT, T. & MANGOLD, D. 2006. New steps in seasonal thermal energy storage in Germany. Ecostock 2006:
The Tenth International Conference on Thermal Energy
Storage. Pomona, NJ, USA.
VAN MEURS, G. A. M. 1985. Seasonal heat storage in the
soil. PhD, T. U. Delft.
WANG, H., ZHAO, J. & CHEN, Z. 2008. Experimental investigation of ice and snow melting process on pavement
utilizing geothermal tail water. Energy Conversion and
489

Thermal, hydraulic and mechanical performances of enhanced grouting

materials for borehole heat exchanger
S. Erol & B. Franois
Universit Libre de Bruxelles, Building, Architecture and Town Planning Dept (BATir), Brussels, Belgium
ABSTRACT: Experimental, numerical and analytical studies are carried out to enhance the thermal performance of vertical ground source heat pump (GSHP) systems by improving the grouting material with the addition
of graphite powder. Several mechanical and thermo-physical tests are performed for two different widely used
commercial grouting materials (i.e. bentonite-based and silica sand-based) and homemade admixtures enhanced
with natural flake graphite, synthetic graphite and expanded graphite. Experimentally to assess the specific heat
exchange rates, the prepared borehole heat exchanger (BHE) probes are operated using two heat pumps in a
1 1 1 m3 sandbox under dry (solid air) condition. The home-made admixture prepared with 5% natural
flake graphite can be considered as an appropriate grout material for BHEs regarding to its rheological and
thermo-physical properties as well as its cost. During the operations, the monitored temperature measurements
in the sandbox are in agreement with numerical simulation and analytical approach prediction. The sandbox
study shows that if the thermo-physical properties of ground is considerably low, the thermal conductivity of
grout has no significant impact on the performance of BHE, the main resisting component in the thermal transfert
being the ground itself.
INTRODUCTION
Geothermal energy is the form of energy that is

extracted from the stored heat ground, and within a
range of 0 to 400 m of depth the stored heat is categorized as shallow geothermal energy. In order to use this
energy, there are varieties of different earth-coupled
heat extraction systems. The closed-loop geothermal
system is one of the mostly used technologies and
its configuration comprises a heat exchanger installed
inside a borehole and a pump that circulates a solution of water or anti-freeze mixture through the buried
pipes. Thus, the heat is transferred from the ground to
the heat carrier fluid. The objective is to maximize the
heat transfer, in order to optimize the performance of
the system.
The GSHP systems for domestic heating and cooling are basically installed in the ground with high
density polyethylene (HDPE) pipes and filled the surrounding empty space mostly with grout inside the
borehole (e.g. in Sweden groundwater can be used as
filling material). In addition to the thermo-physical
and hydro-geological conditions of the ground, the
characteristics of backfill materials, particularly thermal conductivity, may play an important role for the
specific heat exchange rate (Jun et al. 2009; Lee et al.
2011).
In order to improve the specific heat exchange rate
of BHEs, thermal conductivity of the grout material can be enhanced. Particularly, graphite-based
admixtures (Lee et al. 2010; Lee et al. 2011; Delaleux
et al. 2012) have been investigated to enhance the

thermal conductivity of grouting material. Lee et al.
2010 experienced on the suitability of different type of
grouting materials for BHEs. They indicated that with
increasing the content of silica-sand and graphite in an
admixture, the thermal conductivity rises, however, the
viscosity of the admixture also increases. As a result,
by adding 30% graphite to bentonite-based admixture
led to a thermal conductivity of 3.5 W m1 K1 . In
addition, Lee et al. 2011 experienced several in-situ
thermal response tests (TRT) to study the performance
of the vertical GSHP systems with considering various grouting materials. They developed an admixture
containing cement, silica-sand and graphite providing
a thermal conductivity of 2.6 W m1 K1 . Recently,
Delaleux et al. 2012 claimed that by adding less than
15% graphite powder in an admixture, the enhancement of grout thermal conductivity can be led to a
significant increase of the specific heat exchange rate
regarding to present commercial backfilling materials.
Nevertheless, in most of the previous works, other necessary parameters such as permeability, compression
strength and workability, were not determined to proof
the applicability of those admixtures as a grouting
material.
The main objective of the present research is to
focus, not only on the thermal improvement, but also
on the other rheological properties of the grout in
order to guarantee a proper behavior of the BHE.
Several laboratory tests have been performed to determine the suitability of grouting materials for BHE
491
and a small-scale sandbox TRT have been operated

in the laboratory for two different commercial products (bentonite-based and silica sand-based) and one
of the home-made admixtures containing natural flake
graphite. In addition, numerical and analytical studies
are carried out to compare the sandbox test results, and
also to evaluate the specific heat exchange rate and the
thermal resistances of BHE.
2 ADMIXTURE
2.1
Commercial grouting materials
In order to investigate the range of specific heat

exchange rate depending on the typical characteristics of grouting materials, first analyses are carried
out by considering two different commercial products that have been used as backfilling material for
BHEs. One of the considered products is bentonitebased (0.9 W m1 K1 ) and other is silica sandbased (2.3 W m1 K1 ). Those two products have
been chosen for their relatively high difference in thermal conductivity, in order to have a large range of
selection. The objective is to justify how the grouting
materials drive the heat exchange rate.
2.2
Homemade admixtures
The considered graphite types for the pre-analyses

(i.e. mechanical and thermo-physical tests) are listed
by increasing cost: (1) natural flake graphite (TIMREX M100/45150 m), (2) the primary synthetic
graphite with two different grain size distribution
(TIMREX KS150/150 m, TIMREX KS150-600/150
600 m), and (3) expanded graphite powder (TIMREX C-THERM011/2.5% ashes) (TIMCAL 2012).
The components are mixed with various ranges (5
12% graphite, 2440% silica sand (D50 = 260 m),
2445% water, 2028% cement and 07% Ca2+ bentonite), in order to determine the fraction contents
of admixtures. According to the preliminary (visual)
observations, and also regarding to the guidelines
VDI 4640 (Blatt 2 & 3) and Allan & Philippacopoulos (1998; 1999), the proportion of graphite
shall be less than 10% and silica sand should not
exceed 50%, because graphite absorbs large amount of
water and with the addition of silica sand the pumpability of admixture becomes unfeasible due to its
high viscosity (VDI 2001b). Therefore, the amount of
graphite powder is kept 5% in all prepared home-made
admixtures.
durable and frost-proof material (Hermann 2008; Reu

et al. 2011). In order to propose an appropriate grouting
material in accordance with guidelines, the characteristics of grouting materials must be determined in
terms of hydraulic conductivity, uniaxial compression
strength, thermal conductivity and workability.
The allowable permeability of backfilling material
shall be 1 109 m s1 (GSHPA 2011). The permeability has been measured in falling head permeameter.
The considered grouting materials are placed inside
the cells, when they are fresh. Then the samples are left
for curing for 2 weeks under water. Then the falling
hydraulic head in the pipe is measured through 30
days to calculate the permeability. Thereby, the permeability can be evaluated as a function of curing
time.
The uniaxial compression test is the estimation of
the compression strength of a grouting material with
unrestricted horizontal deformation.
Before compression strength tests, the prepared
fresh samples are placed in a humidity room (80%
humidity) under constant temperature at 20 C, and
curing time is set to 10 days and 30 days to observe
how the strength of the materials changes in time.
Thermal conductivity of grouting material is determined with the thermal needle probe device (1.5 mm
diameter and 150 mm length).
While the admixture material is fresh, a thin rigid
stick that has the same diameter then the needle probe
(1.5 mm) is placed in the middle of sample (cylinder
shaped sample d = 100 mm, h = 200 mm). Thereby,
the needle probe can be replaced into the sample when
it is dried. During the thermal conductivity measurement, the surrounded gap between the sample and
the needle is filled with a highly-conductive thermal
paste (5.6 W m1 K1 ) to avoid artifacts due to air
included in the gap.
For determining the pumpability and the plastic viscosity of admixtures characterizing its workability, the
flow cone method, so called Marsh cone is performed.
The specified volume of fresh grout, 1725 mL,
flows through the nozzle of the cone, and the time
of efflux of suspension is measured in seconds. In
Roussel & Le Roy (2005), an empirical method is
described to classify the flowability of different suspensions. The represented criterion is related to several
rheological parameters (i.e. plastic viscosity and yield
stress) which can be calculated depending on the efflux
time, density of grout and cone geometry (e.g. height
of the cone, radius of the nozzle).
4
3 THERMO-PHYSICAL, HYDROLOGICAL
AND MECHANICAL CHARACTERIZATION
OF GROUT MATERIALS
SMALL-SCALE BHE CHARACTERIZATION
4.1 Sandbox description
The principal tasks about the grouting materials are to

provide a good thermal conduction between the pipes
and surrounding ground and to ensure a watertight,
The small-scale TRT experiment is carried out under

dry sand (solid air) condition. The size of the considered sandbox for the experiments is 1 1 1 m3 . The
initial ground temperature is 20 C. The BHEs with a
diameter of 0.135 m are operated with two heat pumps
492
The widely used method to estimate specific heat

exchange rate Q of the BHE is based on the temperature difference between inlet and outlet (in our case,
the sum of the temperature difference of the inflow and
the outflow of each pipe) given as follows:
Figure 1. The illustration shows the procedure of small scale

sandbox TRT operated with two separated heat pumps.
Figure 2. The considered locations of the thermistors for

temperature measurements: a) 7 pieces thermistor along
the y-axis of a BHE with an interval distance of 7.2 cm
b) Thermistors inside the pipes.
that circulate water with temperature fixed at 12 C

(inlet) and at 15 C (outlet), respectively (Fig. 1). In a
real operation, the pump circulates the fluid through
the geothermal probe (100200 m) to have a thermal
gradient between inlet and outlet by the heat exchange
with the ground. In the present test, the 1 m probe
length is not enough to produce a significant temperature difference between inlet and outlet (3 C).
Therefore, the fluid temperatures in the two pipes are
controlled separately.
The outer diameter of the pipes is 0.032 m, and the
flow rate is set to 3.66 104 m3 /s (turbulent flow with
the Reynold number of 1.1 104 ). As the boundary
conditions, we isolated properly the sandbox and the
pipes from the probe to the heat pumps. The operation time is set to the minimum suggested duration
of a TRT, 50 hours (Austin et al. 2000; Gehlin 2002),
because the ambient temperature strongly influences
the temperature distribution inside the sandbox in case
of the longer operations and also after 50 hours, the
radius of influence of the BHE reaches the wall of
the box.
Since the temperature distribution is symmetric in
the sandbox due to conduction dominated heat transfer
system, the temperatures are measured along the y axis
of BHE with the pt-100 precision thermistors. Additionally, two thermistors are placed inside the pipes
where the fluid flow into the BHE, to correct the temperature difference between the heat pump and the
probe (e.g. heat pump set 11.7 C the thermistor read
12 C).
in which f cf is volumetric heat capacity of heat carrier fluid, qf is the flow rate, T1 and T2 denote the
temperature variation of the fluid in the pipes 1 and 2
along the 1 m probe length, L.
However, since the temperature difference of the
fluid between inflow and outflow of each pipe is considerably small (<0.1 C) for a BHE with a 1 m length,
it is not possible to obtain precision measurement
data due to both the sensor accuracy and the resolution of the acquisition device, 0.03 C and 0.1 C,
respectively. Therefore, it appears best to estimate the
specific heat exchange rate with the evaluation of
decreasing sand temperature.
Line source or cylindrical source models can be
adopted to evaluate the thermal characteristics of
the ground by using in-situ TRT data. Most of the
time, those approaches are used to deduce the thermal properties of the ground, controlling the heat
input. However, reversely, it is also possible, knowing the ground properties, to deduce the specific heat
exchange rate through back-analysis. This is what will
be done in chapter 4.2.3 based on the temperature
profile in the sand.
4.2 Analytical solution

The objective to use an analytical solution is to gain
time for long-term prediction of an engineering application, and this implemented analytical approach can
provide us to investigate the near field behavior of a
BHE such an operation of 30 years.
Infinite and finite analytical solutions for BHEs
consider mostly conduction dominated systems
(Ingersoll et al. 1954; Carslaw & Jaeger 1959). However, those assumptions consider a constant heat load
and do not take into account the borehole resistances
to calculate the temperature change in the surrounding
ground. Since we fix the mean fluid temperature to a
certain value, the heat input changes as a function of
time. Therefore, we should impose several backward
computations to estimate the temperature difference
in the vicinity of BHE. First, the borehole resistances
and the heat input rate are analytically calculated as a
function of time. Then the computed heat load, q(t), is
obtained as a function of borehole resistance and mean
fluid temperature.
4.2.1 Borehole resistance
The borehole resistances can be analytically estimated
regarding to its geometry factor and its typical characteristics. It accounts for the convection and conduction
493
pipe resistances and effective borehole wall resistance

(Lamarche et al. 2010):
The total borehole thermal resistance Rb corresponds to the temperature difference between the mean
fluid temperature Tf and the borehole wall temperature Tb divided by the specific heat exchange rate Q.
This total borehole thermal resistance includes pipe
resistance Rpipe (decomposed into convection and conduction resistance Rconv and Rcond , respectively) and
grout thermal resistance Rb . The two pipe resistances,
the convection resistance:
4.2.2 Heat input rate

In the following step, the calculated borehole resistance is taken into account in Eq. 8 to calculate the heat
input rate as a function of time, in order to keep the
mean fluid temperature as a constant in time. Essentially, the equation 8 is used to evaluate the mean
fluid temperature as a function of time by taking line
source method devivated by Carslaw & Jaeger 1959
and adding the effective borehole resistance (Gehlin
2002). In our case, we solve backward this equation to
calculate the heat input rate as a function of time:
where a is the thermal diffusivity, T0 is the undisturbed

soil temperature, and is the Euler constant, that is
typically 0.57721.
in which ri and ro are the inner and outer radius of

pipe, respectively, and h is the heat transfer coefficient
estimated with Nusselt number (Nu), inner diameter
of the pipe di and thermal conductivity of the fluid f .
The third contribution of Eq. 2 is a 2D grout thermal
resistance. There are different equations to calculate
the grout thermal resistance excluding pipe resistance (Hellstrm 1991; Paul 1996; Sharqawy et al.
2009). In particular, Bennet et al. (1987) used the
multipole method. Lamarche et al. 2010 shows that
among other approaches under steady-state and transient conditions, the multipole solution provided the
best estimation with regard to the numerical results.
For single U-shaped pipe, the first-order solution is
given as follows Bennet et al. (1987):
4.2.3 Temperature distribution in the ground

In the last step, the calculated heat input rate as a
function of time can be taken into account for one of
the line source model. We used the moving finite line
source method (MFLS) for a vertical BHE derivated
by Molina et al. 2011 that considers both the axial
effect and groundwater flow. The MFLS agrees well
with numerical results under steady-state and transient
(groundwater flow) underground conditions for short
(>2 hour) and long-term (e.g. 40 years) simulations.
The MFLS method is given as follows:
in which vT is the effective heat transport velocity

depending on the seepage velocity (in our study it is
zero), f (x, y, z, t) is derivated as:
R is the distance to the source placed on z-axis in 3D

Euclidean space and expressed as:
in which 1 = rb /rp , 2 = rb /xc , 3 = rp / 2xc and rb

is the radius of borehole 0.0675m, rp is the external
radius of pipe 0.016 m and xc is the shank space 0.0415
m. is given as follows:
Knowing the experimental temperature distribution, Eq. 9 can be solved backward to deduce q(t),
which is specific heat exchange rate Q (including
thermal transfer and borehole resistance). One of the
temperature measurement in the sandbox at where the
temperature distribution is symmetric can be taken
into account for T (t) to calculate the specific heat
494
RESULTS
5.1 Laboratory test results
Figure 3. 3D model domain with a 1 m length of BHE.

Table 1.
sand.
Characteristics of the pipe, water, air and silica
Parameters
Pipe*
Water**
Air**
Silica sand
(air/solid)
Thermal
conductivity
[W m1 K1 ]
Specific heat
capacity
[J kg1 K1 ]
Density
[kg m3 ]
0.42
0.59
0.024
0.35
2170
4185
1000
820**
960
1000
1.25
1500**
*(Hakagerodur-geothermal 2012), **(Engineering toolbox

2012), Bulk thermal properties of sand with the porosity
of 0.43 [] (Sibelco 2012), measured value.
exchange rate (T at x = 0 m, y = 0.14 m, z = 0.5 m),

and the borehole resistance is estimated according to
the left-hand side of the equation 2 (Tb at x = 0 m,
y = 0.0675 m, z = 0.5 m).
4.3 Numerical model
The objective of numerical model is to understand the
impact of the components on the heat transfer system (i.e. grout, pipe, soil). The study is represented
by a 3D homogeneous model domain (Fig. 3) with
the finite element software program COMSOL Multiphysics. The model domain is set according to the
sandbox size (interior dimensions 1 m 1 m 1 m).
Since the ambient temperature influences the measurement during the operations, the wooden frames with
the thickness of 1.3 cm are also fixed to the model
domain. The mesh is generated using uniform triangle
elements. The BHE is located at the central position
of the model domain.
As the boundary conditions, temperature boundary
condition is applied on the modeled wooden walls
from each side according to the recorded outer wall
temperature (between the isolation material and wood
frame). Other parameters (i.e. inlet, outlet and simulation time) are set according to the sandbox experiment
(Chapter 4.1). For the simulations, groundwater flow is
not considered. The provided parameters of silica sand
shown in Table 1 is set to the model as the bulk thermal
properties of the porous media (air/solid phase).
The fundamental comparison of all variant results can

be found in Table 2. As the reference measurement,
the results of two commercial products are nearly in
agreement with the values provided by the producers. However, during the process of sample preparing,
because of brittleness of C-2 bentonite-based grout,
micro cracks were produced, that could alter the results
for compression strength and thermal conductivity.
Compared the results of home-made grouts with
the reference values of the commercial grouts, permeability, density and compression strength results are
in an allowable range, except the result of admixture
prepared with expanded graphiteA-5 for which permeability 1 109 m/s is not fulfilled, and the lowest
compression strength and density is observed among
other grouts. The reason is the expanded graphite has
lower density than the other graphite powders (e.g.
expanded graphite bulk density = 150 kg m3 , synthetic graphite 150600 m bulk density = 670 kg
m3 ).
The higher fraction of sand in A-1 (40%) and the
larger size graphite grains (> 300 m) which are sunk
into the bottom of A-4 sample caused the sedimentation in those admixtures. Except the admixtures A-1
and A-4, other suspensions are mixed homogeneously.
Compared the Marsh cone test results, the flowability of A-1 and A-4 admixtures are considerably
faster than other home-made grouts, because the wellgraded grain size distribution decreased both the yield
stress and the plastic viscosity of those suspensions.
On the other hand, the efflux of silica sand-based
grout and C-2 bentonite-based w/b = 0.5 is stopped
in the Marsh cone after several drops, even if they are
homogeneous mixtures. The flow time of bentonitebased grouts is decreasing with rising w/b ratio due
to decreasing viscosity. The calculated plastic viscosity results are proportional to the efflux time of all
admixtures and vary depending on the component
characteristics of grouts (e.g. grain size distribution of
components).
Currently, the thermal conductivity results of
only three home-made admixtures and two commercial products are available. Even, the existing
results of home-made admixtures demonstrate that
5% addition of graphite has a significant impact
on the thermal conductivity of grout (e.g. A-1
without graphite = 1.5 W m1 K1 , A-2 with natural
graphite = 2.3 W m1 K1 ).
In Figure 4, after a flat first part of the curve
(corresponding to the closure of the gap between the
loading piston and the sample), uniaxial compression
curves give a straight line that demonstrates the elastic
behavior of the grouts. Then, under high load, nonlinear curve shows plastic behavior. The compression
strength of each material is taken as the peak value of
each curve. When C-2 bentonite-based grouts reach
its elastic limit, then a sudden failure occur that is
the characteristic of brittle response. On the contrary,
495
Table 2.
Measured parameters for two commercial grouting materials and home-made admixtures.
Admixtures
C-1 silica sand-based
C-2 bentonite-based
w/b = 0.5
C-2 bentonite-based
w/b = 0.6
C-2 bentonite-based
w/b = 0.7
C-2 bentonite-based
w/b = 0.8
A-1 without graphite
A-2 with natural
graphite
A-3 with sythetic
graphite 150 m
A-4 with sythetic
graphite
150600 m
A-5 with expanded
graphite
Permeability
[m s1 ]
Density
[kg m3 ] 103
Plastic
viscosity
[Pa s]
Marsh cone
[s/1725 mL]
Compression
strength
[N mm2 ]
Thermal
conductivity
[W m1 K1 ]
<1 109 /6 1010 *

<1 1010 /4.3 1012 *
1.8 / 1.8**
1.7**
No flow
No flow
10 /9.3**/3.8
2.35 /2.3**
<1 1010 /9 1012 *
1.66 /1.66**
0.20
25 1
8.5 /5.2**/3.1
0.95 /0.7**
<1 1010 /8.1 1011 *
1.6 /1.6**
0.13
17 1
<1 1010 /2.8 1010 *
1.55 /1.54**
0.11
14 1
2.8 1012 *
4.4 1012 *
1.75**
1.7**
0.23
0.6
27 1
72 1
8.2**
8**
1.5**
2.3**
2.2 1012 *
1.75**
0.41
48 1
10**
2.5**
3.3 1012 *
1.75**
0.19
22 1
9.8**
>1 103 *
1.2**
0.36
61 1
1.7**
*Measured values after the curing period of 45 days (cured under water), **and Values are the average of two samples and
measured values after the curing period of 30 days and 10 days, respectively (cured under 80% humidity at constant temperature
20 C), values provided by producers, backward calculations based on efflux time, density and the properties of Marsh cone
geomerty (Roussel & Le Roy 2005), radius of nozzle 5 mm, cone angle tan() = 0.253.
for the test after 10 days (c 1) fractured under less

pressure than other samples cured for 30 days (c 2).
5.2 Sandbox experiment results
Figure 4. Comparison of the compression strength results

between all grouts. w/b ratio of bentonite-based grout is 0.6.
c 1 presents the curing period of 10 days and c 2 denotes
30 days. Soil type is dry sand (air solid).
C-1 silica sand-based grout exhibits plastic hardening behavior, when the stress exceeds its elastic limit.
After the peak point of the loaded stress, C-1 silica
sand-based grout fractured. The results of home-made
admixtures (A-1 to A-5) provide nearly identical elastic behavior, except the admixture A-5. Obviously, the
density and the texture of expanded graphite have a
negative impact on the plastic behavior and ultimate
strength.
Concerning the curing period of samples (currently
only for commercial grouts), the samples performed
Three different probes are prepared for the operations.

Two probes are made with the commercial grouts,
C-1 silica sand-based ( = 2.3 W m1 K1 ) and C-2
bentonite-based ( = 0.9 W m1 K1 ), to observe the
influence of grout thermal conductivity on the heat
transfer. In addition, the third probe is made with A-2
admixture containing natural graphite ( = 2.3 W m1
K1 ) to compare with other two commercial grouts,
because based on thermo-hydro-mechanical laboratory analysis this admixture seems to be the most
efficient home-made grout regarding to the cost aspect
and the laboratory test results.
Figures 5 and 6 demonstrate the temperature distribution in the sandbox over distance and over time,
respectively for operated BHE probe prepared with
C-1 silica-sand based grouting material. The room
temperature during the test was in the range of 19.8 C
to 21.6 C. In both figures, the numerical simulation
results are nearly identical respect to the experimental measurements, and the root mean square errors
(RMSE) between numerical and experimental results
are 0.091 and 0.075, respectively. On the other hand,
the analytical solution result is slightly apart from
the experimental measurement (RMSE = 0.154 and
0.114, respectively). The reason is that we cannot
consider the boundary conditions (i.e. the influence
of room temperature) on the analytical solution as
496
Figure 5. Comparison of experimental measurement with

numerical and analytical (Eq. 13) solution results at 50th
operation hour (x = 0 m, y = 0.0675 m 0.5 m, z = 0.5 m)
for BHE probe prepared with C-1 silica sand based grouting.
Figure 6. Comparison of experimental measurement with

numerical and analytical (Eq. 13) solution results over time
(x = 0, y = 0.14 m, z = 0.5 m) for BHE probe prepared with
C-1 silica sand based grouting.
well as we did on the numerical model according to

the temperature measurements on the sandbox wall.
Therefore, a small discrepancy between analytical
solution and the experimental data is observed.
In Figure 6, the plateau view of the experimental
results can be accounted for the acquisition card connected to the thermistors. This problem can be seen
also in Figures 8 and Figure 9.
Figures 7 and 8 show the comparison of temperature results obtained with different grout materials.
In Figure 7, the influence of room temperature on the
sandbox wall can be seen, particularly, on the results of
C-2 andA-2. Similarly, we observe a contrariety in Figure 8 between the results. Regarding to the initial temperature, the bentonite-based probe essentially cannot
affect the temperature difference in the sandbox more
than other two probes due to less thermal conductivity
of grout C-2. Actually, the slight difference between
three curves is mainly due to the ambient temperature
effect. The thermal properties of grouting materials
Figure 7. Comparison of experimental results at 50th operation time hour (x = 0.5 m, y = 0.0675 m 0.5 m, z = 0.5 m).
Figure 8. Comparison of experimental results over time

(x = 0.5, y = 0.14m, z = 0.5 m).
does not have a predominant effect because of the poor

thermo-physical properties of the soil (e.g. bulk volumetric heat capacity of soil = 1.3 106 J m3 K1 ,
thermal conductivity = 0.35 W m1 K1 ) in which the
main resistance for thermal transfer takes place. The
heat exchange is dominated by the soil thermal characteristics rather than the thermal conductivity of
grout and the three grouting materials provides similar performance in term of heat exchange rate, as
demonstrated in Figure 9.
In Figure 9, the specific heat exchange rate and the
borehole resistance are calculated from Eq. 9 and Eq. 2,
respectively (Chapter 4.2). As explained in the previous paragraph, the results demonstrate also that under
dry sand conditions the thermal conductivity of grout
has no important role on the specific heat exchange
rate due to low thermo-physical properties of soil. Considering the borehole resistances, the mean Rb value of
A-2 home-made admixture shown in Figure 9 is lower
than C-1 commercial grout. Since the thermal conductivity of grout is one of the major dependence of the
borehole resistance, the thermal conductivity of A-2
may be higher than C-1. However, according to the
497
GmbH Germany, TIMCAL Ltd. Switzerland and Sika

group Belgium for raw materials.
REFERENCES
Figure 9. Comparison the performance of C-1 silica

sand-based and C-2 bentonite-based and A-2 home-made
grouting materials. Soil type is dry (air solid) sand.
laboratory measurements of grouts, the thermal conductivity of C-1 and A-2 grout materials is identical
(Table 2).
CONCLUSION
The grouting material used for BHEs must guarantee several hydraulic, mechanical and thermo-physical
requirements.The present study provided a wide investigation on the various admixtures with laboratory tests
and comprehensive evaluation of the performance of
BHE probes based on different type of grout materials by the experimental, numerical and analytical
analyses.
Regarding to the laboratory test results, it is conclude that even a small amount of graphite addition
(5%) has a great influence on the thermal conductivity of grout. However, it is not feasable to use all kind
of graphite powders in an admixture used as a backfill
material of BHEs, because different specific characteristics of graphite affect adversely the mechanichal
behaviors of grouts (e.g. low permeability and compression strength). In fact, the home-made admixture
prepared with 5% natural flake graphite can be considered as an appropriate grouting material for BHEs
regarding to the laboratory results and the cost aspect.
Considering the sandbox experiments, if the
thermo-physical properties of ground is considerably
low (e.g. dry soil), the thermal conductivity of grout
has no significant impact on the specific heat exchange
rate. This has been demonstrated by both experimental
and analytical/numerical simulations.
As the further steps to determine the impact of
highly conductive heat transfer system on the performance of BHE, the probes will be operated under
water-saturated sand condition (>2 W m1 K1 ).
ACKNOWLEDGEMENT
The financial support from Walloon Region in
Belgium is profoundly acknowledged. Furthermore,
the authors would like to thank the company Schwenk
Allan, M.L. & Philippacopoulos, A.J. 1998. Thermally conductive cementitious grouts for geothermal heat pumps.
FY 98 Progress Report, BNL 66103.
Allan, M.L. & Philippacopoulos, A.J. 1999. Properties and
performance of thermally conductive cement-based grouts
for geothermal heat pumps. FY 99 Final Report.
Austin, W.A., Yavuzturk, C., Spitler, J.D. 2000. Development of an in-situ system for measureing ground thermal
properties. ASHRAE Transactions. 106(1): 365379.
Bennet, J., Claesson, J., Hellstrm, G. 1987. Multipole
Method to Compute the Conductive Heat Transfer to and
between Pipes in a Composite Cylinder. Notes on Heat
Transfer 3. Lund: Lund Institute of Technology.
Carslaw, H.S. & Jaeger, J.C. 1959. Conduction of Heat
in Solids, second edition. New York: Oxford University
Press.
COMSOL 2012. COMSOL 4.2a, Multiphysics Modeling,
Finite Element Analysis and Engineering Simulations
Software, COMSOL, Inc., Burglington, MA.
Delaleux, F., Py, X., Olives R., Dominguez, A. 2012.
Enhancement of geothermal borehole heat exchangers performances by improvement of bentonite grouts
conductivity. Applied Thermal Engineering 3334:
9299.
Engineering Tool Box 2012. The engineering toolbox: Tools
and basic information for design, engineering and construction of technical applications: Accessed on March
22, 2012, at http://www.EngineeringToolBox.com.
Gehlin, S. 2002. Doctoral Thesis: Thermal Response Test
Method, Development and Evaluation. Lule: Lule University of Technology.
GSHPA 2011. Guideline for closed loop vertical borehole
heat exchanger design, installation and material standards. Milton Keynes: Ground source heat pump system
association National Energy Center.
Hakagerodur-geothermal 2012. Hakagerodour-geothermal:
Handling book for borehole heat exchangers Technical details of geothermal pipes; Model: Geotherm
PE-100. Accessed on March, 12, 2012 by email from
Hakagerodour-geothermal AG.
Hellstrm, G. 1991. Ph.D. Thesis: Ground heat storage thermal analyses of duct storage systems, I. Theory. Lund:
University of Lund.
Herrmann, V. J. 2008. Doctoral Thesis: Ingenieurgeologische Untersuchungen zur Hinterfllung von GeothermieBohrungen mit Erdwrmesonden. Karlsruhe: Karlsruhe
Institute of Technology (KIT).
Ingersoll, L. R., Zobel, O. J., Ingersoll, A. C. 1954. Heat
Conduction with Engineering. Geological and Other
Applications. New York: McGraw-Hill.
Jun, L., Xu, Z., Jun, G. and Jie Y. 2009. Evaluation of heat
exchange rate of GHE in geothermal heat pump systems.
Renewable Energy 34 (12): 28982904.
Lamarche, L., Kajl, S., Beauchamp, B. 2010. A review
of methods to evaluate borehole thermal resistances in
geothermal heat-pump systems. Geothermics 39: 187200.
Lee C., Lee K., Choi H., Choi H. P. 2010. Characteristics of
thermally-enhanced bentonite grouts for geothermal heat
exchanger in South Korea. Science China: Technological
Sciences 53(1): 1238.
Lee, C., Park, M., Nguyen, T., Sohn, B., Choi, J. M., Choi,
H. 2011. Performance evaluation of closed-loop vertical
498
ground heat exchangers by conducting in-situ thermal

response tests. Renewable Energy 42: 7783.
Molina-Giraldo, N., Bayer, P., Blum, P. Zhu, K., Fang, Z.
2010. A moving finite line source model to simulate
borehole heat exchangers with groundwater advection.
International Journal of Thermal Sciences 50: 2506
2513.
Paul, N.D. 1996. M.Sc. Thesis: The Effect of Grout Thermal Conductivity on Vertical Geothermal Heat Exchanger
Design and Performance. Vermillion, SD: South Dakota
University
Reu, M., Proell, M., Koenigsdorff, R. 2011. Quality control
of borehole heat exchanger systems. Espoo: IEA ECES
Annex 21 meeting.
Roussel, N. & Le Roy, R. 2005. The Marsh cone: a test or
a rheological apparatus?. Cement and Concrete Research
35(5): 823830.
Sharqawy, M.H., Mokheimer, E.M., Badr, H.M., 2009.

Effective pipe-to-borehole thermal resistance for vertical
ground heat exchangers. Geothermics 38: 271277.
Sibelco 2012. Sibelco Benelux: Handling book for silica
sand, physical characteristics of sand M-32 one grain size.
Accessed on March, 7, 2012.
TIMCAL 2012. Personal communication 03.07.2012.
Graphite & Carbon Manufacturer TIMCAL Ltd.,
Switzerland.
VDI-Richtlinie 2001a. Thermal use of the underground, Blatt
2. Dsseldorf: Verain Deutscher Ingenieure, VDI-Verlag.
VDI-Richtlinie 2001b. Thermal use of the underground, Blatt
3. Dsseldorf: Verain Deutscher Ingenieure, VDI-Verlag.
499

Technical and economic feasibility study of different thermal energy

supply systems scenarios for multi-residential buildings
Rolf Katzenbach, Frithjof Clauss & David Casini
Technische Universitt Darmstadt, TU Darmstadt Energy Center and Institute
and Laboratory of Geotechnics, Germany
Harald Frchtenicht
Projektentwicklung Frchtenicht GmbH, Babenhausen, Germany
ABSTRACT: The study discussed relates to the techno-economical investigation about updating the energy
supply system of a multi-family residential building built up in the 1960s. This condominium is located in
Frankfurt am Main in Germany, therefore results are strictly connected with the local climate data, regional
soil characteristics, national energy prices and government subsidies. Various alternatives are possible using
renewable energies; in particular the study is focused on shallow geothermal topic and its combination with
solar thermal collectors. The work started simulating the building monthly thermal load and continued with a
renewable energy supply system feasibility study; an analysis of the building thermal insulation is also done.
The borehole heat exchangers for the ground source heat pump plant are calculated using EED software. The
study concludes with a direct comparison of the economic feasibility between all scenarios, looking for the more
suitable and sustainable solution. The results show environmental and economic benefits obtained by a coupled
plant with a ground source heat pump and solar thermal collectors; using underground thermal storage during
the summer season and an auxiliary furnace for winter peak loads.
INTRODUCTION
About 70% of the household energy end-use in Europe

is used for space heating; the number of houses and
their average size are increasing and in the bargain
the total amount of energy used for space heating
is also growing (EEA 2005). Moreover the technical
Energy Efficiency Index (ODEX) in the household
sector decreased by about 23% from 1991 to 2009
in Germany; that means an annual improvement of
almost 1.4% (mostly for electrical appliances and
space heating), but from 2009, however, started a stagnation period (Odyssee 2011). In particular the consumption per dwelling for space heating in Germany
decreased from 2000 to 2009 by 2.16% in average
(Odyssee 2012). The building sector is regarded as
the key to greater energy efficiency developing strategies that permit a refurbishment of existing buildings
and the use of high efficient energy supply systems.
The building sector (including residential, service sector and industrial buildings as well as residential water
heating) accounted in 2004 for 40% of the EUs energy
requirements (IEA 2008). The most residential energy
supply systems are based on fossil fuels; a typical centralized furnace produces heat for the whole building
using water as heat carrier fluid. The most common
fossil fuel used is natural gas, in Germany there are also
a lot of heating generators that work with oil. Though,
a considerable amount of residential water heaters generate hot water by consuming electricity. The overall
efficiency in converting a potential energy of fossil
fuels into electric energy, then into thermal energy
is quite low; also the standard heating generation by
a furnace, the distribution and the regulation has a
quite low efficiency. Heat pumps could be the answer
both for space heating and domestic hot water (DHW)
generators with a high efficiency value. They need
electricity and a high temperature source that could
be air, water or underground to supply a cheap heat,
compared to traditional systems, via low-temperature
delivery systems. Solar thermal plants instead need
just solar irradiation to warm the water used as heat carrier fluid without a sophisticated plant: These systems
can collect and store solar energy just with the use of
collectors (panels), storage tanks and a circulation pipe
system. Coupling ground source heat pumps (GSHP)
with solar thermal panels gives to the whole system the
possibility to increase the overall efficiency. The first
gain obtained is the rise of the seasonal performance
factor (SPF) of the heat pump that means a reduction
in electricity consumption (Hepbasli, A. et al. 2008);
the second one is the possibility to store the surplus
heat during the summer season in the underground to
recharge the surrounding volume with a view for the
winter. Incorporation of sustainable energy sources in
space heating system should be evaluated not only in
501
(improvements and/or changes of the old heat supply

and distribution system) less governmental subsidies.
Starting from this value, annual saving respect to the
pre-existent scenario is added. At the end of the plant
life theres the whole amount of money saved (Ms ).
where CAPEX is the capital expenditure; OPEX is the

annual operational expense; Sn is the n scenario and
the apex A means the first kind of economic valuation.
Figure 1. Monthly solar thermal collectors energy gain per
square meter in the Frankfurt area.
terms of the initial cost of the investment, but also

with respect to the annual energy saving, and the
environmental impact.
2
From the existing knowledge, a reference case was

designed and, starting from these data, different systems were developed and compared.
Using the open-source software eQuest, it was possible to calculate the heat load characterization of
the considered multi-family residential building. Solar
irradiation data for the local area, considering the
monthly typical weather, were calculated online on
IES website (Institution for Environment and Sustainability of the European Commission). Data for the
closed vertical loop connected with the ground source
heat pumps are calculated by Earth Energy Designer
software (EED).
2.1
Typology B) Considering a 15 years loan for annual

budget calculation The second typology of economic
valuation considers an installment loan of the total initial cost (less the government subsidies received) for
15 years with monthly payment.
Installment (I) is calculated with the following
formula:
Solar thermal collectors energy supply

calculation
Global irradiance data are estimated for every 15 minutes during a typical day in every chosen month (IES
2012), irradiation value, considering the given inclination and orientation of the solar thermal panels (35 ,
south oriented), see Figure 1 Energy supply (E) for the
solar thermal plant is calculated with the following
formula:
where ASC is the area occupied by solar collectors; CF (collector conversion factor) is 0.817 (IST
2012); he (heat exchanger efficiency) is 0.08 and L
(miscellaneous losses) is 0.06.
where Lr is the loan interest rate; Ain is the annual

number of installments; Ly is the number of the years.
The whole amount of money saved at the end of
plant life is:
Typology C) Considering inflation, interest rate and

annual fuel cost rise for annual budget calculation
The third typology of economic valuation considers,
in addition to the 15 years loan, the annual inflation
(Inf) and the fuel price escalator (FPR). Regarding the
installment calculation is necessary to put into the formula the real interest rate, which is calculated with the
following formula:
This model also calculates the annual budget considering the inflation rate for the specific scenarios
OPEX:
It also considers the fuel price annual rise taking into

account the annual inflation value:
2.2 Three different grading of economic valuation

Typology A) Simple annual budget calculation
The analysis regards the annual economic balance

during the entire life of the plant that is 25 years.
At the beginning is considered the total initial cost
Normally there is no cooling in multi-family dwellings

in the Frankfurt area, so only the building and water
502
MODEL LIMITATIONS
Table 1. Thermal resistance values (R-value).

Structural
component
Pavement
Last floor
ceiling
External
walls
Table 2. Annual heating energy consumption (base load)

and peak load for the building.
Composition
Thickness
(cm)
Overall R-value
(m2 K/W)
Concrete
Concrete
35
25
1.04
2.80
Masonry,
rendering
45
0.78
Space heating
DHW
Total
Annual Consumption
(MWh/a)
Peak load
(kW)
75.7
13.8
93.3
36.2
2.4
39.0
heating have been analyzed. The study considers

shallow geothermal plants with vertical closed loop
systems which permit to include underground heating storage. Inputs, for every scenario took into
account, were changed as little as possible making the
comparison more accurate and representative.
4
EQUEST INPUT DATA
4.1 Location
The building localization is Frankfurt Main Airport
area (latitude 50.11 N, longitude 8.68 E) and the
pre-processed weather data are taken by eQuest
Software database.
Figure 2. Monthly heating energy consumption.
4.2 Structure
The building considered under study is representative
for a large number of dwellings built during the 1960s.
Its a condominium that has 4 floors above the ground
level and 1 below; the total number of households is
8. The structure occupies 207 m2 with a rectangular
shape (23 m 9 m) and an external perimeter of 64 m.
Every floor is 2.74 m high. The total heated area is
760 m2 , every apartment is about 95 m2 . The below
ground floor is unconditioned as stairways and the
building entrance. The short side of the building is
N-S orientated, so the pitched roof has got the axis
oriented E-W.
4.3
Shells
The structure is divided in 5 shells, one for every

floor, with similar characteristics and properties. The
following Table 1 shows the most important thermal
resistance values for pavement, last floor ceiling and
external walls valid.
5
Figure 3. Monthly heating peak load.
The total energy required for building heating is

carefully increased by a 5% due to high-consumption
tenants. Peak load values and DHW base load values
will be considered without an increase. The DHW contribute to the total annual heating energy consumption
is about 15%.
Therefore the total heating energy required for the
building heated area results to 104.6 kWh/(m2 a) and
the heating intensity is 47.65 W/m2 .
SIMULATION RESULTS
The results of the simulation are the monthly thermal base and peak loads for space heating and DHW.
Space and water heating peak load arent simultaneous
so the total peak load is calculated as the sum of the
maximum monthly space peak load with the average
DHW one.
EED INPUT DATA
EED Software simulates the ground vertical closed

loop that the heat pump needs for space and water
heating. The more important output data is the length
of the whole borehole heat exchanger system for the
economic valuation. Every scenario has got the same
input data: the site ground properties, boreholes and
503
pipe characteristics, the heat carrier fluid and the same

simulation period.
The maximum single well length is considered to
100 m; taking into account German guidelines for
ground coupled heat pumps and direct thermal use of
the underground it is also possible to consider a heat
carrier fluid temperature constriction.
Table 3. Borehole
disposition.
Scenario
S2 S3 S4
S5
6.1
heat
exchanger
(BHE)
spatial
BHE
configuration
Number
of BHE
Space
between
BHE (m)
U-configuration
43
Linear 1 3
Ground properties
The soil at Frankfurt Main Airport area mainly consists of sand. The groundwater table lies 5 m below
the ground surface; therefore the average thermal conductivity for the first 100 m of depth is 2.4 W/(m K).
The average volume-related specific heat capacity is
2.45 MJ/(m3 K). The average ground surface temperature is 8.9 C and the geothermal heat flux is considered
0.07 W/m2 .
The annual interest rate for a 15-years-loan is considered 4% with monthly installments. The inflation rate
is 2% and fuel escalation price is 4%.
6.2
7.2 Fuel and electricity prices
Borehole and pipe
ECONOMIC VALUATION INPUT DATA
7.1 Economic values
The closed loop system has got single-U pipes (DN40,

PN6) with a thermal conductivity of 0.42 W/(m K) and
a contact resistance (outer pipe-filling) of 0.01 m K/W;
the velocity flow rate for the heat carrier fluid is
0.005 m3 /s.
The Borehole has a diameter of 15 cm; the spacing
between in and out pipes is 7 cm; the filling thermal
conductivity is 0.9 W/(m K).
Therefore the Reynolds number is 23.515 and the
effective borehole thermal resistance is 0.1461 m K/W.
Home heating oil price in Germany is 0.0863 EUR/l

and the electricity cost is 0.2614 (Europes Energy
Portal 2012).
6.3
7.4 Ground source heat pump (GSHP) costs
Heat carrier fluid
The heat carrier fluid is water with Monopropylglycole

(25%), its thermal conductivity is 0.475 W/(m K), the
specific heat capacity is 3.930 J/(kg K), the density is
1.033 kg/m3 , the viscosity is 0.0079 kg/(m s) and the
freezing point is 10 C.
6.4
Simulation period
7.3 Oil furnace and electric boiler costs

Furnace initial cost is 100 EUR/kW; the maintenance
cost for a new one is 100 EUR/a, instead both for the
old furnace and the electric boiler it is estimated to
250 EUR/a.
Drilling and grouting cost is 60 EUR/m (BHE length);

for loop pipe is 0.5 EUR/m (Edberg, O. et al. 2011);
for fittings and valves is estimated as 10 EUR/m
(of the BHE length). Heat pump cost is calculated
as 250 EUR/kW (Halfhide, T. et al. 2009) and the
plant maintenance cost is considered 7 EUR/kW (La
Franca, G. 2008). Radiant pavement installation costs
are 45 EUR/m2 .
The simulation period is 25 years and the first month

of operation is August.
6.5 VDI 4640 fluid temperature constraints
8.1
The minimum mean fluid temperature (including peak

load) is 0 C, the maximum is 11 C, including peak
load it is raised to 14 C (VDI4640 2010).
The pre-existing heat supply system is a home heating

oil furnace for the space heating and an electric boiler
for domestic hot water (DHW). Rooms are heated by
classic home convective radiators.
6.6
BHE spatial disposition
The U-configuration (whose area is 27 m 18 m with

8 BHE and 27 m 36 m with 12 BHE) means the wells
disposition is around 3 sides of the building in a U
shape.
SCENARIO DESCRIPTION AND ANALYSIS

RESULTS
Scenario 1: pre-existing heat supply system
8.1.1 System efficiency

The furnace is supposed having an overall space heating efficiency of 71%. The home heating oil fuel
calorific value chosen is 10 kWh/l. The electric boiler
for DHW is supposed having an overall efficiency of
81% (Raimondo, L. 2010).
504
Table 4. Energy supply systems heating efficiency

(Raimondo, L. 2010).
Old oil
furnace system
efficiencies
Heat generation
Heat supplying
Heat distribution
Heat regulation
Overall efficiency
0.80
0.96
0.94
0.96
0.71
DHW electric
boiler
0.95
0.90
0.95
0.81
Figure 5. Monthly peak load by typology of generation
for S2.
Table 5. Energy supply systems heating efficiency for S2
(Raimondo, L. 2010) (Zanchini, E. 2009) (Hepbasli, A. et al.
2008).
GSHP
SPF
Heat generation
Heat supplying
Heat distribution
Heat regulation
Figure 4. Monthly heat supply by typology of generation

for S2.
New oil furnace
4
0.88
0.99
0.97
0.99
0.99
0.97
0.99
Table 6. MIN and MAX carrier fluid temperature for base

and peak load.
8.1.2 Economic valuation

The pre-existing plant is already in use so there
isnt a CAPEX cost; instead OPEX amounts to
14.367 EUR/a. After 25 years the whole cost of the
plant is 359.175 EUR.
Mean fluid T
after 25 years
Base load
Peak load
MIN
( C)
MAX
( C)
T
( C)
8.55
8
3.77
3.72
12.32
11.72
Table 7. Energy CAPEX and OPEX for S2, plus design,

permission, site supervision and assurance costs minus
governmental subsidies (in sum: approx. 15% estimated).
8.2 Scenario 2: GSHP plant (capacity 60% of peak

load) coupled with an auxiliary furnace
In this scenario a GSHP is chosen to replace the old
heat supply system. The heat pump capacity is chosen
to 60% of the peak load which is covered by a new
small oil furnace during the winter season. Thanks
to this coupled system solution its possible to save
money by a shorter whole length of wells for the GSHP
vertical close loop.
8.2.1 Heat supply profile
The capacity chosen for heat pump is 23 kW. The
total amount of heat energy supplied by GSHP is
90.6 MWh/a, which is about the 97% of the total
building annual heat energy demand.
The total heat supplied by auxiliary oil furnace is
2.74 MWh/a and the theoretic capacity is calculated to
16.03 kW.
The installation of a radiant pavement improves the
overall heating efficiency to 0.951 (including supplying, distribution and regulation) (Zanchini, E. 2009).
GSHP
Auxiliary oil furnace
Radiant pavement installation
TOTAL
CAPEX
(EUR)
OPEX
(EUR/a)
40051
1603
34200
75854
6437
380
6817
The overall efficiency for a new oil furnace becomes

0.84; the home heating oil fuel calorific value is
10 kWh/l. The heat pump SPF is considered to 4
(Hepbasli, A. et al. 2008).
8.2.3 EED simulation results
According the the results of the EED simulations, the
BHE length needed for this scenario is 554.9 m. In the
following table are shown the maximum and minimum
temperatures for the vertical closed loop heat carrier
fluid.
505

and peak load.
Mean fluid T
after 25 years
Base load
Peak load
MIN
( C)
MAX
( C)
T
( C)
8.20
8
8.80
7.61
17.00
15.61
Table 9. Energy CAPEX and OPEX for S3 plus design,

governmental subsidies.
for S3.

The GSHP plant costs 40.051 EUR, this value is
safely overestimated confronting it to literature data
(Edberg, O. 2011). CAPEX results 75.854 EUR and
OPEX (including maintenance and fuel consumption)
is 6.817 EUR/a.
8.3
Scenario 3: coupled plant; 40 m2 of solar

thermal collectors, GSHP plant and auxiliary
oil furnace; heat storage during summer season
In this scenario a solar thermal plant is coupled with

a ground source heat pump. The energy supplied by
40 m2 of solar collectors permits to keep turned off
the heat pump almost five months per year (from May
to September). Surplus heat during the summer season is exchanged with the ground loop carrier fluid to
recharge thermally the underground. An auxiliary oil
furnace is also considered just for the winter season
peak load. Thanks to this coupled solution its possible
to save money by a shorter whole length of wells for
the heat pump vertical closed loop.
The monthly heat supply profile is made by the
solar plant, the auxiliary oil furnace and GSHP plant
contribution: the total amount of heat energy supplied by GSHP is 71.46 MWh/a, by auxiliary oil
furnace is 2.72 MWh/a and by solar thermal plant is
19.18 MWh/a; the surplus amount of heat generated
during the summer season is 8.54 MWh/a (Figure 6).
The capacity chosen for the heat pump is 23 kW for
the oil furnace is 13.91 kW.
The installation of a radiant pavement improves the
overall heating efficiency to 0.951 (including supplying, distribution and regulation) (Zanchini, E.
2009). The overall efficiency for a new oil furnace
becomes 0.84; the home heating oil fuel calorific
value is 10 kWh/l. The heat pump SPF is considered
4 (Hepbasli, A. et al. 2008). Solar collectors chosen
have a heat conversion efficiency of 0.817 (Institut fr
Solartechnik 2012); heat exchanger efficiency is 0.8
and miscellaneous losses are expected about 6%.
GSHP
Auxiliary oil furnace
Solar thermal plant
Radiant pavement
installation
TOTAL
CAPEX
(EUR)
OPEX
(EUR/a)
35810
1391
26000
34200
5119
380
172
97401
5671

The BHE length needed for this scenario is 484.8 m.
In the following table are shown the maximum and
minimum temperatures for the vertical closed loop
heat carrier fluid.
The GSHP plant costs 35.810 EUR and solar thermal
plant cost is estimated to approx. 26.000 EUR; the
first value is safely overestimated and the second one
is confirmed by literature data (Edberg, O. et al. 2011).
CAPEX results to 97.401 EUR and OPEX (including
maintenance and fuel consumption) is 5.671 EUR/a.
8.4 Scenario 4: coupled plant; 70 m2 of solar
thermal collectors and a GSHP; heat storage
during summer season
In this scenario, instead of the previous one, has a bigger solar thermal plant thanks to a complete roof area
(south oriented) almost full-covered by solar thermal
panels. The energy supplied by 70 m2 of solar collectors permits to keep turned off the heat pump almost
six months per year (from April to September). The
surplus heat during summer season is exchanged with
the ground loop carrier fluid to recharge thermally the
underground.
The monthly heat supply profile is made by the solar
and GSHP contribution: the total amount of heat
energy supplied by GSHP is 66.75 MWh/a and by solar
thermal plant is 26.61 MWh/a; the surplus amount of
heat generated during the summer is 21.90 MWh/a.
The capacity chosen for heat pump is 35.33 kW.
506
and peak load.
Mean fluid T
after 25 years
Base load
Peak load
MIN
( C)
MAX
( C)
T
( C)
8.15
8
13.81
13.81
21.96
21.81
Table 13. Annual heating energy consumption and peak

load for the building for S5.
Space heating
DHW
Total
Annual consumption
(MWh/a)
Peak load
(kW)
12.6
13.8
27.0
11.2
2.4
12.8

GSHP
Solar thermal plant
Radiant pavement installation
TOTAL
CAPEX
(EUR)
OPEX
(EUR/a)
35810
45500
34200
115510
4900
301
5201
Table 12. Thermal insulation data (Odyssee 2011).
Ceiling
Pavement
External
walls
Windows
(type 1)
Windows
(type 1)
Properties
Before
After

for S5.
Thickness [cm]
Overall R-value [m2 K/W]
Thickness [cm]
Thickness [cm]
Overall U-value [W/m2 K]
35
1.04
35
1.04
45
0.78
3.3
43
8.82
73
6.78
62
9.06
1.2
supplied by 20 m2 of solar collectors permits to keep

turned off the heat pump almost five months (from
May to September). Surplus heat during summer season is exchanged with the ground vertical loop carrier
fluid to recharge thermally the underground.
Overall U-value [W/m2 K]
3.3
1.2

heat carrier fluid.
The GSHP plant costs approx. 35.810 EUR and solar
thermal plant cost is estimated to 45.500 EUR; the
first value is a bit underestimated (maybe for the big
reduction of the whole BHE length caused by coupling with the solar thermal plant) and the second
one is confirmed by literature data (Edberg, O. 2011).
CAPEX results to 115.510 EUR and OPEX (including
maintenance and fuel consumption) is 5.201 EUR/a.
8.5.1 Thermal insulation improvement: eQuest

input data
The eQuest simulation data are the same of the before
simulation a part from some average thermal resistance values, in particular for pavements, ceilings,
exterior walls and windows.
8.5.2 Thermal insulation improvement: eQuest
simulation results
The results of this new simulation are the monthly
thermal base and peak loads for space heating. DHW
load profile remains the same. Space and water heating peak load arent simultaneous, so the total peak
load is calculated as the sum of the maximum monthly
space peak load with the average DHW one.
The total energy required for building heating is
carefully increased by a 5% due to high-consumption
tenants. Peak load values and DHW base load values
will be considered without an increase.
Therefore the total heating energy required for the
building heated area results 17.41 kWh/(m2 a) and the
heating intensity 14.81 W/m2 .
8.5 Scenario 5: building thermal insulation;

coupled plant; 20 m2 of solar thermal
collectors and a GSHP; heat storage
during summer season

The monthly heat supply profile is shown in Figure 7.
In this scenario the building is updated with a thermal

insulation, the heat energy supply system is a coupled
plant: a solar thermal and a GSHP one. The energy

The system overall efficiency is exactly the same of
the Scenario 3.
507
Table 14. MIN and MAX carrier fluid temperatures for base
and peak loads.
Mean fluid T
after 25 years
Base load
Peak load
MIN
( C)
MAX
( C)
T
( C)
1.82
8
11.06
9.54
12.88
17.54
Table 16. Total costs of operation (TOC) in comparison to

the reference scenario S1.

CAPEX
(EUR)
Thermal insulation
GSHP
Solar thermal plant
Radial pavement installation
TOTAL
437532
20000
13000
34200
504732
Scenario
TOC*
(EUR)
Saving compared
to S1 (%)
S1
S2
S3
S4
S5
359171
257657
253786
262861
612917
28%
29%
27%
71%
After 25 years.
9.1 Overall BHE length
OPEX
(EUR/a)
The range of the BHE length needed by every scenario

is really wide; from case to case the percentage for
the weight of this voice on the whole investment cost
varies from 53% for S2 to 37% for S3, to 31% for S4
and just 4% for S5 where the thermal insulation cost
is by far dominating.
1213
86
1299
9.2 Economic evaluations

8.5.5 Thermal insulation improvement: EED
simulation results
heat carrier fluid.
The thermal insulation improvement for the building
gives a great contribution to the CAPEX. Following
the guideline study made by Ray Galvin for thermal
update in existing building (Galvin, R. 2011) it is
possible to investigate some thermal renovation examples taken from architect Benjamin Wimmer. He has
obtained a correlating formula which links the thermal
energy use per square meter after the building refit
and the cost of the unit of kWh saved by the thermal
improvement:
As resulted from eQuest simulation output data the Eth

value is 32.71 kWh/(m2 a), therefore the calculation
gives a Ces value of 0,1875 EUR/kWh/a, that multiplied for 25 years and the building area gives the total
amount of the thermal insulation improvement cost:
437.532 EUR, that are 528,4 EUR/m2 .
CAPEX results to 504.732 EUR and OPEX
(including maintenance and fuel consumption) is
1.299 EUR/a.
CONCLUSIONS
The economic evaluation is shown in Table 16; scenario S5 (thermal insulation improvement) is not
economic.
With configuration S3 it is possible to save about
29% of the overall cost in comparison to the reference
scenario S1. Scenarios S2 and S4 show a similar cost
savings for scenario S2 the pay-back time is about 13
years.
9.3 Environmental advantages for S3
Thanks to the S3 configuration its possible to save
about 28.000 kg of carbon dioxide annually. The specific CO2 emission from electricity generation plants
for public power supply is 0.51 kg CO2 /kWh net
(AGEB 2001), instead for kilowatt hour of thermal
energy produced by home heating oil fuel is 0.256 kg
CO2 /kWh (DEFRA 2007).The result after the building
energy supply system upgrade is reducing the annual
carbon dioxide emission by about 70%.
9.4 Final remarks
The presented feasibility study gives a first and important impression about the methodology how to evaluate
the energetic rehabilitation of existing mulit-family
residential buildings. This is europeanwide the sector with the largest energy consumption. It is clear that
there are additional activities necessary in research and
development in order to optimize the technologies and
the methodologies to save energy in the building sector
for sustainability.
REFERENCES
Considering all scenarios is possible to underline some

aspects of the different chosen kinds of investments.
AGEB AG Energibilanzen e.V. 2011. Jahresbericht.

Energieverbrauch in Deutschland im Jahr 2011.
508
BMU Bundesministerium fr Umwelt, Naturschutz und

Reaktorsicherheit (German Federal Ministry for the Environment, Nature Conservation and Nuclear Safety). 2011.
Guidelines on the funding of measures for the use of renewable energies /market incentive program, March 11th.
DEFRA Department for Environment Food and Rural
Affairs. 2007. Act on CO2 calculator: public trial version,
data methodology and assumptions paper.
EEA European Environmental Agency. 2005. Households
consumption and the environment. EEA Report. ISSN
17259177.
EEP Europes Energy Portal. 2012. August 13th.
http://www.energy.eu/.
Edberg, O.; Barker, N. & Stambaugh, J. 2011. Review of
technical information on renewable heat technologies.
Oxfordshire: Department of Energy and Climate Change.
AEA Group.
Galvin, R. Thermal upgrade of existing homes in Germany:
the building code, subsidies and economic efficiency.
CSERGE Working paper EDM 09-11. ISSN 0967-8875.
Halfhide, T.; MacDonald, S.; MacDonald, J. & Titcomb, S.
2009. A feasibility study of geothermal heating and
cooling. Colgate University.
Hepbasli, A. & Kalinci, Y. 2008. A review of heat pump heating system. Renewable and Sustainable Energy Reviews.
IEA International Energy Agency. 2008. Case studies in the
residential sector. ISBN: 978-92-64-04214-8.
IES Institution for Environment and Sustainibility, European Commission. 2012. Solar irradiation
online data, Joint Research Center. August 13th;

http://re.jrc.ec.europa.eu/pvgis/apps/radmonth.php.
IST Institut fr Solar Technik. 2012. Online Collector
Catalogue, Seasonal Performance Factor. August 13th;
http://www.solarenergy.ch/index.php?id=111&L=6.
La Franca, G. 2008. Il condominio geotermico. s.l.: RCI n 10:
pp. 6672.
Odyssee. 2011. SAVE-ODYSSEE Project on energy efficiency indicators. Energy efficiency profile: Germany.
Odyssee. 2012. Free energy indicators in Europe
Households. August 13th; http://www.odyssee-indicators.
org/online-indicators/.
Raimondo, L. 2010. Riqualificazione energetica di edificio residenziale a Rivalta. Seminario sulla riqualificazione energetica degli edifici esistenti. February 12th,
Turin [IT].
VDI 4610 Verein Deutscher Ingenieure. 2012. Energieeffizienz betriebstechnischer Anlagen. 2012, October.
VDI 4640 Verein Deutscher Ingenieure. 2010. Thermische
Nutzung des Untergrundes. 2010, June.
Zanchini, E.; Morini, G.L. & Terlizzese T. 2009. Progetto
di un sistema con pompe di calore accoppiate al terreno
e pannelli solari per un complesso di edifici residenziali a emissioni zero. 2009 September, 8th International Conference on Sustainable Energy Technologies,
Aachen [DE].
509

Long term performance of heat exchanger piles

C.G. Olgun, S.L. Abdelaziz & J.R. Martin
Virginia Tech, Blacksburg, VA, USA
ABSTRACT: Heat exchanger piles utilize the constant temperature and the thermal storage capacity of the
ground for heating and cooling of buildings. Sustainable use of the ground as a renewable energy source depends
on the seasonal energy load balance. One of the critical factors for the sustainable operation of heat exchanger
piles is that a constant temperature of the ground is maintained over seasons. The entire soil mass can be gradually
heated up or cooled down if the energy demand is unbalanced. This paper presents the findings on the longterm performance of heat exchanger piles and their efficiency for areas where the demand is non-symmetrical.
Analyses have been performed to investigate the long-term performance of heat exchanger piles. The analyses
simulated 30 years of pile operation and resulted in significant findings for long-term performance of heat
exchanger piles under different climatic conditions.
INTRODUCTION
Relatively constant temperature of the ground and its

thermal storage capacity can be exploited for heating and cooling of buildings. Traditionally, geothermal
boreholes have utilized this concept for space heating
and cooling. Recently, this ground-coupling concept
has been expanded from geothermal borehole systems
to the use of building foundation elements as heat
exchangers.
Ground temperature remains constant after a depth
of about 6 meters (20 feet) as the near surface soils
act as a thermal insulator as the ambient temperatures
vary seasonally. The value of the ground temperature depends on the regional climate and reflects
the mean annual temperature at that specific location (Kusuda and Achenbach, 1965). In most regions
this constant temperature is about 1024 Celsius
(5075 Fahrenheit).
Heat exchanger piles are designed to access and
exploit the relative constant temperature of the ground
and its energy storage capacity for efficient heating and cooling of buildings. Heat exchanger piles
were first developed in Austria in the 1980s, but have
only recently begun seeing common usage worldwide (Brandl, 2006). In this hybrid system, geothermal
circulation loops are integrated into the deep foundation elements, such as piles, piers, or drilled shafts
that are already designed to provide structural support. Fluid is circulated through the tubes in the heat
exchanger piles. The heat energy is fed into the ground
for cooling in the summer and withdrawn from the
ground for heating in the winter. The temperature
differential between the ground and the outside temperature acts as an energy pathway to harvest stored
ground energy for bringing the building temperature
to comfort zone levels.
Sustainable operation of heat exchanger piles or

any other thermoactive foundation element relies on
maintaining the constant temperature of the ground.
Ground temperature changes can directly affect the
heat exchange capacity and efficiency of operations. It
is critical to evaluate how the heat exchanger pile operation over the life cycle of the structure will change the
temperature gradients around the pile. This is particularly important at regions where the respective heating
and cooling energy demands are not balanced.
Numerical analyses have been performed to investigate the long-term performance of heat exchanger
piles. Energy demand analyses were performed to
derive the base energy load for a typical commercial building at three different climatic environments.
These loads these were applied to a heat exchange pile
to simulate heat exchange operations using a numerical model. The analyses considered thirty years of
heat exchange operations and resulted in significant
findings for long-term performance of heat exchanger
piles under different climatic conditions. This paper
summarizes the numerical analysis approach and the
fundamental results. The main emphasis is given to
practical implications and the effect of energy demand
is investigated for different climatic conditions.
LONG-TERM PERFORMANCE OF HEAT

EXCHANGER PILES
Heat exchanger piles are an alternative energy technology designed to access and exploit the relatively
constant temperature of the ground and its energy
storage capacity for efficient heating and cooling of
buildings. In this concept, the deep foundation elements for structural support are used conjunctively
511
Figure 2. Long-term temperature change around heat

exchanger piles and evolution of temperatures in the ground
(adapted from Ouyang and Soga, 2010).
Figure 1. (a) Schematic showing a building supported by

heat exchanger piles (b) Typical configuration of a heat
exchanger pile with integrated circulation loops.
as heating/cooling elements, as illustrated in Figure 1. As shown, the piles contain circulation tubes
and act as heat exchangers, as heat energy from the
superstructure is circulated through the tubing with
water/antifreeze mix. Heat energy is fed into and withdrawn from the ground for cooling in the summer
and heating in the winter respectively. Heat exchange
is performed with a heat pump. Of particular importance, energy cost savings in typical buildings could
be as much as 80% (Hamada et al., 2001). The added
cost of geothermal loops for heat exchange is relatively small because heat exchanger piles are used
where deep foundation elements are already planned
for structural support.
Ground-source heat pump systems are advantageous over air-source heat pump systems because they
exploit the ground with a more favorable baseline
temperature for heating and cooling of buildings. For
example, as the ambient temperature in Blacksburg,
VA is 2 C (28 F) during the winter, the ground with
an undisturbed temperature of 13 C (55 F) presents
a heat source for raising the room temperature to the
comfort zone levels. A similar temperature differential between the ambient temperature and the constant
ground temperature also provides the potential for
building cooling in the summer. Therefore it is critical
to maintain the stable temperature of the ground over
seasons for long-term sustainability of heat exchange
operations.
The balance of heating and cooling between consecutive episodes ensures the sustainable utilization of
the ground as a renewable energy source. A balanced
thermal exploitation helps offset the thermal effects
of each respective cooling and heating season. Under
certain circumstances heat exchange operations can

generate anisotropic thermal fields in the ground as
shown in Figure 2. This is particularly pronounced
if the heat energy extracted in the winter for space
heating (i.e. ground cooling) is not balanced in comparison to the heat energy injected into the ground in
the summer for cooling purposes (i.e. ground heating).
It is necessary to evaluate how the heat exchange
operations will change the temperature gradients
around the pile over the life cycle of the structure. It is
quite possible that the heat exchange operations over
seasons can alter the temperature fields around the
heat exchanger piles especially when energy demand
is non-symmetrical over seasons. Long-term progression of temperatures around a heat exchanger pile
has implications in terms of the efficiency of heat
exchange operations. Ground can progressively cool
when the energy demand is heating dominated in
colder climates and on the contrary it can progressively
get warmer in climates where energy demand is cooling dominated in warmer climates. It is also possible
that the load carrying capacity of the pile can be compromised due the potential temperature induced effects
on soil compressibility around the pile. Similarly, if
imbalanced thermal loads are induced over long time
periods, and/or if key soil parameters are altered by
these loadings, the foundation system might undergo
differential movements as well. This study focuses on
the long-term performance of heat exchange operations and efficiency of these operations at different
climatic environments.
NUMERICAL ANALYSIS OF LONG-TERM

HEAT EXCHANGE OPERATIONS
Energy demand and temperature progression analyses

were performed in an effort to investigate the sustainable utilization of the ground as a renewable thermal
resource.An office building with four rooms is utilized
512
Figure 3. Layout of the office building used in the energy

demand analyses.
to estimate the seasonal energy demand for heating

and cooling. Layout of the office building is shown
in Figure 3. Energy demand and the estimated ground
thermal loads were further used to model long-term
heating and cooling of the ground around the heat
exchanger piles.
3.1 Energy demand analyses
Building energy demand was estimated for three locations of general interest for long-term performance
of ground coupled heat pump systems; 1) A location where heating and cooling loads are reasonably
balanced (Charlotte, NC), 2) Energy demand is heating dominant (Chicago, IL), and 3) Energy demand is
cooling dominant (Austin, TX). Energy demand calculations utilized the seasonal climate at each specific
location as well as the building properties, heat energy
utilization patterns of the end-users (Abdelaziz, 2013).
Annual temperatures at these locations are shown
in Figure 4. As seen, the mean annual temperature in
Charlotte, NC is 16.0 C with a moderate to warm climate. We anticipate the energy demand in Charlotte to
be seasonally balanced. On the other hand the mean
annual temperature in Chicago, IL is 8.9 C with a
colder climate. Mean annual temperature in Austin,
TX is 20.5 C with a much warmer climate. Chicago
is selected such that the colder climate at this location corresponds to a case where the heating demand
is more dominant where the ground will be used predominantly for heating purposes in the winter, whereas
the cooling in the summer is less pronounced. On the
contrary the warm weather in Austin will be representative for cases where the energy demand is cooling
dominant where the ground will potentially be heated
progressively.
Energy demand analyses were performed for this
office building for the three locations. These analyses considered the ambient temperature variations at
these locations as well as energy consumption patterns
for space heating and cooling. Realistic parameters
Figure 4. Estimated seasonal energy demand curves at

three selected locations and the representative sine wave
approximations.
were used to characterize the building. Therefore,

even though this is a generic analysis, it captures the
important trends that govern the explored concepts
of seasonal energy use at different climates. Ground
thermal loads were calculated by distributing the total
thermal load over the total geothermal loop length
such that the maximum ground thermal load was about
50 W/m. Details of the energy load calculations can be
found inAbdelaziz (2013) and the results are presented
in Figure 5 for these three cities.
It is seen that the duration and amplitudes of heating
and cooling episodes in winter and summer are similar
for Charlotte.A slightly higher cooling load is apparent
from the reported ground thermal loads. Even though
the energy demand is fairly symmetrical between winter and summer in Charlotte, the operational energy of
a geothermal heat pump induces additional energy to
the system and results in the development of a heat
injection bias for the ground thermal load.
513
Figure 6. Cross section of the heat exchanger pile used in

the finite element analyses.
will help us investigate thermal gradients around a heat

exchanger pile for these climatic conditions.
3.2 Numerical simulation of heat exchange

operations
Figure 5. Estimated seasonal energy demand curves at

three selected locations and the representative sine wave
approximations.
The colder climate in Chicago results in an energy

demand dominated by heating in the winter. As seen,
the duration of the heating episode in the colder season
spans over 7 months. This imbalance between heating
and cooling loads presents a potential for progressive
cooling of the ground. On the contrary, the energy
demand is cooling dominant in Austin, representing
the other side of the energy imbalance spectrum. The
cooling dominated nature of the energy demand in
Austin is exemplified by the higher amplitude and
longer duration of the cooling episode compared to
short and low amplitude annual heating cycle. Heat
exchange operations in Austin are likely to result in the
progressive heating of the ground. Collectively these
cases represent different energy demand patterns and
A series of two dimensional finite element analyses were performed using the approach outlined by
Lazzari et al. (2010). Numerical modeling of the heat
exchange operations was performed using COMSOL
(2011). The heat exchanger pile used in the analyses
is 50 cm (20 inches) in diameter, within a homogeneous, isotropic soil medium. The pile was assumed
to have double 32 mm (1.25 inch) PEX tubes placed
on opposing sides. Figure 6 shows the cross section
of the pile used in the analyses. The center-to-center
distance between the upward and the downward sides
of the tubes is 10 cm (4 inches). A free boundary is
assigned at the external boundaries of the model. The
extent of the model was selected such that no thermal
interaction occurs across the external boundaries.
Heat exchange operations over time were simulated by specifying a time dependent energy injection
or extraction rate within the pile using the estimated
ground thermal loads for the office building at the
selected locations.
It is computationally expensive to use the actual
thermal loads to perform the long-term analyses over
30 years of heat exchange operations. Therefore, we
developed an equivalent sine wave approach to represent the ground thermal load for long-term analyses
and applied the annual sine wave consecutively for a
duration of 30 years to simulate long-term operations.
Overlay of the sine wave approximations with
respect to the seasonal energy demands at the selected
cities are shown in Figure 5. The equivalent sine wave
provides a smoother thermal load curve that can be
analyzed with a much larger time step compared to the
30-minute timestep of the analysis with actual thermal load. The equivalent sine wave is selected such
that it results in ground temperatures that match those
from the actual ground thermal load profile. Details
514
Figure 7. Computed pile surface temperatures for 1-year

of heat exchange operations at the three locations using
the actual ground thermal loads and the representative sine
waves.
of the sine wave approximation method are presented

in Abdelaziz (2013).
Validity of the sine wave approximation was confirmed by comparison of finite element analyses using
the 1-year simulations that utilized the thermal loads
based on the actual energy demands and the representative sine waves. The results are shown in Figure 7
where the temperatures at the pile surface are plotted
for the analyses that used the actual ground thermal
loads and the equivalent sine wave loads. The temperatures computed using the equivalent sine wave are in
very good agreement with the results from the model
with the actual thermal loads. This confirms the validity of the equivalent sine wave approach as the actual
thermal loads can be represented with equivalent sine
waves. As mentioned above, the equivalent sine wave
approach allows a much more efficient computational
Figure 8. Progression of ground temperatures around the

heat exchanger pile for long-term operations in Charlotte, NC.
effort by making it possible to perform long-term

analyses over an extended duration.
The equivalent sine waves are further used to perform the finite element analyses over 30 years to
simulate long-term performance for the three selected
locations. The results of the long-term analyses are
shown in Figure 8 where long-term progressions of
temperatures for Charlotte are presented. It can be
seen that the ground temperatures at the pile face
stay relatively unchanged at the end of 30 years of
heat exchange operations. The temperature at the pile
face starts from the initial ground temperature of 16 C
and fluctuates with seasonal episodes of heating and
cooling. Similarly ground temperatures at 1-meter and
2-meters from the pile center are also shown and these
results demonstrate that the temperature fluctuations
515

heat exchanger pile for long-term operations in Chicago, IL.

heat exchanger pile for long-term operations in Austin, TX.
occur at smaller amplitudes at increasing distances

from the pile.
It is seen that there is a slight increase of ground
temperatures at the end of 30 years due to the above
mentioned effect of the additional heating induced by
the geothermal heat pump operation. This additional
heat plays a role to imbalance towards a net increase
of building cooling demand. Nevertheless Charlotte
represents a reasonably well balance between heating
and cooling demands. The temperature fluctuations
at this location are due to the ground heating in the
summer and ground cooling in the winter which apparently offset each other after each seasonal cycle. This
shows that, relative and opposite effects of heating and
cooling episodes are critical for offsetting the temperature induced changes for a sustainable heat exchange
operation.
Ground temperatures as a result of long-term heat

exchange operations in Chicago are shown in Figure 9.
Ground progressively gets cooler in Chicago where
the energy demand is heating dominated where an
unbalanced amount of heat energy is extracted from
the ground over seasons. On the contrary the ground
gets progressively warmer at Austin, as the energy
demand is predominantly cooling dominated as shown
in Figure 10. These two cases where energy demand is
imbalanced represent two opposite end of the energy
demand scenarios. Induced temperature changes at
these two cities show similar but opposite trends, each
reflecting the nature of the energy imbalance for the
corresponding climate.
These results indicate that the seasonal balance
of energy demand during heat exchange operations
can be critical for progression of temperatures around
516
an energy pile. Any significant changes to the initially stable ground temperatures can result in the
system to lose efficiency after several years of operation. It is also quite noteworthy to indicate that
soil compressibility and strength can also be affected
with temperature changes and therefore can affect the
capacity of the energy pile and/or result in unanticipated pile head movements. This study underlines
the potential for such temperature changes as a result
of long-term operations. However any changes to
soil strength/compressibility due to these temperature
changes are beyond the scope of this paper.
4
CONCLUSIONS
Heat exchanger piles are an alternative energy technology designed to access and exploit the relative
constant temperature of the ground and its energy
storage capacity for efficient heating and cooling of
buildings. Heat exchanger piles are finding broader
use around the world and one of the fundamental
challenges is to address their long-term performance
within a sustainable framework. Heat exchange operations can generate anisotropic thermal fields in the
especially if the energy demand is unbalanced over
seasons.
It is critical to maintain the stable temperature of
the ground over seasons for long-term sustainability of
heat exchange operations. A series of energy demand
analyses coupled with finite element analyses were
performed to investigate the long-term performance of
energy piles and progression of temperatures within
the ground around. The analyses indicate that the
nature and degree of temperature progression around
an energy pile is directly related to the seasonal energy
demand. It is seen that temperature changes induced
to the ground are minimal for cases where respective energy demands during winter and summer are
balanced as evidenced by the analyses for Charlotte,
NC. The findings have implications for long-term heat
exchange efficiency of energy piles and can also have
potential effects on soil behavior around the pile.
It is informative to note how sustainable heat
exchange operations are linked to the seasonal energy
demand at different environments. It underlines the
fact that the heat exchange operations can lose
efficiency over time unless preventive measures are

taken. These could include a recharging approach to
seasonally balance the ground temperatures or the
development of sustainable approaches to balance the
energy demands of different structures within an urban
setting.
ACKNOWLEDGEMENTS
This material is based upon work supported by the
National Science Foundation under grant CMMI0928807. This support is greatly appreciated. Any
opinions, conclusions or recommendations expressed
herein are those of the authors and do not necessarily
reflect the views of the National Science Foundation.
REFERENCES
Abdelaziz, S.L. (2013) Behavior of energy piles: deep foundations used as heat exchangers, Ph.D. Dissertation,
Virginia Polytechnic Institute and State University, under
preparation.
Abdelaziz, S.L., Olgun, C.G. and Martin, J.R. (2011) Design
and operational considerations of geothermal energy
piles, Proceedings of Geo-Frontiers 2011: Advances in
Geotechnical Engineering: ASCE Geotechnical Special
Publication, No. 211, J. Han, and D.E. Alzamora, Eds,
pp. 450459.
Brandl, H. (2006). Energy Foundations and other thermoactive ground Structures, Gotechnique. Vol. 56, No. 2,
pp. 81122.
COMSOL (2011) Introduction to COMSOL Multyphysics:
version 4.2, Reference Manual and Tutorial, COMSOL
Inc., Burlington MA.
Hamada, Y., Nakamura, M., Ochifuji, K., Nagano, K. and
Yokoyama, S. (2001) Field performance of a Japanese
low energy home relying on renewable energy, Energy
and Buildings, vol. 33, no. 8, pp. 805814.
Kusuda, T. and Achenbach, P.R. (1965) Earth temperature and thermal diffusivity at selected stations in United
States, American Society of Heating, Refrigerating and
Air-Conditioning Engineers (ASHRAE) Transactions,
Vol. 71, No. 1, pp. 6175.
Lazzari, S., Priarone, A. and Zanchini, E. (2010) Long-term
performance of BHE (borehole heat exchanger) fields
with negligible groundwater movement, Energy, Vol. 35,
No. 12, pp. 49664974.
517

A device for studying simultaneous heat and moisture

movement through soils
B. Hanumantha Rao
Assistant Professor, School of Infrastructure, IIT Bhubaneswar, Bhubaneswar, India
ABSTRACT: Studies on the simultaneous movement of heat and moisture in soils gain a significant importance
in many engineering applications. This paper presents usefulness of a simple device (i.e., a thermal probe) that
can be employed to investigate the simultaneous movement of heat and moisture in kaolin. The device, which
has fabricated in-house, is simple to use, handy and less expensive. It imposes a line heat source thermal flux
into the soil mass that creates a thermal gradient and as a result, moisture movement occurs from higher to lower
temperature regions. It has been observed that the device is very efficient and quite useful in studies related
to simultaneous movement of heat and moisture in soils. Present study, further, also demonstrates the distinct
advantage of the electrical properties to measure the moisture content of a fine-grained soil, in a nondestructive
and noninvasive way.
INTRODUCTION
Simultaneous heat and moisture movement in soils

due to induced thermal gradients bear a considerable
importance in many engineering applications such as
design of buried electrical cables (King and Halfter,
1982), energy conservation schemes (Ewen, 1988),
nuclear waste disposal facilities (Davies and Banerjee,
1980; Zhao, 1992), underground thermal energy storage systems (Eklof and Gehlin, 1996; Gabrielsson
et al., 2000), ground heat exchanger (Roth et al. 2004),
and oil and gas pipelines (Slegel and Davis, 1977),
etc. Excess heat will be generated when these systems are functional. This may result in heat-up of the
surrounding soil mass that creates temperature gradients between high and low heat regions. Consequently,
movement of moisture takes place within these regions
in the soil mass because of the elevated temperature
gradient. However, excess generation of heat may also
cause thermal instability of the soil mass, which influences many engineering properties of soils that vary
as a function of the moisture content (Abdel-Hadi
and Mitchell 1981; Brandon 1989). Hence, it becomes
essential to establish simultaneous heat and moisture
movement characteristics of soils for ensuring safety
of the engineering systems.
Several experimental (Abdel-Hadi and Mitchell
1981; Thomas and Ferguson, 1999; Krishnaiah and
Singh 2003), analytical (Basha and Selvadurai, 1998;
Chen and Ledesma, 2006), mathematical, and numerical models (Prat, 1986; Radhakrishna et al., 1984;
Freitas et al., 1996; Thomas and Li 1997; Wang et al.,
2004) have been developed by earlier researchers
either to determine or to simulate the simultaneous
heat and moisture transfer phenomenon in the soil

mass. However, most of the studies are dealing with
analytical and numerical models and confined, that
too, explaining the fundamental concepts involved it
(Moya et al. 1999). In addition, limited efforts were
devoted towards experimental investigations to establish as well as to understand the phenomenon of
simultaneous heat and moisture movement in the soil
mass. This may be attributed to lack of quick and
convenient techniques for direct measurement of soil
moisture content, precisely. As a result, researchers
have been resorting to conventional oven dry methods
for determination of soil moisture content. Because,
conventional methods adopt to destructive sampling
procedures, obtaining reliable results are difficult due
to errors associated with loss of moisture content during sample collection. Studies also demonstrate that
the most important parameters that affect the rate and
the amount of moisture movement are imposed thermal gradient sources and the soil type (Abdel-Hadi
and Mitchell, 1981; Moya et al. 1999; Krishnaiah and
Singh, 2003).
This paper presents usefulness of a simple device,
a thermal probe, which can be employed for investigating simultaneous heat and moisture movement
characteristics of a fine-grained soil (kaolin). Thermal
probe, which generates a line heat source to impose
thermal gradient in a fine-grained soil, is simple to
use, handy and less expensive. In addition, efforts were
also made to demonstrate the applicability of electrical
properties (i.e., voltage) for measuring the soil moisture content, directly, that too in a non-destructive and
noninvasive way.
519
Table 1. Physical, Chemical and Mineralogical Characteristics of the Soil.

Property
Value
G
Particle size distribution
characteristics
Sand (%)
Silt (%)
Clay (%)
Atterberg limits
W L (%)
W P (%)
I P (%)
W SL (%)
US classification*
Minerals present
Chemical composition
SiO2
Al2 O3
Fe2 O3
TiO2
CaO
K2 O
MgO
P2 O5
SrO
Na2 O
MnO
2.63
46
54
54
27
27
17
CH
Kaolinite, Illite
43.5
33.5
1.5
3.3
0.4
0.07
0.7
0.03
0.00
0.17
0.05
Figure 1. Arrangement of the point electrodes and thermocouples in the soil sample.
ASTM D-2487
EXPERIMENTAL INVESTIGATIONS
Kaolin, which is commercially available, was used in

the present study. The physical characteristics such
as specific gravity, G, with the help of an UltraPycnometer, (Quantachrome, USA), particle size distribution and consistency limits (liquid limit, WL ,
plastic limit, WP , plasticity index, IP , and shrinkage
limit, WSL ) of the soil was determined following the
guidelines provided by ASTM D 5550, ASTM D 42263, ASTM D 4318 and ASTM D 427, respectively, and
the results obtained are presented in Table 1.
The mineralogical and chemical composition of
kaolin was determined by employing X-ray Diffraction Spectrometer (make Phillips 2404, Holland)
and X-ray Fluorescence setup (make Phillips 1410,
Holland), respectively, and results obtained are presented in Table 1.
3 THERMAL FLUX IMPOSITION
TEST SET-UP
A test set-up, fabricated in-house and developed by
Rao and Singh (2010) was used for thermal flux imposition studies. The set-up mainly consists of a thermal
probe, two concentric hollow cylinders (inner and
outer) and a pair of cover and bottom plates. Thermal probe acts as the line heat source of constant heat
flux, Q (Gangadhararao and Singh, 1999; Krishnaiah
and Singh, 2003). The inner cylinder acts as a mould
for preparing the soil sample and is perforated in

radial direction. Due to application of the thermal gradient, the moisture from the soil sample permeates
through these perforations and gets collected in the
outer cylinder.
A 15 mm diameter hole is made in the centre of the
top cover plate through which the thermal probe can be
fitted, as depicted in Fig. 1. The bottom plate contains
two concentric circular grooves, which help inner and
out cylinders to get tight fitted on the bottom plate.
Four pairs of electrodes, located opposite to each other
on top and bottom plates at a radial distance of 6, 15,
25 and 35 mm (in y-direction), respectively, were fitted
with 20 mm long stainless steel electrodes at 10 mm c/c
spacing, as depicted in Fig. 1.
As such, electrodes fitted on top and bottom plates
form a pair of point electrodes (33 mm tip-to-tip distance, d) across which the voltage change due to
expulsion of water in the radial direction, only, was
measured using the soil conductivity meter (Kumar
and Singh 2004).
In addition, the cover plate has been provided with
several 3 mm diameter holes to fit thermocouples,
which are placed at a radial distance of 3.5 (on the surface of the thermal probe), 10, 20, 30 and 40 mm, from
the probe. For preparing the sample, the two cylinders were fitted to the base plate and the soil slurry
of uniform consistency (L/S = 0. 75) was poured into
it. Intermittently, the set-up was tamped to remove
the entrapped air, if any. The top plate was fixed
to the inner cylinder and thermal probe, electrodes
and thermocouples were installed.
Following the methodology proposed by Rao and
Singh (2010), preliminary thermal flux imposition
tests were conducted at different intensities of heat
flux, Q, values (i.e., 50.87, 68.53, 112.67, and
138.6 W/cm, respectively, which correspond to a voltage of 1.4, 1.6, 2.0, and 2.2 V, respectively). Results
obtained from preliminary tests are depicted in Fig. 2.
As depicted in the figure, the variation of temperature, , and voltage, V, measured across a pair of an
520
Figure 3. The variation of V across different electrode positions over prolonged periods of thermal flux imposition in
the sample.
Figure 2. The variation of (a) voltage, and (b) temperature with time corresponding to EP-1 for different heat fluxs
intensities.
electrode position-1 (EP-1 in x-direction) with time, t.

Similar results were obtained for remaining electrode
positions. However, for the sake of brevity results
obtained across EP-1, only, was presented in this study.
In general, it can be observed from the figures that with
increase in Q, V (values recorded in ve numerals) and
increases. However, after approximately five hours of
continuous exposure to heat flux, perturbation in both
V and values, which indicates the thermal instability
of the soil mass, can be noticed. This indicates that to
avoid the thermal instability of kaolin, exposure period
should not exceed beyond five-hour duration. With this
in mind, a five-hour period of heat imposition has been
established as a single cycle of thermal flux imposition. As such, multiple single cycle of thermal flux
impositions tests were conducted by applying a voltage of 2.2 V until the soil specimen attains significant
dry state.
4
As depicted in Figs. 3 and 4, respectively, results of

V and measured across different pairs of electrode
positions after exposure to several multiple single
cycle of thermal flux imposition tests. It can be
noticed from Fig. 3 that for a given time, t, change
Figure 4. The variation of temperature recoded by different

thermocouples for the imposed thermal flux on the sample.
in V between two electrode positions increases with

increase in radial outward distance. Also, change in
V between any two successive single cycle of thermal flux imposition indicates soil has achieved a
certain new state, which is mainly due to the expulsion of moisture within the soil sample at a given
location. This movement of moisture leads to increase
of soil resistance and hence, increase of V with time.
Similarly, built-up at different electrode positions
is increasing with the increase in their position (i.e.,
TC0 > TC1 > TC2 > TC3 > TC4). However, builtup recorded at the surface of the thermal probe (i.e.,
TC0) is prominently higher than its value recorded at
remaining thermocouple positions.
This indicates heat dissipation with an increase in
radial outward distance. Incidentally, all electrodes
positions exhibit a maximum value of V equals to
2700 mV, which is the measure of driest possible
state of a soil sample at approximately 530 hours. It
can be noted that initial temperature (27 C) is almost
same at the beginning of every repetitive single cycle
of thermal flux imposition test, which indicates regain
521
Figure 5. Conceptual diagram showing phenomenon of

simultaneous heat and moisture movement in a soil.
of an ambient condition by the soil specimen before

the next cycle of thermal flux imposition begins.
Fig. 5 demonstrates the conceptual mechanism of
simultaneous heat and moisture movement in a soil
mass. It has been reported that the transportation of
heat from higher temperature locations to lower temperature locations occurs via the combination of all
phases of soil medium such as solid (soil grains,
denoted as A), liquid (pore water, denoted as B), and
gas (vapor, denoted as C). As depicted in the figure,
conduction is the dominant mechanism of movement
of heat especially through solid phase, whereas, conduction coupled with advection found to be principle
mechanisms for the movement of heat through liquid
and gas phases. However, moisture movement in the
soil mass is predominantly by advection phenomenon.
It can also be noted that when the soil is in a fully
saturated state, transportation of moisture occurs in
the form of bulk fluid flow because of all pores get
filled with water. However, when it becomes unsaturated state, attains after exposure to several numbers
of single cycle of thermal flux imposition tests,
movement of moisture takes place mostly by film capillary flow. But, when the soil state is either saturated
or unsaturated, advection mechanism largely controls
movement of moisture in the soil mass. It can also be
noted that because of continuous exposure to thermal
gradients some amount of water will be transformed
into vapor water, which moves from higher to lower
temperature regions by the process of vapor diffusion.
Further, attempts were made to convert the measured voltage into moisture content by employing the
following relationship (Rao and Singh, 2010):
where and are constants, V c and V s correspond

to the critical voltage and the voltage exhibited by the
fully saturated specimen, respectively. Parameters ,
, V c and V s should be dependent on the type of the
soil and for soil WC used in the present study their
Figure 6. The variation of moisture content obtained from

the measured voltage with time.
values are equal to 23, 3.5, 2590 mV and 110 mV,

respectively.
By inputting the data presented in Fig. 3 and
using Eq. 1, moisture content of the soil sample was
computed and results are depicted in Fig. 6. It can
be observed from the figure that moisture content
decreases with increase in time and for a given time
moisture content decreases with increase in radial
outward distance.
CONCLUDING REMARKS
The study demonstrates usefulness and applicability

of a simple device, a thermal probe, which can be
employed efficiently to investigate the simultaneous
movement of heat and moisture in fine-grained soils.
It has been observed that thermal instability, when used
the present developed thermal probe, may occur in a
soil mass after approximately five-hours of continuous exposure to heat flux. Also, attempts have been
made to highlight the distinct advantage of electrical
properties to measure moisture content of a soil mass,
that too in a nondestructive and noninvasive way.
REFERENCES
Abdel-Hadi, O.N. & Mitchell, J.K. 1981. Coupled heat
and water flows around buried cables. Journal of the
Geotechnical Engineering Division 107(11): 14611487.
ASTM D 422-63, 1994. Standard test method for particle
size analysis of soils. Annual Book of ASTM Standards,
Vol. 04.08, ASTM International, West Conshohocken,
PA, USA.
ASTM D 427-93, 1994. Test method for shrinkage factors
of soils by mercury method. Annual Book of ASTM
Standards, Vol. 04.08, ASTM International, West Conshohocken, PA, USA.
ASTM D 4318-93, 1994. Standard test method for liquid
limit, plastic limit and plasticity index of soils. Annual
Book of ASTM Standards, Vol. 04.08, ASTM International, West Conshohocken, PA, USA.
522
ASTM D 5550-00, 2000. Test method for specific gravity of soil solids by gas pycnometer. Annual Book of
ASTM Standards, Vol. 04.08, ASTM International, West
Conshohocken, PA, USA.
Basha, H.A., & Selvadurai, A.P.S. 1998. Heat-induced moisture transport in the vicinity of a spherical heat source.
Int. J. Numer. Anal. Meth. Geomech. 22: 969981.
Chen, G.J. & Ledesma, A. 2006. Coupled heat and moisture
flow in unsaturated soil for a radioactive waste repository.
Unsaturated Soils-2006, 24772487.
Davies, T.G. & Banerjee, P.K. 1980. Constitutive relationships for ocean sediments subjected to stress and temperature gradients. Report UKAEA/2/80 Department of Civil
and Structural Engineering, University College, Cardiff.
Eklof, C. & Gehlin, S. 1996. TEDa mobile equipment for
thermal response test. Masters thesis 1996:198E Sweden:
Lule University of Technology.
Ewen, J. 1988. Thermal instability in gently heated unsaturated sand. Int. J. Heat Mass Transfer 31: 17011710.
Freitas, D.S. Prata, A.T. & De Lima, A.J. 1996. Thermal
performance of underground cables with constant and
cyclic currents in presence of moisture migration in the
surrounding soil. Trans. Power Delivery, IEEE 11(3):
11591170.
Gabrielsson, A. Bergdahl, U. & Moritz, L. 2000. Thermal
energy storage in soils at temperatures reaching 90 C.
Journal of Solar Energy Engineering 122: 38.
Gangadharao, M.V.B.B. & Singh, D.N. 1999, A generalized
relationship to estimate thermal resistivity of soils. Can.
Geotech. J. 36(4): 767773.
Kings, S.Y. & Halfter, N.A. 1982. Underground power cables,
London, Longman.
Krishnaiah, S. & Singh, D.N. 2003. A methodology to determine soil moisture movement due to thermal gradients.
Experimental Thermal and Fluid Science 27: 715721.
Moya, R.E.S. Prata, A.T. & Cunha Neto, J.A.B. 1999.

Experimental analysis of unsteady heat and moisture
transfer around a heated cylinder buried into a porous
medium. Int. J. Heat Mass Transfer 42: 21872198.
Rajeev, K.P. & Singh, D.N. 2004. Instrumentation and testing methodology for detecting Cl- contaminants in soils.
Journal of Testing and Evaluation, ASTM 32(2): 8187.
Rao, B.H. & Singh, D.N. 2010. Application of thermal flux
for establishing soil-water characteristic curve of kaolin.
Geomechanics and Geoengineering: An International
Journal 5(4): 259266.
Roth, P. Georgiev, A. Busso, A. & Barraza, E. 2004. First in
situ determination of ground and borehole thermal properties in Latin America. Renewable Energy 29: 19471963.
Slegel, D.L. & Davis, L.R. 1977. Transient heat and mass
transfer in soils in the vicinity of heated porous pipes.
Journal of Heat Transfer 99: 541621.
Thomas, H.R. & Ferguson, W.F. 1999. A fully coupled
heat and mass transfer model incorporating contaminate
gas transfer in an unsaturated porous medium. Comput.
Geotech. 24(1): 6587.
Thomas, H.R. & Li, C.L.W. 1997. An assessment of model
of heat and moisture transfer in unsaturated soil. Geotechnique 47(1): 113131.
Wang, J. Christakis, N. Patel, M. Cross, M. & Leaper, M.
2004. A computational model of coupled heat and moisture transfer with phase change in granular sugar during varying environmental conditions. Numerical Heat
Transfer Part A: Applications 45(8): 751776.
Zhao, J. 1992. Geohydrological and thermal aspects of deep
underground waste disposal. Proc. of the Second Int.
Conf. on Environmental Issues and Waste Management in
Energy and Minerals Production, Balkema, Rotterdam,
669676.
523

Soil thermal behavior of a horizontal ground source heat system

H.R. Thomas, B.D.P. Hepburn & M. Sedighi
Geoenvironmental Research Centre, Cardiff University, Cardiff, Wales, United Kingdom
ABSTRACT: This paper presents a field scale experimental setup and initial results from a horizontal ground
source heat system installed in Powys, Mid-Wales, UK. The design of the experimental setup is detailed along
with the remote monitoring techniques employed. In addition, an explanation is given of the approach adopted in
the management of the logged data. Initial results from this field experimental work are presented, depicting the
preliminary ground temperature distribution. The results from the experimental site improve the understanding
of the grounds thermal response due to heat extraction within such systems, leading to improvements in the
design of ground source heat systems.
INTRODUCTION
The scope of this work focusses of closed horizontal

ground source heat (GSH) systems which can be used
to generate sustainable space/water heating. The systems fundamentally consist of a buried pipe (or ground
loop), within which a circulatory fluid passes, absorbing thermal energy from the surrounding ground.
Despite the growing popularity of horizontal GSH systems, there is limited information available on ground
performance within operating systems (DECC 2010).
Current available data predominantly focuses on the
mechanical aspects of the systems and the so called
Coefficient Of Performance (COP). Globally there are
some examples of research which has reported the
ground temperature surrounding horizontal GSH systems. Experimental measurements were carried out at
a horizontal GSH system in Elszig, Turkey with a focus
on the system performance (Esen et al., 2005). Wu et al.
(2010) reported both ground temperature and climatic
data at a horizontal system in Oxfordshire, UK.
This lack of data leads naturally to an incomplete
understanding of how the ground responds to thermal
energy extraction from GSH systems. This also results
in limited guidance when designing and implementing
such monitoring schemes. The monitoring arrangement presented in this paper aims to provide a basic
template which may eventually be used to assist future
designs.
Within the current Sustainable Earth Energy
research project (Seren) at the Geoenvironmental
Research Centre, field scale instrumentation of a horizontal GSH system has been undertaken at a site
located in Powys, Wales, UK. A substantial investigation into the ground thermal behavior has been undertaken, incorporating a large number of in-situ sensors.
In addition, a weather station has been installed at
the site to measure the influencing climatic variables
and a site investigation was undertaken to obtain

the physical properties of the ground. Details of the
experimental design are presented in this paper along
with the remote monitoring techniques employed.
The logged data has been structured and managed with a developed data management tool. Initial
ground temperature data recorded at the site are also
presented.
1.1 Experimental site monitoring/installation
The experimental site is located at Ffynnon Gynydd,
Powys, Wales, UK. A horizontal GSH system was
installed as the dominant heating system for a large,
newly built farm house constructed at the site. The
ground source heat system compromises five horizontal ground loops with a total length of approximately
750 m. These five loops were evenly spread across a
40 m wide by 70 m long adjacent south facing field.
The loops were installed at an approximate depth of
2 m, employing a parallel arrangement. A polyethylene
pipe with an internal diameter of 40 mm was used for
all ground loops. The mechanical aspect of the system
compromised of a 16 kW Dimplex heat pump, which
was installed within a plant room in the house. Figure 1
presents a schematic diagram of the GSH system as
installed.
Prior to the installation, the field was covered with
long grass. Post installation, the field returned to the
same condition and use. Additional details of the
ground characteristics are provided in section 1.3. A
series of thermistor strings were installed to monitor the ground temperature surrounding the horizontal
GSH system and the adjacent ground. The installed
thermistor strings focus on one of the five ground
loops. An overall plan schematic showing the longitudinal spacing of the thermistors can be found in
Figure 1.
525
Figure 1. Plan schematic of the horizontal ground source heat system, weather station and longitudinal position of the
thermistor cross-sections installed at the experimental site.
Figure 2. Spatial arrangement of thermistors at the detailed (a) and intermediate (b) cross-sections installed at the
experimental site (facing in the northward direction).
Thermistors measuring the ground temperature

were installed in eight cross-sections. The crosssections were evenly spaced at 10 meter intervals along
the loop length (see Figure 1). Two types of configurations have been adopted in the arranging of the
thermistors. The arrangements include i) a detailed
arrangement utilized in cross-sections AA and BB and
ii) an intermediate arrangement utilized in crosssections A, B, C, D, E and F. As Figure 2 shows, the
detailed thermistor arrangement incorporated a larger
number of thermistors thereby allowing for a more
detailed examination of the ground behavior at these

cross-sections.
Based upon the expected in-situ ground and climatic conditions an RST Instruments Thermistor
String was selected which has a temperature reading
accuracy of 0.1K within the anticipated operational
range and conditions at the site.
In addition to the ground monitoring, climatic variables capable of influencing the thermal behavior of
the soil were identified and monitored at the site. A
weather station was installed at the location specified
526
in Figure 1. The ambient air temperature, solar radiation, relative humidity, wind speed and rainfall are
all monitored at the station. The sensors were selected
based upon the operational ranges, durability and manufacturer specified accuracy, similar to the process
previously highlighted with regards to the thermistors.
Facilities to monitor the system flow rate and fluid
temperature differential across the heat pump were
also installed. The thermal energy extracted by the
GSH system can be calculated using these two system parameters in conjunction with the fluids specific
heat capacity. The system flow rate was measured at a
single point located close to the heat pump unit. Two
fluid temperature sensors were located immediately
either side of the heat pump unit in order to measure
the temperature differential across it.
1.2 Data monitoring and transmission
Sensor readings from the ground, climatic and system
sensors are logged every 15 minutes. A multiplexer
unit was used in combination with a data-logger in
order to simultaneously collect all 120 sensors outputs
(112 ground temperature, 3 system and 5 climatic).
The multiplexer unit was located above the manifold chamber, while the data-logger was located in
the plant room (see Figure 1). The multiplexer unit
was positioned at the base of the monitored ground
loop to reduce the required wiring. The connecting
wire between the multiplexer and data-logger units
was housed in a conduit approximately 1 m beneath
the ground.
The monitoring system was designed to incorporate remote communication techniques which allow
for live communication and remote data downloading
off-site. This communication capability is an important system specification due to the distance between
the experimental site and University campus. The logger was connected to a GPRS (General Packet Radio
Service) modem, allowing the system to be contacted
via the mobile phone network. It is worth note that
the successful installation of such systems depends on
the available mobile network coverage for the intended
data usage at the site.
1.3 Site investigation and soil properties
During the ground sensor installation, a series of
soil samples were retrieved throughout the site, from
depths ranging between 0.5 and 2.2 m. The samples
were analyzed at Cardiff Universitys soil laboratory
in accordance with relevant British Standards methods (BS 2002a and BS 2002b). The soil properties
including moisture content, porosity, bulk unit weight,
particle size distribution and mineral content were
determined. The thermal conductivity and specific
heat capacity were also measured using an in-situ
thermal probe (Decagon Devices, 2012).
Based upon the sample results, the ground can be
classified as medium gravel-medium silt at the southern end of the site and coarse silt fine gravel at the
Table 1.
Measured physical properties of soil.
Property
Average
Standard
Deviation
Thermal conductivity, W m1 K1
Specific heat capacity,
J kg1 K1
Porosity
Bulk Density, Mg m3
Moisture content, %
2.3
1172
0.44
108.7
0.31
1.83
13.1
0.1
0.25
4.3
northern end of the site according to the British Standard classification. This difference in soil classification over a distance of approximately 70 m highlights
the heterogeneity of the ground at the site. Table 1
presents the average physical properties of the soils
along with the calculated standard deviations.
A full topographical survey using Global Positioning System (GPS) technology was undertaken to map
the surrounding area; thermistors, ground loop positions and the location of retrieved samples. The manufacturers specifications of the device used within the
survey stated a horizontal accuracy and precision of
3 mm and a vertical accuracy and precision of 5 mm,
sufficient for the current study.
2 ANALYSIS OF RESULTS
2.1 Database development
It is planned that data collection at 15 minute intervals
will continue throughout the duration of the monitoring scheme, which is anticipated to last a minimum 3
years. This regime will result in a minimum of 105,120
collection intervals yielding 12,614,400 items of data
over the 3 year period. Based on this there is a clear
need for a suitable system that can effectively record,
structure, check and analyze the incoming data. A
data management tool was developed using Microsoft
Server 2012 and written in C#.NET, using LINQ technology. Incoming raw data files remotely downloaded
from the data logger are read, and converted, by the
application into SQL transient data tables. Once read
the data can be interrogated in a variety of ways, checking that no duplications or data holes exist. A master
table containing all site data from the investigation
period can then be updated.
The up-to-date database can be subsequently used
to efficiently analyze the recorded data. A graphical
user interface (GUI) was developed using Windows
Presentation Foundation (WPF) in order to allow the
user to supply specific data constraints (see Figure 3). The input constraints are used to generate
SQL queries, allowing specific data to be analyzed by
date range, temperature range, spatial position or time.
The data sets can be subsequently exported to spreadsheet if necessary, or graphical software, as a .CSV
file (Comma Separated Value), and in Excel format
directly.
527
Figure 3. Screenshot of the GUI used to set analysis constraints.
2.2
Initial results
The installation of the ground instruments and ground

investigation was undertaken in September 2011. The
weather station, multiplexer and data logger were
installed after this date.The system has been fully functioning and remotely logging data continually since
May 2012. The horizontal GSH system was first used
by the occupants in September 2012, upon completion
of the new build property. The ground data recorded
between May and September 2012 therefore represents
the undisturbed conditions, where no thermal energy
was being extracted.
Initial results from the experimental site are presented in the following sections. Figure 4 shows
ground temperature and fluid flow data between the
8th September and 7th November 2012. The plotted
ground temperature data was recorded by two thermistors located in intermediate cross-section A, sensors
denoted 1 and 2 in Figure 2b. The two thermistors are
located vertically in-line with the so called cold leg of
the ground loop, which transports the circulatory fluid
on the outward half of the circulatory fluid path (left
hand side ground loop pipe depicted on Figure 2b). The
two thermistors are located at depths of 2.23 m and
1.51 m beneath the ground surface respectively. The
flow of the circulatory fluid (per 15 minute interval) is
also plotted. The flow data represents the usage of the
GSH system. A combination of the flow and ground
temperature data therefore allows the thermal behavior
of the ground to be inspected, as the system is used.
Figure 4 shows that the fluid flow within the GSH
system increased over the observation period. This
indicates that the system was increasingly used by the

house occupants, coinciding with the onset of winter.
As expected, the ground temperature over the same
period decreased in time. Thermistor 1, located immediately adjacent to the ground loop reported a lower
temperature than thermistor 2.
Between September 24th and 28th no system flow
was recorded due to a system fault. During the
same period, the ground temperature recorded by
both thermistors increased. This indicates that the
gener al decrease in temperature over the inspection
period presented was due to thermal energy extraction from the GSH system and not a change in the
climatic conditions. Beyond September 28th, Figure 4 shows the ground temperatures continued to
decrease.
A contour plot showing the temperature distribution at the detailed cross-section AA, (referred to in
Figures 1 and 2a), is presented in Figure 5. The contour plot has been developed based upon data recorded
on November 7th 2012, corresponding with the final
data points presented in Figure 4. The location of the
sensors and the raw temperature data are plotted on the
contour plot along with the location of the intersection
points of the ground loop (represented as solid white
circles).
The lowest ground temperature was recorded adjacent to the cold leg of the ground loop. The data
shows that the horizontal thermal influence of the
ground loop on the ground extended to approximately
1.75 m from the system origin (mid-point between the
two ground loops shown). It is noted that the heating
528
Figure 4. Ground temperatures as recorded by thermistors 1 and 2 at intermediate cross-section A and the recorded system
flow between the 8th September 2012 and the 7th November 2012.
Figure 5. Contour plot showing the ground temperature distribution at detailed cross-section AA on 7th November 2012.
system had only been in full use for approximately one

week at the time of inspection.
Considering a slightly longer time span than presented above (i.e. from September 8th to December
2nd 2012), a series of observations can be drawn

from the data. The lowest ground temperature recorded
at the previously highlighted thermistors was 6.0 C
on the 1st December, at which time the undisturbed
529
ground temperature at a similar depth was 10.01 C,

4.1 C higher. On December 2nd the horizontal thermal
influence of the ground loop extended to approximately 3 m from the system origin. It can therefore be
said that the observed horizontal ground thermal influence due to heat extraction increased by approximately
1.25 m in the 25 days between 7th November and 2nd
December 2012. For the inspection period between
the 8th September and 2nd December a total energy
extraction of 7,826 kWh was calculated. This concurs
with the expected figure indicating that the fluid temperature and flow data can be used to calculate the
thermal energy extraction.
thermal influence of the ground loop on the ground

grew throughout the observation period.
ACKNOWLEDGEMENTS
The work described in this paper has been carried
out as part of the GRCs (Geoenvironmental Research
Centre) Seren project, which is funded by the Welsh
European Funding Office (WEFO). The financial
support is gratefully acknowledged. The authors also
wish to thank Dr A. Davis from WDS Green Energy
as the industrial partner of the project.
REFERENCES
CONCLUSIONS
A field scale study of a horizontal ground source heat

system has been described in this paper. The methodology adopted to study the in-situ behavior of the ground
due to heat extraction has been explained. The work
described was accompanied by a geotechnical ground
investigation, which indicated the high heterogeneity
of the soil characteristics and physical properties. The
techniques employed to remotely monitor and download the recorded data from the site has been described
in conjunction with a data management tool designed
to structure and interrogate the logged data.
Initial results presented from the study of the ground
behavior during the initial heat extraction period
indicate that the experimental setup is performing
as designed. The remote communication techniques
and data management tool are also fully functioning as designed. The ground data recorded shows
that the ground temperature decreases as thermal
energy is extracted. Reasonable total energy consumption, based upon the anticipated heating load up until
December 2nd 2012, was monitored. The horizontal
British Standards Institution (BSI), 2002a. BS 14688-1:2002

Geotechnical investigation and testing Identification and
classification of soil, Part 1: Identification and description. London, BSI
British Standards Institution (BSI), 2002b. BS 14688-1:2002
Geotechnical investigation and testing Identification and
classification of soil, Part 2: Principles for a classification.
London, BSI
Decagon Devices. 2012. KD2 Thermal Probe technical specifications [Online] Available at:, http://www.decagon.com/
products/environmental-instruments/Thermal-PropertiesInstruments/kd-2-pro/ [Accessed: 19/12/12]
Department of Energy and Climate Change (DECC). 2010.
Getting warmer: a field trial of heat pumps. London
Esen, H., Inalli, M. & Esen, M. 2007. Numerical and experimental analysis of a horizontal ground-coupled heat pump
system, Building and Environment, 42: 11261134.
Gnuplot. 2012. Official gnuplot documentation [Online]
Available at:http://www.gnuplot.info/ [Accessed: 7/12/12]
Wu. Y., Gan, G., Verhoef, A., Vidale, P.L. & Gonzalez,
R.G. 2010. Experimental measurement and numerical
simulation of horizontal-coupled slinky ground source
heat exchangers, Applied Thermal Engineering, 30:
25742583.
530
Natural and anthropogenic bio-chemical

processes within soils and rocks

Mechanical and hydraulic effects of deep roots planting on slope stability

M. Cecconi, V. Pane & P. Napoli
University of Perugia, Italy
C. Zarotti
Prati Armati S.r.l., Opera (MI), Italy
ABSTRACT: Slope vegetation is currently finding wide applications all over the world to mitigate erosive
phenomena and improve slope stability. An innovative natural technology for slope protection has been recently
developed in Italy; this uses only natural perennial grass plants with deep roots and allows operating in different
climatic areas. The technology appears promising with regard to shallow slope instability. The mechanical effects
of vegetation result from the root/soil interaction processes, while the hydraulic effects derive from the reduction
of soil water content enhanced by plant transpiration and root uptake. An original model is proposed in the paper
to take into account the mechanical effects on the soil shear strength; the main assumption of the model regards
the function developed for the root area ratio. The research sets as a final goal the quantitative assessment of
both mechanical and hydraulic effects induced by roots on slope stability.
INTRODUCTION
Over the years and in different world areas, many techniques of slope protection and re-naturalization have
been developed. Slope vegetation represents a quite
innovative technique for the control and mitigation of
water erosion phenomena and, contextually, may play
a positive role in slope stability by simply considering
that the plant roots act as special soil reinforcement.
This aspect of soil vegetation gains much importance
provided that herbaceous deep-roots with high tensile strength properties are well seated into the soil.
The role of vegetation in slope protection from
erosion phenomena has been studied and documented throughout experimental investigations.
Several approaches based on theoretical models,
physical or empirical models have been proposed
in the literature for the quantitative evaluation of erosion (soil loss). Among these, it is worth to cite the
Universal Soil Loss Equation USLE (Wischmeier,
1976; Wischmeier and Smith, 1978). Amongst the
consolidated techniques aimed at reducing soil and
rock erosion developed in the years 195060 and
involving the use of manufactured products such as
geonets, geogrids, fascines, special interest is to be
paid to herbaceous plants characterized by deep roots
with great length (3 m) and extraordinary tensile
strength which may contribute to reduce erosion.
In particular, the aerial portion of such herbaceous
grassy plants dissipates most of kinetic energy of rain
drops, smoothing their erosive action; moreover, even
when the plants are seasonally dried up, an important fraction of rain flows above the aerial portion
of the vegetation in case of intense rainfalls. This
eco-technique, obtained by seeding perennial herbaceous plants directly into the soil, is of simple and
fast installation and does not require any maintenance. Recent studies, also supported by botanists,
agronomists, naturalists and geologists, have highlighted the ability of many herbaceous species to
effectively contrast erosive phenomena, even in bare
and sterile soils where other common species do not
succeed to vegetate.
The implantation of grassy species, made of deep
and thin roots with large tensile strength, may represent
an interesting technique also for the improvement of
slope equilibrium conditions, limitedly to shallow and
potentially unstable soil masses. In order to assess the
additional important role played by slope vegetation,
the soil-roots interaction needs to be specifically analyzed from a mechanical and a hydraulic point of view.
Some of these aspects are dealt with in the following
paragraphs. The subject is complex since several phenomena are involved in the soil/vegetation interaction
and their study requires specific skills in various fields
such as hydraulics, agronomy, soil physics, in addition
to soil mechanics. From a theoretical point of view,
the mass balance equations must be respected in the
position of the problem, by taking into due consideration the concurring phenomena of soil evaporation,
plant transpiration, water runoff along the slope, water
infiltration into the soil and water uptake from roots.
In fact, the root system may affect the hydrologic balance, due to the capacity of the aerial plant system
to reduce water infiltration and soil water content;
the latter reduction causes a further increase in soil
shear strength. The Authors have recently started a
comprehensive study aimed at numerical modeling
533
the soil-root interaction from both mechanical and

hydraulic points of view. The research is motivated
by the need of improving the comprehension of such
concurring phenomena.
The next paragraph is devoted to specifically analyze the mechanical effects of deep roots planting on
slope stability, while the hydraulic effects will be dealt
with in the following sections.
can be successfully fitted through power functions of

the type:
where parameters a and b only depend on the plant

species (Cecconi et al., 2012).
2.1
SOIL-ROOTS MECHANICAL
INTERACTION
As well-known from the specialized literature, the

roots system generally favors the increase of soil
shear strength within the rooted layer (Wu, 1976;
Waldron, 1977; Gray and Leiser, 1989; Gray and Sotir,
1996; Wang and Lee, 1998; Bischetti, 2000; Qi and
Hu, 2006). This improvement mainly depends on two
different processes:
the positive role of plant roots acting as a mechanical soil reinforcement;
the beneficial influence upon the hydrologic balance of the involved area, due to both the capacity
of the aerial plant apparatus to capture (and redirect)
part of the rainfall, and to the ability of the whole
plant system to absorb water from the surrounding
soil and transfer it to the atmosphere through transpiration (root water uptake). The latter mechanism
may yield to an increase of suction and, as a consequence, of the soil shear strength (Vanapalli et al.,
1996; Wan et al., 2011).
In the 70s, Wu (1976) and Waldron (1977) have
proposed a simple mechanical model for the single
root embedded into the soil, assuming that the root
behaves like a linear cylindrical fiber of high tensile
strength.
According to this approach, the increase of soil
shear strength offered by the root, r , is a function of the root tensile strength, t r , the soil friction
angle, , and the ratio between roots cross section, Ar ,
and the rooted soil cross section, Ars . The generalization of the problem faced for the single root to
the whole root-system permeating the soil, results into
Equation (1):
where k is an non-dimensional empirical factor which

is assumed to depend on the plant specie. The root tensile strength, t r , entering Equation (1) can be evaluated
through experimental tensile tests. Data obtained from
tests on several variants of perennial grassy species,
mainly belonging to the botanical families of Graminacae and Leguminosae (Bonfanti and Bischetti,
2009) clearly show that t r sensibly decreases with
increasing root diameter, d, and may attain extremely
high values (tens of MPa) for a root diameter of about
0.51.5 mm. For such roots, the experimental data
Root area ratio
In Equation (1), the quantity Ar /Ars represents the ratio

between the rooted-area, Ar , and the rooted soil crosssection area, Ars . The ratio Ar /Ars is also denoted as
root area ratio, RAR and typically decreases with
depth; this occurrence is generally verified despite
of the complex geometry of the root system which
presents a great variability depending on plant species,
soil properties and profile, but also climatic and environmental conditions (e.g.: Abe and Ziemer, 1991;
Feddes et al., 2001; Osman and Barakbah, 2006; Zuo
et al., 2004; Preti et al., 2010). Due to such intrinsic
variability, the evaluation of the root area ratio requires
careful attention.To give an example, Preti et al. (2010)
have proposed for both Ar (z) and ratio RAR(z) an
exponentially decreasing function with depth z.
In this study, the evaluation of root area ratio for
herbaceous deep-roots is based on the geometrical
model schematically shown in Figure 1 and originally
proposed by Cecconi et al. (2012) for fasciculate root
systems. The geometry consists of a truncated cone
with opening angle, (generally varying in the range
1015 ), surface radius, r, and maximum radius R
attained at maximum depth z r,max . Every single root
has diameter, d i , and cross section, ar,i . The maximum
value of the root area ratio, RARmax , is found at the
ground surface. At any depth z from the ground table,
the rooted area, Ar (z) may be given by:
In Equation (3), the number of roots n(z, d i ) varies with

depth and root diameter; in fact, usually some of the
roots do not reach the maximum depth z r,max .
Thus, in order to evaluate the RAR profile with
depth, quantities Ars and Ar (z) have to be calculated.
From a numerical point of view, Ar (z), has been calculated by firstly dividing the maximum root depth,
z r,max , into 25 layers 1030 cm in thickness (depending on z r,max ), and then assigning, for each layer,
m classes of different diameter. To this aim, a random function generates the root number n(z, d i ) for
each discretized layer and root diameter. Maximum
values of n are approximately set for each meter of
depth, by simply considering that for a single fasciculate root-system the number of roots decreases with
depth. From the available experimental evidence, reasonable values of n vary from nmax = 40 in the top
layer, down to nmin = 10 at 23 m of depth. Then, by
considering that, presumably, the number of fasciculate root systems in a 1 m2 soil area is about 10 30,
534
Figure 1. Geometrical model for fasciculate root systems

(from Cecconi et al., 2012).
the maximum rooted area Ar , at ground table (z = 0),

is approximately Ar,0 = 2000 mm2 for each 1 m2 soil
area.
At this point, for simplicity, it is convenient to
assume a constant rooted soil volume, Ars ; to do that,
the truncated cone is assimilated to a rooted soil
cylinder of equivalent radius R*, i.e.:
In a 1 m2 reference rooted soil area, as the one

considered in a stability analysis performed with the
infinite slope method (see 2.2), one obtains:
Values of Ar,0 and Ars yield to ratio RAR at ground

table, RARmax
= 0.2%, in agreement with other data
available in the literature for similar herbaceous deep
roots (Bischetti, 2000).
As an example, Figure 2 shows for Eragrass species
deep-roots the calculated distribution (Fig. 2a) of
rooted area Ar (z) and the corresponding estimated
RAR profile (Fig. 2b). Although the longest roots could
Figure 2. Fasciculate deep roots of Eragrass species:

numerical results showing a) the distribution of Ar (z) with
depth z and b) the obtained RAR profile.
even reach 3 m of depth, the RAR value becomes negligible at much smaller depths (1.52 m). The analytical
function proposed by Preti et al (2010) to describe the
RAR profile is also plotted in Figure 2b:
having assigned RARmax = 0.2% and b = 1.5 m (average rooting depth); a relatively poor agreement is
found between the two distributions. On the contrary,
the RAR(z) profile proposed herein appears to be better
described by the following function:
with c = 0.6 m, f = 1.3.
535
Figure 3. Single root crossing a shear surface: angle of shear

distortion (from Gray and Ohashi, 1983).
2.2
Shear strength increase from deep-roots
The maximum increase in soil shear strength provided

by the roots, r (z), has been derived by extending the
application of Equation (1) to a heterogeneous roots
system as follows:
In Equation (7), t r,i is the root tensile strength (for a

root diameter d i ) and is the angle of shear distortion
of a single root crossing a potential shear surface (see
Figure 3).
In any case, in order to use Equation (7), a minimum
root length, l min , is required to avoid the occurrence
of slippage before root tensile failure (see Waldron,
1977). In fact, the roots embedment must be sufficiently large, so that the frictional resistance at
the soil/root interface could exceed the tensile root
strength and prevent pull-out of the root itself. By
assuming each single root to a cylindrical elastic fiber
(Gray & Leiser, 1989), the minimum root embedment
l f is given by Equation (8):
Figure 4. Effects of root reinforcement on slope stability;

profiles: a) r vs. depth z; b) safety factor SF vs. z.
friction angle and cohesion c , is given by:
If root lengths are shorter than l min , the root will

slip or pull-out before tensile failure could occur. In
Equation (8), pr is the maximum shear stress at the
soil/root contact.
In this work, equation (7) has been applied in order
to quantify the mechanical effects of embedded deep
roots on slope equilibrium conditions. To this aim,
slopes with relatively shallow (11.5 m) soil coverings underlain by a stiffer stratum may provide a good
case study; this problem can be considered as onedimensional and can be modeled through the infinite
slope method.
By taking into account the mechanical effects of the
deep roots, the safety factor SF r for a soil cover with
where is the slope angle, while z and Dw are respectively the depth of the potential shear surface and the
distance between this depth and the water table, if
present (Dw = 0). In the following, the simple case
of a root-permeated slope of cohesionless pyroclastic
soils above water table (classified as sands, = 38 ,
c = 0) is considered. The average slope angle is rather
large, = 30 , and the root systems consist of Eragrass
species, with an average root diameter d = 0.66 mm
536
(d min = 0.24 mm, d max = 1.08 mm) and a maximum

root depth z r,max = 3 m.
The results of the slope stability analysis are shown
in Figure 4. The calculated RAR(z) profile shown in
Figure 2b leads through Equations (7, 9) to the
profiles r and SF r plotted in Figures 4a and 4b
(green triangles). The favorable root effect is put in
evidence in Figure 4b, when the calculated values of
the factors of safety are compared to those pertaining
to a slightly cemented soil (c = 5 kPa, 10 kPa), in the
absence of roots; the vertical line denotes the critical
value SF = 1.35 obtained for c = 0.
The mechanical effect of roots leads to a noticeable
increase of SF: in particular, such increase is comparable in the upper 12.5 m to the one induced, by
an increase of cohesion (for a not-rooted soil) of about
10 kPa; at larger depths the effects of roots becomes
less noticeable for engineering purposes. Large values
of SF calculated at very small depths (<0.75 m) have
not been plotted, due to their scatter and their small
practical significance.
3
SOIL ROOT INTERACTION:

HYDRAULIC EFFECTS
The analysis of the hydraulic effects in the soil-root

interaction is very complex to understand and model.
In this section, a position of the problem is outlined,
on accounting for different occurring phenomena of
plants evapo-transpiration, water infiltration into the
soil, water runoff along the slope. Following Blight
(2003), the soil water balance equation in the vadose
zone can be symbolically written as:
where, over the reference time t, R is the total rainfall,

IC represents the rainfall intercepted by vegetation,
RO is the runoff, ET indicates the evapo-transpiration
from the shallow subsoil, S is the water stored in the
soil and RE represents any water flow through the
vadose zone. The term L in Equation (10) is introduced
by Blight (2003) to take into account any inaccuracies
in the measurements or in the position of boundary
conditions.
The complete evaluation of Equation (10) requires
available data for each of the six terms. The problem
is thus very complex and, generally, Equation (10) is
used to find one of the cited quantities, starting from
the knowledge of the remaining ones.
3.1
Interception
When soil vegetation is made of deep herbaceous

roots, like those examined in the present study
(e.g. Eragrass fasciculate roots), the quantity IC
requires to be better investigated. For example, this
quantity can be evaluated from the simple relationship
proposed by Morgan & Rickson (1995):
where is the slope angle and CC is the aerial percentage of vegetation cover which, for herbaceous plants,
can very high. Within the above quantity IC, a portion of the intercepted water, IC store , is stored on the
leaves and may later evaporate, while another portion,
denoted in the literature as temporarily intercepted
throughfall, TIF, reaches the ground as stem-flow, SF,
or leaf drainage, LD, i.e:
With similar considerations, in the soil water balance equation the following quantities could be made
explicit, for example, in the form:
where ET 1 is the actual evapotranspiration, ET 2

derives from the water stored on the leaves and then
conveyed to the atmosphere; RO1 represents the runoff
of water reaching directly the soil and RO2 the induced
runoff from water temporarily intercepted by vegetation and then released to soil through stemflow or leaf
drainage.
For the considered vegetation type (grassy deep
roots), the interception storage IC store , which may
attain its maximum value IC max , can be evaluated
through the following equation (Morgan & Rickson,
1995; Merriam, 1973):
where Rcum is the cumulative measured rain in the reference time; however, for grassy covers IC max does not
exceed 1.22.5 mm. From Equation (11) and (12a) it is
then possible to compute the TIF amount of water and
evaluate the stemflow SF according to (van Elewijck,
1988):
where is the average angle of the plant stems with

respect to ground. Finally, the leaf drainage LD can be
simply calculated through Equation (12b). In particular, for herbaceous deep-roots, the angle can be very
small, rendering TIF approximately equal to the leaf
drainage LD.
3.2 Water uptake
For the evaluation of the left side of the mass balance
Equation (10) globally representing the water infiltration into the subsoil different approaches can be
followed: empirical equations, physically-based simplified models and theoretical models including those
537
based on the Richards equation applied to the unsaturated soil. When the water uptake by transpiring roots
from the surrounding soil is also taken into account,
the Richards equation (1931) can be reformulated
according to the equation proposed by Mathur and Rao
(1999):
in which is the volumetric water content, h is the

total hydraulic head and K(h) is the soil hydraulic
conductivity. The quantity S(z, t), denoted as the sink
term (volumetric water content removed per unit time)
mainly depends on the soil water content, on the root
density profile and the potential transpiration rate
(Feddes et al., 1976; Prasad, 1988).
Many sink term functions have been developed in
the literature. Some of them are based on the assumption that the rate of transpiration may equal the whole
water uptake throughout the RAR profile, i.e.:
where TR is the plant transpiration rate and C(z, t) is a

weighing factor depending on the root length density
function, water content profile and more generally soil
water retention properties.
In this study, the function adopted for C(z, t) is the
one proposed by Selim and Iskandar (1978):
Figure 5. Calculated root length density function vs. RAR

(Eragrass species).
based on the daily reference evapotranspiration

(ET 0 ):
where the k coefficients respectively quantify the transpiration capacity of the plant throughout the growth
period, and the soil evaporation capacity as a function of the last rainfall event and soil vegetation. The
product k cb ET 0 represents the daily transpiration TR
[mm/d], see Eq. 16.
4 AN EXAMPLE OF STABILITY ANALYSIS
FOR A ROOT-PERMEATED SOIL
in which L(z) [cm/cm3 ] is the root length density

function.
As a first assumption, L(z) has been simply calculated by dividing the maximum root depth, z r,max ,
into 25 layers of 1030 cm in thickness (hi ), and then
assuming the incidence of a roots number n(z, di ) for
the total layer thickness. By so doing, one obtains:
Similarly to the cited procedure adopted for the evaluation of the RAR(z) profile, a random function generates
the root number n(z, di ) for each discretized layer and
root diameter. As an example, the obtained distribution
of L(z) is plotted as a function of RAR(z) in Figure 5.
Finally, looking again at Equation (10), the evapotranspiration (ET ) term entering the right side of the
equation depends on vegetation type, climatic conditions, soil characteristics, and it is certainly complex
to calculate; this challenge is beyond the scope of this
study. It is sufficient to remind that for a quantitative
assessment of evapo-transpiration the FAO-PenmanMonteith method (Allen et al., 1998) is often used,
This paragraph is devoted to illustrate the results

obtained from a numerical example of application,
aimed at highlighting the mechanical and hydraulic
effects of deep roots planting on slope stability conditions. The case study pointed out in previous section
2.2 of a root-permeated slope ( = 30 ) of pyroclastic soils ( = 38 , c = 0) is again considered. The
root systems consist of Eragrass species (average root
d = 0.66 mm and z r,max = 3 m). For this example, the
assumed initial distributions with depth of soil suction
(s) and volumetric water content () are those shown
with open symbols in Figure 6a and 6b, respectively.
The high value of initial water content, near to initial
saturated conditions, is consistent with the high porosity of pyroclastic covers (
=70%). Once the position of
the problem is stated, in order to assess with the above
approximations the mechanical and hydraulic effects
of roots on slope stability it is possible to follow the
calculation steps below:
538
compute the daily transpiration rate through the

FAO-Penman-Monteith method;
estimate the root water uptake with Equations (16),
(17), (18) and (19). By so doing, a new volumetric
water content profile, (z), is obtained, as the one
plotted in Figure 6a with full symbols. Note that
in this example, (z) represents an average daily
value;
Figure 7. Increase of safety factor (SF): root water uptake

effects (blu circles) and coupled hydraulic and mechanical
effects (green triangles) of deep roots on slope stability.
Figure 6. Effects of root water uptake on a) volumetric water

content and b) suction profiles (sandy soil, Eragrass root
species).
the definition of a proper soil-water retention curve

( = f (s)) for the considered soil (e.g.: Fredlund &
Xing, 1994; Van Genuchten, 1980) allows the evaluation of a modified suction profile from the
rooted-soil, as shown in Figure 6b;
finally, an extended Mohr-Coulomb failure criterion for unsaturated soils (e.g.: Vanapalli et al.,
1996) allows the evaluation of the shear strength
of the unsaturated soil, thus also accounting for the
presence of soil vegetation.
For the case study at hand, by modeling the slope
through the infinite slope method, the safety factor
SF r for a soil cover can be written as follows:
Figure 8. Example of slope vegetation through deep-roots

planting (Central Italy): a) before planting; b) 18 months later.
in the diagram to plot the data. The hydraulic relative

increase of safety factor, SF h /SF, is referred to the
upper scale while the total changes in the safety factor (mechanical + hydraulic, SF/SF) are plotted in
relation to the bottom scale. Note that, although both
profiles decrease with depth, the two scales in abscissa
differ of two orders of magnitudes. This implies that,
for the considered problem, the prevailing effect of
roots on slope equilibrium is the mechanical one.
5
The numerical results of the slope stability analysis
are shown in Figure 7, which shows the increase of
the safety factor, SF, with respect to that calculated
for the not-rooted soil. Two different scales are used
CONCLUDING REMARKS
Erosion phenomena and surface slope instability may

be effectively mitigated by vegetation with deep root
systems. Such technology consisting of seeding
539
Gray, D. H., Sotir, R. B. 1996. Biotechnical and Soil Bioengineering Slope Stabilization: A Practical Guide for Erosion
Control. John Wiley and Sons.
Mathur, S., Rao S., 1999. Modeling water uptake by plant
roots. Journal of Irrigation and Drainage Engineering,
125(3), 159165.
Merriam, R.A. 1973. Fog drip from artificial leaves in a fog
wind tunnel. Water Resources Research, 9, 15918.
Morgan, R.PC. and Rickson, RJ. 1995. Slope stabilization
and Erosion Control: a bioengineering approach, E &
FN SPON, Chapman & Hall.
Osman, N., Barakbah, S.S., 2006. Parameters to predict
slope stabilitysoil water and root profiles. Ecol. Eng. 28,
9095.
Qi, G. and Hu, L. 2006. Study on mechanism and application
of slope protection with vegetation, Chinese Journal of
Rock Mechanics and Engineering, 25(11): 22202225.
Prasad, R. 1988. A linear root water uptake model. Journal
of Hydrology. 99, 297306.
Preti, F., Dania, A., Laio, F. 2010. Root profile assessment
by means of hydrological, pedological and above-ground
vegetation information for bio-engineering purposes.
ACKNOWLEDEGMENTS
Ecological Engineering 36 (2010) 305316.
Richards, L.A. 1931. Capillary conduction of liquids in
porous mediums. Physics, 1, 318333.
The Authors wish to acknowledge Ing. C. Panzani
Selim, H. M., and I. K. Iskandar. 1978. Nitrogen behavior in
(PratiArmati, Italy) for his useful suggestions.
land treatment of wastewater: A simplified model. In State
of Knowledge in Land Treatment of Wastewater, 171179.
Hanover, N.H.: Cold Regions Research and Engineering
REFERENCES
Lab.
Van Elewijek, L., 1988. Influence of leaf and branch slope
Abe, K. and Ziemer, R. 1991. Effect of tree roots on shear
on stemflow amount. British Geomorphological Research
zone: modeling reinforced shear stress, USDA Pacific
Group Symposium on Vegetation and geomorphology,
Southwest Research Station, Forest Service, USA.
Bristol, UK.
Allen, R.G., Pereira L.S., Raes, D., Smith M., 1998.
Van Genuchten, M.Th., 1980. A closed-form equation for
Crop evapotranspiration: guidelines for computing crop
requirements. irrigation and drainage Paper 56, FAO,
Soil Sci. Soc. Am. Journal, 44, 892898.
Rome, Italy, p. 326.
Bischetti, G.B. 2000. Quantificazione delleffetto dellapparato Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E., Clifton, A.W.
1996. Model for the prediction of shear strength with
radicale sulla stabilit dei versanti. Journal of Agricultural
respect to soil suction. Canadian Geotechnical Journal,
Engineering, 2, 7081.
33: 379392.
Blight, G. E. 2003. The vadose zone soil-water balance and
Waldron, L.J. 1977. The shear stress resistance of roottranspiration rates of vegetation. Gotechnique, 53, No. 1,
permeated homogeneous and stratified soil. Soil Sci. Soc.
5564.
Am. J., 41:843849.
Bonfanti, F., Bischetti, G.B. 2009. Resistenza a trazione delle
Wan, Y., Xue, Q., Zhao, Y. 2011. Mechanism study and
radici e modello di interazione terreno-radici. Rapporto
numerical simulation of vegetation affecting the slope
interno (in Italian), Istituto di Idraulica Agraria, Universit
stability, Electronic Journal of Geotechnical Engineering,
degli Studi di Milano.
16: 742751.
Cecconi, M., Pane V., Napoli P., Cattoni E. 2012. Deep roots
Wang, K. and Lee, C. 1998. Brief mechanical analyplanting for surface slope protection. Electronic Journal
sis of bioengineering techniques for slope protection,
of Geotechnical Engineering, 17U: 28092820.
Chinese Journal of Rock Mechanics and Engineering,
Feddes, R. A., Kowalik, P., Malinka, K. K., and Zaradny, H.
17(6): 687691.
1976. Simulation of field water uptake by plants using a
Wischmeier, W.H. 1976. Use and Misuse of the universal soil
soil water dependent root extraction function. Journal of
loss equation, Journal of Soil and Water Conservation, 31:
Hydrology, 31, 1326.
n.1, 59.
Feddes, R.A., Hoff, H., Bruen, M., Dawson, T., de Rosnay,
Wischmeier, W.H., Smith, D.D. 1978. Predicting rainfall
P., Dirmeyer, P., Jackson, R.B., Kabat, P., Kleidon, A.,
erosion losses: A guide to conservation planning. Agr.
Lilly, A., Pitman, A.J., 2001. Modeling root-water uptake
Handbook, n. 537, USDA/Science and Education Adminin hydrological and climate models. Bull. Am. Met. Soc.
istartion, US. Govt. Printing Office, Washington DC.,
82 (12), 27972809.
USA.
Fredlund D.G., Xing A., 1994. Equations for the soil-water
Wu, T. H. 1976. Investigations of landslides on Prince of
characteristic curve. Canadian Geotechnical Journal, 31,
Wales Island, Alaska, Geotech. Eng. Rep 5, Columbus,
521532.
OH: Dept. Civ. Eng., Ohio State Univ. 94 p.
Gray, D. H., Ohashi, H. 1983. Mechanics of fiber reinZuo, Q., Jie, F., Zhang, R., Meng, L.A., 2004. Generalforcement of sand. Journal of Geotechnical Engineering
ized function of wheats root length density distributions.
Division ASCE. Vol. 109, No. 3, pp. 335353.
Vadose Zone J. 3.
Gray, D. H., Leiser, A. T. 1989. Biotechnical slope protection and erosion control. Krieger Publishing Company.
Malabar, Florida.
high-strength roots of perennial herbaceous plants

appears to be effective, simple and maintenance-free.
The increase in soil shear strength mainly depends
on the mechanical reinforcement induced by roots;
however this positive role is limited to quite shallow slope covers (12 m). A further contribute to soil
shear strength, although rather small, is given by water
uptake from roots and consequent increase of soil suction. A typical example of the described technique is
given in Figure 8, showing a steep slope of weathered
basaltic rock, before and after the grass roots planting. From the figure it is clear that the herbaceous
species have effectively prevented surface erosion and
re-naturalized the upper portion of the slope, despite
of the lithological and morphological site conditions
which could appear unfavorable to root embedding and
growing.
540

Chemo-mechanical weathering of calcarenites: Experiments & theory

M.O. Ciantia & C. di Prisco
DICA Politecnico di Milano, P.za L. da Vinci, Milano, Italy
R. Castellanza
DISAT Universit degli Studi di Milano Bicocca, p.za della Scienza, Milano
ABSTRACT: Calcarenite is a very common natural soft rock strongly affected by weathering processes that
markedly reduce its mechanical properties with time. Cliffs and underground cavities formed in calcarenites are
frequently affected by intense erosion phenomena and unexpected collapses. These latter are usually the final
result of a very complex hydro-chemo mechanical process taking place at the micro-scale which can be detected
and analyzed by means of field and laboratory experimental test campaigns. In general, test results mainly put
in evidence: a) a marked and instantaneous reduction in strength and stiffness when macro-pores are filled with
water, b) a slow successive reduction in strength and stiffness occurring in a very long period of time; c) a more
pronounced weakening of the material in case of wetting and drying cycles. In the present work we will discuss
i) some experimental results used to identify the physics of the processes taking place at the micro-scale and ii)
we will describe the main features of a multiscale strain hardening-chemical softening elasto-plastic constitutive
model recently conceived by the authors to simulate the hydro-chemo-mechanical process.
INTRODUCTION
Formations of calcarenite are very common in many

regions of Mediterranean shores. These are affected by
weathering processes induced by a chemo-mechanical
degradation of the rock mass in contact with water.
In Apulia the number of karstic and manmade caves,
excavated within the underground environment, is
remarkably high, thus producing significant risk conditions for the population and the local economy. A
large number of failures of both natural underground
caves and anthropic cavities (Parise and Lollino
(2011)) have been registered in the recent decades. The
Grotta Palazzese cave, located in Polignano a Mare
(BA) (Figure 1) is an exceptional example, where failure could affect both the coastal cliff and a large part
of the historical city. Several factors can be instrumental in the changes of rock masses around caves
prior to failure. For karstic caves, the changes mainly
depend on the chemo-mechanical processes inside the
rock mass and therefore generally develop in the long
term. For man-made caves (Figure 2), instead, failures
frequently occur just few years after the end of the
extraction works (Hutchinson et al., 2002; Castellanza
et al., 2008) as a consequence of: i) the low strength of
the rock, ii) variations of either geometrical or loading conditions around the cave, iii) creep phenomena,
(iv) degradation, with time of the rock properties due
to weathering processes (Andriani and Walsh, 2006,
Ciantia et al., 2013).
In order to cope with weathering processes affecting the calcarenites of either natural or anthropic
Figure 1. Example of Apulian rocky coastal cliff morphology: Grotta Palazzese in Polignano a Mare (BA).
structures, the authors have implemented the following research activities: i) an experimental study for
defining at both micro and macro scales the weathering processes; ii) a theoretical study for developing a
constitutive model suitable for reproducing the previously identified degradation processes; iii) a numerical
study for making available a numerical tool for analsing boundary value problems. In the present work a
synthesis of both experimental and theoretical studies
541
Figure 2. a) Sinkhole in Gallipoli due to the collapse

of an anthropic cave (Polimeno, 2007); b) partial collapse
underground cavities in Canosa di Puglia.
is reported. The entire conceptual path is detailed in

Ciantia (2013).
2
Figure 3. Typical thin-sections (PM) obtained at an optical

microscope where diagenetic bond (DG) and depositional
bond (DP) are clearly visible.
EXPERIMENTAL STUDY
Calcarenites are sedimentary soft rocks formed by the

cementation of calcareous grains originally unbonded.
The grains, with an organogeus origin, are calcareous
shells. At the death of the animals the shells are accumulated at the bottom of the same marine environment
in which they were formed.
In the present study, three different sampling locations, approximately far 100 km among them, in
the Apulian Calcarenite formation has been selected:
Canosa di Puglia (label CP), Gravina in Puglia (label
GP) and Polignano a Mare (label PM).
2.1
Microscale investigations
In order to individuate the possible micro-scale mechanisms responsible of weathering processes, before
performing any test, the first mandatory goal is to
define the material as it is in its current state. Consequently a complete set of micro-investigations have
been performed to define the microstructure both in
2 and 3D and quantify the porosity of the material by means of different methods. The following
techniques were employed: X-ray Micro-ComputerTomography (MCT), Scanning Electron Microscope
(SEM) analyses and Mercury Intrusion Porosimetry
(MIP).
In Figure 3 a typical thin-section obtained from
samples hardened with epoxy resin are photographed
by means of a high resolution camera connected to an
optical microscope, while in Figure 4 a Scanning Electron Microscope image of a thin section together with
punctual chemical analyses (EDS) showing that grain
and bonds are made of calcite (98%) is reported.
By observing the material at the micro scale
two different types of bonds between grains can be
recognized:
i) the diagenetic bonds (DG), identified as the bonds
formed during diagenesis. During the sedimentation process the grains themselves are filled with
the same calcite crystals forming the diagenetic
Figure 4. Typical SEM image (PM) and EDS chemical

analyses of a Calcarenite.
bonds. For this reason usually these bonds, at the

polarized light microscope, appear totally similar
to the grains (black arrows in Figure 3);
ii) the second type of bonds, named depositional bonds
(DP), is composed by less densely packed material
formed by small calcite micrograins held together
by microscopic calcite crystals. This explains the
lighter color of these bonds when observed with
an optical microscope (white arrows in Figure 3).
Often, a bond is the result of the superimposition of
these two types of micro-structures, as the surface
of diagenetic bonds is usually covered by a layer of
loosely packed microcrystals.
As is shown in Figure 4, SEM observations, due
to the high resolution, show that both depositional
and diagenetic bonds are formed by microcrystals
and EDS analyses allow us to state that the chemical
composition of these crystals is CaCO3 .
To visualize in three dimensions the structure of
the material, in Figure 5 SEM picture of the natural microstructure are reported. A very complex and
highly porous structure is evident: the surface of both
grains and bonds is coated by a layer of calcite microcrystals. By looking at the sequence a-b-c-d in Figure 5
542
Figure 7. Schematic structure of bonds and a) saturation

induced STD of DP bonds and b) dissolution driven LTD of
DG bonds.
Figure 5. SEM images of 3D microstructure with different
magnifications.
Figure 6. Porosity values calculated by means of the standard weighing method, MIP, 2D and 3D image analysis for
the eleven calcarenites analysed.
a sort of fractal structure can be recognized: the

microstructure in Figure 5a seems to be coincident
with the microstructure in Figure 5d.
By using the previously mentioned methods, porosity has been calculated for all the material mentioned
above and the corresponding results are displayed in
Figure 6 by means of comparative histograms. Since
all the methods give similar results, the mean value is
used as the final representative porosity value. In general, the calcarenite is characterized by a high porosity
in the range between 4050%, that is a void ratio close
to 1 and by a specific gravity of 2.73.
According to the microscale investigations, the
weathering debonding process taking place when
the rock interacts with water can be described by
means of two microscopic elementary mechanisms
(see Figure 7):
i) the short term debonding (STD) is the result of
the saturation process: water penetrates through the
porous structure, damages the depositional bonds
and the powder composing the DPs falls instantaneously into suspension. The driving variables
of STD process are therefore both the degree of
saturation S r and the amount of powder entrapped
inside the porous matrix forming the powder bonds
(see Ciantia 2013). The latter, hereafter indicated as
sus , is defined as the ratio between the variation in
suspended powder mass and the initial amount of
powder mass:
sus is constrained by the inequality 0 = sus = 1.

When sus = 1, STD process is completed: that is,
all the powder is suspended.
ii) the long term debonding (LTD) is the result of
the chemical dissolution of diagenetic bonds taking place when calcarenite is flooded by water for
a long period of time. With time, the bonds forming the structure of the soft rock are progressively
weathered until the material is transformed into a
calcareous soil.
The process evolution is governed by both the rate
of the dissolution reaction and the ionic composition
of the bulk fluid. The driving variable of LTD is the
accumulated relative mass removal of calcite from the
solid (or reaction progress):
also dis is constrained by the inequality 0 dis 1

(De Groot, 1966). When dis = 1, the reaction is
completed: that is, all the calcite is dissolved.
2.2
Macroscale chemo-mechanical tests
The experimental macroscale test campaign has been

performed to evaluate the hydro-chemical effects on
543
Figure 8. a) UC (Uniaxial Compression) and (b) BR

(Brazilian) tests on CP calcarenite under dry and wet conditions; c) UC strength loss upon wetting vs. porosity for
all the studied calcarenites and d) comparison between the
tensile and compressive strength losses upon wetting.
the mechanical behavior of calcarenites caused by

both STD and LTD debonding mechanisms.
The short term debonding process effects (STD)
have been mainly investigated by performing conventional UC (uniaxial compression) and BR ( indirect
tensile or Brazilian compression) tests on a set of
dry (S r = 0) and wet (S r 1) samples. In Figure 8a,
b axial stress-strain and tensile stress-displacement
curves concerning CP calcarenite samples are illustrated, respectively. In Figure 8c the dependence on
porosity of the %UC strength loss upon wetting is
shown, whereas in Figure 8d, the tensile and compressive strength losses are compared. It is hence
evident that by increasing the material porosity, the
role of depositional bonds seems to become more
important and that STD seems to damage the material
isotropically.
LTD process has been analyzed by performing two
different experimental test series: In test series (a) the
samples are first weathered by means of an acid solution. An acid solution with an average pH = 2.8 has
been used as was proposed by Castellanza and Nova
(2004) to accelerate the process. In this initial phase,
the dissolved mass M is the measured variable and
by means of eq. (2) dis can be evaluated. Subsequently,
the samples are tested under uniaxial compression up
to failure. In Figure 9a the relative stress-strain curves
are shown, whereas in Figure 9b the dependence of
both strength and stiffness on dis is illustrated.
In test series (b) unconfined creep tests under controlled water flux conditions have been performed. The
sample in its dry state is uniaxially loaded up to a fix
value a of vertical stress and subsequently submerged
in water for an imposed time period t (Figure 10).
In its wet state the load is increased up to a and kept
Figure 9. UC tests on pre-weathered specimens: a) stress

strain curves of UCT; b,c) normalized uniaxial compression
strength and stiffness vs. accumulated dissolved mass index.
Figure 10. Picture and hydraulic circuit of the uniaxial

weathering test device (U-WTD).
constant under a controlled water flux (40 l/h) until the

long term debonding induces failure.
In Figure 11, the experimental data obtained by performing a test for which a = 1 MPa, a = 1.5 MPa
(about 90% of the wet ultimate strength for the tested
calcarenite) and t = 2 minutes, are illustrated. In
water 8340 hours are necessary to dissolve 2.9% of
the initial mass and to cause failure. In Figure 11b
both the development of axial strains induced by LTD
544
Figure 12. Bond-representative elementary volume.
Figure 11. U-WTD: experimental results on GP under

water flow.
(phase AB) and the change in stiffness (phase BC) due

to LTD are evident.
under saturated conditions, by omitting the standard

apex notation.
By considering the macro representative elementary volume (REV) as a periodic repetition of both
diagenetic and depositional bonds we can obtain the
following simplified upscaling relation:
3 THEORETICAL STUDY
As was illustrated above both strength and stiffness of
calcarenite are markedly affected by either the falling
into suspension of depositional bonds or chemical dissolution of diagenetic bonds. Independently of which
occurring of the two mechanisms, the macroscopic
strength and deformability variations experienced by
the material can be mainly interpreted on a microstructural viewpoint as a reduction in the bond sectional
area. In contrast, for the sake of simplicity purely
mechanical processes can be micro-structurally interpreted as processes affecting both the deformability
and the strength of the material constituting the bonds
(Gens & Nova, 1993; di Prisco, 1992, Nova et al,
2003).
To define a function describing at the macro-scale
the evolution of the strength of a material experiencing
either STD or LTD processes, the simplified geometry
of Figure 12 can be taken into consideration.
As was previously suggested, the tensile force
transmitted by the bond is proportional to the crosssectional area of the bond itself (Ab ) and to the tensile
strength (evaluated at the micro-scale) ():
where X is an up-scaling constant depending on the

geometrical configuration of the microstructure and
pt is the intersection at the macroscale of the yield
locus with the isotropic axis (Figure 13). In what it
follows, for the sake of simplicity, the state of stress,
both at the micro ad macro scales, is described, even
where n1 and n2 stand for the number of diagenetic and

depositional bonds in the macro-REV along the considered direction, respectively; whereas AMACRO
is the
rev
surface area of the macro-REV perpendicular to the
considered direction. Due to the hypothesis of periodic
repetition of the bond REV, it results that AMACRO
correrev
sponds (according to Figure 12) to the cross-sectional
area of a grain Agr (of radius R0 ) taken n times, where
n represents the number of grains forming the macro
REV. Eq. (4) can thus be recast as
where y1 depends exclusively on the amount of dissolved mass (dis ) since diagenetic bond area can
change for chemical dissolution effects (LTD). On the
other hand y2 depends on both the amount of powder
forming DP (sus ) and the degree of saturation S r . By
indicating with Y the sum of the diagenetic (y1 ) and
depositional (y2 ) bond contributions, the macroscale
tensile strength can be recast as,
where Y is the micro scale weathering function, Y is

the macro weathering function ranging between 0 and
545
hence using eq. (8), we obtain that
and finally, by substituting eqs. (10) and (6) into eq. (7),
the chemo-mechanical coupled hardening law results:
As was mentioned above, the micro scale weathering function results to be the sum of two distinct
functions y1 (dis ) and y2 (sus , Sr ). By assuming a
circular cross sectional area for the bonds, the two
ratios appearing in eq. (5) result:
where the first and the second lines of eq. (12) represent the counterpart of strength provided by the
diagenetic and depositional bonds, respectively. By
interpreting for the sake of simplicity the dissolution
as an isotropic thinning of the calcarenite, the variation
of the volume of the solid can be expressed as,
Figure 13. a) Homothetic shrinkage of yield locus induced

by STD and LTD and (b) zoom of tensile range.
1 and P t is a scalar function monotonically decreasing with plastic deformations. The rate of macroscopic
tensile strength results
where d dia (t) represents the thickness of the dissolved

layer at a given instant of time (Figure 13) and A
corresponds to the reacting surface area. Using the
definition of the solid matter density:
and by substituting eq. (2) and (14) into eq. (13) we

obtain:
Macroscopic observations made on calcarenite by
Lagioia & Nova (1995) and Nova et al. (2003) show
that the variation of tensile strength due to purely
mechanical effects depends on both volumetric and
deviatoric plastic deformation rates. The relation they
proposed reads,
and since it fits in a satisfactory manner experimental

data, it is used, by means of a down-scaling procedure, to determine the micro stress ()
and its rate ( ).
In fact, considering only mechanical effects, eq. (7)
results
where A is the specific surface area in cm2 /cm3 of

the material. By indicating with dis,cr , the value of
dis for which all bonds dissolve (d dia = r 0 ) and by
substituting eq. (15) into eq. (12):
for dis dis,cr , while y1 = 0 for the granular state

(dis > dis,cr ). Analogously in case of STD:
546
where (i) the critical value of suspended mass is equal

to unity (sus,cr = 1) and (ii) the dependence on the
degree of saturation is hidden inside the evolution of
sus , i.e. sus = sus (S r ). The micro scale weathering
function can be thus expressed as:
A downscaling procedure has been used to determine

r2
n1
r2
n2
the two multiplying terms ( 0,dia

and 0,sus
) in eq.
R20 n
R20 n
(18). The two terms can be evaluated using experimental results on an intact material (dis = 0) under both
dry (sus = 0) and wet (sus = 1) conditions. In these
two conditions, the micro-scale weathering function
(eq. (18)) results:
where Y d and Y w represent the micro-weathering

functions under dry and wet conditions, respectively.
By introducing eqs. (6) into eq. (19), we obtain:
Figure 14. Calibration of weathering function Y (xdis ,

xsus , Sr ): 3D view of experiments and weathering function
(data from Ciantia 2013).
where dis is a directly measurable quantity strictly

depending on the kinetics of dissolution (Ciantia &
Hueckel, 2013).
By assuming the capillary effects to be negligible with respect to the importance of the depositional
bonds, it becomes possible to derive the second term
between parentheses of eq. (18) phenomenologically
(Figure 14). The data reported in the previously cited
figure were in fact obtained by performing uniaxial
compression tests on partially saturated samples.
The final result is (Figure 15)
where Y d and Y w represent the macro-weathering

functions under dry and wet conditions, respectively.
From Figure 14b and assuming an omothetical hardening for the yield function, we can thus write
for dis dis,cr (bonded material) and

d
w
where c0
and c0
are the uniaxial compression
strength of the material under dry and wet conditions,
respectively. Substituting eq. (21) in (20) we finally
derive:
and by substituting eq. (22) into eq. (18) and using

eq. (6), the macro weathering function reads
for dis > dis,cr (granular material). The interpolation

function defining the dependence of Y on the saturation index is likely to be independent of the porosity
of the material, whereas P(sus ) is a function depending on the amount of suspended mass (sus ) present
in the porous rock at a given time. For a closed system in which the total powder mass remains constant
(no dissolution is admitted), P = 1, while if the system
is open, the problem becomes coupled to the transport
processes of water. Andriani and Walsh (2007) showed
that with wetting and drying cycles for open systems
(system where suspended mass can be expelled from
the sample) the increase in strength upon redrying is
not fully recovered.
547
the suspended powder mass and the saturation index

are related to the evolution of the unique hardening
parameter chosen to summarize the mechanical effect
of weathering on the considered material.
REFERENCES
Figure 15. Phase diagram for calcarenite.
The extended hardening law can now be expressed

in terms of macroscale variables by substituting eq. (6)
in eq. (11). The final result is:
CONCLUDING REMARKS
In this paper the problem of conceiving an ad hoc

constitutive model for simulating the weathering of
calcarenites is approached. To achieve such a goal,
the authors have decided to employ the theoretical framework of strain hardening elasto-plasticity
and in particular the strategy already introduced by
some authors essentially based on the idea of starting from standard soil elasto-plastic strain hardening
constitutive relationships and modifying both the yield
function and the hardening rule.
The microstructural experimental evidence put in
evidence the presence of two different types of bonds
among particles characterized by a different value
of density and by a different way of interacting
with water: the depositional bonds and the diagenetic
bonds.
In this paper a sort of simplified upscaling/
downscaling procedure has been proposed to take into
account the process of debonding induced by both the
saturation of the material and the dissolution of diagenetic bonds caused by the ionic composition of the
bulk fluid.
This has allowed to the authors to define a suitable
hardening rule in which the cumulated dissolved mass,
Andriani, G.F., Walsh, N., 2006. Physical properties and

textural parameters of calcarenitic rocks: qualitative and
quantitative evaluations. Eng. Geol. 67, 515.
Andriani, G. F. & Walsh, N. 2007. The effects of wetting
and drying, and marine salt crystallization on calcarenite
rocks used as building material in historic monuments.
Geological Society, London, Special Pubblication 271,
179188.
Castellanza, R. & Nova, R. (2004). Oedometric Tests on
Artificially Weathered Carbonatic Soft Rocks. Journal of
Geotechnical and Geoenvironmental Engineering ASCE
130, n7, 728739.
Castellanza, R., Gerolymatou, E., Nova, R., 2008. An attempt
to predict the failure time of abandoned mine pillars. Rock
Mech. Rock Eng. 41 (3), 377401.
Ciantia, M. O. 2013. Multiscale hydro-chemo-mechanical
modelling of the weathering of calcareous rocks: an
experimental theoretical and numerical study. PhD thesis,
Politecnico di Milano, Italy.
Ciantia, M. O. & Hueckel, T. 2013. Weathering of submerged
stressed calcarenites: chemo-mechanical coupling mechanisms, Gotechnique, [10.1680/geot.SIP13.P.024].
Ciantia, M. O., Castellanza, R., di Prisco, C. & Hueckel, T.
2013. Experimental Methodology for Chemo-mechanical
Weathering of Calcarenites. 331336 (online ISBN-9783-642-32492-5)
De Groot, S. R. 1966. Termodynamics of irriversible
processes, Amsterdam: North Holland.
Di Prisco, C., Matiotti, R. & Nova, R. (1992). A mathematical model of grouted sand behaviour, Proc. NUMOG IV,
Swansea, UK, Pande, G. and Pietrusczczak, S., Balkema,
Rotterdam, 2535.
Hutchinson, D.J., Phillips, C., Cascante, G., 2002. Risk considerations for crown pillar stability assessment for mine
closure planning. Geotech. Geol. Eng. 20, 4163.
Parise, M. & Lollino, P. 2011. A preliminary analysis of
failure mechanisms in karst and man-made underground
caves in Southern Italy. Geomorphology 134, 132143.
Gens, A. & Nova, R. 1993. Conceptual bases for a constitutive model for bonded soils and weak rocks, Hard
Soils-Soft Rocks, Athens, Greece, Anagnostopoulos and
others, Balkema, Rotterdam, 485494.
Lagioia, R. & Nova, R. (1995). An experimental and theoretical study of the behavior of a calcarenite in triaxial
compression, Gotechnique 45, n4, 633648.
Nova, R., Castellanza, R. & Tamagnini, C. (2003). A constitutive model for bonded geomaterials subject to mechanical
and or chemical degradation. Int. J. Num. Anal. Meth.
Geomech. 27, n9, 705732.
Polimeno, A. 2007. Il crollo di via Firenze in Gallipoli.
lintervento dei vigili del fuoco, Geologi e Territorio,
4-2006/1-2007, 1319 (in Italian).
548

Chemo-mechanical behaviour of a low activity clay

A. Ferrari, P. Witteveen & L. Laloui
Laboratory for Soil Mechanics, Swiss Federal Institute of Technology Lausanne, Switzerland
ABSTRACT: This paper presents the results of an experimental program on the chemo-mechanical behaviour
of mixtures of a low activity clay (illite powder) and sodium chloride solutions at different concentrations. A
correlation between the ionic concentrations of the pore liquid and the osmotic suction of the soil is proposed
in order to present and discus the results in terms of osmotic suction. The experimental programme consists of
oedometric tests involving changes in the vertical stress and in the osmotic suction. The experimental results point
out a correlation between the osmotic suction and the yield stress. The experimental results are used to extend
a constitutive framework for soils in order to take the observed chemo-mechanical couplings into account. The
numerical model has been calibrated for the illite using the parameters obtained through tests under mechanical
loading at a constant osmotic suction and validated using more elaborate stress paths.
INTRODUCTION
Several environmental and civil engineering applications (such as the salinification of groundwater, wellbore stability problems and the storage of
nuclear wastes) require a deep understanding of the
chemo-mechanical behaviour of soils. Despite the
vast number of situations in which couplings between
chemistry and mechanics occur, the available experimental results are quite limited and the available
constitutive models rely on limited evidence. Moreover, available experimental results are essentially
focused on highly active clays.
Clay aggregates exhibit a negative electrical charge
all along their outer boundaries, leading clays to
attract positively charged ions from the pore liquid
and the diffuse double layer to be constructed (Gouy,
1910; Chapman, 1913; Verwey & Overbeek, 1948;
Mitchell & Soga, 2005). A high electrical potential
results from the high concentration of cations and
the low concentration of anions at the clay boundaries. The electrical potential causes the clay to act
as a non-ideal semi-permeable membrane. The resulting osmotic processes have been readily described in
the literature (e.g., Bolt, 1955; Bolt & Miller 1956;
Mesri & Olson 1970, 1971; Mitchell et al. 1973,
Sridharan & Venkatappa Rao 1973, 1979; Marine &
Fritz, 1981; Fritz, 1986; Barbour & Fredlund, 1989;
Loret et al., 2002).
The migration of ions or water can affect the
behavior of clays. As a consequence of the chemical
concentration gradient, ions can penetrate in the pores,
due to the negatively charged clay aggregate boundaries that will attract the positively charged cations.
As a consequence, the inter-aggregate repulsive forces
decrease, leading the soil volume to decrease (chemical consolidation). If the higher concentration is on
the outside of the considered soil mass, water flows
out of the soil leading to an overall volume reduction
(osmotic consolidation). These two main phenomena
may occur in parallel. Several authors use different terminology for these phenomena; Barbour & Fredlund
(1989) mention osmotic consolidation and osmotically induced consolidation, whereas Kaczmarek &
Hueckel (1998) discuss chemical consolidation and
chemico-osmotic consolidation.
Consolidation due to a difference in chemical
composition of the pore water was investigated by
Fernandez & Quigley (1985, 1991) and later by
Hueckel (1997) and Boukpeti et al. (2004). They studied the influence of a single organic contaminant on
the compressibility properties of Sarnia clay, which
is a Na-montmorillonite with traces of Ca-smectite.
Hueckel (1997) and Boukpeti et al. (2004) mainly
modelled the evolution of the yield limit observing
that the yield stress decreases as the contaminant
concentration increases.
Di Maio (1996), Di Maio & Fenelli (1997) and Di
Maio & Onorati (1999) described a series of experiments with the aim of evidencing and quantifying the
aforementioned consolidation phenomena on a bentonite, yielding large deformations (up to 15% due to a
change in pore liquid only). Constitutive models based
on those results were proposed by Loret et al. (2002),
Gajo & Loret (2003), Gajo et al. (2002), Gens (2010)
and Guimares et al. (2007).
With increasing interest in the geomechanical applications in which the chemistry of the pore liquid plays
a major role, advanced chemo-mechanical frameworks are increasingly important. This study aimed
549
to extend the experimental database on the subject of one-dimensional consolidation induced by a

change in the chemical composition of the pore liquid for a low-activity clay. Experimental results from
an extended laboratory programme are presented.
A coupled elastoplastic, chemo-mechanical model is
then proposed and used to model the experimental
results. In contrast to previous studies, where the (ion)
concentration was considered as the governing variable, the present model is defined for changes in
osmotic suction, which allows the results to be presented and discussed independently of the pore liquid
components.
EXPERIMENTAL PROGRAMME
Experimental evidence was collected on (i) the evolution law for the yield limit stress with respect to
the pore liquid chemistry, and (ii) deformation due
to chemical solicitations (Witteveen et al. 2013). The
programme involved a series of chemo-mechanical
experiments in which the osmotic suction was controlled through the sodium chloride concentration in
the pore water.
2.1 Tested material and specimens preparation

Solutions of sodium chloride (NaCl) in distilled
water at different ion concentrations (0.06.0 M), were
mixed with a non-swelling illite powder. The tested
illite has a liquid limit wl = 0.54 and a plastic limit
wp = 0.30. The specific gravity is G s = 2.65. The
grain-size distribution presents a sand-size fraction of
12%, a silt-size fraction of 21% and a clay-size fraction of 67%. Because of the fairly low activity of the
illite, the material is quite different from that in other
studies reported in the literature.
2.2
Measurement of the osmotic suction
A correlation between the osmotic suction () and ion

concentration was obtained first. Filter paper method,
using Schleicher and Schuell no. 589/2 40.5 mm filter papers was used to measure the matric and total
suction of the clay powder-pore liquid mixtures. The
suction values of the filter paper were calibrated with
respect to the water content for the matric suction
(difference between the pore air and the pore water
pressures) and the total suction (matric suction plus
osmotic suction) separately.
Mixtures were prepared with distilled water at
different water contents; additional mixtures were prepared with a 4.0 M solution at different water contents.
Additionally, the total suction of clay powder-pore
liquid mixtures prepared at different NaCl target concentrations (distilled water, 0.5, 1.0, 2.0, 4.0 and 6.0 M)
and at different water contents was determined using
a chilled-mirror dew-point psychrometer (Leong et al.
2003; Cardoso et al. 2007). The water content of all
Figure 1. Applied stress paths for the oedometer tests

involving changes in osmotic suction and mechanical
loading.
mixtures was determined after the tests by oven drying

at 105 C.
2.3 Chemo-mechanical oedometric tests

Chemo-mechanical oedometric tests involving changes
in the vertically applied mechanical stress and osmotic
suction were conducted. The specimens were prepared
by compacting a clay powder-pore liquid mixture
directly in the oedometer rings (60 mm in diameter and 10 mm high), targeting an initial void ratio
of e0 = 0.950.02. The initial water content was
w0 = 0.390.05, corresponding to fully saturated conditions. The typical lading time for mechanical increments was 24 hours.
The osmotic suction and vertically applied mechanical stress (corresponding to the vertical effective
stress) were the independent driving variables. The
experimental programme was designed such that
the influence of osmotic suction on the mechanical
behaviour could be researched first. To this end, a
series of seven oedometer tests at different but constant values of osmotic suction of the pore liquid were
conducted. Two of the samples were prepared with distilled water; the five others were prepared with one of
the aforementioned solutions.
The influence of changes in osmotic suction was
investigated through two experiments in which there
was an incremental concentration change at different
but constant mechanical stress levels (Fig. 1). Two
oedometric samples were prepared with distilled water
and loaded mechanically up to different vertical stress
levels (v = 30 kPa and v = 77 kPa). The chemical
loading was achieved by replacing the oedometer cell
fluid with an electrolyte of higher osmotic suction and
monitoring the deformation of the sample until it stabilised after two to four days. The surrounding solution
550
Figure 3. Results of four oedometric tests under mechanical

loading at different osmotic suctions.
Figure 2. Total and matric suction measurements on mixtures of illite powder and either distilled water or a 4.0 M
sodium chloride solution.
was frequently renewed to assure the ion concentration

in the sample. A saturated sodium chloride solution
was applied in the last step. Afterwards, these chemically loaded samples were subjected to the remaining
loading increments of the conventional mechanical
oedometric test.
The test results show that the osmotic suction measured for the specimens prepared with distilled water
is greater than zero as a consequence of the dissolution of ions from the clay powder, as the illite can
exchange cations. An average initial osmotic suction
0 = 0.61 MPa is assumed.
The results of the suction measurements with both
the filter paper method and dew-point psychrometer
allowed to establish the following empirical relationship for the osmotic suction (in MPa) in a mixture of
illite and a sodium chloride solution as a function of
the ion concentration (c) of that solution:
3 TEST RESULTS
3.2 Chemo-mechanical oedometric tests
3.1 Suction measurements
Fig. 2 depicts the results of the filter paper method and
the psychrometric readings for the mixtures prepared
with distilled water and with the 4.0 M solution. The
total suction [MPa] and matric suction sm [MPa]
are plotted against the water content of the mixtures of
illite powder with either distilled water or a 4.0 M NaCl
solution. In the distilled water mixtures, the matric
suction is two orders of magnitude smaller than the
total suction. This effect is even larger (three to four
orders of magnitude) for the mixtures prepared with
the salt solution because the presence of salt increases
the osmotic component of the suction, while the matric
suction remains fairly constant. These results indicate that the presence of salt in the pore water does
not influence significantly the matric suction. Thus,
the measurements in terms of total suction are subsequently assumed to be representative of the osmotic
component and in the following the matric suction
will be neglected. However, this assumption might be
different for other types of clay and salt because the
interaction between clay and salt might be of a different
nature.
The results of four of the tests under mechanical loading are plotted in Fig. 3. The results show that the
initial stiffness of the material decreases as the osmotic
suction increases. The yield stress decreases as the
osmotic suction increases while the slopes of the normal consolidation line (NCL) and unloading-reloading
line (URL) do not change significantly with respect to
the osmotic suction.
The evolution of the yield stress with the osmotic
suction is depicted in Fig. 4. The trend is similar to
the one proposed for the preconsolidation pressure
as a function of the concentration of one (organic)
contaminant by Hueckel (1997) and Boukpeti et al.
(2004). The following mathematical form is used to
fit the results:
where c is the vertical yield stress [kPa], is the

chemical evolution factor [], is the current osmotic
suction [MPa], 0 = 0.61 MPa is the osmotic suction
551
Advanced Constitutive Model for Environmental

Geomechanics (Laloui et al., 2010). The current
extension, ACMEG-C: Chemical Effect, is able to
account for changes in osmotic suction as well as the
evolution of the elastic and plastic material parameters
and the plastic deformation induced by these changes.
ACMEG-C is a soil plasticity-based model for
chemo-mechanical formulations. A complete presentation of the model can be found in Witteveen et al.
(2013). In the following, the main features of the model
are recalled. The total strain increment, dij , to be
divided into chemo-elastic, deij , and chemo-plastic
components,
The constitutive model is conceived in terms of the

following conjugate stress-strain variables:
Figure 4. Correlation between the vertical yield stress and

osmotic suction.
where p is the mean effective stress, q is the deviatoric

stress, v is the volumetric strain, q is the deviatoric
strain, J 2D and I 2D are the second invariants of the
stress and strain tensor, respectively.
4.1.1 Chemo-hypo-elasticity
The elastic strain is decomposed into chemo-elastic
volumetric (dev ) and deviatoric (ded ) strains. They are
computed as follows:
The hypo-elastic moduli K and G are given by

Figure 5. Results of the oedometric tests under mechanical
and chemical loading.
for the samples prepared with distilled water and c0

is the vertical yield stress at 0 . The results of the
two tests including chemical loading increments are
depicted in Fig. 5. In both cases, the clay underwent
consolidation due to a change in the osmotic suction.
4 ACMEG-C: A CHEMO-MECHANICAL
CONSTITUTIVE MODEL
4.1
Constitutive model
The experimental results of the oedometric tests are

used to formulate a constitutive framework with the
aim of reproducing and quantitatively predicting the
observed chemo-mechanical behaviour. This framework is based on the Hujeux model (Hujeux, 1979).
Existing extensions of it are known under ACMEG
where K ref and G ref are the reference bulk and shear
moduli of the material, respectively, determined at a
reference mean effective stress pref , and ne is a material
parameter.
According to the experimental findings, the reference bulk modulus K ref is assumed to evolve with
the osmotic suction. The following power function is
proposed:
where K ref ,0 is the reference bulk modulus for a test at

a reference osmotic suction.
552
4.1.2 Chemo-plasticity
In the Hujeux models plasticity framework the total
plastic strain increment is a linear combination of
two irreversible processes: isotropic and deviatoric
mechanisms. Both mechanisms have their own yield
function, f iso and f dev , respectively. The total plastic strain increment is the sum of the partial plastic
p,k
strain increments, dij , for each mechanism k:
The isotropic chemo-plastic mechanism is defined in

the p -plane and is expressed by
where p is the current effective stress, pc is the mean

yield stress and r iso is the degree of plastification
(mobilised hardening) of the isotropic yield limit.
Using this expression, a smooth transition from the
elastic to the elasto-plastic behavior can be represented. This implies a progressive evolution of the
activated plastic strain according to the bonding surface theory, as presented by Dafalias and Herrmann
(1980) and Franois and Laloui (2008).
During loading, r iso is a hyperbolic function of
the volumetric plastic strain induced by the isotropic
p,iso
mechanism v (Hujeux, 1979) and is expressed as
follows:
where cr is a material parameter.

The yield limit stress is a function of the activated volumetric plastic strain and osmotic suction.
Considering the empirical relationship (Eq. 2) and
assuming that the functions for the vertical effective
yield stress and the mean effective yield stress are
of the same form, the yield limit is defined by the
apparent preconsolidation stress as:
where pc0 is the preconsolidation stress at a reference

osmotic suction 0 , is the plastic stiffness modulus,
p
v is the total plastic volumetric strain (produced by the
two mechanisms), is the material parameter defining the shape of the isotropic yield limit with respect
to the osmotic suction and is the current osmotic
suction.
The plastic stiffness modulus (the inverse of the
p
slope of the linear function v -log pc ) was assumed
dependent on the osmotic suction in order to compensate the elastic stiffness dependency on the osmotic
suction, and ensuring a constant slope for the normal
compression line. The following logarithmic correlation is assumed:
where is the chemical evolution factor [], is

the current osmotic suction [MPa], 0 is a reference
osmotic suction and 0 is the plastic modulus at 0 [].
Therefore, inserting Eqs (14) and (13) in Eq. (10),
the detailed expression of the isotropic yield function
becomes
where the material parameters 0 and cr (within the

expression for r iso ) express the evolution of mechanical hardening and and express the evolution of
chemical hardening.
The isotropic behaviour is modelled with an associated flow rule as follows:
The deviatoric chemo-plastic mechanism is an

extension of the original Cam-Clay model, as presented by Roscoe & Burland (1968) and Hujeux
(1979). The deviatoric part of the constitutive model
is presented in Witteveen et al. (2013).
The volumetric and the deviatoric mechanisms are
p
linked through the hardening variable v . Therefore, a
fully coupled behaviour can be represented. The yield
stress limit depends on this parameter, and it appears
in both expressions of the yield functions. This causes
the isotropic yield surface to stretch when the deviatoric yield surface is expanded and vice versa, as shown
in Fig. 6.
4.2 Numerical simulations

The reference elastic bulk modulus, the reference
shear modulus and the plastic potential parameter were
determined considering the results of a mechanical
oedometric test on a sample prepared with distilled
water:
553
Figure 6. Representation of the isotropic and deviatoric

mechanisms of the chemo-mechanical constitutive model
ACMEG-C.
Figure 8. Stress paths of the oedometric test with a chemical loading step at a vertical stress of approximately 30 kPa,
including the evolution of the yield curve.
Figure 7. Simulation of three of the oedometric tests under

mechanical loading.
assuming that K 0 = 1 sin ,
Figure 9. Simulation of the oedometric tests under mechanical loading and a chemical loading step at a vertical mechanical stress of approximately 80 kPa using the
chemo-mechanical constitutive model ACMEG-C.
where pref = 1 MPa is the reference mean effective

stress, e0 is the initial void ratio [], C c is the compression index [], C s0 is the slope of the oedometric
curve upon first loading [] and K 0 is the lateral stress
ratio for one-dimensional strain []. The shear strength
angle of the tested illite is = 25 (obtained from
complementary triaxial tests). The other parameters
of the mechanical part of the model were determined
by fitting the curve of the test results.
Curve fitting processes on the results of the oedometric tests under mechanical loading were used to
fit the parameters defining the chemo-mechanical

behavior. The chemical form factors , and are
determined using plots of the oedometric modulus,
yield stress and plastic stiffness modulus, respectively,
against the osmotic suction. These parameters were
used to simulate three of the other oedometric tests
under mechanical loading. The results of the simulation are plotted in Fig. 7. The model results correspond
well with the experimental results.
The calibration of the numerical model has been
verified with two tests under both mechanical and
554
chemical loading. The simulations were then run for

stress paths C1 and C2 (Fig. 1), including a chemical loading step at approximately 30 kPa and 80 kPa
in the first and second cases, respectively. The results
for the stress path C2 are presented and discussed in
the following. Fig. 8 presents the stress path; Fig. 9
depicts the results of the simulations. The model and
experimental results correspond well at least until the
chemical loading. The vertical deformation associated to the chemical loading is slightly overestimated.
It is worth to mention that this deformation is modelled solely through the dependency functions of the
oedometric modulus, yield stress and plastic stiffness
modulus. Both the model and the experiment show
that the chemical loading brought the specimen to a
normal-consolidated condition. The slope of the plastic deformation and its order of magnitude are well
predicted.
5
CONCLUSIONS
This work aimed at extending the experimental evidence on the volumetric response of soils subjected
to a change in the chemical composition in the pore
liquid. In contrast with previous studies, a low activity
clay was considered. The osmotic suction, controlled
through the sodium chloride concentration in the pore
water, is considered to be the governing variable.
The laboratory programe aimed at establishing the
relationship between the osmotic suction and the composition of the pore liquid. A laboratory programme
was conducted. Subsequently, oedometric test were
carried out involving changes in the mechanical stress
and in the concentration of the solutions surrounding the specimens. Results showed that the oedometric
modulus upon first loading and the vertical yield stress
decreased as the osmotic suction increased, whereas
the compression and swelling indices were independent of the suction. The findings allowed an extension
of the ACMEG constitutive framework to be developed. ACMEG-C: Chemical Effect is able to take
changes in osmotic suction into account, as well as
the evolution of elastic and plastic material parameters
and the plastic deformation induced by these changes.
The model was calibrated with one oedometric test on
a sample prepared with distilled water and considering
the oedometric test results under mechanical loading.
The model was validated through the two tests including a chemical loading path. These stress paths show
that a large part of the chemo-mechanical oedometric
test can be predicted well.
REFERENCES
Barbour, S.L. & Fredlund, D.G. (1989). Mechanisms of
osmotic flow and volume change in clay soils. Can.
Geotech. J. 26, 551562.
Bolt, G.H. (1955). Physico-Chemical Analysis of the Compressibility of Pure Clays. Gotechnique 6, No. 2, 8693.
Bolt, G.H. & Miller, R.D. (1956). Compression studies of
illite suspensions. Proc. Am. Soc. Soil Sci. 19, 285288.
Boukpeti, N., Charlier, R. & Hueckel, T. (2004). Modelling

contamination of clays. In: Stephanson, O. (Ed.), Coupled Thermo-Hydro-Mechanical-Chemical Processes in
Geo-Systems Fundamentals, Modelling, Experiments
and Applications (Elsevier Geo-Engineering Book Series,
Volume 2). Elsevier Ltd., Oxford (United Kingdom),
523528.
Cardoso, R., Romero, E., Lima, A. & Ferrari, A. (2007).
A comparative study of soil suction measurement using
two different high-range psychrometers. In: Schanz,
T. (Ed.) 2nd Int. Conf. on Mech. of Unsat. Soils,
Weimar (Germany). Springer-Verlag Berlin (Germany),
pp. 7993.
Chapman, D.L. (1913). A contribution to the theory of
electrocapillarity. Philos. Mag. Ser. 6, Vol. 25, Is. 148,
475481.
Dafalias, Y. & Herrmann, L. (1980). A bounding surface
soil plasticity model. In: Pande, G.N. & Zienkiewicz,
O.C. (Eds.) Proc. Int. Symp. Soils under Cyclic Transient
Loading. Balkema, Rotterdam (Netherlands), 335345.
Di Maio, C. (1996). Exposure of bentonite to salt solution:
osmotic and mechanical effects. Gotechnique 46, No. 4,
695707.
Di Maio, C. & Fenelli, G.B. (1997). Influenza delle interazioni chimico-fisiche sulla deformabilit di alcuni terreni
argillosi. Rivista Italiana di Geotecnica 1, 695707.
Di Maio, C. & Onorati, R. (1999). Prove di laboratorio:
Influenza della composizione del liquido di cella. Rendiconti del XX Convegno Nazionale di Geotecnica, Parma,
8794.
Fernandez, F. & Quigley, R.M. (1985). Hydraulic conductivity of natural clays permeated with simple liquid
hydrocarbons. Can. Geotech. J. 22, 205214.
Fernandez, F. & Quigley, R.M. (1991). Controlling the
destructive effect of clay-organic liquid interactions by
application of effective stresses. Can. Geotech. J. 28,
388398.
Franois, B. & Laloui, L. (2008). ACMEG-TS: A constitutive model for unsaturated soils under non-isothermal
conditions. Int. J. Num. Anal. Meth. in Geomechanics 32,
19551988.
Fritz, S.J. (1986). Ideality of clay membranes in osmotic
processes: a review. Clays and Clay Minerals 34, No. 2,
214223.
Gajo, A. & Loret, B. (2003). Finite element simulations of
chemo-mechanical coupling in elastic-plastic homoionic
expansive clays. Comput. Methods Appl. Mech. Engrg.
192, 34893530.
Gajo, A., Loret, B. & Hueckel, T. (2002). Electro-chemomechanical couplings in saturated porous media: elasticplastic behaviour of heteroionic expansive clays. Int. J.
Solids Struct. 39, 43274362.
Gens, A. (2010). Soil-environment interactions in geotechnical engineering. Gotechnique 60, No. 1, 374.
Guimares, L. Do N., Gens, A. & Olivella, S. (2007). Coupled Thermo-Hydro-Mechanical and Chemical Analysis
of Expansive Clay Subjected to Heating and Hydration.
Transport in Porous Media 66, 341372.
Gouy, G. (1910). Constitution of the electric charge at the
surface of an electrolyte. J. Phys. 9, 457467.
Hueckel, T. (1997). Chemo-plasticity of clays subjected to
stress and flow of a single contaminant. Int. J. Num. Anal.
Meth. Geomech. 21, 4372.
Hujeux, J.C. (1979). Calcul numrique de problmes
de consolidation lastoplastique. Ph.D. Thesis, Ecole
Centrale de Paris, Paris (France).
Kaczmarek, M. & Hueckel, T. (1998). Chemo-Mechanical
Consolidation of Clays: Analytical Solutions for a
555
Linearized One-Dimensional Problem. Transport in

Porous Media 32, 4974.
Laloui, L, Nuth M. & Franois, B. (2010). Mechanics of
unsaturated soils, in Mechanics of unsaturated geomaterials, Laloui, Lyesse (ed.) John Wiley & Sons, 2010,
pp. 2951.
Leong, E.-C., Tripathy, S. & Rahardjo, H. (2003). Total suction measurement of unsaturated soils with a device using
the chilled-mirror dew point technique. Gotechnique,
Vol. 53, No. 2, pp. 173182.
Loret, B., Hueckel, T. & Gajo, A. (2002). Chemomechanical coupling in saturated porous media: elasticplastic behaviour of homoionic expansive clays. Int. J.
Solids Struct. 39, 27732806.
Marine, I.W. & Fritz, S.J. (1981). Osmotic Model to Explain
Anomalous Hydraulic Heads. Water Resources Research
17, No. 1, 7382.
Mesri, G. & Olson, R.E. (1970). Shear strength of montmorillonite. Gotechnique 20, No. 3, 261270.
Mesri, G. & Olson, R.E. (1971). Consolidation characteristics
of montmorillonite. Gotechnique 21, No. 4, 341352.
Mitchell, J.K., Greenberg, J.A. & Witherspoon, P.A. (1973).
Chemico-Osmotic Effects in Fine-Grained Soils. In:
Mitchell J.K. & Idriss I.M. (Eds.), Selected Papers of
James K. Mitchell. ASCE, Reston VA (United States of

America), 398413.
Mitchell, J.K. & Soga, K. (2005). Fundamentals of Soil
Behavior. John Wiley & Sons, Inc., Hoboken, NJ (United
States of America).
Roscoe, K.H. & Burland, J.B. (1968). On the generalized stress-strain behaviour of wet clay. In: Heyman, J. & Leckie, F.A. (Eds.) Engineering plasticity,
Cambridge University Press, Cambridge (United Kingdom), 535609.
Sridharan, A. & Venkatappa Rao, G. (1973). Mechanisms
controlling volume change of saturated clays and the role
of the effective stress concept. Gotechnique 23, No. 3,
359382.
Sridharan, A. & Venkatappa Rao, G. (1979). Shear strength
behaviour of saturated clays and the role of the effective
stress concept. Gotechnique 29, No. 2, 177193.
Verwey, E.J.W. & Overbeek, J.Th.G. (1948). Theory of
the stability of lyophobic colloids. Elsevier Pub. Co.,
New York, NJ (United States of America).
Witteveen, P.J., Ferrari, A. & Laloui, L. (2013). An experimental and constitutive investigation on the chemomechanical behaviour of a clay. Geotechnique 63, No. 3,
244255.
556

Exploiting physico-chemical modification of soils in closed face tunnelling

S.A. Jefferis
Environmental Geotechnics Ltd, Banbury, UK and The Department of Engineering Science,
University of Oxford, Oxford, UK
A.S. Merritt
Geotechnical Consulting Group LLP, London, UK
ABSTRACT: When tunnelling in unstable grounds it is necessary to remove the cut soil from the face whilst
maintaining face stability. This can be achieved by slurrifying the excavated soil and using a circulating fluid to
remove it from the face. In another system, earth pressure balance tunnelling, the cut soil is continuously formed
into a paste which is extracted via a screw conveyor in which it must build a plug. To support the face, the slurry
or paste must seal the face as it must be pressurised sufficiently to counter the soil and groundwater pressures.
Typically, bentonite slurries are used for slurry tunnelling but polymer systems can offer advantages. For earth
pressure balance machines, soil conditioning additives are more complex and varied; many different materials
are used including foams. This paper gives an overview of the problems associated with analysing the behaviour
of tunnelling additives and invites further research.
INTRODUCTION
If a tunnel is excavated in unstable ground with an

open faced tunnel boring machine (TBM), soil and
groundwater may flood in and the ground ahead of
the machine may collapse preventing further work,
causing surface settlement and possibly sinkholes. It
is therefore necessary to support the face and prevent
groundwater ingress into the tunnel. Typically this is
done by sealing the face of the TBM into the ground
and circulating a slurry through it to remove the cut soil
(slurry tunnelling, see Figure 1), or by forming the cut
soil into a paste which is removed from the face with a
screw conveyor operated so as to maintain a pressure
at the face (earth pressure balance, EPB tunnelling, see
Figure 2). For very open grounds with high groundwater pressures which cannot be sealed by either a slurry
or an EPB paste, the recently developed water control
process may be employed; the soil is extracted from
the machine via a screw and an inline crusher connected to a slurry box from which it is removed by
slurry circulation (Storry et al., 2013). However, the
process has limitations; the face must be stable without a supporting pressure acting on an EPB paste or
slurry filter cake.
This paper will focus on slurry and EPB systems
and the physical/chemical modification of the excavated soils, slurries and pastes that may be required to
optimise machine performance. It is intended to introduce the topic to those unfamiliar with the problems
of operating TBMs and the associated slurries and soil
conditioning agents. It may be read in conjunction with
Milligan (2000) which provides a substantial review of
Figure 1. Schematic of a slurry TBM.
Figure 2. Schematic of an EPB TBM.
the slurry and EPB tunnelling processes and their associated additives though there have been significant
developments since then.
1.1 Terminology
As this paper is addressed to those concerned with the
chemical and physical behaviour of soils, a few words
of explanation of tunnelling terminology are appropriate. The face of a tunnelling machine is that region
where the slowly rotating cutter-head comes into contact with the soil to be excavated. The head will be
557
dressed with an array of cutting tools appropriate to

the ground to be excavated (often a compromise as,
along a drive, a machine may have to pass through
many ground types). The cut soil passes through the
cutter-head into a pressurised chamber from which it
is removed either in a slurry or as a paste. For EPB
TBMs, conditioning additives may be injected, through
a rotary coupling, and thence into the face via injection points mounted on the cutter-head and/or into the
chamber behind the cutter-head and/or the screw conveyor. Mixing bars may be mounted on the back of
the cutter-head to promote mixing of the cut soil and
additives in the chamber but even with mixing bars the
degree of mixing will be poor. EPB TBMs are not mixers and cannot be mixers the energy demand would
be enormous and progress slow.
The cut soil from excavations is typically referred
to as spoil but this implies that it is a waste under
European legislation, it is the intention of the holder
that defines whether a material is a waste (Potter
and Jefferis, 2005). As excavated material should be
re-used whenever possible it is suggested that the term
spoil is avoided and in this paper the excavated material
will be referred to as soil, cut soil or arisings.
2 THE SLURRY TUNNELLING PROCESS

In a slurry tunnelling machine, the face of the machine
is sealed and a water-based slurry is pumped into the
chamber at a pressure sufficient to counter the groundwater pressure and support the ground at the face. The
injected slurry becomes mixed with the excavated soil
and the soil laden mix is pumped from the machine
to a separation plant at the ground surface where the
cut soil is removed and the slurry returned to the face.
Typically, slurryTBMs are used in coarser soils, ideally
those with relatively low fines contents.
In the separation plant, coarse materials are
removed on vibrating screens. Fine sands and coarse
silts usually are removed with combined hydrocyclone-dewatering screen units (the solids in the
slurry are concentrated into the hydrocyclone underflows and then dewatered on the screens). In soils
containing a significant amount of fine silt and clay
it is necessary to employ a centrifuge or filter press
for their removal. If presses are used, and often with
centrifuges, the slurry must be flocculated before
treatment. Typically an organic flocculant will be used
often in conjunction with an inorganic coagulant.
In addition to the role as a carrying fluid the slurry
must stabilise the face and permit the soil to be
removed from it in the separation plant. The principal
requirements for the slurry are:
in coarse grounds it must block the soil pores at

the face so that there is no excessive slurry loss.
Typically this is achieved by using a gelling additive such as bentonite but it also may be achieved
with strongly pseudoplastic fluids (e.g. polymer
solutions);
in finer grounds the slurry must develop a cake to

limit filter loss to the ground, that is, it must limit
filtration of water from the slurry into the soil so that
sufficient pressure is applied at the face to maintain
its stability. If filter loss is not controlled by a low
permeability cake at the face the applied pressure
will be dissipated over some depth into the ground
which may allow ravelling of the ground into the
face;
the rotation of the cutter-head continuously removes
the filter cake and so its filter loss control action is
limited as the resistance of a cake increases with its
thickness and hence its time under pressure;
in general the face will be sealed by a combination
of penetrated slurry and filter cake developing on it;
the slurry must carry the cut soil from the TBM
to the separation plant without unacceptable settlement of solids in the connecting pipeline which may
be from hundreds to thousands of metres long and
do so at an acceptable pumping pressure;
the slurry should be lubricious and limit abrasion of
the pipe walls by the solids in the slurry;
any additives in the slurry should not adversely
affect the performance of the separation plant.
Ideally additives should improve its performance,
for example, by inhibiting the break-up of cut soil
lumps so that they can be removed on the coarse
screens.
For many grounds, a sodium bentonite clay at a concentration of about 2 to 5% is used to form the slurry
as bentonite has good fluid loss control characteristics.
However, it is sensitive to degradation of properties
by cement and chemical contamination. Furthermore,
bentonite slurries, because of their viscous and gelling
properties, increase pipeline pumping pressures.
2.1
Polymer fluids
Bentonite is therefore not always the ideal slurry forming material and alternatives have been tried. For
example, in the 1980s, two major tunnels on the Cairo
Wastewater Project were driven with a bentonite free
slurry. Two machines were used and the tunnels passed
through, gravels, sands and clays, often with mixed
strata in the face. After a considerable programme of
research, a slurry system based on an organic polymer,
sodium carboxymethyl cellulose (CMC) was developed. The range of polymers available at that time was
somewhat limited and CMC was used as it can have
a synergistic effect with clays including native clays
in excavated soil. This synergism markedly reduces
fluid loss and modestly develops slurry viscosity (and
so improves soil carrying capacity). Furthermore its
pseudoplastic rheology means that pumping pressures
may be lower than with bentonite slurries (which
although thixotropic can still show relatively high yield
stress).
Typically, the CMC was used at a concentration of
0.05 to 0.2% in a slurry which could contain over 50%
soil solids when leaving the TBM and 5 to 10 % solids
558
when returned after soil separation only fines as a

separation plant will remove the coarser material (all
concentrations by weight of water in the slurry).
For slurry tunnelling, CMC has three disadvantages: it is biodegradable; its effect is reduced by
cement contamination (calcium ions from cement can
precipitate CMCs as can magnesium or other polyvalent ions), and it binds to clay so that it is removed
with any clay at the separation plant (this loss is
inevitable for any material that sorbs onto or develops
synergy with the cut soil).
Experiments with different CMCs showed that high
viscosity grades were more cost effective than low
viscosity grades in developing viscosity and reducing
fluid loss. On the Cairo project a biocide was added
to the slurry to prevent bacterial degradation and agar
gel plates were used to check for bacterial fouling.
Today polymers are available which are more resistant to bacterial attack and biocides are seldom used
though on occasion they should be especially for the
complex bentonite-polymer fluids that are used in horizontal directional drilling which must be considered
in a separate paper.
Polymers other than CMC are now available but
bentonite-free slurry systems have yet to be widely
adopted, though they have been successfully used for
the construction of bored piles across the globe (Lam,
2011). For large diameter tunnels bentonite based slurries are the norm and indeed it takes some courage to
move from a proven but less than ideal system such as
bentonite for TBMs which may be over 14 m in diameter with slurry flow rates of over 1000 m3 /hr. For small
diameter machines where face support is less critical
and flow rates are much lower, experimentation with
other slurry systems is less risky and in practice water
is often used again potentially a sub-optimal cutting,
stabilising and carrying fluid.
with sufficient fines it may be easy to form a plug in

the screw conveyor but in others additives are essential.
Requirements are that:
3 THE EPB TUNNELLING PROCESS

In earth pressure balance tunnelling the face of the
machine is sealed and the removal of cut soil from it
is controlled with a screw conveyor so that the chamber is pressurised and thus the face kept stable and
the groundwater held back. The excavated soil plus
additives, if required, must form a cohesive, flowable
material which not only seals the face and prevents the
ground collapsing onto the cutter-head which would
lead to excessive cutter-head torque, but also builds a
plug in the screw conveyor with sufficient cohesion
that it can develop the required pressure in the chamber and thence the cutter-head. The screw is a highly
specialised piece of equipment. On a large machine it
may be over 1 m in diameter and if high pressures are
to be dropped over 15 m long. The screw is driven by a
variable speed drive to enable control of the pressure.
Traditionally, EPB TBMs were used in soils containing
at least 30% fines, but developments in soil conditioning have expanded their application range to coarser
soils with less fines (Thewes, 2007). In some grounds
the cut soil is converted into a workable material that can build a plug a conditioned soil.
The required material may be likened to a nonsetting concrete. A wide range of particle sizes/cut
soil lump sizes can be accommodated provided
a coherent, non-segregating pseudo-concrete is
formed. Indeed the concrete slump test is often
used to assess conditioned granular soils (the slump
test gives an indication of stability, flowability and
compactibility all important parameters for workability). Typically slumps in the range 100 to
200 mm are recommended (Peila et al. 2009, Merritt
et al. 2013);
the plug material should be able to accommodate
water from the ground as grounds which yield large
amounts of water may have to be excavated. Some
management of water may be achieved by increasing the amount of additives used. However, plug
formation may not be possible in coarse, high water
content soils regardless of the amount of additive
used and additional measures such as prior grouting
of the ground along the tunnel may be necessary;
any additive should be easily dispersible into the soil
as the mixing action within the machine or the screw
will be limited. As already noted, TBMs should
not be designed as mixers as this would require
excessive power;
additives should be lubricious and not increase the
cutter-head torque. Ideally they should reduce it;
additives must be sufficiently fluid to pass through
the rotary coupling on the cutter-head;
additives should act sufficiently rapidly that their
effects occur within the intended section of the TBM
and/or the screw;
the properties of the conditioned soil should be
robust and insensitive to realistic changes in the
composition of the mixture (e.g. soil grading, water
content, additive concentration).
Bentonite slurries at about 510% solids content

have been used as conditioning additives as they give
good lubricity and can provide some cohesion. However, at 5 to 10% a bentonite slurry has little capacity
to manage further water and in wet grounds it may
be impossible to form a plug to control ground water
inflow. Bentonite slurries also increase the water content of the cut soil in the TBM and that discharged
from it.
On a major sub-sea tunnel project in the 1990s, the
soil was a loose wet silty material with typically only
a low clay content. It thus showed little cohesion. A
dilute polyacrylamide solution with a powerful flocculating action was used to bind and work with the
small amount clay that was present to build cohesion.
This polyacrylamide, when used at a solution concentration of 0.2 to 1%, was easily mixed into the soil
to give a spongy workable mass. The polymer sorbed
559
onto the fines and water contents over 200% could

be sustained with the low clay soils without the system degenerating into a soup (without polymer the
maximum water content was about 60%).
A further feature of some polyacrylamide-soil
blends is that clay becomes much easier to handle,
loses its stickiness and no longer adheres to metal
surfaces. This can help to reduce blockages within a
TBM (see discussion on stickiness). However, the clay
no longer adheres to stones in the soil which become
clean, clay free and potentially more abrasive.
At concentrations over about 1% the behaviour of
the particular polyacrylamide-soil mix changed. The
polyacrylamide solution was no longer easily miscible with soil but coated it (possibly because all the
exposed sorption sites were rapidly exhausted at the
higher solution concentration) so that the mix took
on the appearance of frog spawn with lumps of soil
in a stringy, slimy mass (one of many demonstrations
that merely increasing polymer concentration will not
ensure effective conditioning of difficult soils).
In certain circumstances a slimy coating can be an
advantage but on that particular project it was a problem as the soil was carried from the screw conveyor
to muck trains via an inclined belt conveyor and the
slimy soil slid back down the belt into the tunnel invert
from whence it had to be dug out by hand. Having to
shovel many tons of slimy soil can make tunnel operatives less than well-disposed to those researching TBM
additives!
For a particular set of soil and ground conditions,
the optimum additive formulations can be assessed
by laboratory trials but the behaviour in the tunnel may not follow that in the laboratory. Too often
unexpected complications occur and experience of
additive-machine interactions becomes crucial.
Furthermore the TBM environment is very complex and there will be many constraints other than
soil conditioning. It is important that any additives
proposed for use in a machine have robust and predictable effects. Performance which is marginal or
sensitive to specific conditions in the laboratory will
be unacceptable in tunnel operations.
4 THE GROUND TO BE TUNNELLED
Tunnels may pass through any type of ground, rock,
gravel, sand, silt and clay. A single tunnel may at
sections of the drive intersect all of these materials.
Furthermore, the face of a tunnel may contain mixed
strata and the relative proportions of the various soil
types may vary along a drive, for example, if the strata
are inclined at an angle to the line of the tunnel. The
groundwater conditions (pressure, permeability, salinity) also have a major impact on the tunnelling process
and additive selection.
The choice of tunnelling machine type often has to
be a compromise dictated by a few of the soil types
present in the face. For example, in a mixed face
drive with coarse sand and clay, a slurry machine may
be required to manage the sand despite the fact that

the clay may disperse into the slurry and require the
addition of filter presses or centrifuges to the slurry
separation plant (with the result that the separation
plant my cost more than the TBM).
For EPB machines in mixed faces of gravels, sands
and clean silts, treatments may be largely limited to
changing the physical properties of the porewater,
for example, with viscosifying agents or foam so as
to provide some apparent cohesion, reduce permeability and generally improve the workability of the
soil. If high porosity soils are encountered or there
are high water inflows, water absorbent (swelling)
polymers may be introduced for localised water
management.
For EPB and slurry tunnelling in soils containing some clay, there is potential for both chemical
and physical clay-additive interactions. In general one
should not aim to physically or chemically modify all
the cut clay or other soils (i.e. it should not be the
intention to break down all lumps of cut soil to their
individual component particles). This would be simply uneconomic because of the additive demand and
require more soil-additive mixing than can be achieved
in a typical tunnelling machine. As a consequence,
when assessing EPB conditioning additives in laboratory trials, it is important to consider the degree of
mixing of the soils and additives. Well mixed homogeneous laboratory samples can help to inform an
understanding of soil-additive interactions. However,
the differences between well homogenised laboratory
samples and in-TBM conditioned soils can be significant. The optimum procedure is to develop additives
that form a cohesive plastic paste whilst at the same
time maintaining much of the soil as intact lumps, as
large as can be handled by the EPB screw.
For slurry tunnelling the aim will be to keep cohesive soils in as large lumps as practicable and research
is in progress on the controls on lump degradation.
Fines tend to disproportionately increase pumping
pressure and are difficult to remove in the separation
plant. Additives should modify the surface of cut soil
lumps and inhibit their swelling so as to achieve the
required behaviour in the soil handling system.
When undertaking laboratory tests it is important
to test relevant soils those from boreholes along
the planned tunnel drive. Pure clay minerals such as
bentonite (montmorillonite) and kaolinite which are
often available in soils laboratories are seldom, indeed
effectively never, found in tunnel drives. Tests with
pure minerals may provide insights into fundamental interaction mechanisms but should not be taken as
indicative of in-tunnel behaviour.
Furthermore, in the field, clays are likely to have
mixed exchangeable cations, often calcium and magnesium whereas laboratory clays often are supplied
in the sodium form. The prevalence of calcium and
magnesium on natural clays is because at the rather
low ionic concentrations typical of natural groundwaters (excluding those subject to seawater mixing etc.)
calcium and magnesium (or other polyvalent ions)
560
are preferred over sodium (or other monovalent ions)

on exchange sites. Unfortunately many additives
including low and high molecular weight polymers
and also foaming agents are anionic with sodium or
another monovalent ion as the cation. If such additives are added in proportions significant in relation to
the number of exchange sites on exposed surfaces (i.e.
the sites at and near the surfaces of cohesive soil/clay
lumps) significant cation exchange may occur. Typically the introduction of monovalent ions in place
of polyvalent ions will increase the liquid limit of a
clay and thus, at constant water content, increase its
strength. Such exchange also may make clays more
expansive. However, ion exchange can be relatively
slow so that effects may be most evident after soils
have left the TBM and during subsequent handling.
SOIL HANDLING THE WHOLE SYSTEM
It should not be forgotten that additives can affect and

may be required to affect the cut soil after its discharge
from a TBM or from a slurry separation plant. For
example, the soil discharged from the screw of an EPB
machine may drop onto a conveyor belt, on which it
must remain without tending to slough or slide off and
thence to rail cars for transport out of the tunnel to a
temporary stockpile for onward transport to the final
location for re-use or disposal. For major infrastructure works, onward movement by rail is often preferred
to avoid the disturbance created by lorry traffic. At all
stages from the tunnel face to final resting place the
soil must be manageable and additive suppliers should
consider the whole system and not just the TBM. For
example, it can be very difficult to remove a congealed and compacted mass of cohesive clay from a rail
car. Problems also can be encountered with conveyor
systems, where different conditioning additives can
make clays either very sticky or very slippery.
ENVIRONMENTAL CONSTRAINTS
The arisings from tunnelling works must not contain

any materials which prevent their safe handling and
re-use or disposal. Thus the arisings and any additives
used must comply with local health and safety and
environmental regulations. Health and safety should
not be a major constraint as suppliers will be aware
of requirements for safe practice though consideration
regarding use in the confined space of a tunnel will be
important.
Environmental constraints can vary significantly
between jurisdictions and may depend on the final
disposition of the arisings re-use or landfill. Note
that a European directive bans the disposal of liquid
waste to landfill and similar constraints may apply in
other jurisdictions. The definition of what is a liquid
(i.e. at what point, strength or other property, a paste
becomes a liquid) may vary between jurisdictions.
Problems also can arise if the re-use or disposal is

under a separate contract from the tunnelling works.
There can be conflicts between the properties that are
required and/or can be achieved within a TBM and its
associated plant and the requirements for the chosen
management route for the arisings. For example, moisture contents for the latter may be set unrealistically
low (a further source of potential misunderstanding is
that in soil mechanics moisture contents are quoted by
dry weight whereas in the minerals and material handling industries moisture contents are usually by total
(wet) weight).
Environmental constraints also will require consideration of the potential for additives to penetrate into
the ground ahead of and around the cutter-head and
impact on the groundwater. There also will be potential for harm as a result of spills of bulk chemicals in
transit, storage and use and these risks will need to be
managed.
As the particular chemical species or blends used in
additives may be proprietary to the suppliers, obtaining the necessary data for environmental assessments
can raise problems and require the employment of
independent third party consultants (independent of
supplier, user and regulator) who, under appropriate confidentiality agreements, undertake the necessary assessments. If insufficient environmental data is
available on any component of an additive then specific testing may be necessary testing which can be
expensive and take many months. In effect, this can
restrict the choice of additive components to those
that are known to and accepted by the local regulators. Environmental compliance should not be taken
for granted and sufficient time should be included in
the project programme. Assessments should include
all components of additives, carrier fluids such as oils
as well as the active ingredients.
INJECTION OF ADDITIVES
In general additives must be fluids liquids, fine slurries or foams. Injection of powders or coarser solids
into the cutter-head is difficult as additives must pass
through a rotary coupling. There is also the problem
of mixing powders into soils within the tunnelling
machine.
Addition as fluids can complicate the addition of
water-active additives such as super water-absorbent
polymers which act by sorbing and so removing free
water. If these are premixed with water their action
is lost. It is therefore necessary to add them as
suspensions in a carrier oil (likely vegetable derived).
Additives may not have an immediate effect, especially if their action requires sorption onto soil surfaces
or chemical reaction. It is therefore important that they
are added sufficiently in advance of the location at
which their effect is required. Addition at the face provides the maximum residence time in the machine and
the maximum opportunity for mixing of the cut soil
and additives. If effects are required at the face or
561
for injection directly into the screw, additives must be

rapid acting.
SPECIFIC ADDITIVES
In the following sections a brief overview is given of

some major additive types. The role of the additives is
considered for both EPB and slurry machines.
8.1
Stickiness management agents
Stickiness in the context of TBM operation is the

propensity for clay soils to stick to the internal parts of
the machine including the cutter-head and the chamber. Severe stickiness can lead to complete clogging
of the cutter-head, chamber and/or other parts of the
machine. It may be impossible to clear a clogged
machine mechanically and the clogging material may
have to be dug out manually, possibly under compressed air to exclude groundwater and maintain face
stability.
Various products are available that aim to reduce
stickiness and by different mechanisms; however, it
has been said (though not by the authors) that stickiness is an intractable problem and that no additives
are effective for the prevention of clogging. This is
obviously the perfect challenge to researchers, and
recent laboratory studies have shown potential developments and insights (e.g. Zumsteg et al., 2013). There
is also a significant literature from the oil well drilling
industry. However, problems with stickiness can arise
for many reasons including, for slurry TBMs, failure to sufficiently separate clay so that the slurry is
returned to the machine as a heavy viscous fluid incapable of flushing soils from the solid surfaces. Also
additives, when tried, may be added at insufficient
concentration and insufficiently frequently to maintain
the necessary active concentration in the system it
is not always appreciated that many clay management
additives work by sorption onto soils. Active concentrations therefore decline in use (and thus with time)
and especially rapidly if clay soils disperse to fine particles and do not remain as coarse lumps greater
amounts of additive will be required for dispersed soils
than those that can be kept as coarser lumps.
8.2
Foams
These are very widely used in EPB tunnelling and

there is a substantial literature on soil gradings and
groundwater pressure regimes in which foams can
be used (e.g. Thewes, 2007; Peila, 2009). Foaming
agents can be based on products such as alkyl sulfates
and sulfonates (these are merely examples; there are
very many surfactant types ranging from household
detergents to application specific products). As well
as acting as surfactants, these products can be very
powerful dispersing agents so powerful that the first
author has observed kaolin dispersed to a size such that
it can be washed through qualitative grade laboratory

filter paper.
Other additives such as polymers and dispersing
agents may be included in foam products to modify the properties of the foam (e.g. improve foam
stability). Foams are not used in slurry machines,
indeed foaming can be a substantial problem. If the
slurry turns to foam, pumps will cavitate and hydrocyclones used for de-sanding and de-silting the slurry
also may mal-function.
Some cement grout formulations contain materials that act as surfactants and surfactants can be
deliberately included in chemical grouts where the
components are otherwise immiscible (for example,
silicates hardened with immiscible esters note it
has been suggested, Jefferis (2003), that for these
the ester-alkali reaction, known as a saponification,
is responsible for foaming but this does not appear
always to be the case).
8.3
Dispersing agents
These may be low to medium molecular weight polyacrylamides, lignosulfonates etc. Comparable materials may be used as superplasticisers for concrete.
When used in EPB tunnelling, dispersing agents can
soften clays and in heavy clay soils this can help to
form a paste in the screw. Some are held to reduce
clay stickiness. However, to achieve these effects it is
likely that extra water will have to be added and there
may be logistical problems if large amounts of water
are required. In granular soils of low clay content, dispersing agents can help by breaking up clay aggregates
so providing more fines for paste production. In this
role, initial dispersion of a clay present at the face and
flocculation of the released fines to stiffen the paste at
the screw could, in theory, be beneficial though difficult in practice on a TBM which is already a massively
complex machine to operate. Further work on the use
of interacting and synergistic additives is required.
Dispersing agents must be used with considerable
care as they also can promote stickiness. Dispersion
increases the fines fraction and thus tending to make
the soil more cohesive though this may be offset by
adding water. Dispersed clay soils, as noted above, may
take-up water more rapidly than untreated soils if it
is available. This can make it difficult to achieve the
low moisture content arisings which are required for
some re-use and/or disposal situations.
In slurry machines, dispersing agents may be added
to the return slurry if it has become too thick as a
result of excess clay content or flocculation of clay in
the slurry. This may be particularly necessary when
bentonite is used to form the slurry, as is typical for
large diameter machines but not for smaller machines,
for example, in pipe jacking.
Thickening of a clay slurry may be the result of salts
in the ground, for example, in near-shore environments
though the effects of seawater are often quite modest.
Bentonite slurries also can be thickened by cement
when tunnelling through cement grouted ground or if
562
cement from back-grouting of the tunnel lining reaches

the head. Such thickening is best treated with sodium
bicarbonate which precipitates calcium ions as calcium carbonate (sodium carbonate is not effective in
this role).
8.4 Dispersion inhibitors
These would seem to be a very important class of
additive which have yet to be properly exploited in tunnelling. As considered herein, a dispersion inhibitor is
not a flocculant. For flocculation to occur clay particles
must be able to move and take up open structured floc
arrangements. The aim of a dispersion inhibitor is to
keep the clay lumps, as cut in the head of the machine,
largely intact. Dispersion inhibitors have been very
effectively exploited in piling. Large diameter piles
can be bored under polymer fluids which so effectively
inhibit dispersion that the soil-slurry separation plant
which is normally required with bentonite slurries can
be omitted with substantial cost savings and a much
reduced site footprint. With such polymer fluids the
limited amount of soil that becomes mixed into the
fluid can be removed by using the pile bore as a settling column (unfortunately not possible in tunnelling)
and simple settling tanks (Jefferis et al. 2009).
8.5 Viscosifiers
For EPB tunnelling in coarse soils, these may be used
to viscosify the pore-water to make the soil more
workable and effective at plug-building in the screw.
However, viscosifiers, even at high concentrations,
cannot convert all wet granular soils to a plug building
materials there must be sufficient fines to develop a
coherent system and not just create a mass of coarse
particles suspended in a viscous slime.
In slurry TBMs, viscosifiers have been added to
the slurry circuit to limit the rate loss of slurry to open
grounds. However, increasing gel strength, for example, by increasing the bentonite concentration can be
more effective, though as is always the case, bentonite
and other additives can have only a finite effect. Physical blocking with the cut soil itself (if it generates
appropriate size fractions difficult to predict and
manage) and possibly swollen super water absorbent
polymers may be more effective.
8.6 Filter loss control agents
Filter loss control that is the control of the loss
of water from a paste or slurry to the ground in
the face can be an issue for both EPB and slurry
machines. However, it is of most concern in slurry
machines where excessive fluid loss can lead to face
instability especially when a machine is stopped and
at such times it is important that the head is filled with
a slurry with good filter loss control properties.
The issue of filter loss is complex and cannot be
addressed analytically in this paper. However, it is
important to recognise that filtration resistance develops with time under filtration (flow resistance is
proportional to thickness/permeability). In the face of
a TBM the cake is continuously removed by cutting
tools so that the average cake age may be measurable
in seconds. As a consequence the filtration resistance
of any slurry penetrated into the ground ahead of
the machine can make an important contribution to
fluid loss control (i.e. the flow resistance of the penetrated slurry rather than that of the cake at the soil
surface). Typically the cake plus penetrated slurry
formed with a bentonite slurry is found to provide sufficient fluid loss control. There are many additives that
can further reduce filter loss including carboxymethyl
cellulose, modified starches and xanthan gum. However, all of these can be rapidly biodegraded unless
treated to modify susceptibility. Degradation can be
very rapid once a significant biologically infection has
developed.
As a general note, when using systems containing biodegradable additives it is very important that
all tanks etc. are emptied and cleaned between uses.
Any degradable material remaining in equipment may
allow a reservoir of infection to develop.
8.7
Super water absorbent polymers
These are typically cross-linked polyacrylamides

which swell and absorb water but do not dissolve into
solution. They can be very effective in removing water
from the cut soil in an EPB machine. However, overdosing, and it can be difficult to assess the required
dose, can lead to very substantial stiffening of a soil
and even stalling of the machine. They must therefore
be used with considerable care.
Water absorbent polymers also can be used to draw
water from and so stiffen over-wet arisings from EPB
and slurry tunnelling and other works though they
do not, of course, change the overall evaporable water
content
Super water absorbent polymers are not appropriate
for general and continuing use. Rather they are agents
of last resort to manage brief, localised high water
incidents.
8.8
Other additives
The above is an overview of some of the basic functions required of or achievable with TBM additives. A
review of suppliers literature shows that many other
additives are available to perform functions other than
those listed above. However, for some of these functions it is less than clear how the effects are defined
and how they can be quantified.
9
CONCLUSIONS
Slurry and soil conditioning agents remain an underresearched area in the academic arena despite the
enormous volume of practical experience from tunnelling projects across the World. Some very useful
563
research work has been published but much remains to

be done. There are many research opportunities; developing the fundamental understanding of soil-chemical
interaction mechanisms, quantifying their effects on
properties relevant to tunnelling and developing appropriate test procedures. However, a few words of caution
are necessary. Before carrying out research, ensure
that the work is focused on useful properties. It is very
easy dramatically to change the properties of clays
by adding materials such as dispersants and organic
polymers but are the resulting properties robust and
useful? Other constraints on research are:
the treatments must be economic;

all products must be acceptable for human health
and safety and the environment both in the tunnel
locale and at the site of final re-use or disposal;
tests on pure clay minerals such as kaolin or bentonite may give false positives. The behaviour of
real soils can be much more complex;
typically, natural clays have calcium and/or magnesium ions in the exchange sites. Exchange of
these ions with sodium from additives may increase
strength at constant moisture content and promote
swelling if water is available;
research should be carried out on soils with real
gradings not narrowly graded laboratory materials;
in the geotechnical literature, clay-chemical interactions are often analysed in terms of simplistic
concepts of flocculation, and hypothesised effects
on clay diffuse double layers etc. There are great
opportunities for deeper analyses of the coupled
phenomena in clay-additive-machine interaction.
These will enable new generations of treatment
materials.
Finally, in addition to constraints on materials, it

is important to note that it is difficult to carry out
controlled research on tunnelling machines because:
for EPB soil conditioning, although numerous laboratory studies have been reported, no formalised
standard laboratory test procedures currently exist;
development of test procedures and field assessment of the validity of results from them is required;
the soil in the face will be continuously varying it
is rare that soil is uniform across the full face of a
machine for significant distances along a drive;
individual TBM drivers may have particular preferences for additive use and thus results may vary
between drivers;
research is seldom the major priority on an operational machine where production is likely to be
key;
if any additive fails tunnelling may be brought to a
halt and/or the machine may become clogged;
as is so often the case, the principal opportunity
for developments may come when a machine is in
trouble and production is being delayed. Necessity
remains the mother of invention.
REFERENCES
Jefferis, S.A. 2003. Long term performance of grouts and the
effects of grout by-products, Proc. ASCE Specialty Conf.,
Grouting and Ground Improvement, New Orleans.
Jefferis, S.A., Troughton, V. & Lam, C. 2009. Polymer systems for fluid supported excavations, Proc. Geotechnical
Issues in Construction, CIRIA Report X513, pp. 712.
Lam, C. 2011. Properties and Applications of Polymer Support Fluids in Geotechnical Engineering. DPhil thesis,
University of Oxford.
Merritt, A.S., Jefferis, S.A., Storry, R.B. & Brais, L.P.
2013. Soil Conditioning Laboratory Trials for the Port of
Miami Tunnel, Miami, Florida, USA. To be published in,
G. Anagnostou & H. Ehrbar (eds) World Tunnel Congress
2013 Geneva Underground the way to the future!
Milligan, G. 2000. Lubrication and soil conditioning in
tunnelling, pipe jacking and microtunnelling: A state-ofthe-art review, The Pipe Jacking and Tunnelling Research
Group.
Peila, D. Oggeri, C., & Borio, L. 2009. Using the SlumpTest to
Assess Behaviour of Conditioned Soil for EPB Tunnelling.
Environmental and Engineering Geoscience, Vol. XV, No.
3, pp. 167174.
Potter, A. & Jefferis, S.A. 2005. Management of process
arisings from tunnels and other earthworks: a guide to
regulatory compliance, The Pipe Jacking and Tunnelling
Research Group.
Storry, R.B., Brais, L.P. & Pascual, P. 2013. A Geotechnical Challenge at the Limit, a Case History, Port of Miami
Tunnel, Miami, Florida, USA.To be published in, G.Anagnostou & H. Ehrbar (eds), World Tunnel Congress 2013
Geneva Underground the way to the future!
Thewes, M. (2007). TBM Tunnelling Challenges redefining
the state of the art. Collection of Keynote Lectures, ITAAITES World Tunnel Congress, Prague, 2007, pp. 1521.
Zumsteg, R., Plotze, M. & Puzrin, A. 2013. Reduction of the
clogging potential of clays: new chemical applications and
novel quantification approaches. Geotechnique 6, No. 4
pp. 276286.
564

Combining biomineralization with chemically enhanced drainage in soils

P. Minder & A.M. Puzrin
ETH Zurich, Institute for Geotechnical Engineering, Zurich, Switzerland
ABSTRACT: This work focuses on the potential combination of microbial soil improvement methods with
chemically enhanced drainage in fine grained soil. The efficiency of carbonate precipitation with Bacillus pasteurii cultures was evaluated on laboratory scale soil samples containing clay minerals. Microbial carbonate
precipitation successfully increased uniaxial compression strength even in fine-grained soils with a minimum
hydraulic conductivity of 5E-9 m/s. During treatment, permeability is reduced by one order of magnitude and
can decline below the required limit. In order to avoid falling below the threshold it is proposed to improve in
advance hydraulic properties of the soil by using a technique of chemically enhanced drainage. In this study
permeability increase of more than one order of magnitude was achieved for soils containing smectite clay minerals. Combination of the two methods allows expanding the possible range of applicability of biomineralization
toward less permeable soils.
INTRODUCTION
Artificial cementation of soils by addition of chemicals is widely used (e.g. in road construction for
hydraulic stabilization of the deck). These methods
usually require mechanical mixing of the treated soil,
yield highly alkaline soil chemistry and are limited
to close surface application (Hammes & Verstraete,
2002).
Microbially induced precipitation of calcite (CaCO3)
provides cementation of soil particles under flowthrough conditions. Hammes & Verstraete (2002) have
shown that calcite precipitation is governed by four key
parameters. The formation of calcite depends on
the concentration of calcium in the solute,

the available carbonate,
the pH of the environment and
the availability of nucleation sites.
Microorganisms can locally influence these parameters by their metabolism, so that calcite precipitation
occurs distributedly and not locally at the injection
point of calcium solutions. A homogeneous distribution of an active bacteria culture can therefore provide
homogeneous cementation of soil under flow-through
treatment.
Bacillus pasteurii bacteria are especially appropriate to provide a controlled precipitation of calcite.
They produce large amounts of the urease enzyme that
enables them to use urea (CO(NH2 )2 ) as energy and
nitrogen source. The decomposition of urea simultaneously increases the pH in the vicinity of the bacteria
and releases carbonate (after Whiffin et al., 2007).
The bacteria surface itself can serve as nucleation site
(DeJong et al., 2006).
In sands with high hydraulic conductivity homogeneous cementation was obtained and correlation
between calcite content and shear strength could be
assessed (Harkes et al., 2008).
In this project the method of biomineralization
is evaluated with respect to its application in less
permeable soil.
2.1 Conceptual design

On a conceptual level, this study assesses the potential of combining microbial methods to increase shear
strength with chemical methods to increase permeability of soils. The former requires a sufficient
hydraulic accessibility of the open pore space in order
to place and feed microorganisms in soil. This is naturally not given in fine-grained soil, where already
a small percentage of clay can reduce the fluid flow
drastically.
The first step in this study consists of the placement
of bacteria and inducing precipitation of carbonate in
a soil with large aggregated clay minerals (Fig. 1).
The goal of this step was to investigate, whether in the
presence of clay minerals an increase in shear resistance can be achieved by biomineralization. In order
to keep the cultivation of the bacteria unaffected by
any chemical treatment, the soil used in this step was
a quartz illite mixture, where large illite clay particles
are chosen to represent an aggregated smectite phase.
In a second step, it is then investigated whether the
hydraulic conductivity allowing for successful biomineralization can be produced by means of a permanent
565
Figure 1. Cementation of clay particles by microbial

induced precipitation of carbonates.
Figure 3. Disassembled sample cell with the two removable

sidewalls for uniaxial testing.
Figure 2. Transformation of swellable clay platelets with

large diffuse double layers to aggregated clay particles.
chemical modification of the sealing smectite clay

phase (Fig. 2).
Many of the chemically enhanced drainage methods proposed in the literature are not compatible with
the use of microorganisms e.g. one of the most efficient way to increase permeability by several orders of
magnitude (Fernandez & Quigley, 1985) uses liquid
hydrocarbons, which are toxic to the microorganisms established in biomineralization methods and
in general classified as environmental pollutants.
Based on its specific interaction with the clay surface
(Pltze & Kahr, 2008) the organic salt guanidinium
was chosen as chemical agent to increase permeability. It is soluble in water and reacts with the clay at
low concentrations. Biodegradation of guanidinium
is reported for species occurring in natural surface
water (Mitchell, 1987). Different mixtures of smectite clay (Calcigel, Sd-Chemie AG, Germany) with
quartz grains (sand < 0.3 mm, Cook Industrial Minerals Pty. Ltd, Perth, Australia and silty sand, Millisil
B1, Sihelco AG, Birsfelden, Switzerland) were subjected to standard oedometer test both in their natural
and their modified form. The analysis of the timesettlement curve was used to investigate the efficiency
of the chemical improvement. Additionally its effectiveness under flow-through conditions was tested in
one set of permeameter tests.
Both steps are addressed individually in this study
and finally evaluated on their potential combination.
2.2 Experimental setup for biomineralization

Biomineralization experiments were operated with
microorganisms of the species Bacillus Pasteurii (also
classified as Sporosarcina Pasteurii). Cultivation was
conducted under aerobic batch condition in a medium

containing 20 g/l urea and 8 g/l nutrient broth. Inocculation of the samples was performed by transferring 5
volume percent of a freshly grown culture to new nutrient solution. Sterile handling techniques were applied
to keep the cultures pure and guarantee reproducibility.
The feasibility of biomineralization in clay was
evaluated by performing flow-through experiments
with bacteria suspensions and nutrient media in soil
columns. The soil was inocculated with B. Pasteurii by
mixing the soil with bacteria suspension prior to sample reconstitution. Supply of nutrient and calcium was
provided by flushing the soil with an optimized pore
water solution as proposed by DeJong et al. (2006).
Adjustment of nutrient solution to a pH 6.5 with sterilized HCl provided stable solutions without calcite
precipitates prior to bacteria exposure. The increase
in shear strength due to cementation was subsequently
assessed by uniaxial testing of biomineralized and natural control samples. Since exposure to air of the clay
rich samples would lead to drying and negative pore
water pressures the uniaxial tests were performed in
an oil bath.
The sample cell for the biomineralization experiments was based on an oedometer cell extended by two
half tubes (Fig. 3), which were removable to provide a
laterally unsupported sample for uniaxial testing.
Two biomineralization samples of an illite-quartz
soil were operated with different pumping schemes.
As active B. Pasteurii cultures raise the pH of the
pore fluid permanently, the change in pH of influent
and effluent medium was taken as rough indication
for bacteria activity. In both samples the effluent pH
exceeded 9.5 after the second day of treatment with
nutrient solution.
Intermitting pumping led to growth of the bacteria into the tubing and eventually to failure to supply
the nutrient into the soil. Constant flushing resulted in
less premature precipitation in the supply tubing. Nevertheless high backpressure was generated after two
days of feeding. The peristaltic pump was not able to
maintain the flow rate. Therefore the feeding technique
was changed to injection of media directly into the soil
with two needles. The injection point was moved by
1 cm per day. After a feeding time of nine additional
days the sample was sheared at a displacement rate of
10 mm/h. An untreated control sample was tested at
the same shear rate.
566
Figure 4. Uniaxial shear behavior of a microbially impoved

soil sample and an unmodified control sample for comparison. The shear strength is improved and the failure occurs
more brittle.
RESULTS
3.1 Biomineralization in soils with low permeability

The local injection of solution into the soil with needles
allowed supplying the nutrient to the microorganisms homogeneously. This finally led to the expected
cementation of the soil particles throughout the sample. During undrained uniaxial loading a higher failure
load was detected compared to the sample without
biomineralization (Fig. 4) and the failure occurred
more brittle. The presence of clay minerals did not
impair the cementing effect of microbial calcite
precipitation.
On the way to this result several additional qualitative and quantitative observations were made. A
considerable decrease in permeability was observed
for all biomineralization samples. The average permeability of the soil decreased from 4.8E-8 m/s before
the experiment to 3.1E-9 m/s after biomineralization.
These average values were obtained with constant
head experiments and are due to the heterogeneity of the biomineralized samples only very rough
estimates for the local decrease.
The reduced hydraulic conductivity led to a significant reduction of fluid low and a localization of the
calcite precipitation in the vicinity of the inlet. Without moving the injection point, the development of a
high backpressure and the growth of the bacteria into
the tubing would eventually lead to termination of the
treatment. Once the hydraulic conductivity declined
below 5E-9 m/s, it was no longer feasible to feed the
bacteria homogeneously with nutrients.
In order to provide an adequate flow rate under such
conditions, it is required to use either high hydraulic
gradients or long treatment time with low-activity bacteria. Both adaptions will result in unsatisfying side
effects in the field.
In practical applications increasing the hydraulic

gradient to achieve sufficient flow rates will result in
locally very high pore-water pressures close to the
injection points. For the treated soil layer the factor of safety for e.g. slope stability can therefore
be reduced significantly during the injection phase.
Since the slopes requiring soil improvements are
mostly already close to failure this additional exposure during treatment is generally not acceptable. High
injection pressures additionally bear the potential to
produce hydraulic fractures, which will lead to pronounced preferential flow and eventually to failure to
homogeneously mineralize the soil layer.
Conversely, accepting low pressures only will
require a fundamental adjustment of the used nutrients and microorganisms. Feeding abundant nitrogen
reducing soil bacteria (e.g. Castellaniella denitrificans, van Paassen et al., 2008) instead of injecting
new microorganisms together with low concentrated
nutrients is theoretically an option, but the prolonged
duration of treatment will conflict with economical and temporal requirements for soil improvement
methods.
In conclusion, the experiment have shown, that
biomineralization is possible for fine-grained soil with
a sufficiently high permeability. There is however a
risk that during treatment the permeability will fall
below the required limit. The most promising solution
to overcome this obstacle is to increase the hydraulic
conductivity in advance, so that once the process has
started fluid flow will not drop below the required
minimum. In the next step we therefore investigate the
potential of chemically enhanced drainage.
3.2 Increase of permeability in smectite/quartz
mixtures
The analysis of the time-settlement curves revealed
that permeability was increased for all samples with
the chemically modified clay compared to the unmodified samples. In mixtures with sand the permeability
was increased to such an extent, that the data could not
be evaluated as the settlement occurred almost immediately. Hence, only the data of silty sand mixed with
clay is presented, where both sample sets showed the
expected time dependent settlement according to the
1-dimensional consolidation theory.
Figure 5 shows the derived permeability at different stress levels for a clay content of 10%. The initial
difference between modified and unmodified samples
at low stresses is almost perfectly maintained during
loading. For close to surface applications (effective
vertical stress < 100 kPa) the increase is one order
of magnitude and the permeability is elevated to the
targeted range of 1E-8 m/s. The relative improvement
slightly drops during compaction of the sample by
higher stresses.
In the case of a higher clay content a more pronounced decrease during loading is observed for both
the absolute permeability and the relative improvement (Fig. 6). Here it is informative to compare the
567
Figure 5. Hydraulic conductivities during oedometric loading for 10% clay content. The permeability of the chemically
modified soil is increased over the whole range compared to
the unmodified raw material.
Figure 6. Hydraulic conductivities during oedometric loading for 20% clay content. During compaction the difference
between modified and raw material is reduced.
permeability at a given void ratio (Fig. 7) as well.

An initially more than 50 times higher permeability
is observed prior to compaction at a void ratio of 0.68.
The modified soil structure features thus more effective flow paths for circulation than the raw material at
the same void ratio. Due to the large plastic deformations, this difference reduces significantly upon
compaction.
Also under flow-through conditions the chemical
enhanced drainage maintained is effectiveness. Figure 8 compares the different treatment methods and
shows that even the structurally undisturbing delivery
of the chemical via the pore fluid lead to the targeted
increase above 1E-8 m/s.
The improvement achieved by chemically enhanced
drainage methods reached values above 1E-8 m/s
for both types of addition (mixing and flushing).
Figure 7. Comparison of hydraulic conductivity for different void ratios. During compaction both void ratio and
permeability are decreased. After large plastic deformations
the relative improvement is reduced.
Figure 8. Hydraulic conductivity over time during permeameter testing. The sample flushed with the guanidinium
solution features a gradual increase of permeability in the
course of the treatment.
Only after significant mechanical straining, the newly

formed structure is destroyed. The flow rate feasible
at this elevated permeability should be sufficient in
order to supply the microorganisms with nutrients for
carbonate precipitation.
CONCLUSION
The qualitative and quantitative findings of the performed experiments allow drawing four main conclusions about biomineralization with respect to its
application in fine-grained soils.
Firstly, microbial carbonate precipitation successfully increases shear strength above a certain limiting
568
permeability even in fine-grained soils. In this study

the minimum hydraulic conductivity was 5E-9 m/s.
It could be shown that the presence of clay minerals
did not impair the strength improvement.
Secondly, during the process of biomineralization
the precipitated calcite reduces the hydraulic conductivity. The experimental data indicated a decrease of
one order of magnitude. If during this process the
permeability declines below the limiting threshold,
incomplete and heterogeneous soil improvement will
be the result.
Thirdly, a measure to expand the application of the
microbial method to less permeable soils is to increase
the hydraulic conductivity in advance by one order of
magnitude, so that during treatment the required fluid
flow can be maintained with reasonable pore-water
pressures.
Finally, it could be shown, that with the chemical enhanced drainage technique an efficient method
to increase hydraulic conductivity exists. Increase in
hydraulic conductivity of more than one order of magnitude was achieved for soils containing smectite clay
minerals.
There is a reasonable likelihood that the chemical
method works in soil with other clays minerals as well.
Further research is recommended in order to assess
the optimum range of applicability of the combined
bio-chemical method with respect to initial hydraulic
conductivity, soil composition and microorganism.
ACKNOWLEDGMENTS
The authors would like to thank Dr. M. Pltze and
Dr. J. Tront for their help during chemical or microbiological laboratory testing respectively. The part of
this work on biomineralization has been supported by

the ASTRA/VSS grant VSS2010/502 Road-landslide
interaction: Applications.
REFERENCES
DeJong, J. T., Fritzges, M. B. & Nsslein, K. (2006). Microbially Induced Cementation to Control Sand Response to
Undrained Shear. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 132 (11): 13811392.
Fernandez, F. & Quigley, R.M. (1985). Hydraulic conductivity of natural clays permeated with simple liquid
hydrocarbons. Canadian Geotechnical Journal, 22(2):
205214.
Hammes, F. & Verstraete, W. (2002). Key roles of pH and
calcium metabolism in microbial carbonate precipitation.
Reviews in Environmental Science & Bio/Technology,
Vol. 1: 37.
Harkes, M. P., Booster, J. L., van Paassen, L. A., van Loosdrecht, M. C. M. & Whiffin, V. S. (2008). Microbial
induced carbonate precipitation as ground improvement
method bacterial fixation and empirical correlation
CaCO3 vs strength. 1st international conference BioGeoCivil Engineering, Delft, the Netherlands.
Mitchell, W. R. (1987). Microbial Degradation of Guanidinium Ion. Chemosphere, Vol. 16 (5): 10711086.
Pltze M. & Kahr G., Diagnostic intercalation in clay
minerals use of guanidine carbonate. Proc. of the 4th
MECC, Mineralogia, 33(2008): 132.
van Paassen, L. A., Daza, C. M., Staal, M., Sorokin, D.
Y. and van Loosdrecht, M. C. M. (2008). In situ soil
reinforcement by microbial denitrification. 1st international conference BioGeoCivil Engineering, Delft, the
Netherlands.
Whiffin, V. S., van Paassen, L. A. & Harkes, M. P. (2007).
Microbial Carbonate Precipitation as a Soil Improvement Technique. Geomicrobiology Journal, Vol. 24 (5):
417432.
569
Characterization of polluted sites and related aquifers

Characterization of LNAPL distribution in whole domains subject to

precipitation by the simplified image analysis method
G. Flores
Graduate School of Engineering, Kyoto University, Kyoto, Japan
T. Katsumi, T. Inui & H. Ramli

Graduate School of Global Environmental Studies, Kyoto University, Kyoto, Japan
ABSTRACT: The Simplified Image Analysis Method (SIAM) is a laboratory technique devised to assess the
three-phase dynamic saturation of water and Non-Aqueous Phase Liquids (NAPLs) in granular soils. This method
is based on an extension of the Beer-Lambert Law of Transmittance that establishes a linear relationship between
the saturation of water (S w ), NAPL (S o ), and their corresponding average optical densities (Di ), for pictures
taken using two consumer-grade digital cameras attached to two band-pass filters (wavelengths = 450 and
640 nm). In this study, we use SIAM to characterize the migration of diesel (density = 0.865 g/cm3 , viscosity
= 7 mm2 /s), a Light NAPL (LNAPL), when subject to different precipitation conditions (precipitation intensity
i = 6.6 and 63.2 mm/h) in a one-dimensional column (3.5 3.5 50 cm) filled with Toyoura sand (particle
density S = 2.65 g/cm3 , equivalent grain size D60 = 0.196).
INTRODUCTION
Non-Aqueous Phase Liquids (NAPLs) pose significant contamination risks to the groundwater when
released in the vadose zone (Mercer & Cohen 1990,
Capiro et al. 2007). Remediation of these releases in
an efficient and cost-effective way should be guided
by field data interpreted by numerical models using
the appropriate assumptions (Kechavarzi et al. 2000).
To verify the accuracy of these models, laboratory
tests should be run and precise saturation information
should be obtained, especially under the dynamic conditions usually present in nature (Lenhard & Parker
1987, Fagerlund et al. 2007, Flores et al. 2011). This
study introduces the Simplified ImageAnalysis Method
(SIAM) as a tool to measure water and light nonaqueous phase liquids (LNAPLs) saturation distributions in whole domains when evaluating the effects of
precipitation on LNAPL migration in porous media.
2 THEORETICAL BACKGROUND
Figure 1. Radiation I 0 is attenuated to I t by an absorbing

solution of concentration c moles per liter.
where Di is the optical density, a numerical constant,

b the length of the path, c the number of moles per liter
of absorbing solution, I 0 is the initial radiant power,
and I t the transmitted power (MacAdam 1981, Iizuka
1987, Skoog et al. 2007).
For digital images, the average optical density Di is
defined for the reflected light intensity as:
2.1 Image analysis

The Beer-Lambert Law of Transmittance states that
when a beam of monochromatic radiation I 0 strikes a
block of absorbing matter perpendicular to a surface,
after passing through a length b of the material, its
power is decreased to I t as a result of absorption:
where N is the number of pixels contained in the area of

interest and, for a given spectral band i, d ji is the optical
density of the individual pixels, I rji is the intensity of the
reflected light given by the individual pixel values, and
I 0ji is the intensity of the light that would be reflected
by an ideal white surface (Kechavarzi et al. 2000).
573
It has been shown (Flores et al. 2011) that the

Beer-Lambert Law of Transmittance establishes a
linear relationship between optical density and the
concentration of a dye:
where D0 is the optical density of a solution of unit

concentration, and Dt the optical density of a solution
of concentration c. This linear relationship was experimentally corroborated by Kechavarzi et al. (2000) and
by Flores et al. (2011).
When two cameras with two different band-pass filters (wavelengths = i and j) are used, and when water
and NAPL are mixed with dyes whose predominant
color wavelengths are also i and j, we obtain two different sets of linear equations that can be solved for
S w and S o :
This is the base of the Multispectral Image Analysis

Method: the calculation of two correlation equations
via calibration tests using small samples and their
subsequent use to determine water and NAPL saturation values (S w and S o ) on larger three-phase
(air/water/NAPL) domains (Kechavarzi et al. 2000).
2.2
Simplified Image Analysis Method (SIAM)
The Multispectral Image Analysis Method relies on

the use of different band-pass filtersflat by designto
allow digital cameras to capture the light reflected by
the studied system on one particular wavelength each.
Since these band-pass filters are designed for parallel
light, they behave in a different way according to the
angle of incidence of the reflected light.
Noting that the relative position between the camera and the domain remains constant throughout the
tests, the different angles of incidence for the reflected
light are also constant. This means that instead of one,
the system behaves as if there existed dozens of small
(and different) band-pass filters that remain fixed on
position for the whole duration of the test, requiring
the preparation of a different set of calibration equations for each small assumed band-pass filter, which
would be impractical.
Observing that each Equation (4) represents a plane,
and that only three non-collinear points are needed
to define one, a careful choice of points will provide
the needed set of equations for each mesh element.
For these studied conditions, the best points are those
located in the three extremes of the plane:
S w = 0%; S o = 0% Dry Sand

S w = 0%; S o = 100% LNAPL Saturated Sand
S w = 100%; S o = 0% Water Saturated Sand
Figure 2. Experimental setup.
optical density values for each mesh element of the

studied domain can then be calculated and compared
to the corresponding ones for all three cases, and a
matrix of correlation equation sets can be obtained,
each one corresponding to each mesh element, that
can be solved for S w and S o :
where m and n are the dimensions of the matrix, [Di ]mn

and [Dj ]mn are the values of average optical density
of each mesh element for wavelengths i and j;
00
[D00
i ]mn and [Dj ]mn are the average optical density of
10
each mesh element for dry sand; [D10
i ]mn and [Di ]mn
01
and
[D
]
for
for water saturated sand; and [D01
]
mn
mn
i
i
NAPL saturated sand. This is the base of the Simplified
Image Analysis Method (SIAM).
3
EXPERIMENT
3.1 Equipment and materials

A 3.5 3.5 50 cm column with a transparent glasswall was designed to study the behavior of a
LNAPL, affected by precipitation, in a Toyoura sand
(S = 2.65 g/cm3 , D60 = 0.196, e = 1.1) column with
the aid of the Simplified Image Analysis Method.
The wetting fluid was blue-dyed water (Brilliant Blue
FCF, 1:10000) and the non-wetting fluid was reddyed diesel 2 ( = 0.865 g/cm3 , = 7 mm2 /s, Sudan
III, 1:10000), a LNAPL considered in the pollution
inventory lists of Australia, Canada, the UK, USA and
other countries (Australia DSEWPC 1999, Environment Canada 2010, UK Environment Agency 2011,
US EPA 2011).
3.2 Calibration
If the domain is filled with sand under each limit

condition, and later photographed, all elements of the
matrix will share the same conditions. The average
This test required three calibration pictures with each

camera: each camera took a picture of the column filled with dry sand, with water saturated sand,
and with LNAPL saturated sand. These six pictures
574
10
01
00
correspond to [D00
450 ]mn , [D450 ]mn , [D450 ]mn , [D640 ]mn ,
10
01
[D640 ]mn , and [D640 ]mn .
3.3 Experiment
The column was initially filled with Toyoura sand
fully saturated with water. The test was divided in four
stages: water drainage (t = 0 to 6 h), LNAPL infiltration (t = 6 to 54 h), precipitation (t = 54 to 55 h), and
stabilization (t = 55 to 91 h). To study the effects of
precipitation on the migration of LNAPL, two experiments were run with different precipitation intensities:
i = 6.6 and 63.2 mm/h.
Drainage. The water tank was quickly lowered
55 cm from its original position to a height of
h = 5 cm (5 cm below the bottom of the column),
and the water inside the column was allowed to drain.
Air was allowed to freely infiltrate the sand column
through a porous stone located on top of it. The top
of the column was open to avoid producing a vacuum
effect. This stage took 6 hours.
LNAPL infiltration. During this stage, 25 ml of
LNAPL were infiltrated from the top of the column
through the porous stone. The system was let stabilize
for 2 days. This stage took 48 hours.
Precipitation. To simulate precipitation, a pump was
set to deliver water to the system through the porous
stone, at two different rates. For the first experiment,
the rate was 6.6 mm/h. For the second one, 63.2 mm/h.
This stage took 1 hour.
Stabilization. After the precipitation, the system
was let stabilize while the behavior of LNAPL was
observed via the SIAM. This stage took 36 hours.
Two digital pictures of the column were taken simultaneously every hour using two cameras, one with a
450 nm band-pass filter and the other with a 640 nm
band-pass filter. Both cameras were set to manual
mode, and all the pictures were acquired with the same
aperture, shutter speed, and white balance settings.
The cameras were remotely controlled (using Nikon
Camera Control Pro 2 software) to avoid vibrations
and displacement. The two 500 W floodlights were
turned on 30 seconds prior to taking each picture and
turned off 30 seconds afterwards to avoid changing
the temperature of the column. Room temperature was
maintained at 20 C and humidity at 70%.
3.4
Figure 3. Water (top) and LNAPL (bottom) saturation with

time, at different heights, for precipitation i = 6.6 mm/h.
Computational analysis
All pictures were exported from NEF format (Nikon

proprietary RAW version files) to TIFF format
(Tagged Image File Format) using Nikon ViewNX
1.5.0. The TIFF images were analyzed with an ad-hoc
program written in MATLAB 2007a.
Using the six calibration pictures obtained in 3.2,
the average optical density matrices [D450 ]mn and
[D640 ]mn were calculated for each picture taken during
the test, and the water and LNAPL saturation matrices
([S w ]mn and [S o ]mn ), which correspond to each picture
following the procedure described in Section 2.2, were
obtained by solving Equation (5).
Figure 4. Water (top) and LNAPL (bottom) saturation with

time, at different heights, for precipitation i = 63.2 mm/h.
Figure 3 and Figure 4 plot the values of water and

LNAPL saturation (S w and S o ) at different heights
575
Figure 5. Water (top) and LNAPL (bottom) saturation

distribution with time, for precipitation i = 6.6 mm/h.
(h = 5, 10, 15, 20, 25, 30, 35, 40, 45, and 50 cm),
for both experiments (i = 6.6 and 63.2 mm/h).
While Figures 34 allow us to observe with a great
level of detail the changes with time in the values
of water and LNAPL saturation (S w and S o ) at different heights, they still hide the general behavior of
the LNAPL migration due to precipitation. To better
observe this behavior, and thanks to the continuous
recording of the S w and S o values at all heights (not
just at the heights plotted in Figures 34) that the Simplified Image Analysis Method allows, we will plot
all data using a XY contour graphic that will help us
recognize the migration behavior of the studied contaminant. In Figures 56, the horizontal axis represents
time, the vertical one, height, and the tone of the grey
scale represents saturation values.
To observe the behavior of the LNAPL during the
precipitation stage (t = 54 to 55 h), we can zoom in
the XY contour graphic to the appropriate time. Figure 7 allows us to compare the effects of the different
precipitation intensities (i = 6.6 and 63.2 mm/h) in the
LNAPL migration behavior.
From the comparison of Figures 36 we can observe
the differences in the behavior of our LNAPL when
subject to different precipitation intensities. The general behavior during the first 54 h is similar for both
experiments due to the similar characteristics of the
first two stages. Differences in saturation values can
be attributed to slight differences during the packing
processes.
Figure 7 helps us observe how both precipitation
intensities push LNAPL downward, despite our contaminant having a density lower than that of water. It is
also clear how the stronger intensity (i = 63.2 mm/h)
pushes the LNAPL 5 cm deeper than the lighter one
(i = 6.6 mm/h), both in its upper and lower limits.
Figure 6. Water (top) and LNAPL (bottom) saturation

distribution with time, for precipitation i = 63.2 mm/h.
Figure 7. LNAPL saturation distribution with time,

between t = 54 and 55 h, for precipitation i = 6.6 (top) and
63.2 mm/h (bottom).
A small LNAPL bulb is separated from the main group

by the precipitation in the stronger case, but most of
it is pushed up by the water during the stabilization
stage, as can be observed in Figure 6. After the stabilization, the bottom limit of the LNAPL is similar
for both cases, but the upper limit remains approximately 5 cm lower, which means that the LNAPL has
a higher average concentration after being subject to
the stronger precipitation, than the lighter one.
From Figure 3 (bottom) and Figure 4 (bottom) we
can calculate the values of maximum and residual
576
the residual LNAPL saturation value can be roughly

predicted from the maximum saturation reached at
each point over time, with a 0.3:1 approximate relationship between them.
This study focused only on the effects of precipitation intensity in the migration behavior of LNAPL.
Further study will be done on the effects of LNAPL
volume and precipitation length.
REFERENCES
Figure 8. Relationship between maximum and residual

LNAPL saturation values.
LNAPL saturation (Figure 8). We can observe a linear

trend between both values, irrespective of the intensity
of the rain. This means that the residual LNAPL saturation value can be roughly predicted from the maximum
saturation reached at each point over time. Since the
residual saturation value is, by definition, lower than
the maximum saturation, this trend has a slope lower
than 1:1 (in the particular case of our study, 0.3:1).
5
CONCLUSIONS
This study has verified that the Simplified Image

Analysis Method (SIAM) can be used as an effective laboratory technique to measure the saturation of
both water and LNAPL in full domains, when subject
to dynamic conditions such as precipitation of varied
intensities.
In this particular research, two precipitation intensities were studied: i = 6.6 and 63.2 mm/h. It was
observed that the precipitation forces the LNAPL to
migrate downward, to different depths that depend on
the intensity. Buoyancy forces push LNAPL upward
again after the precipitation has finished, and the bottom limit will be relocated to approximately a similar
depth, irrespective of the precipitation intensity. The
upper limit, however, will remain lower in the case
of the higher precipitation, causing that the LNAPL
region will have a slightly higher average saturation.
A linear relationship was found between the maximum and residual saturation values of LNAPL, irrespective of the precipitation intensity. This means that
Australia DSEWPC 1999. Substance list and thresholds:

National pollutant inventory.
Capiro, N. L., Stafford, B. P., Rixey, W. G., Bedient, P. B. &
Alvarez, P. J. J. 2007. Fuel-grade ethanol transport and
impacts to groundwater in a pilot-scale aquifer tank. Water
Research, 41, 656664.
Environment Canada 2010. National pollutant release
inventory. Canada Gazette.
Fagerlund, F., Illangasekare, T. H. & Niemi, A. 2007.
Nonaqueous-phase liquid infiltration and immobilization
in heterogeneous media: 1. Experimental methods and
two-layered reference case. Vadose Zone J, 6, 471482.
Flores, G., Katsumi, T., Inui, T. & Kamon, M. 2011. A simplified image analysis method to study LNAPL migration
in porous media. Soils and Foundations, 51, 835847.
Iizuka, K. 1987. Engineering optics, New York, SpringerVerlag.
Kechavarzi, C., Soga, K. & Wiart, P. 2000. Multispectral
image analysis method to determine dynamic fluid saturation distribution in two-dimensional three-fluid phase
flow laboratory experiments. Journal of Contaminant
Hydrology, 46, 265293.
Lenhard, R. J. & Parker, J. C. 1987. Measurement and prediction of saturation-pressure relationships in three-phase
porous media systems. Jnl. of Contaminant Hydrology, 1,
407424.
MacAdam, D. L. 1981. Color measurement, theme and
variations, New York, Springer-Verlag.
Mercer, J. W. & Cohen, R. M. 1990. A review of immiscible
fluids in the subsurface: Properties, models, characterization and remediation. Journal of Contaminant Hydrology,
6, 107163.
Skoog, D. A., Holler, F. J. & Crouch, S. R. 2007. Principles of instrumental analysis, Belmont, CA, Thomsom
Brooks/Cole.
UK Environment Agency 2011. Pollution inventory substances: Pollution inventory.
US EPA 2011. Toxics release inventory (TRI): Toxic chemical
list.
577

Interaction of metal adsorption and solubilization in a lateritic clay

S.G. Gabas
Technology and Science Center of Mato Grosso do Sul Federal University, Campo Grande, Brazil
M.E.G. Boscov
Polythecnic of Sao Paulo University, Sao Paulo, Brazil
ABSTRACT: The investigation aimed to assess minerals and mechanisms responsible for pollutant retention in
lateritic clay in order to understand metal speciation. Compacted specimens were initially submitted to diffusion
tests carried out with solutions of Cd and Pb at pH values of 3.0, 5.5 and 8.0, and subsequently to the BCR
sequential extraction procedure. Diffusion tests showed relevant retention of both metals. Cadmium showed
more affinity for the exchangeable fraction of the soil (kaolinite) and lead for the iron oxides and hydroxides.
Moreover, there was an important contribution of metal concentration from soil minerals: some metals are
solubilized from the soil while others are adsorbed. These coupled phenomena could be a very important factor
of chemical mobilization in groundwater of otherwise stable iron, aluminum and other elements such as arsenic.
This study offered evidence that these reactions should be taken into account in studies concerning tropical soils.
INTRODUCTION
Extensive research has been conducted on pollutant

transport and retention in soils, either to understand the
involved mechanisms (Rowe 1988; Rowe et al. 1888;
Shackelford 1989, 1991, 1993) or to apply this knowledge in contamination prevention or in remediation
(Roehl & Czurda 1998; Kugler et al. 2002; Stuemer
et al. 2008).
The most frequent contaminants in the environment are pathogenic microorganisms, inorganic chemicals, organic chemicals and radioactive nuclides
(Aswathanarayana 1995). Among the inorganic chemicals, some toxic metals (Hg, Cd, Pb, Cr, Ni) have been
continually accumulated in the environment.
Mobility of inert, dissolved contaminants in porous
media can be described by the advective/dispersive
transport equation, which in one-dimensional form
may be represented as:
where l is the directional coordinate in the flow direction, ul is the solute interstitial velocity (u = v/n, v is
the Darcy velocity and n is the total effective soil porosity), Dl is the hydrodynamic dispersion coefficient in
the flow direction and C is the solute concentration.
When sorption of contaminants by soil grains takes
place, solute migration may be expressed as:
where Rd is the retardation factor, which represents all

sorption processes that lead to solute immobilization.
Most studies of soil contamination by metals focus
on the evaluation of transport equation parameters
such as diffusion coefficient and retardation factor,
whereas few have investigated how heavy metals are
actually fixed to the soil barrier (Roehl & Czurda 1998;
Li 2003; 2006; Kugler et al. 2002).
Metal sorption by soil particles is controlled by
interactions between solid and liquid phases and
involves different processes, i.e. adsorption (nonspecific and specific), ion exchange reactions, complexation reactions and precipitation (Sposito 1989;
Stumm 1992; McBride 1994). These processes differ by the type of chemical bonding and the strength
of fixation of the metal to the soil particle, both
important regarding long-term immobilization and
remobilization.
Non-specific adsorption is considered a reversible
reaction and normally takes place in permanent charge
clay minerals, such as smectites and vermiculite.
Specific adsorption is considered stronger than nonspecific adsorption (Sposito 1989) and occurs at
variable-charge minerals surfaces, like the edges of
individual kaolinite platelets, iron and manganese
oxides and aluminium hydroxide, and organic matter.
These minerals have charge deficiencies that are pH
dependent, resulting from the dissociation of hydroxyl
ions (OH ) and water molecules (H2 O) attached to
their surfaces. The strength of specific adsorption is
due to ionic and covalent bonds.
Sequential extraction procedures as proposed by
the Community Bureau of Reference (European
579
Commission 2001) allow for the assessment of metal

speciation in soils. In spite of the fact that most chemical extraction methods are reagent dependent (Meers
et al. 2007) and not phase-specific (Quevauviller et al.
1995), they make it possible to determine operationally
defined groups of metal fractions in the soil. Critical
reviews and improvements in extraction procedures
are presented by Davidson et al. (1998, 2004) and
Gomz-Ariza et al. (1999, 2000). Sequential extraction applications to environmental issues are discussed
by Kluger et al. 2002, Nel et al. 2007, Doelsh et al.
2008 and Petrucci et al. 2011.
2
2.1
MATERIAL AND METHODS

Material
A residual lateritic clayey soil of diabase from the State

of Sao Paulo (southeast of Brazil) was investigated.
Lateritic clays cover large extensions of Brazil and are
often used as clay liners for landfill sites as well as in
dam and pavement construction due to their adequate
geotechnical properties when compacted, such as high
shear strength, low compressibility, low expansibility
in the presence of water, and low hydraulic conductivity (Boscov 1997). The physicochemical characterization of the soil determined by standard methods is
presented in Table 2.
2.2
Diffusion tests
Diffusion tests were carried out with soil specimens

compacted at optimum water content of Proctor standard effort inside the diffusion cells (Figure 1), saturated with distilled water by capillarity, and finally
topped with cadmium and lead solutions. Tests were
carried out in a temperature controlled (22 2 C)
room during 90 days. During the tests, samples of
the reservoir solution were periodically collected for
element determination.
At the end of the tests, cells were dissembled, and
soil specimens were extruded and cut into five slices
of equal thickness. The soil slices were submitted to
a pressure of 25 MPa in order to obtain the interstitial
solution for chemical analysis. Metal adsorption in the
top slice of the soil specimens was also evaluated using
a sequential extraction procedure.
Cadmium (Cd) and lead (Pb) were chosen for the
investigation because of their different mobility in soils
(Kugler et al. 2002; Li 2006). Concentration of test
solutions was 3 mg L1 for Cd and 100 mg L1 for Pb,
values based on the limits prescribed for soil quality according to Brazilian environmental legislation
(Brazil 2009). For each metal, three solutions were prepared at pH of 3.0, 5.5 and 8.0, which were adjusted
by the addition of HNO3 and NH4 OH.
Diffusion blank tests, herein called reverse diffusion tests, were also carried out with solutions at
the three pH values without metal concentration in
order to evaluate metal migration from the soil to
Figure 1. Diffusion test: (a) assemblage; (b) initial and final

concentrations in the reservoir and in the soil pore water.
the reservoir solution. Additional blank tests were

executed to monitor pH during diffusion time.
2.3 Sequential extraction procedure
The top slice of each soil specimen submitted to diffusion, after being squeezed by the application of a
25-MPa pressure, was submitted to sequential extraction. The compressed soil was broken up with a
hammer and crushed to powder in an agate crusher.
Three replicates of each sample, weighted after homogeneous mixing, were submitted to the sequential
extraction procedure.
Three extraction steps, proposed by BCR (European
Commission 2001), were used to define fractions that
retain metals. Stability of bonds increases and mobility of metals decreases in the following order: Step 1,
Step 2, Step 3 and Step 4 (Table 1).
2.4 Chemical analyses
Chemical analyses of cadmium, lead, iron and aluminum were carried out by inductively coupled plasma
580
Table 1. The three-step BCR sequential extraction procedures.

Fraction
Target phase
Procedure
Step 1
Exchangeable
0.11 mol/L acetic acid
Step 2
Step 3
Reducible
Oxidable
Soluble species, carbonates, cation

exchange sites
Iron and manganese oxihydroxides
Organic matter and suphildes
Step 4
Residual
Occluded in remaining minerals
Table 2.
0.5 mol/L hydroxylammonium chloride at pH 1.5

Hydrogen peroxide followed by 1.0 mol/L
ammonium acetate at pH 2
Aqua regia
Physicochemical characterization of the soil.
Property
Result
Method
pH (Distilled water)
Organic matter content (%)
Cation exchange capacity (mmolc /kg)
5.3
2.5
56.6
Point of zero charge

Grain-size distribution (%)
4.3
0
6
27
14
59
45
30
15
0.25
26.3
15.7
1 109
ASTM D4972
Walkley-Black Method (CAES, 2006)
Ion exchange resin
(Rodella and Alcarde, 1994)
Titulometric method
ASTM D422
Coarse sand
Medium sand
Fine sand
Silt (%)
Clay (%)
Liquid limit, wL (%)

Plasticity limit, wP (%)
Plasticity index, PI (%)
Skempton activity, A
Optimum water content (%)
Maximum dry unit weight (kN/m3 )
Compacted hydraulic conductivity (m/s)
(optimum water content, Proctor energy)
Chemical composition
Mineralogical composition
ASTM D4318
ASTM D4318
ASTM D4318
A = PI/C (C = clay content)
ASTM D698
ASTM D698
Constant-head, flexible-wall permeability test
Si: 31.1%; Fe: 24.4%;

Al: 23.9%; Ti: 5.2%;
Mg: 0.17%; Mn: 0.14%
Quartz, ilmenite, hematite,
magnetite, gibbsite and
traces of vermiculite
atomic emission spectrometry (ICP-AES) (Atom

Comp Series 800, Jarrel-Ash Division, USA) and
inductively coupled plasma mass spectrometry (ICPMS) (Element, Finnigan MAT, Bremen, Germany).
Reservoir samples collected in the first 15 days
of diffusion were analyzed by ICP-AES due to the
initial high concentrations, and henceforth by ICPMS. Extracts of the sequential extraction tests were
also analyzed by ICP-MS; analytical calibration curves
were prepared with the extraction solutions used in
each step, i.e. same matrix. Iron and aluminum for all
extracts were analyzed by ICP-AES.
A reference certified material (RCM), BCR 701, was

used to evaluate the quality of chemical extraction
and analysis. The same extraction solutions were used
for the RCM and for the soil sequential extraction
tests. Total Cd, Cu, Cr, Ni, Pb and Zn, which are the
metals controlled by the certified method (European
X-ray difractometry
Magnetic treatment 0.5A
Glycol + heating at 490 C + FeO treatment
Commission 2001), were determined. Reagent blanks

were also analyzed.
3
3.1
RESULTS
Physicochemical characterization
Results of physicochemical and geotechnical characterization and the related methods are presented in
Table 2.
3.2
2.5 Quality assurance
X-ray fluorescence
Diffusion tests
Blank diffusion tests showed that solution pH does

not remain constant during the experiment (Figure 2),
tending with time to the pH of the natural soil (5.3),
and that there is metal diffusion from the soil to the
reservoir solution. Metal concentrations in the reservoir as a function of time and in the soil pore water in
the end of the tests are indicated in Table 3.
581
Figure 2. pH of reservoir solutions along diffusion tests.

Table 3. Metal concentrations at the diffusion tests
(g L1 ): reservoir solution (Rs) and soil pore water (Spw).
Metal
Cadmium
pH
Rs Initial
Rs Final
Spw Layer 1 (Top)
Spw Layer 2
Spw Layer 3
Spw Layer 4
Spw Layer 5 (Bottom)
2.83
3.12
0.12
23.1
1.29
0.26
0.04
0.10
5.65
2.95
0.29
33.2
1.46
0.22
0.11
0.15
Lead
8.02
3.37
0.37
134
4.44
0.95
0.50
0.09
2.98
112
0.14
0.61
0.27
0.23
0.16
0.10
5.73
101
0.14
9.30
1.00
0.60
0.68
0.24
7.90
98.2
0.13
5.60
1.22
0.78
0.60
0.27
Initial cadmium concentrations in the reservoir

were approximately 3 g L1 , whereas after 90 days
of diffusion values of 0.12, 0.29 and 0.37 g L1 were
measured, respectively, for pH 3.0, 5.5 and 8.0. These
low final reservoir concentrations indicate expressive
occurrence of diffusion during the 90 days of the tests.
Cadmium that migrated from the reservoir to the
soil specimen was partly retained by the soil grains
and partly diffused downwards through the specimen.
Concentrations in the soil pore water indicate the pollutant front entering the soil, with higher values in
the top layer and a background value (0.10 g L1 )
at the bottom layer. Initial lead concentrations in the
reservoir were about 100 g L1 , decreasing to near
0.14 g L1 in the end of the test. The pollutant migration front in the soil specimens is also evident by the
observation of the concentrations in the soil pore water
of each layer.
3.3
Sequential extraction procedure
Cadmium was mainly adsorbed at Step 1 (exchangeable) and secondly at Step 2 (Fe and Mn oxides)
(Figure 3). Lead was markedly adsorbed at Step 2
(iron oxides), secondly at Step 1 (exchangeable), and
slightly at Step 3 (sulphides and organic matter) at pH
8.0 (Figure 4). However, characterization tests results
indicate low concentration of organic matter and no
sulphides in the soil sample. This suggests that lead
could have been readsorbed from Step 2 to Step 3, as
already noticed by some authors (Sutherland & Tack
2003; Davidson et al. 2004).
Figure 3. Cadmium distribution in the top layer of the soil

specimen after the diffusion test: (a) metal concentration;
(b) proportion.
Iron and aluminium solubilization was observed

during cadmium and lead diffusion tests. Concentrations of iron and aluminium were also quantified
for the samples from all diffusion tests, as shown in
Figures 5 and 6.
The relative distribution of the extracted mass of
iron for the diffusion tests with cadmium (Figure 5)
and lead (Figure 6) solutions is different from that
observed in the reverse diffusion tests (Figure 7): iron
oxides phase and contributes with a higher percentage
whereas the exchangeable phase decreases significantly in importance (Figure 7a). For aluminium, in
the other hand, the decrease in iron extraction from the
exchangeable phase corresponds to an increase in step
3 (organic matter) (Figure 7b). Although not expected
in Step 3, iron extraction has also been observed
by other authors and can be related to readsorption
(Davidson et al., 2004) and to the non specificity of
the extractors (Gonalves et al., 2004).
The total amount of iron and aluminum extracted
from the soil is also different for reverse diffusion tests
and cadmium and lead diffusion tests (Figure 8). Iron
and aluminium extraction was higher in the samples
undergone reverse diffusion tests than in those submitted to cadmium and lead diffusion tests for pH 3 and
5.5, and a slightly smaller for pH 8.. The difference
between the results from samples of reverse and direct
582
Figure 4. Lead distribution in the top layer of the soil

specimen after the diffusion test: (a) metal concentration;
(b) proportion.
Figure 6. Metal extraction from lead diffusion tests: (a) iron;

(b) aluminum.
Figure 5. Metal extraction from cadmium diffusion tests:

(a) iron; (b) aluminum.
diffusion tests was more remarkable for iron than for

aluminium.
Attention should be called to the fact that cadmium,
a mobile and easily exchangeable cation, was considerably adsorbed specifically in the iron oxides: 30 to 40%
of the total adsorbed mass is not subject to immediate
and reversible adsorption (Figure 3b). This behaviour
is even more impressive for lead, as 70 to 80% of
the total mass was specifically adsorbed (Figure 4b).
It should be considered that BCR procedures could
underestimate the proportion of heavy metal bound to
this fraction (Doesch et al. 2008). Anyway, the results
may be significant for the feasibility of remediation
techniques such as soil flushing or bioremediation,
inasmuch as they directly affect desorption behaviour.
They could also explain the remarkable adsorption
capacity of lateritic soils when compared to their low
cation exchange capacity, indicating that the contribution of variable charges associated to iron oxides is
relevant. Some environmental implications for metal
adsorption into manganese oxides are discussed by
Landrot et al. 2012.
Results indicate that solute transport and chemical equilibrium speciation are coupled processes that
should be considered in research experiments and in
practical applications. Mathematical models that simulate simultaneous processes, such as proposed by
Darban et al. 2010, are not common yet, but should
be developed and improved.
583
CONCLUSIONS
Cadmium and lead adsorption occur in compacted lateritic soil as indicated by the diffusion tests results.
As expected, metal diffusion is pH dependent, but pH
does not remain constant during diffusion tests.
Sequential extraction tests indicate that metal
adsorption occurs also at variable charge minerals,
as kaolinite and iron oxides. Cadmium is mainly
adsorbed at the exchangeable fraction of the soil
(kaolinite) and lead at iron oxides. It is remarkable that 30 to 40% of the total mass of cadmium,
which is usually a mobile ion, is adsorbed at iron
oxides by a stronger bonding than that of the generally considered immediate and reversible non-specific
adsorption. Specific adsorption is even more noticeable for lead, since 80% of the total mass of this metal
was retained by iron oxides.
Iron and aluminium solubilization during diffusion
and sequential extraction tests indicate the occurrence
of coupled phenomena that might happen between
solutions and clay liners.
Although sequential extraction procedures have
some disadvantages, i.e they are complex and timeconsuming, they are useful to evaluate possible interactions between solutes and soils and to prevent
undesired metal liberation from liners or remediation
structures.
REFERENCES
Figure 7. Metal extraction from reverse diffusion tests:
(a) iron; (b) aluminum.
Figure 8. Metal extraction in diffusion tests: (a) iron;

(b) aluminum.
Aswathanarayna, U. 1995. Geoenvironment an introduction.

Rotterdam/Brookfield: A.A. Balkema.
Boscov M.E.G. 1997. Contribuio ao projeto de sistemas de
conteno de resduos perigosos utilizando solos laterticos. So Paulo. Escola Politcnica, Universidade de So
Paulo.
Brazil. 2009. National Council of the Environment. Resolution n.420.
Darban, A.K.; Yong, R.N.; Rajav, R. 2010. Coupled chemical speciation-solute transport model for prediction of
solute transport in clay buffers. Applied Clay Science, 47:
127132.
Davidson, C.M.; Duncan, A.L.; Littlejohn, D.; Ure, A.M.;
Garden, L.M. 1998. A critical evaluation of the three-stage
BCR extraction procedure to assess the potencial mobility
and toxicity of heavy metals in industrially-contaminated
land. Analytica Chimica Acta, 363: 4555.
Davidson, C.M.; Hursthouse, A.S.; Tognarelli, D.M.; Ure,
A.M.; Urquhart, G.J. 2004. Should acid ammonium
oxalate replace hydroxilammonium chloride in step
2 of the revised BCR sequential extraction protocol
for soil and sediment? Analytica Chimica Acta, 508:
193199.
Doelsch, E.; Moussard, G.; Saint Macary, H. 2008. Fractionation of tropical soil-borne heavy metals comparison of
two extraction procedures. Geoderma, 143: 168179.
European Commission. 2001. The certification of the
extractable contents (mass fractions) of Cd, Cr, Cu, Ni,
Pb and Zn in freshwater sediment following a sequential
extraction procedure BCR-701. Report EUR 19775 EN.
77p.
Gmez-Ariza, J.L; Girldez, I.; Snchez-Rodas, D.; Morales,
E. 1999. Metal readsorption and redistribution during the
584
analytical fractionation of trace elements in oxic estuarine

sediments. Analytica Chimica Acta, 399:295307.
Gmez-Ariza, J.L; Girldez, I.; Snchez-Rodas, D.; Morales,
E. 2000. Metal speciation extraction procedure optimized for heavily polluted and iron oxide rich sediments.
Analytica Chimica Acta, 414:151164.
Gonalves, M.A.; Nogueira, J.M.F.; Figueiras, J.; Putnis, C.V.;
Almeida, C. 2004. Base metals and organic content in
stream sediments in the vicinity of a landfill. Applied
Geochemistry, 19:137151.
Kugler, H.; Ottner, F.; Froeschi, H.; Adamcova, R.;
Schwaighofer, B. 2002. Retention of inorganic pollutants
in clayey base sealings of municipal landfills. Applied
Clay Science, 21: 4558.
Landrot, G.; Ginder-Vogel, M.; Livi, K.; Fitts, J. P.; Sparks,
D.L. 2012. ChromiumIII oxidation by three poorly cristallyne manganese oxides. 2. Solid phase analyses. Enviromental science and technology. 46: 1160111609.
Li, L.Y. 2003. Multi-component of heavy metal contaminants
adsorptivity and compatibility onto variable charge clay
mineral. Clay Science, 12, N 2: 7280.
Li, L.Y. 2006. Retention capacity and environmental monbility of Pb in soils along highway corridor. Water,
Air, and Soil Pollution, 170: 211227, Springer, DOI:
10.1007/s11270-005-9002-0.
Petrucci, E.; Montanaro, D.; Merli, C. 2011. Sequential
extraction analysis provides decision-making tools for the
use of contaminated sediments. Chemistry and Ecology,
27, Supplement, February: 107118.
Quevauviller, Ph; Maier, E.A.; Griepink, B. 1995. Quality
assurance for environmental analysis method evaluation within the Measurements and Testing Programme.
Amsterdam: Elsevier, 649p.
McBride, M.B. 1994. Environmental Chemistry of Soils.
New York: Oxford University Press.
Meers, E.; Du Laing, G.; Unamuno, V.; Ruttens, A.;
Vangronsveld, J.; Tacka, F.M.G.; Verloo, M.G. 2007.
Comparison of cadmium extractability from soils by
commonly used single extraction protocols. Geoderma,
141: 247259.
Nel C.; Soubrand-Colin, M.; Piquet-Pissaloux, A.; Bril, H.

2007. Mobility and bioavailability of Cr, Cu, Ni, Pb,
Ni and Zn in a basaltic grassland: comparison of selective extractinos with quantitative approaches at different
scales. Applied Geochemistry, 22: 724735.
Roehl, K.E.; Czurda, K. 1988. Diffusion and solid speciation
of Cd and Pb in clay liners. Applied Clay Science, 12:
387402.
Rowe, R.K. 1988. Contaminant migration through
groundwater- the role of modeling in the design of
barriers. Eleventh Canadian Geotechnical Colloquium.
Canadian Geotechnical Journal, 25: 778798.
Rowe, R.K.; Caers, C.J.; Barone, F. 1988. Laboratory determination of diffusion and distribution coefficients of
contaminants using undisturbed clayey soil. Canadian
Geotechnical Journal, 25: 108118.
Shackelford, C.D. 1993. Contaminant transport. In DANIEL,
D.E. (Ed.). Geotechnical Practice for waste disposal.
Londres: Chapman & Hall.
Shackelford, C.D. 1989. Diffusion of contaminants through
waste containment barriers. Transport Research Record,
1219: 169182.
Shackelford, C.D.; Daniel, D.E. 1991. Diffusion in saturated soils: II- results for compacted clay. Journal of
Geotechnical Engineering, ASCE, 117 (3): 467484.
Sposito, G. 1989. The Chemistry of Soils. New York: Oxford
University Press.
Stuemer, M. M.; Boscov, M. E. G.; Oliveira, E. 2008. Metal
transport parameters of a gneiss saprolitic silty soil for
liner design. Journal of Geotechnical and Geoenvironmental Engineering, 134(1): 417554.
Stumm, W. 1992. Chemistry of the solid-water interface: processes at the mineral-water and particle-water interface
in natural systems. New York: John Wiley.
Sutherland, R.A.; Tack, F.M.G. 2003. Fractionation of Cu,
Pb and Zn in certified reference soil SRM 2710 and
SRM 2711 using the optimized BCR sequential extraction procedure. Advances in Environmental Research, 8:
3750.
585

Factors affecting heavy metal leaching from excavated rocks

with natural contamination
T. Inui, T. Katsumi & A. Takai
M. Kamon
Kagawa National College of Technology, Takamatsu, Japan
ABSTRACT: When excavation works are executed at stratums that naturally contain heavy metals due to their
geologic histories, time-saving and reliable assessment of contamination potential by the metals and metalloids
in excavated soil/rock is required. Leaching of heavy metals from excavated rocks is induced as the results both
from physical factors (crushing, exposure to water, temperature) and chemical factors (oxidation, soil-water
interaction, pH change, etc.). This paper addresses long term leaching behavior of arsenic and lead affected
by these factors, which were determined by outdoor exposure tests conducted on the excavated rock samples
weathered outdoors for more than two years. In addition, several time-saving laboratory tests for characterizing
the contamination potential of arsenic and lead were validated by comparatively assessing the results of the
outdoor exposure tests, to develop some simple evaluation methods of contamination potential by metals from
natural resources.
INTRODUCTION
In Japan, soils and rocks often contain some metals

such as lead and arsenic with relatively low concentrations due to not anthropogenic contamination, but
their geologic histories (Geological Survey of Japan,
AIST 2004). These constituents are fundamentally stable and immobile under the depositional environment.
However, leaching of these constituents is triggered by
continuous exposure to water and oxygen when excavation works are executed in such soil/rock layers. It is
well known that acid drainage is generated mainly by
interactions between water, oxygen, and sulfide minerals in soil/rock (Equations 1), and often increases the
leaching concentrations of these metals.
Thus, development of a testing methodology to estimate the long-term leaching potential and behavior of
heavy metals from natural resources has been a great
geoenvironmental challenge in Japan.
Excavated materials containing heavy metals,
even from natural resources, should be properly
stored/reused in the ways which prevent adverse environmental impacts due to leaching and migration of
heavy metals. For example, denser compaction and/or
installation of earthen covers are employed in Japan to
reduce the contaminant release. In evaluating the environmental impact due to the intake of contaminated
groundwater, it should be verified that the environmental impact are negligible or within acceptable level
by predicting the leaching and transport of contaminants of concern (COCs) in subsurface. Generally,
some regulatory batch leaching tests are performed to
characterize the leaching potential of the COCs. The
actual leaching concentration/amount and the relative
flux of COCs into the surrounding environment, which
depend on the given field condition, are influential to
the environmental impact (Inui et al. 2007). However,
these regulatory leaching tests do not provide enough
information to estimate these parameters. Thus, some
protocols to estimate the in situ leaching behavior with
a combined use of several leaching tests have been
developed these days (e.g. Kosson et al. 2002). However, when typical leaching tests are employed for rock
materials, sample preparation including sampling and
crushing is likely to influence the leaching amount of
trace elements (Inui et al. 2010). In addition, effects
of oxidation, caused by exposure to water and oxygen
after excavation, on the leaching behavior should be
considered since the acid rock drainage due to dissolution of sulfide minerals in rock is one of the main
mechanisms to promote the leaching of heavy metals.
This manuscript addresses the results of more than
27-month outdoor exposure tests for five rock samples
to simulate their in situ leaching behavior. The rock
samples contain certain concentrations of lead and
arsenic from natural resources, which are expected to
587
be released if they are exposed to weathering. In addition, leaching amounts/characteristics of trace metals
in outdoor exposure tests were assessed comparatively with those in various conventional laboratory
tests, which include the total trace metal content
test, conventional leaching tests, and accelerated oxidation tests, to discuss the factors affecting heavy
metal leaching from excavated rocks with natural
contamination.
2
2.1

Rock samples
Five different rocks materials were used. They were

collected at construction sites such as excavation and
tunneling, which were located in mountainous areas
far from industrial and urban areas in Japan. Thus, it
can be judged that heavy metals contained are derived
from natural resources. Five rock samples consist of
black shale, andesite, and three mudstones (MS-1, 2
and 3). They were all stored with their natural water
contents under a temperature of approximately 20 C
in sealed condition to prevent the exposure to oxygen and water. Table 1 shows their appearances and
the total contents of As and Pb as well as chemical
compositions determined by the X-ray fluorescence
analysis. According to the Japanese guideline (Ministry of Land, Infrastructure, Transport and Tourism,
Japan 2010), for excavated soils and rocks containing
more than 9 mg-As/kg or 23 mg-Pb/kg from natural
resources, their contamination potentials should be
assessed prior to reuse/landfill. All the samples used
in this study contain As, and only black shale contains
Pb, higher than these criteria.
Figure 1 shows pH values of rock samples for distilled water and H2 O2 solutions with 3% and 30%
concentrations. pH values for H2 O2 solutions were
widely employed to evaluate the possible pH change
of geomaterials when they are oxidized under weathering condition. Geomaterials with pH < 3.5 for 30%
H2 O2 solution are considered to have acidification
potentials (Japanese Geotechnical Society 2009). In
this study, 3% H2 O2 solution was also used to evaluate the acidification potential against weaker oxidation
Table 1.
effects. Black shale and andesite are fundamentally

acidic even for distilled water, and three mudstones
are alkaline rocks without any oxidation effects. However, pH value of MS-2 for 30% H2 O2 solution was
drastically reduced to 2.8, and pH values of MS-3 for
both 3% and 30% H2 O2 solutions were lower than
3. From these results, the acidification potentials of
mudstones are in the order of MS-3 > MS-2 > MS-1.
Hattori et al. (2003) concluded that the composition
ratio of S (sulfur) to Ca (calcium) in molecular concentration is a good index of the acidification potential,
and the rock with this ratio higher than 1 will generate
the acid drainage. In this study, the rock samples with
higher acidification potentials (black shale, andesite
and MS-1) have the S/Ca ratio higher than 1, according
to chemical compositions shown in Table 1.
2.2
Outdoor exposure test
Outdoor exposure tests have been conducted for five

rock samples since October 2009. This manuscript
presents the results obtained until February 2012
(about 27 months). Each sample was crushed into 9.5
to 37.5 mm in grain size (2.0 to 9.5 mm in diameter
for mudstone 1 due to its high crushability). 4.0 kg
Figure 1. pH of each rock sample for distilled water and

H2 O2 solutions.
Chemical properties of rock samples used in this study.
588
of each sample with its natural moisture content was

packed in a cylindrical container with a cross section
of 0.05 m2 with dry density of 2.0 Mg/m3 for black
shale, 1.1 Mg/m3 for andesite, 1.2 Mg/m3 for MS-1
and 1.6 Mg/m3 for MS-2 & MS-3. The specimens were
exposed to the rainfall and atmosphere throughout the
test (Figure 2). Rainfall intensity and percolation volume were continuously monitored. The leachate flowing through each specimen was periodically collected
for chemical analyses.
An
atomic
absorption
spectrophotometer
(Shimadzu, AA-6800) with a hydride generator and an
electro-thermal atomization system was used to determine the concentration of As, Pb and other solutes in
the leachate. pH, Eh, electrical conductivity (EC) , and
temperature were also monitored.
3
3.1

pH change of leachate by weathering
Figure 3 show pH profiles for five rock samples with

time. Cumulative percolation volumes after 27 month
outdoor exposure were varying from 16 to 19 in the liquid to solid ratio (mL/g-sample), due to the difference
in permeability of the specimens.
Leachate generated from the black shale and the
andesite was acidic. pH values ranged between 2.1 and
2.5, and 2.4 ant 4.0, respectively. For the andesite, pH
values were gradually decreased 10 months after. This
tread can be explained well by the result of the accelerated oxidation test using H2 O2 solutions (see Figure 1).
For the andesite, slight but obvious pH drops were
observed against the accelerated oxidation by both
3% and 30% H2 O2 solutions. Thus, pH values were
decreased after a certain period of outdoor weathering
due to gradual oxidation effect. For the MS-1, which
was proved to have the least acidification potential
among five samples by the accelerated test (Figure 1),
pH values were stabilized at pH7 to 10. However, for
the MS-2 and MS-3, which allowed the acidification
for H2 O2 solutions, the irregular pH drops occurred,
and acid drainage were generated in outdoor exposure
tests.
Comparing pH values with those monitored in laboratory accelerated tests, pH values in the batch test
using 30% H2 O2 solution was lowest, and pH for 3%
H2 O2 solution was a little lower than or similar to
those in the outdoor exposure test. This indicates that
30%H2 O2 solution is more influential than outdoor
exposure in more than two years in terms of acidification, and accelerated oxidation using 3% H2 O2
solution is almost comparable to outdoor exposure in
a few years.
From these observations, it can be concluded that
pH changes against 3% and 30% H2 O2 solutions could
classify acidification potentials reasonably under the
weathered condition, however the acceleration by 30%
H2 O2 solution possibly overestimate the acidification
progress in outdoor even for two years.
3.2 Leaching behaviors
Figure 2. View of outdoor exposure test.
Figure 3. Profiles of pH values for five rock samples in

outdoor exposure tests.
Figures 4 show profiles of percolation volume, EC,

Eh, temperature and concentrations of As and Pb of
the leachate sample with time.
For the black shale (Figure 4(a)), the leachate was
basically acidic as shown in Figure 2, and As concentrations were higher than 0.1 mg/L in the first three
months, then decreased with time and stabilized at
around 0.02 mg/L. Temperature rising during summer
(JulySeptember) were followed by slight decrease in
pH and Eh, and increase in EC and leaching concentrations. This is because the sample were subjected to
the wet and higher temperature condition in summer
and dissolution of minerals was promoted due to the
oxidation processes. Leaching of Pb was also detected
but with much lower concentrations than As.
In Figure 5, pH and Eh values of the leachate collected in both outdoor exposure tests and laboratory
batch leaching tests for black shale are plotted on
the pH-Eh/pe (electron activity) diagram of dominant
forms of As and Fe in As-Fe-S-H2 O system (Zhu &
Merkel 2001). In the outdoor exposure test, higher
leaching concentrations of As were observed when pH
or Eh values were relatively low except in a early testing stage, where dominant forms of iron and arsenic
are Fe2+ /Fe3+ and H3AsO3 (arsenous acid) respectively according to Zhu & Merkel (2000). When lower
leaching concentrations were observed (see the circle in Figure 5), the dominant form of As is expected
589
Figure 4. Profiles of leaching concentrations, EC, water temperature and percolation volume in outdoor exposure tests.
590
Figure 6. Comparison of As leaching amounts obtained in

27 months outdoor exposure test and laboratory leaching
tests.
Figure 5. pH and Eh values observed in outdoor exposure
tests and laboratory leaching tests combined with pH-Eh diagram of dominant forms of As and Fe in As-Fe- S-H2 O system
(Zhu & Merkel 2001).
to be HAsO4 , which is more easily absorbed to iron

compounds and less mobile than H3AsO3. It is considered that these changes in As forms affected by Eh
and pH values triggered the As leaching with higher
concentrations in wet and higher temperature period.
Testing results for the andesite (Figure 4(b)) had
similarities with those for the black shale. At the initial stage, EC values and As leaching concentrations
were relatively high and then gradually decreasing
and stabilized. However, 10 months after, with gradual lowering of pH values, they were getting higher
particularly in summer as observed in black shale.
As mentioned before, the andesite sample was gradually oxidized during the outdoor exposure test, and
leaching of As and Pb continued even after 1 years
passed.
A consistent leaching behavior was confirmed
among three mudstone samples (Figure 4(c) to (e)).
Although some irregular pH drops were observed as
mentioned in 3.1, alkaline leachate were basically generated in most of testing period. Leaching of As and
Pb associated with pH drops was not observed, and
As leaching was observed under a relatively strong
alkaline condition. This is because As in the mudstones were likely to be absorbed to iron/manganese
oxide/hydroxide minerals, which is easily released
if OH concentration in porewater becomes higher,
unlike the black shale from which As in sulfide
minerals was dominantly released due to oxidation.
3.3
Comparison of leaching amounts
Figures 5 shows As leaching amounts of the black

shale in the outdoor exposure test and several laboratory leaching/characterization tests (Inui et al. 2013).
Laboratory tests included a typical batch leaching test,

which is regulated by Japanese Ministry of the Environment for testing the soil quality, tank leaching test,
and column leaching test. In the batch and column
leaching tests, the rock sample were crushed until
its percentage passing at 2 mm and 4.75 mm became
100%, respectively. The sample subjected to accelerated oxidation, which was simulated by storing the
crushed sample in an incubator under 80% O2 and
100% humidity condition in 200 days, was prepared
for the batch leaching test. In addition, leaching concentrations during the aforementioned pH measurement using H2 O2 solutions were analyzed to evaluate
the effects of accelerated oxidation.
In Figure 6, leaching amounts of As from unit
weight of the rock sample are plotted with the cumulative volume of solvent or percolated water contacting
with the sample during the tests, which is represented
by the liquid to solid ratio (L/S).
In the outdoor exposure test, the leaching amount at
L/S = about 10 reached 0.84 mg/kg, which is slightly
larger than those in the batch leaching test as well as
the accelerated oxidation tests, which were conducted
with L/S = 10. Considering that EC values of leachate
of the black shale were highest among all the rock
samples, the chemical equilibrium was achieved in the
closed batch system and dissolution of As was likely to
be limited. Column leaching test gave 10 times larger
leaching amount than the outdoor exposure test, probably because a crushed sample (<4.75 mm in diameter)
was used, and the permeant was continuously renewed
in the column leaching test. Considering the effect
of acidification/oxidation, the accelerated oxidation in
an incubator provided a much larger leaching amount
than that in the outdoor exposure test, which indicates
that oxidation in one of the most influential factors and
its effects on leaching should be properly evaluated in
laboratory characterization tests.
Although detailed comparison of leaching amounts
for other four rock samples are presented in Inui et al.
591
(2013), the main findings for the comparison are that

the leaching amount of As obtained in the conventional batch leaching test can be a good index of the
in situ leaching amount until L/S = 10, particularly
for the rock samples with less acidification potentials
like mudstones. The sample with accelerated oxidation provided the leaching amount for the safe side in
the batch leaching test, probably because the sample
used in the batch test was subjected to both crushing
and oxidation.
CONCLUSIONS
This manuscript presents the testing results on long

term leaching behavior of heavy metals contained in
five rock samples due to their geologic histories. Outdoor exposure tests were conducted for more than two
years to assess factors affecting the in situ leaching
behavior under weathered condition. In addition, laboratory characterization tests were validated as a tool
to estimate the long term leaching characteristics of
trace metals by comparatively assessing the results
with those obtained in the outdoor exposure tests. The
important findings obtained can be summarized as
follows:
1) In outdoor exposure tests, oxidation gradually
occurred due to exposure to water and oxygen.
As observed for the acidic rocks (black shale and
andesite), leaching of As contained in sulfide minerals was caused by the oxidation. In addition, As
leashing with relatively higher concentrations were
observed in wet and higher temperature periods
since the dominant form of As was changed into
mobile HAsO
4 , affected by pH and Eh changes.
2) pH changes against 3% and 30% H2 O2 solutions
in the accelerated oxidation test could classify
acidification potentials well under the weathered
condition, but the acceleration by 30% H2 O2
solution possibly overestimate the acidification
progress in outdoor even for two years.
3) The leaching amount of As obtained in the conventional batch leaching test with the liquid to solid
ratio of 10 can be a good index of the field leaching amount at an equivalent infiltration volume
except the cases of rock samples with high leaching
concentrations of other chemicals, since the chemical equilibrium may limit the leaching of trace
metals in the batch test. The accelerated oxidation
tests can simulate the outdoor leaching amount for
the safe side in most rock samples.
REFERENCES
Geological Survey of Japan, AIST. 2004. Elemental Distribution in Japan Geological Map of Japan, National
Institute of Advanced Industrial Science and Technology.
Hattori, S., Ohta, T. and Kiya, H. 2003. Engineering geological study on exudation of acid water from rock mucks
Evaluation methods of rocks at the Hakkouda Tunnel near
mine area, Jour. Japan Soc. Eng. Geol., 43 (6), 359371,
in Japanese.
Inui, T., Kamon, M., and Katsumi, T. 2007. Suitability of
the solid waste utilization in geotechnical applications
from a viewpoint of environmental risk, Geoenvironmental Engineering, Geotechnical Special Publication No.
163, S.E. Burns, P.J. Culligan, J.C. Evans, P.J. Fox, K.R.
Reddy, N. Yesiller (eds.), ASCE, on CD.
Inui, T., Katsumi, T., Katayama, M. and Kamon, M. 2010.
Effects of friability and grain size on the leaching of
heavy metals in excavated rock materials, Environmental Geotechnics for Sustainable Development, M. Datta
et al. (eds.), Tata McGraw Hill, New Delhi, 730733.
Inui, T., Katsumi, T., Takai, A. and Kamon, M. 2013. Evaluating the long-term leaching characteristics of heavy metals
in excavated rocks. 18th International Congress on Soil
Mechanics and Geotechnical Engineering, accepted for
publication.
Japanese Geotechnical Society. 2009. Test Method for pH of
Suspend Soils, JGS 0211, Japanese Geotechnical Society,
Tokyo.
Kosson, D.S., van der Sloot, H.A., Sanchez, F., and
Garranrants,A.C. 2002.An integrated framework for evaluating leaching in waste management and utilization of
secondary materials. Environmental Engineering Science,
19 (3), 159204.
Ministry of Land, Infrastructure, Transport and Tourism,
Japan (MILT). 2010. Technical Manual on the Countermeasures against Soils and Rocks Containing NaturalDerived Heavy Metals in Construction Works (Draft), 10,
in Japanese.
Zhu, Y. and Merkel, B.J. 2001. The dissolution and solubility
of Scorodite, FeAsO4 2H2 O: Evaluation and simulation
with PHREEQC2, Wiss. Mitt. Inst. fur Geologie, TU
Bergakedemie Freiberg, Germany, 18, 112.
592

Novel techniques for simulating and monitoring impact of contaminants

on geoenvironment
K.R. Iyer, P. Pathak & D.N. Singh
IIT Bombay, Powai, Mumbai, India
ABSTRACT: This paper presents details of the techniques, which are indigenous, need based, economical
and developed by researchers at IIT Bombay, for investigating the impact of hazardous and toxic wastes on
geoenvironment. Such wastes, which inherently contain heavy metals, are being disposed of by various industries,
thermal/atomic power stations and research facilities. The heavy metals leach out of the waste matrix, in due
course of time due to their interaction with the geoenvironment, and contaminate it. Under these circumstances,
the techniques presented in this paper have been found to be quite helpful in simulating and monitoring the
spread and fate of contaminants in the geoenvironment. This paper also emphasizes the need of adopting these
techniques for geomaterial characterization when it comes in contact with contaminants.
INTRODUCTION
Increase in population over the years, rapid urbanization and incessant industrial growth have led to
huge utilization of various forms of energy (thermal,
nuclear) for various strategic, industrial and research
applications. This situation is also instrumental in
generation of large quantities of hazardous and toxic
wastes (viz., solid, liquid and gaseous), which cause
significant impact on the geoenvironment mainly due
to leaching of heavy metals present in them (IAEA,
2004; Naidu et al., 2010).
The interaction between the geomaterial and the
contaminant is a complex phenomenon and is influenced by various attributes of contaminants such as
its chemical activity, elevated temperature and the
presence of radionuclides, and the presence of particulate and multi-phase geomaterial. Hence, monitoring
the influence of disposed waste on the geomaterial
and on the geoenvironment becomes quite difficult
(Gurumoorthy & Singh, 2004a,b; Gurumoorthy &
Singh, 2005, Rakesh et al., 2009; Naidu et al., 2010;
Rao et al., 2009; 2013, in press). However, it becomes
quite important to develop suitable instrumentation
and methodologies for the laboratory/field simulation
of geomaterial-contaminant interaction and monitoring the spread of contaminant in the geoenvironment.
In this context, earlier researchers have proposed
various laboratory and field instrumentation techniques for characterization of geomaterial (viz., soil,
rock mass etc.) for safe execution of environmental and geotechnical engineering projects, which have
been adopted by various agencies (ASTM International, USA, Bureau of Indian Standards, British Standards, Eurocode, Environmental Protection Agency,
Federal Highway Administration, USA, etc.) as code
of practice for sampling and testing in laboratory
and insitu conditions. The mechanical properties such

as strength, compressibility and hydraulic conductivity can be studied for understanding the macrocharacteristics of geomaterial, by employing these
techniques. However, there is a lack of techniques
which can be employed for simulating the influence
of different energy fields (viz., mechanical, chemical,
thermal, electromagnetic and radiation) on the geomaterial, by studying various micro-level mechanisms
that govern the geomaterial-contaminant interaction
(viz., diffusion, sorption and desorption, coupled heat
and moisture migration, and heat and contaminant
transport). Also, even well-established manufacturers
and reputed equipment suppliers are reluctant to support development of specific instruments as per the
research needs due to inertia to adopt new concepts,
lack of interest, economics of the production, etc.
Under these circumstances, it becomes necessary
to develop alternate instrumentation techniques, on
case to case basis, which can be employed to understand/simulate the short- and long-term response
of the geomaterial (soil, rock, ash deposits, etc.)
when they interact with contaminants. In this context,
when various agencies approached the researchers at
IIT Bombay, India, to obtain solutions for safe disposal, utilization and proper containment of the waste,
several instrumentation techniques were developed.
Details of such instrumentation, which are primarily,
need based, indigenous and economical are discussed
in the following.
2
2.1
DEVELOPMENT OF NOVEL TECHNIQUES

Mineralogical alteration
Due to geomaterial-waste interaction, there are

chances that the geomaterial may get mineralogically
593
Figure 1. Setup for contaminant transport using centrifuge

tests.
Plate 1. The zeolitization unit.
altered over a period of time. A good example is interaction of geomaterials (viz., flyash containing SiO2 )
with alkali (NaOH, KOH) at elevated temperatures,
wherein their physical, chemical and mineralogical
characteristics get altered, resulting in the formation of
a value added product like zeolite. Such an interaction
occurs naturally in lagoons due to wet disposal (i.e.,
slurry disposal) of fly ash from thermal power stations
into lagoons over a period of time (Kolay & Singh,
2002; Jha & Singh, 2011), which have several industrial applications (Kolay & Singh, 2001). In order to
understand the formation of zeolites from flyash and
assess the quality of product formed, a zeolitization
unit (refer Plate 1) was developed. The unit utilizes
a water bath and reflux system, for achieving alkali
activation of ash by hydrothermal process, to form
different types of zeolites. This unit also helps in estimating zeolitization potentialof geomaterials, which
is quite useful for deciding upon its appropriate usage
and engineering applications (Kolay & Singh, 2000;
Jha & Singh, 2012).
2.2 Accelerated physical modelling

Accelerated physical modeling (or Centrifuge modeling) has been demonstrated to be quite useful for
accelerating the contaminant transport mechanism
and hence can be an effective alternative to field
scale studies for modelling geoenvironmental engineering problems (Kumar & Singh, 2012). However,
the centrifuge modeling for simulating the diffusion
phenomena has not received much attention. In this
context, the modelling of the diffusion of ions (read
contaminants) through soils using a geotechnical centrifuge has been conducted (refer schematic diagram
in Figure 1), which helps to explain the contaminant
Plate 2. The (a) laboratory and (b) field thermal probes,

and (c) thermal property detector.
transport mechanism through soils (Kumar & Singh,

2012).
The experimental results were validated with results
obtained from the finite-element-based softwares
SEEP/W and CTRAN/W (GeoStudio 2007), and it has
been demonstrated that centrifuge modelling is quite
useful for simulation of the diffusion in soils (Sreedeep
et al., 2003).
2.3 Thermal probes

With an objective to simulate the response of geomaterial to thermal flux (heat induced due to the
contaminants) and to study the mechanism(s) of coupled flow (heat induced moisture and contaminant
migration), thermal probes (refer Plate 2) were fabricated (Singh et al., 2003; Hanumantha & Singh, 2010).
The thermal response (viz., thermal resistivity, thermal
594
Plate 3. Setup for heat migration using centrifuge tests.
diffusivity and effusivity, and specific heat) of different geomaterial, have been established for different
dry density and moisture content, by employing these
probes (Dalinaidu & Singh, 2004, Krishnaiah & Singh,
2004a; Padmakumar et al., 2012).
Krishnaiah et al. (2004) and Das et al. (2008) have
determined the porosity of rock samples and concrete, respectively by the application of thermal flux.
Krishnaiah & Singh (2004b, 2006) have investigated
the influence of accelerated gravity on thermal properties of the geomaterial, in a geotechnical centrifuge,
by employing the instrumentation depicted in Plate 3.
Further, the migration of heat in soils has also been
studied and the variation of temperature with depth
over a period of time could be monitored by employing
a thermal probe installed in centrifuge model, and the
migration of heat in field could be estimated by applying scale factor for time, 1/N2 (Krishnaiah & Singh,
2004b).
2.4
Soil contaminant detector
In order to determine the presence and concentration

of contaminants in geomaterial, the soil contaminant
detector (SCD) setup, was developed and is depicted in
Plate 4. The presence and movement of contaminant(s)
in the soil mass (depicted in Plate 5) can be detected
by sending a low frequency sinusoidal signal into it
and recording the differential output voltage, OV, in
mV, across each pair of the electrodes. Hence, concentration of contaminant at different locations (wherever
the electrodes are placed diametrically opposite to each
other), can be detected along the length of the sample
(Kumar & Singh, 2005).
2.5
Plate 4. The soil contaminant detector.
Diffusion cells
One of the major challenges faced by planners and

engineers is the safe handling and disposal of lowlevel and high-level radioactive wastes in geological
formations (IAEA, 2004). The pores in these formations create a path for the ground water movement
and thereby enhance the probability of dissolution and
migration of radionuclides from the radioactive waste
matrix in the geoenvironment.
Plate 5. The soil columns with contaminants placed

from top.
In order to understand the mechanism(s) of contaminant transport (diffusion and/or advection) in the
porous media, Gurumoorthy & Singh (2004a, b, 2005)
and Witthser et al. (2006) have conducted experiments using different ions in their active as well
as inactive forms. Diffusion cells, depicted in Figures 2 and 3 have been developed for monitoring
and establishing the response of intact and fractured
rock samples, respectively. Rao & Singh (2008) have
developed a diffusion cell (refer Fig. 4) that facilitates
accelerated diffusion through the intact rock mass.
A diffusion cell was fabricated (Kumar & Singh
2004, 2005; Sreedeep & Singh, 2008), as depicted
in Figure 5 for monitoring the diffusion characteristics of the fine-grained soil, which can be quantified
by determination of the effective diffusion coefficient
(Sreedeep & Singh, 2008; Rakesh et al. 2009). In the
case of the diffusion cell, there will be diffusion of
salts (viz., NaCl or SrCl2) from contaminated soil
(CS) to uncontaminated soil (US) with time, due to
595
Figure 5. The diffusion cell for soil.
Figure 2. Diffusion cell for intact rock.
Figure 6. Details of the lysimeter.
the soil can be related (from electrodes inserted in

the soil sample) to the diffusion characteristics of a
soil-contaminant system.
Figure 3. Diffusion cell for fractured rock.
2.6 The lysimeter
Figure 4. Diffusion cell for accelerated test.
the concentration gradient. The basic principle of this

methodology is that the electrical impedance of the soil
is inversely proportional to the concentration of contaminant. Hence, the electrical impedance response of
Laboratory column studies (under normal and accelerated gravity environments) and numerical modeling
have been employed by previous researchers to study
the contaminant transport (Rao et al., 2013, in press)
through them. However, these methods suffer from
various limitations, particularly, due to difficult in
reproducing the in-situ soil conditions and boundary
conditions, in a very small model. Under these circumstances, an in-situ lysimeter (Figure 6) has been found
to be quite useful for conducting the investigations to
obtain hydraulic conductivity of fully or partially saturated contaminated/uncontaminated soils (Rao et al.,
2009; 2013, in press). These lysimeters can easily be
installed in the soil mass, without disturbing its state,
and are large enough to be representative of insitu soil
conditions. Details of the lysimeter setup are depicted
in Figure 6.
596
Table 1.
Details of the instrumentation.
REFERENCES
Symbol
Description
A
GWT
R
Access tube
Moisture detection probe
Ground water table
Spiked soil
To simulate contaminated
soil with Tritium
Suction
(30, 40)
sampler
(10, 30)
(r,z)
(10, 20)
(30, 10)
Mild steel ring For creating a control volume
of approximately 1 m3
TDR probe
For determining in-situ moisture
content of the control volume
Undisturbed
Control volume of the soil
soil
Bhat, A.M., Rao, B.H. & Singh, D.N. 2007. A Generalized Relationship for Estimating Dielectric Constant of
Soils, Journal of ASTM International Published Online:
15 August 2007, DOI: 10.1520/JAI100635.
Dalinaidu, A. & Singh, D.N. 2004. A Generalized Procedure for Determining Thermal Resistivity of Soils.
International Journal of Thermal Sciences 43(1): 4351.
Das, B.B., Singh, D.N. & Pandey, S.P. 2008. Some
Investigations for Establishing Suitability of Watsons
Strength-porosity Model for Concrete, Journal of ASTM
International, Published online: 02 January 2008, DOI:
10.1520/JAI101345.
Gurumoorthy C. & Singh, D.N. 2004a. Diffusion of Iodide,
Cesium and Strontium in Charnokite Rock Mass. Journal of Radioanalytical and Nuclear Chemistry 262(3):
639644.
Gurumoorthy C. & Singh, D.N. 2004b. Experimental
Methodology to Assess Contaminant Diffusion in Rock
Mass. Journal of Environmental Monitoring and Assessment 91: 277291.
Gurumoorthy, C. & Singh, D.N. 2005. Centrifuge Modeling
of Diffusion Through the Rock Mass. Journal of Testing
and Evaluation, ASTM 31(1): 562568.
Hanumantha, Rao,B. & Singh, D.N. 2010. Application of
Thermal Flux for Establishing Soil-Water-Characteristic
Curve of Kaolin. Geomechanics and Geoengineering: An
International Journal 5(4): 259266.
IAEA. 2004. Safety assessment methodologies for near surface disposal facilities, results of a co-ordinated research
program, Review and Enhancement of Safety Assessments
Approaches andTools,Vol. 1, InternationalAtomic Energy
Agency, Vienna.
Jha, B. & Singh, D.N. 2011. A Review on Synthesis, Characterization and Industrial Applications of Flyash Zeolites.
Journal of Materials Education 33(12): 65132.
Jha, B. & Singh, D.N. 2012. Zeolitization Characteristics
of a Flyash from Wet-and Dry- Disposal Systems. Acta
Geotechnica Slovenica 2012/2: 6371.
Kolay, P.K. & Singh, D.N. 2000. Effect of Zeolitization on
Compaction, Consolidation and Permeation Characteristics of a Lagoon Ash, Journal of Testing and Evaluation,
ASTM 28(6): 425430.
Kolay, P.K. & Singh, D.N. 2001. Effect of zeolitization on
physicochemico-mineralogical and geotechnical properties of lagoon ash. Canadian Geotechnical Journal 38(5):
11051112.
Kolay, P.K. & Singh, D.N. 2002. Characterization of an Alkali
Activated LagoonAsh and itsApplication for Heavy Metal
Retention. FUEL 81(4): 483489.
Krishnaiah, S. & Singh, D.N. 2004a. A device for Determination of Thermal Properties of Soil. Journal of Testing
and Evaluation, ASTM 32(2): 114119.
Krishnaiah, S. & Singh, D.N. 2004b. Centrifuge Modelling of
Heat migration in soils. International Journal of Physical
Modeling in Geotechnics 4(3): 3947.
Krishnaiah, S., Singh, D.N. & Jadhav, G.N. 2004. A Methodology for Determining Thermal Properties of Rocks. Int.
Journal of Rock Mechanics and Mining Sciences 41:
877882.
Krishnaiah, S. & Singh, D.N. 2006. Determination of
Thermal Properties of Soils in a Geotechnical Centrifuge. Journal of Testing and Evaluation, ASTM 34(4):
319326.
Kumar, P, R. & Singh, D.N. 2004. Instrumentation and Testing Methodology for Detecting Cl- Contaminants in Soils.
Journal of Testing and Evaluation, ASTM 32(2): 8187.
S1
S2
S3
S4
MS
T
U
Purpose
2.7 Application of impedance analysis

Impedance analysis has been employed for measuring the complex electrical properties (viz. permittivity,
conductivity and impedance) of various geomaterial
(Bhat et al., 2007; Rao et al., 2007; Sreedeep & Singh,
2008) and has been employed to study contaminant
transport mechanisms and characterization of unsaturated soils. Efforts are on to develop a methodology
which will facilitate easy and quick determination
of distribution coefficient of soil-contaminant system (Pathak et al., 2013a;b, unpubl.) by employing
an Impedance analyzer.
CONCLUDING COMMENTS
Various instrumentation techniques and methodologies have been developed by researchers at IIT
Bombay for simulating and monitoring impact of
contaminants on geoenvironment. These techniques
facilitate simulation of geomaterial-contaminant interaction and contaminant transport, which are important
for obtaining suitable solutions to storage and safe disposal of hazardous and toxic wastes in geomaterial,
and thus minimizing their effect on the geoenvironment. Although, most of these techniques have been
observed to be functioning quite satisfactorily in the
laboratory scale simulation, extensive investigations
must be carried out, on different types of geomaterial, under in-situ conditions, to demonstrate their
generality and suitability.
ACKNOWLEDGEMENT
Authors are thankful to the reviewers and editorial
board members of various Journals and funding agencies, and examiners of theses and dissertations, for
shaping up their research and development work.
Critical comments from readers are welcome to help
us in shaping up future research.
597
Kumar, P.R. & Singh, D.N. 2005. A Novel Technique

for Monitoring Contaminant Transport Through Soils.
Environmental Monitoring and Assessment 109(13):
147160.
Kumar, P.R. & Singh, D.N. 2012. Geotechnical Centrifuge
Modeling of Chloride Diffusion through Soils. International Journal of Geomechanics, ASCE 12(3): 327332.
Naidu, A.D., Rao, B. H., Shanthakumar, S. & Singh D.N.
(2010) Determination of Distribution Coefficient of Geomaterials and Immobilizing Agents, Canadian Geotechnical Journal 47: 11391148.
Padmakumar, G.P., Srinivas, K., Uday, K.V., Iyer, K.R.,
Pathak, P., Keshava, S.M. & Singh, D.N. 2012. Characterization of Aeolian Sands from Indian Desert. Engineering
Geology 139/140: 3849.
Pathak, P., Singh, D.N., Pandit, G.G. & Rakesh, R.R. 2013a.
Determination of Distribution Coefficient: A Critical
Review. International Journal of Environment and Waste
Management, Revised and Communicated.
Pathak, P., Singh, D.N., Pandit, G.G. & Apte, P.R. 2013b.
Establishing Sensitivity of Distribution Coefficient on
Various Attributes of Soil-contaminant System. Journal
of Hazardous, Toxic and Radioactive Waste, Revised and
Communicated.
Rakesh, R.R., Singh, D.N. & Nair, R.N. 2009. A Methodology for Simulating Radionuclide Diffusion in Unsaturated
Soils. Geotechnical and Geological Engineering 27(1):
1321.
Rao, B.H., Bhat, A.M. & Singh, D.N. 2007. Application of
Impedance Spectroscopy for Modeling Flow of AC in
Soils. Geomechanics and Geoengineering: An International Journal 2(3): 197206.
Rao, B. H. & Singh, D.N. 2008. Determination of Diffusion Characteristics of Intact Rock Mass: A Critical
Evaluation. GeotechnicalTesting Journal 31(6): 490502.
Rao, B.H., Sridhar V., Rakesh R.R., Singh, D.N., Narayan,
P.K. & Wattal P.K. 2009. Application of In-situ Lysimetric studies for determining Soil Hydraulic Conductivity,
Geotechnical and Geological Engineering 27: 595606.
Rao, B.H., Sridhar, V., Rakesh, R.R., Singh, D.N., Narayan,
P.K. & Wattal, P.K. 2013. Modeling Radioactive Contaminant Transport in Soils, International Journal of
Environment and Waste Management, In Press.
Singh, D.N., Devid, K. & Naidu, A.D. 2003. Fabrication of
Thermal Probes for Estimation of Soil Thermal Resistivity, Journal of Testing and Evaluation, ASTM 31(1):
6572.
Sreedeep S., Berton, C., Moronnoz, T. & Singh, D. N.
2003. Centrifuge and Numerical Modeling of Contaminant Transport Through the Unsaturated Silty Soil, ISSMGE International Conference on From Experimental
Evidence towards Numerical Modelling of Unsaturated
Soils, September 18/19, Bauhaus-Universitt Weimar.
Sreedeep, S. & Singh, D.N. 2008. A Novel Technique for
Studying Diffusion of Contaminants in Fine-Grained
Soils, Geomechanics and Geoengineering: An International Journal 3(3): 199209.
Witthser, K., Dalinaidu, A. & Singh, D.N. 2006. Investigations on Diffusion Characteristics of Granite and Chalk
Rock Mass. Geotechnical and Geological Engineering
24(2): 325334.
598

Studying the effects of contamination on the geotechnical properties

of clayey soil
M.O. Karkush
Civil Engineering, Baghdad University, Iraq
A.T. Zaboon
Construction and Building Department, University of Technology, Iraq
H.M. Hussien
Civil Engineering, Baghdad University, Iraq
ABSTRACT: The present study describes the geotechnical behavior of synthetically contaminated soil. Physical, chemical, and mechanical properties were compared with the geotechnical properties of intact soil. The soil
samples, disturbed and undisturbed, were obtained from Al-Khadymiya district, which is located at the north
west of Baghdad city in Iraq. Four different types of contaminants were used: kerosene, ammonium hydroxide,
lead nitrate and copper sulphate. The natural soil samples were contaminated synthetically by soaking in isolated
pans containing a solute of water and contaminant for a period of 30 days. The contaminants were mixed with
distilled water in two percentages 10 and 25% of the dry weight of the clay soil sample. The results showed
that these contaminants have significant effects on the geotechnical and chemical properties of the soil. The
contaminants causing an increase in Atterbergs limits (except samples contaminated with lead nitrate), maximum dry unit weigh (d,max ), initial void ratio (eo ) (except samples contaminated with lead nitrate), compression
index (Cc ), swelling index (Cr ), and collapse potential (CP ). Also, the contaminants caused a decrease in specific
gravity (Gs ) and optimum moisture content (opt ) (except samples contaminated with ammonium hydroxide),
coefficient of vertical consolidation (Cv ), and cohesion between soil particles (c).
INTRODUCTION
The problem of soil contamination is recently escalating due to increasing the industrial wastes such as
petroleum hydrocarbon, organic solvents, and heavy
metals as well as the intensive use of agricultural fertilizers. The occurrence of contaminants in soil above
a certain level cause a deterioration or change of some
geotechnical properties of soil. Soil contamination can
be considered as the presence of man-made chemicals or other alteration in the intact soil environment.
Soil contamination typically arises from the rupture
of underground storage tanks and pipes, application
of fertilizers and pesticides, percolation of contaminated surface water to subsurface strata, leaching of
wastes from sanitary landfills or direct disposing of
industrial wastes to the soil.
Iraq has several thousand of contaminated sites
resulting from a combination of general industrial
activities, military activities and post-conflict damage
and looting (UNEP Report 2005). The purpose of this
work is to measure the impact of different types of contaminants on the chemical and geotechnical properties
of soil.
External contaminants penetrating a soil mass
through wet or dry precipitation which behave
differently with regard to each soil type according

to the absorption properties, texture, density, humidity, and other factors. This is due to the fact that
these properties are not homogeneously developed in a
certain soil and change largely with stratified layers
of soil.
Abdul Rasool (1999) studied the effects of oil contamination on the geotechnical properties of disturbed
soil samples and concluded that oil contamination
significantly affects these properties.
Al-Layla & Al-Kiki (2006) studied the effects of
four types of liquid industrial wastes on the geotechnical properties of collapsible clay soil (swelling soil)
and concluded that these contaminants lead to an
increase in the swelling pressure and a decrease in the
compressibility of soil.
Rahman et al. (2010) studied the effects of oil
contamination on the geotechnical properties of soil.
The soil samples were artificially contaminated with
oil percents 4, 8, 12 and 16% of the dry weight of
intact soil. The oil contamination causes significant
decrease about 1939% in plastic and liquid limits.
The undrained shear strength, cu was 126 kPa for natural soil had dropped to very close values of 35 kPa at
oil content of 16%. Also, a reduction in permeability
was observed as a result of the oil contamination.
599
Zulfahmi et al. (2012) studied the effects of amang

(tin tailing) which is commonly found in the vicinity
of disused mining area on the water quality, landscape
and mechanical properties of clay soil. The contaminated soil samples were prepared by adding 5, 10
and 20% of tailing, based on dry weight of the soil.
The results showed that Atterbergs limits (liquid &
plastic limits) were decreased and the maximum dry
density increased while optimum moisture content
decreased with the increase in the percentage of tailing added to the soil samples. Also, the permeability of
contaminated soil increased with the increase in tailing contents ranging from 19.8 cm/hr to 23.8 cm/hr,
while the undrained shear strength, Cu, decreased from
646 kPa (5% of tailing) to 312 kPa (20% of tailing).
The effects of hydrochloric, phosphoric and sulphuric acids on the compaction characteristics of black
cotton soil at different percentages had been studied by
Umesha et al. (2012). Atterbergs limits, compaction
characteristics and unconfined compressive strength
tests were carried out. The results showed that acid
contamination mixed with soil have adverse effects on
the geotechnical properties of soil.
2
2.1
EXPERIMENTAL WORK
Soil sampling and classification
The soil used in this study was obtained from

Al-Khadymiya district in Baghdad city, Iraq. The disturbed soil sample was obtained from open sloped
excavation to a depth of 2.25 m from the natural ground
surface which is located below the ground water table.
The disturbed soil samples were put in plastic bags,
while the undisturbed soil samples were covered with
tighten wax and then transported to the laboratory with
Table 1.
Designation of tested soil samples.
Soil sample
symbol
S0
S1 & S2
Natural soil sample

Soil samples contaminated with 10% &
25% of kerosene respectively
25% of ammonium hydroxide respectively
25% of copper sulphate respectively
25% of lead nitrate respectively
S3 & S4
S5 & S6
S7 & S8
Table 2.
Description
index label. The natural soil was classified visually as

silty clay of high plasticity (CH) according to unified
classification soil system (USCS).
2.2 Materials used
After soil sampling and visual classification, both disturbed and undisturbed soil samples, were air dried
and divided into nine parts. Eight of these soil samples were soaked with four types of chemical solutions
which were used as soil contaminants. Each contaminant was added in two percentages (10 & 25)% by
weight of the dry soil sample. The designation of soil
samples tested in the present study are given in Table 1.
Four different types of contaminants representing three groups of contaminants were used in the
present study: (1) kerosene was used as an example of petroleum hydrocarbon wastes, (2) ammonium
hydroxide was used as an example of agricultural
fertilizer, and (3) copper (II) sulphate and Lead(II)
nitrate were used as examples of heavy metals.
The difficulty of the presence of lead and copper as pure elements and the low solubility of these
elements, lead to use the most common compounds
of these elements, such as copper sulfate and lead
nitrate. The chemical composition and properties of
the contaminants used in the present study are given
in Table 2.
2.3 Soaking process
Both disturbed and undisturbed soil samples, were
soaked in eight individual plastic covered containers
for 30 days. The soil samples were covered with chemical solution which consist of the contaminant and
distilled water. Each contaminant was added with two
percentages (10 & 25)% of the dry weight of soil sample. The solution of contaminant and distilled water
covered the soil sample. The distilled water was added
to help in the spreading and mixing of contaminant
into the soil which cant happen without the presence
of water.
3
3.1 Chemical analysis

Chemical tests were conducted on both natural synthetically contaminated soil samples to determine the
chemical composition of soil samples. The results of
chemical tests are given in Table 3.
Chemical composition and properties of used contaminants.
Contaminant
Chemical symbol
Density kg/m3
Molar mass g/mol
Solubility in water**
Kerosene
Ammonium hydroxide
Copper(II) sulfate
Lead(II) nitrate
CxHy*
NH4 OH
CuSO4 5H2 O
Pb(NO3 )2
780
910
3603
4530
35.04
249.68
331.2
Immiscible
Miscible
320 g/L
52 g/100 mL
*x and y ranged (46) for industrial kerosene, and **at 20 C
600
3.2 X-ray diffraction tests

X-ray diffraction tests were conducted on both natural and contaminated soil samples. The results showed
that the clay minerals consist of montmorillonite, palygorskite, mica and kaolinite, while the nonclay minerals consist of quartz, feldspar, calcite and dolomite.
The d-spacing of minerals is the distance between different planes of atoms in the crystal lattice. Mostly
the values of d-spacing remain constant and do not
affected by the type of contaminants, but there were
noticeable variation in the values of d-spacing for
montmorillonite which ranged (13.5914.73) and
mica which ranged (9.910.16) .
Kerosene caused the removal of organic matter from
interlayer of montmorillonite. There are no significant
changes in the peak intensity of calcite in the sheet
of both natural and contaminated soil samples with
kerosene. This is probably due to three possibilities:
(1) the low acidity effect of kerosene on calcite, (2)
the calcite is highly crystalline which resists the effects
of kerosene, and (3) the highly buffering capacity of
studied soil (calcareous soil) that resists any change in
soil pH.
In soil samples S3 and S4 , the results of x-ray
diffractions showed a decrease in the diffraction line
of 14.03 , which reflects the collapse in montmorillonite layers and indicate that the montmorillonite in
studied soil is a highly charged layer, so the collapsing
of montmorillonite layer increases with an increase of
NH4 concentration in the soil. Samples S5 , S6 , S7 and
S8 have shown no effect for heavy metals on the clay
mineral composition of these samples which reflect
the inactivity of chemical reactions between clay minerals and heavy metals. These reactions need special
conditions of high temperature and pressure.
3.3
3.4 Atterbergs limits

The results of Atterbergs limits tests, liquid and plastic
limits, and plasticity index (PI) are given in Table 5.
Particle-size distribution
The particle-size distribution of soil samples were

tested according to unified soil classification system,
USCS. Soil samples classification results are given in
Table 4.
The grains of tested soil samples become finer
after contamination except in samples S5 , S6 , S7 , and
S8 which built up coarser particles. There are two
main cementing agents between soil particles, organic
Table 3.
matter and calcium carbonate (CaCO3 ). These agents

have the ability to increase both the sticking between
soil particles and stability of soil aggregate.
Kerosene increases the percentage of fine particles (size < 0.002 mm) due to the ability of kerosene
to dissolve the organic matter between soil particles (kerosene is organic-solvent). This phenomenon
influences the stability of soil aggregate and leads to
separate clay particles from soil aggregates (Balsdent
et al. 1987).
In ammonium contamination, many studies have
been carried out (Kadduo 1960, Essa 1999,
Muhammad 2007). It was found that most of montmorillonite in Iraqi soils are inherited from mica, highly
layer charged and able to fix (K+1 ) between interlayer of mineral. The soil contaminated with (NH4 )
has a higher amount of montmorillonite, (NH4 ) ion
will enter vacant spaces between interlayer (K-vacant
spaces) and cause collapsing of the mineral layers. The
process will increase the separating of clay particles
from soil aggregates.
In soil samples contaminated with heavy metals,
the electronegative is an important factor in determining which of the trace metals chemisorbs on minerals
with the highest preference. The more electronegative
metals form the strongest covalent bonds with (O)
atoms on any particular mineral surface.
The divalent metal (Cu) has a preference to
adsorbed on clay surfaces more than (Pb) and create a cation bridge between clay particles, leading to
increase the stability of aggregates. The results showed
that Cu and Pb increase the percentages of grains
coarser than 0.002 mm in the soil sample.
Table 4.
Particle-size distribution of soil samples.
Soil sample S0
Silt (%)
Clay (%)
USCS
S1
S2
S3
S4
S5
S6
S7
S8
48 40 36 44 42 52 60 50 56
52 60 64 56 58 48 40 50 44
CH CH CH CH CH CH CH CL CL
Results of the chemical tests.
Soil sample
SO3 %
Cl1 %
CaCO3 %
OM %
pH
TSS %
Gypsum %
CEC*
S0
S1
S2
S3
S4
S5
S6
S7
S8
0.14
0.07
0.08
0.09
0.03
0.73
3.21
0.06
0.05
0.11
0.09
0.09
0.10
0.09
0.09
0.10
0.11
0.09
33.8
35.5
35.5
36.00
35.20
32.00
28.80
32.40
37.10
0.28
0.37
2.20
0.20
0.19
0.32
0.43
0.20
0.39
8.39
8.65
8.30
8.91
8.97
8.32
8.56
8.68
8.51
0.45
0.21
0.26
0.30
0.10
1.75
7.15
0.26
0.20
0.29
0.11
0.13
0.15
0.05
1.32
6.21
0.10
0.09
35.81
37.68
36.84
32.37
30.92
32.97
36.53
31.46
32.85
*CEC is the cation exchange capacity of minerals in meq/100 g at pH of 7.
601
According to the obtained results, the liquid limit

and plastic limit were increased in all samples with
the increase of percentage of contaminant added to
the soil sample, while the plasticity index remains
approximately constant except, S1 and S2 .
The synthetic contamination of soil causes variation
in liquid and plastic limits, which may resulting due to
the variation in the capacity of exchangeable cation
and chemical reactions between the chemical composition of contaminant and mineral composition of
soil. These actions cause inherent variations of structure and composition in the lattice of the clay mineral
itself.
Kerosene causes removing of the organic matter
form interlayer of montmorillonite and creating voids.
This may cause an increase in LL and PL, while in
the case of ammonium contamination, the collapse in
montmorillonite layers may cause an increase in LL
and PL while the plasticity index remained constant.
In the case of heavy metals, there were variations in
the adsorbed cation composition which causes slight
variation in Atterbergs limits.
3.5 Specific gravity and unit weight of soil

The results of specific gravity (Gs ), the in situ bulk
unit weight and moisture content tests are given in
Table 6. Also, the maximum dry unit weight and optimum moisture content of soil samples were tested
according to (BS:1377: 1975, test 13). The results of
aforementioned tests are given in Table 6.
It can be notice that the specific gravity decreases
with increasing kerosene and ammonium hydroxide
contents. This is due to the low densities for these
chemical compounds. The specific gravity increases
with increasing copper sulphate content which may
be due to the high density and solubility of this contaminant. The lead nitrate causes a decreasing of the
specific gravity due to the solubility and chemical
reactions between lead nitrate and clay minerals.
The results of compaction tests indicated that optimum moisture content was increased in soil samples
which causes a decrease in the maximum dry unit
weight. This decrease may be due to the flocculation
of soil particles, as well as the increasing in organic
materials.
Table 5. Atterbergs limits of soil samples.
3.6 Consolidation test

Soil sample
S0
S1
S2
S3
S4
S5
S6
S7
S8
LL (%)
PL (%)
PI (%)
53
23
30
65
26
39
67
26
41
56
26
30
57
27
30
53
24
29
54
25
29
49
22
27
47
22
25
The results obtained from one dimensional consolidation tests (oedometer) are given in Table 7.
The initial void ratio (except soil samples S7 and
S8), compression index, and swelling index increased
with an increasing percentage of contaminant content, while the coefficient of vertical consolidation
decreased with increasing the contaminant content.
In the case of kerosene contamination, the kerosene
causes removing of organic matter and decreasing the
d-spacing which leads to an increase in the compression index and the void ratio. In case of ammonium
contamination, the collapse in montmorillonite interlayers and highly layer charge may be the reason of
increasing the compression index and void ratio.
In case of copper and lead contamination, the
increase in (Cc) may be due to the basis charge of
the particles and the nature of the fluid which affects
the adsorbed cations, or the increase may be due to the
dissolved salts (Grim 1968).
The clay mineral composition affects the recompression behavior of soil specimen. In the case of
Table 6. Results of specific gravity, unit weight and

moisture content of soil samples.
Soil
sample
Gs
d
kN/m3
t
kN/m3
opt
%
d,max
kN/m3
S0
S1
S2
S3
S4
S5
S6
S7
S8
2.74
2.64
2.63
2.72
2.67
2.75
2.77
2.69
2.63
14.88
14.00
13.85
14.73
13.82
13.42
12.82
14.72
14.42
32.57
34.65
34.93
31.32
35.38
37.49
43.51
31.84
33.05
19.72
18.87
18.69
19.35
18.71
18.45
18.40
19.42
19.13
17.57
16.80
15.87
17.89
18.80
17.42
15.93
16.38
15.63
17.30
17.38
17.43
17.45
17.62
17.77
17.82
17.92
18.23
Table 7.
Compressibility and permeability properties of soil samples.
Soil sample
eo
Cc
Cr
Pc kPa
mv m2 /kN
Cv mm2 /min
k 106 m/sec
S0
S1
S2
S3
S4
S5
S6
S7
S8
0.805
0.849
0.859
0.812
0.896
1.010
1.118
0.792
0.791
0.104
0.215
0.241
0.172
0.176
0.196
0.295
0.234
0.215
0.001
0.050
0.055
0.039
0.058
0.064
0.063
0.070
0.070
100
90
90
150
150
150
90
100
75
0.302
1.435
0.578
0.377
0.361
0.426
0.725
0.529
0.454
44.70
10.88
6.03
22.43
11.04
8.98
8.17
15.60
11.44
2.21
2.55
1.57
1.38
1.65
1.63
1.97
1.35
1.85
602
montmorillonite clay soil, the individual particles are

brought substantially in contact by almost complete
relative permanent removal of interlayer water. Also, if
the recompression caused only the partial dehydration
of the montmorillonite particles, very large volume
changes will be expected if the load is reduced and the
clay is again wetted.
In the kerosene contaminated soil samples, the
increase in the swelling index may be attributed to
Table 8.
Results of collapse tests.

Collapse potential, Cp %
Soil
sample
Single
collapse
Description
Double
collapse
Description
S0
S1
S2
S3
S4
S5
S6
S7
S8
0.01
0.0
0.11
0.0
0.24
0.20
0.55
0.02
0.41
NP1
NP
NP
NP
NP
NP
NP
NP
NP
1.3
2.8
3.2
3.7
3.8
1.4
7.4
1.6
2.7
MT2
MT
MT
MT
MT
MT
T3
MT
MT
1: no problem, 2: moderate trouble, & 3: trouble
the entrance of the water between interlayers of montmorillonite (after dissolving organic matter), while in
ammonium contamination, the increase in swelling
index may be due to the expansion of montmorillonite
after collapse of mineral layers and entering the (NH4 )
ion in vacant site.
3.7 Collapsibility tests
A set of laboratory tests were conducted on both natural and contaminated soil samples to determine the
magnitude of one-dimensional collapse potential that
occurs when unsaturated soils are inundated with fluid
using single and double oedometer tests (Jennings &
Knight 1975). The collapse potential is defined as the
ratio of the change in the specimen void ratio upon
inundation to its initial void ratio before loading. The
results of collapse potential tests and classification of
soil samples are given in Table 8 and Figure 1.
The results of collapse potential obtained from double oedometer tests is completely differ than that
obtained from single oedometer tests. The results of
single collapse tests showed that 10% of kerosene
or ammonium hydroxide does not affect the collapse
potential, while 25% of kerosene or ammonium causes
an increase of the collapse potential due to the increase
in volume of voids.
Figure 1. Results of double collapse tests.
603
The soil samples that are contaminated with heavy

metals have shown two different behaviors. Copper
causes a decrease in the void ratio and lead causes an
increase in the void ratio, as compared with natural soil
at the same stress level. Also, it is important to note
Table 9.
3.8 Strength tests

The shear strength depends on the clay and nonclay
mineral composition, grain-size distribution, shape of
the grains, and the arrangement of the grains with
respect to each other (Grim 1968).
Results of direct shear tests.
Soil sample
S0
S1
S2
S3
S4
S5
S6
S7
S8
c (kPa)
(deg)
65
28
57
30
30
30
45
29
14
40
54
24
27
40
55
25
52
25
Table 10.
that 25% of copper and lead does not exhibit different

behavior than 10% of the same contaminants.
3.8.1 Direct shear strength test

The results of direct shear tests represented by the
parameters, cohesion and angle of internal friction are
given in Table 9.
Classification of soil samples according to the soil sensitivity.

Undisturbed sample
Remolded sample
Soil sample
peak kPa
peak kPa
c kPa
Sensitivity
value %
Description*
S0
S1
S2
S3
S4
S5
S6
S7
S8
44.64
29.53
19.62
24.62
24.62
18.93
24.03
17.07
14.81
39.32
49.8
41.94
28.83
34.08
39.32
28.83
34.08
36.7
34.63
19.52
17.27
21.39
24.13
19.23
14.81
13.54
12.56
41.94
49.8
52.42
36.7
39.32
34.08
49.8
41.94
42.94
22.32
14.77
9.81
12.31
6.16
9.47
12.02
8.54
7.41
1.3
1.5
1.1
1.2
1.0
1.3
1.3
1.4
1.1
LS
LS
LS
LS
LS
LS
LS
LS
LS
*Low sensitivity
Figure 2. Results of unconfined compression test.
604
According to the results of direct shear tests, the

cohesion of soil decreased with increasing the percentage of contaminants, while the values of angle of
internal friction showed an arbitrary variation.
Table 11.
Results of triaxial compression tests.
Soil sample S0
c (kPa)
S1
S2
S3
S4
S5
S6
S7
S8
135 50 45 125 110 85 35 35 20
3.8.2 Unconfined compression test (UC)

Unconfined compression tests were conducted on
undisturbed and remolded soil specimens to obtain the
cohesion and sensitivity of clay soil. The sensitivity is
the ratio of the strength of the soil in an undisturbed
state to the strength of the remolded state at the same
moisture content. The results of unconfined compression tests represented by cohesion and sensitivity are
given in Table 10 and Figure 2.
Figure 3. Results of triaxial test.
605
The presence of contaminants in soil causes a

decrease in the cohesion of soil samples. The amount
of reduction in the cohesion of soil samples becomes
greater as the percentage of contaminant content
increases, except the soil sample that are contaminated with copper, where an increase in copper content
reduces the effect of contaminant on the cohesion
of soil.
The presence of contaminant also causes a decrease
in the sensitivity of the soil, except soil samples contaminated with 10% of kerosene, which causes an
insignificant increase in the sensitivity of soil. The
thixotropic regain in strength seems to be at a maximum in the clay soil samples with the moisture content
nearly equal to the liquid limit. When the moisture content decreases below the liquid limit, the thixotropic
regain also decreases which causes a decrease in the
sensitivity of contaminated soil samples.
3.8.3 Triaxial compression test (UU)
The results of (UU) test are given in Table 11 and Figure 3. According to the results obtained from testing,
there is a decrease in the cohesion value of contaminated soil samples no matter the type of contaminant.
Also, the reduction in cohesion value increases with
an increase in the contaminant content in the soil
sample. In the kerosene contaminated samples, the
removal of organic matter, which considered most
important cementing agent between soil particles, may
causes the sharp reduction in shear strength (cohesion). Also, contamination of soil samples causes
significant reduction in the deviatoric stress. There
is no significant effect to the kerosene content on
the deviatoric stress and cohesion beyond 10% of the
dry soil sample weight. The cohesion and deviatoric
stress is not affected by contamination of soil with
ammonium.
The contamination of soil samples with copper and
lead caused a significant reduction in the cohesion
and deviatoric stress. The reduction in cohesion and
deviatoric stress may be due to the dissolution of
some salts that cause broken bonds between soil
particles.
CONCLUSIONS
The presence of soil contaminant has different effects

on the physical, chemical and mechanical properties
of soil. The effects mainly depend on the type and
percentage of contaminant in the soil. The soil properties of contaminated soil samples have influenced
significantly with the increase of contaminant percentage. The ammonium hydroxide has less effect on
the soil properties in compared with kerosene, copper
sulphate, and lead nitrate.
Kerosene and ammonium hydroxide make the particles finer than in its natural condition, but copper
sulfate and lead nitrate make the particles coarser.
The liquid limit and plastic limit are increased in all
samples except soil samples contaminated with lead

nitrate.
The dry and bulk densities of soil samples are
decreased with an increasing the percentage of contaminants added to the natural soil, while, d,max
increases and opt decreases except samples contaminated with ammonium hydroxide.
The initial void ratio (except samples contaminated
with lead nitrate), compression index and swelling
index are increased with increasing the percentage of
contaminant added to the natural soil, while the coefficient of consolidation is decreased with an increase
of the contaminant content.
The collapse potential increases slightly with
increasing the percentage of contaminants for soil
samples tested by single collapse test. However, there
was noticeable change in the collapse potential for
sample tested by double collapse test which may be
resulted from increasing TSS and gypsum content.
Accordingly, the double collapse can be considered
more reliable than the single collapse because in the
last test, the collapse potential was specifically measured at a stress level of 200 kPa which may be not
correct to be adopted.
There is an extreme reduction in shear strength
parameters with an increasing the percentage of contaminant content for all soil samples. Contamination of
soil also causes a decrease in the sensitivity value, but
does not change the classification of soil sensitivity.
REFERENCES
Abdul Rasool, H. 1999. Some geotechnical properties of oil
contaminated soil. M.Sc. thesis, college of engineering,
university of Baghdad.
Al-Layla, M. T. H. & Al-Kiki, A. M. A. 2006. Study the
Effect of industrial wastes on the engineering properties
of swelling clayey soils. Journal of Tikrit for engineering
science, 13 (2):5361. (in Arabic).
Annual book of ASTM standards 2003. Soil and rock;
building; stone; peats.
Balesdent J., Mariotti A., & Guillet B. 1987. Natural 13C
abundance as a tracer for studies of soil organic matter
dynamics. Soil biology and biochemistry, Vol.19:2530.
Banat, K. M. 1980. The principles of clay minerals. Baghdad
university press. (in Arabic).
Essa S. k. 1979. Chemical and mineralogical studies on
some soils of Shahrazor area. M. Sc., thesis, university
of Sulaimani, college of agriculture.
Grim, R. 1968. Clay mineralogy. McGraw-Hill book Co.
Jennings, J.E. & Knight, K. 1975. A guide to construction on or with materials exhibited additional settlement
due to collapse of grain structure. Proc. of 6th regional
conference for Africa on soil mechanics and foundation
engineering, Durban, South Africa: 99105.
Kadduo N. S. 1960. Clay mineralogy of some alluvial
soils of Iraq and Dubuque silt loam and underlying
dolomitic limestone of Wisconsin. Ph.D. thesis, university
of Wisconsin Madison.
Muhammed D. A. 2007. Interaction between organic matter and smectite minerals in some Iraqi soils of different
agricultural use. Ph.D., thesis, university of Sulaimani,
college of agriculture.
606
Rahman, Z. A., Hamzah, U. & Taha, M. R. 2010. Influence

of oil contamination on geotechnical properties of basaltic
residual soil. American Journal of Applied Sciences 7 (7):
954961, ISSN 15469239.
Umesha, T. S., Dinesh, S. V. & Sivapullaiah, P. V. 2012.
Effects of acids on geotechnical properties of black
cotton soil. International Journal of Geology, Issue 3,
Volume 6.
UNEP report, 2005. Assessment of environmental Hot

Spots in Iraq. Report ISBN 92-807-2650-1, Job No.
DEP/0755/GE, Imprimerie Chirat, France.
Zulfahmi, A. R., Wan Zuhairi, W. Y., Raihan, M. T., Sahibin,
A. R., Wan Mohd Razi, I., Tukimat, L., Siti Nur Syakireen,
Z. & Noorrulakma, A. 2012. Influence of Amang (tin
tailing) on geotechnical properties of clay soil. Sains
Malaysiana 41(3): 303312.
607

Batch equilibrium test methodology for estimating the retardation

factor (Rd ) using a hexachlorocyclohexane (HCH) solution
L.S. Macedo, F.P. Mano & G. Mondelli
Institute for Technological Research of the State of So Paulo, So Paulo, Brazil
ABSTRACT: This paper presents part of a geo-environmental site investigation of a contaminated site, located
in Santo Andr, State of So Paulo, Brazil. The adsorption capacity of the HCH (Hexachlorocyclohexane)
isomers by the local natural soil, using the batch equilibrium test was studied. The batch equilibrium test method
allows the construction of the adsorption isotherms, for estimating parameters, like the partition coefficient (Kd )
and the retardation factor (Rd ). The test methodology was consisted by three phases: elaboration of the pollutant
solution; determination of the equilibrium time; and determination of the sorption isotherms. The preliminary
results shown that the HCH isomers for the study site is less adsorbed by the soil than expected, presenting low
Kd and Rd values. Also, the interpretation of these results can allow better understanding the behavior of the
contamination plume and how each HCH isomer interacts with the physical environment.
INTRODUCTION
The best known mechanisms of the pollutant transport

in porous media are: advection, mechanical dispersion, diffusion and hydrodynamic dispersion (e.g.
Bear, 1972; Tindall et al, 1999; Fetter, 1999; Boscov,
2008). The retention capacity of the soils influences
the mobility of the pollutants and, specifically sorption properties are very influent when hydrophobic
contaminant occurs in the environment. Sorption processes include adsorption, chemisorption, absorption
and ion exchange.
The mobility of the pollutants in porous media
is associated with the physical-chemical characteristics of the compounds, and with the geological,
geotechnical, geochemical, hydrogeological and climate properties of the site. The study of the mobility
of inorganic compounds, like contamination caused
by heavy metals, has been more common than studies
where organic compounds occur. For organic compounds, some difficulties to understand their behavior
in the natural media are presented during the site investigation, mainly when analytical and interaction tests
with natural soil are needed.
Generally, the organic compounds have low solubility. This fact limits obtaining the pollutant solutions
with known concentrations of the compounds of interest to be used during the laboratory tests, as column,
diffusion and batch equilibrium tests. Also, the background concentrations of the pollutant solution need
to be higher than the detection and the quantification
limits of the analytical laboratory equipment. Problems with handling, toxicity besides waste and effluent
disposal are also very common in geochemistry and
Environmental Geotechnics laboratories.
The adsorption process can be separated in two

groups: physical adsorption, as ion-exchange reactions and van der Waals forces; and chemical adsorption, involving short-range chemical valence bonds
(Yong et al, 2012).
The main characteristics of the soils that directly
influence the retention capacity of the pollutants
are: mineral constituents, mainly clays types, specific
surface, organic matter content, fluid pH, etc.
The parameter that better represents the sorption
mechanism is the distribution coefficient, as known as
partition coefficient (Kd ). It indicates the affinity of the
chemical compound to be attached on the soil particles
surface. Kd can be obtained using batch equilibrium
tests in laboratory.
Hexachlrocyclohexane (HCH C6 H6 Cl6 ) is an
industrial chemical compound of the extended family
of the organochlorines, known as persistent organic
pollutants (POPs). The process of synthesizing HCH
generates a mixture of several isomers in various proportions. Isomers of HCH have the same molecular
formula, but with different atomic arrangements and
physicochemical properties. The pesticide product, in
the past commonly found on the market, is the technical grade HCH (t-HCH), a mixture of isomers, as
the -HCH (alpha), -HCH (beta), HGH- (delta) and
-HCH (gamma), also known as lindane, the most
toxic among them. Table 1 lists the physicochemical
properties of HCH isomers. Table 2 presents different
solubility values for each HCH isomer, which usually
vary a lot, depending on the laboratory temperature
and bibliography.
The objective of this paper is to present the development of a methodology for the batch equilibrium
test using HCH. As part of a great geo-environmental
609
Table 1.
Physicochemical properties of HCH isomers.
Isomer
-HCH
-HCH
-HCH
-HCH
CAS-Number
Molecular weight
Melting point ( C)
Boiling point ( C)
Density (g/cm3 )
Water Solubility* (mg/L)
Vapor pressure (mmHg)
Henry constant* (atm.m3 /mol)
Log Kow*
319-84-6
290.8
158.0
288.0
1.9
2.0E+0
4.5E5
6.70E6
3.80
319-85-7
290.8
309.0
288.0
1.89
2.4E1
3.6E7
4.40E7
3.78
319-86-8
290.8
141.5
288.0
3.14E+1
3.52E5
4.14
58-89-9
290.8
112.5
323.4
1.87
7.3E+0
4.2E5
5.14E6
3.72
*25 C. (Mackay, 2006; INCHEM, 2012; RAIS, 2012; EPI Suite, 2012)
Table 2. Water solubility for HCH isomers (mg/L).
Temp. ( C)
-HCH
-HCH
-HCH
-HCH
-HCH
1
Tindall
et al. ATSDR
(1999) (2005)
WHO
(1992)
WHO Weil
(2004) (1974)
25
2.0
0.24
37.5
31.4
71.14
28
2.0
1.5/0.22
20
717
2.0/69.81
0.24
6.8
2.0
11.04/78.81
25
2
0.2
7.8
31.4
41.4
28 C; 2 20 C
Figure 1. Geographical location of the study site.
site investigation program, this study will allow understanding the HCH isomers behavior in a natural soil
sample from a contaminated site in Brazil.
STUDY SITE
The study site is located in the municipality of Santo

Andr, State of So Paulo, Brazil (Fig. 1), where studies for the development of remediation technologies
for HCH contamination have been carried out. In 1987,
about 30,000 tons of soil/sediment contaminated with
HCH from the rectification of a stream was dumped
on an area of about 72,545 m2 . This stream flowed
close to a factory that produced HCH pesticide until
the 1980s. In 1989, seven trenches were constructed
with compacted soil and PVC (polyvinyl chloride)
geomembrane, involving the top and the bottom of
these hazardous wastes. Presumably, the waste used to
be directly disposed onto the natural land, susceptible
to weathering, before its disposal in the trenches.
The construction of these trenches was based in
an engineering project, approved at the time by the
local environmental agency. Nevertheless, with the
increasing knowledge acquired over the last twenty
years regarding new geosynthetic materials and Environmental Geotechnics projects design to prevent the
contamination caused by waste disposal, it is suspected
that the surrounding of the trenches is contaminated.
Because of this fact, the Institute for Technological
Research of the State of So Paulo (IPT) has been

investigating the area since 1999, as requested by the
owner of the area.
Besides HCH, the waste disposed in the trenches
has inert and non-inert compounds in their composition, including heavy metals and other volatile and
semi-volatile organic compounds (VOCs and SVOCs).
Geologically, the site is located at the sedimentary
Basin of So Paulo (Tertiary Age). It is characterized
by clays and sands lying on a crystalline basement that
encompasses very old metasedimentary rocks (Precambrian crystalline rocks). Quaternary sediments
(sand, clay and other unconsolidated sediments) have
been laid on the metasedimentary rocks.
At west of the seven trenches, there is a large residential area, constructed in 2000, which in the past
belonged to a slaughterhouse. The depth of the groundwater table in the trenches area varies from 0.37 to
6.18 m. There is a stream in the north of the trenches,
which flows north-westwards, which presents high
some HCH isomers concentrations upstream and
downstream of the trenches. Figure 2 presents a photo
showing the residential area and the location of the
trenches, viewed from the east side of the trenches.
These facts show the complexity of the contamination problem on this site, which involves many
variables, demanding a very detailed and complete
site investigation to better understand how, where and
610
Table 3. HCH isomers concentrations found for the pure

BHC pesticide.
Figure 2. Residential area and location of the trenches.
when the contamination occurs due to a possible leak

from the PVC trenches.
Chemical analysis were performed on the groundwater and the surficial soil surrounding the trenches,
which indicated the presence of HCH and other compounds in concentrations above the potable and intervention values limits imposed by the State of So
Paulo. Some VOCs were also identified by IPT (2009)
at the site, such as BTEX, chlorobenzene (mono) and
2,3,4 trichlorophenol.
Since the local contamination was confirmed, IPT
was asked to detail the geo-environmental site investigation, in order to define the contamination plume
extension and evolution. Also, it was necessary mapping other possible contamination sources, beyond the
trenches (IPT, 2010). For this, a mobility study including experimental laboratory methods was performed
at the site, as part of a detailed phase of the site
investigation.
DEVELOPED METODOLOGY
The batch equilibrium test assesses the soil ability to

remove chemical compounds of a specific pollutant
solution. It is performed with previous chosen disturbed soil sample, thus increasing the surface contact
of the solution with the soil particles. There is no
hydraulic flow and diffusion mechanism during the
test, considering that each mechanism acts by itself.
The methodology was based on the procedures described by ASTM D4646/03 and ASTM C1733/10.
Basically, the test consists of a mixture of a previously determined volume of a solution with known
concentration of one or several chemical compounds
of interest, with a predetermined quantity of the chose
disturbed soil. This mixture is rotated for a certain
time, called equilibrium time.
Different soil/solution ratios can be tested. Normally, soil mass remains the same for different concentrations of the pollutant solution. Also, it is possible
to maintain the same concentration of the pollutant
HCH
Concentration (mg/kg)
Concentration (%)
26,000
15,000
16,000
14,000
71,000
36.62
21.13
22.53
19.72
100.0
solution and varying the soil mass. For HCH isomers,

which have very low solubility, to achieve the maximum concentration for each isomer in deionized water
is a chemistry challenge, so, varying the soil mass
is an interesting choice. The solution is mixed for a
defined time interval and is then separated from the
soil particles (adsorbent), and the liquid phase is analyzed. At the end of the test, it is possible to build
the sorption isotherms. Some modifications of this
methodology were made to allow the applicability to
the HCH isomers, as described in detail follows.
3.1 Preparation of the pollutant solution

To obtain the pollutant solution, a pure waste of the
BHC pesticide (popular name of the HCH) was used.
The BHC pesticide was analyzed for concentrations
of the different HCH isomers, having 7% of the total
HCH in their formula, as shown by the results of the
chemical analyses in Table 3.
The use of the BHC pesticide in this study differs of
others studies consulted (Rood et al, 1994; Sprynskyy
et al, 2008; Wadaskar et al, 2010), since those studies use standard pollutants salts during the preparation
of the solutions. However, our use of the commercial
product seems to be closer to the real problem.
For elaboration of the pollutant solution, it was considered that the HCH isomers have been leaching from
the waste to the environment, naturally. Therefore, we
decided to perform only one solubilization of the BHC
pesticide compounds in deionized water, followed by
filtration to remove the remaining insoluble particles
from the water. This process was based on Brazilian
Standard ABNT (Brazilian Association of Technical Standards) 10.004/2004, for waste classification
and characterization, using the test methodologies
described by ABNT NBR 10.005/04 and 10.006/04
ABNT NBR, for leaching and solubilization tests,
respectively.
The solubilization test, according to ABNT NBR
10.006/04, allows the solubilization of compounds
present in a mass of waste at 23 C (+/2 C) in a
liquid medium. This was the preliminary method used
as a base to prepare the pollutant solution.
The final concentrations of the pollutant solution
were expected to be close to the maximum solubility
degree of each isomer in the water, so that their analytical analysis results were accurate enough. Table 2
611
Figure 3. Weighing the BHC pesticide.

Figure 4. Mixing the BHC pesticide with deionized water.
shows that the -HCH isomer is more soluble, followed by the -HCH and -HCH. -HCH is practically
insoluble in water.
The definition of the mass of pesticide used was
based on the solubility of the HCH found in the literature (Table 2). Also in the results presented in the
Table 3 for the pure BHC pesticide. To calculate the
mass of the BHC pesticide to be used for each isomer
we considered what the best mass/liquid ratio would be
to achieve the maximum solubility concentrations for
each isomer in deionized water. Using a safety margin,
a value of 7.5 g/L was established.
The preparation of the pollutant solution was performed getting duplicate extractable solution and the
following procedure:
In a ceramic cup, 7.5 g of BHC pesticide was

weighed (Figure 3). This as mixed with 1.5 L of
deionized water (solid/liquid ratio of 1:200), using a
borosilicate glass flask of 1.8 L of volume capacity
(Figure 4);
the mixture shake for 18 hours at 23 C (+/2 C)
in a rotary shaker with 28 rpm (Figure 5) (ABNT
NBR 10.005/04);
the mixture rested for six 6 days at 23 C (+/2 C),
completing a period of 7 days of the BHC pesticide in contact with deionized water (ABNT NBR
10.006/04);
after rest, the solution remained in an oven at 32 C
(+/2 C) for 5 hours;
the solution was shaken for 5 minutes on a rotary
shaker with 28 rpm (ABNT NBR 10.006/04);
the fluid extraction was carried out using a pressurized filtration apparatus with cellulose ester membrane of 0.45 m mesh (ABNT NBR 10.006/04),
leaving only the BHC pesticide on the filter
(Figure 6);
The prepared solution was conserved in amber
flasks and sent to laboratory of chemical analysis,
which was stored at 4 C (2 C) (Figure 7). Until a
maximum of 14 days after the extraction, the solution was analyzed using a gas chromatography with
electron capture detector (USEPA, 1996).
Figure 5. Shaking the mixture.
Figure 6. Pressurized filtration of the BHC pesticide.
612
Table 5.
Physicochemical properties of the soil sample.
Sample ID
PM-14B
Depth (m)
Moisture content (w) (%)
Organic Matter Content (%)
Clay (%)
Silt (%)
Sand (%)
Soil specific dried mass (g/cm3 )
Soil specific natural mass (g/cm3 )
Clay minerals predominate
7.50 a 12.50
22.00
1.70
29.00
17.00
54.00
1.75
2.05
Montmorillonite/
kaolinite
5.66
19.65
35.00
0.55
25.00
pH
Eletrical condutivity (S/cm)
Porosity n (%)
Void ratio e
CEC (meq/100 g)
batch equilibrium test. A total 42 L of the solution were

produced to be sufficient for the different phases of
the batch test. At least 1 L was necessary to determine
the initial concentrations of the solution, and also for
each stage of the experiment. This fact was the mainly
difficult found to carry out the test.
Figure 7. Amber flask for preservation of the ready solution.

Table 4. Desired concentrations versus reached concentrations for the HCH pollutant solution.
HCH
Desired
concentration (mg/L)
Reached
concentration (mg/L)
2.00
0.241.50
6.8037.50
2.0031.40
11.0472.40
3.13
0.27
6.50
11.15
21.01
3.2 Soil sample

The disturbed soil sample was collected with liner
using the direct push system. The sample was representative of the deeper layer of the geological and
geotechnical profile that could be impacted by the
HCH contamination plume. The sample was dried
and sieved in 2.0 mm mesh. Table 5 presents the
physicochemical properties of the soil sample tested.
As described above, some modifications were carried out in the procedures of the leaching and solubilization tests as determined by ABNT. The direct
analysis of the pure BHC pesticide (Table 3), shown
that the concentrations of the HCH isomers are much
higher than each isomer level of solubility. It is known
(unless the temperature range) that the final concentrations of the solution extracted not to depend on
variations that can occur in the solubilization process.
Since when the maximum isomer solubility is reached,
the solubilization process ends.
Comparing the desired concentrations with concentrations achieved in the pollutant solution (Table 4),
isomers -HCH and -HCH were observed to be in the
desired range of solubilization, when compared with
the data presented in the Table 2. -HCH concentration
was slightly below the desired range, but also acceptable. For the -HCH, the concentration of this isomer
was considered close to the solubility suggested in the
literature and also acceptable.
With this pollutant solution prepared, it was possible
to analyze the retention capacity of the HCH isomers
for a natural soil collected in the study site, through the
3.3 Determining the equilibrium time

The determination of the equilibrium time is no longer
provided in ASTM D4646-03 (2008). The 24 h time is
suggested to make the test convenient and to minimize
the microbial degradation. Because of the particular
behavior and characteristics of the HCH isomers, this
phase of the test was chosen to better evaluate the
sorption behavior of this contaminant.
For the determination of the equilibrium time a
soil/solution ratio of 1:4 (or 50 g of soil for 200 mL of
the pollutant solution) was adopted. The samples were
placed to shake horizontally at 29 rpm, with controlled
temperature of 22 C (2 C). Different shaking times
were used: 2, 4, 8, 24, 48, 72 and 168 hours. Each
condition (time sample) had two duplicates and one
blank. For this configuration, 12 flasks of 200 L were
necessary for each point, because the analytical laboratory asked for at least 800 mL to analyze isomers of
HCH by gas chromatography. So, for the different 7
equilibrium times tested, a total of 84 amber flasks of
200 mL were prepared.
Due to the size of the horizontal shaker, the 84 flasks
were shaken separately for 4 batches of 24 flasks each.
613
After the shaking for each pre-determined equilibrium

time, the solid and liquid phases were separated using a
centrifuge with 1250 rpm for 20 minutes and filtration
in a filter paper with 0.1 m mesh.
The resulting fluid extract was stored in a refrigerated chamber at 4 C (2 C) and sent to the laboratory
for chemical analysis. The 14 day expiration date for
the HCH samples was respected.
At the end of this step, the degree of sorption (S)
values were calculated for each isomer, as described
by the Equation 1. Charts of S versus equilibrium time
were constructed. According to the ASTM D4646-03
(2008), Kd can be calculated using Equation 2.
Figure 8. Degree of sorption versus equilibrium time.
where:
Ci is the initial concentration of the solute defined

by the medium concentration of the blanks, in
mg/mL;
C is the final concentration of the HCH isomer in the
fluid extracted after the equilibrium time in contact
with the soil sample;
V is the total solution volume for each 4 flasks,
in mL;
M is the total mass of the soil for each 4 flasks,
in g.
The equilibrium time is defined as the time when S
presents constant behavior of the compound of interest. When the compound does not achieve equilibrium,
it is recommended to use 24 h in the adsorption step
of the test, as recommended by the ASTM D4646 03
(2008).
3.4
Sorption isotherms construction
Once the equilibrium time is defined, the sorption step

can be initiated. In this step, the time will be constant
for all samples and the mass of the soil (adsorbent)
will vary. Soil/solution ratios of 1:4, 1:10, 1:20, 1:50
and 1:100 (or 50 g, 20 g, 10 g, 4 g and 2 g of soil, for
200 mL of the pollutant solution) are recommended.
For each condition of soil/solution ratios, 2 duplicates
and 1 blank were obtained. Thus, a total of 60 flasks
were prepared and 15 chemical analyses for each HCH
isomer were carried out.
The following procedures are the same as described
for the equilibrium time determination, such as shaking, centrifugation, filtration and preservation.
After the sorption step, it is possible to construct the
sorption isotherms for each isomer, which shows the
relationship between the amount of the solute adsorbed
by the soil particles and the concentration of the solute
remained (equilibrium concentration), at a constant
temperature.
The sorption isotherms can be obtained by fixing

the soil/solution ratio and by varying the concentration of the solution applied to the soil (CSI isotherm
Constant Soil Isotherm). Also, it is possible to fix
the concentration of the solution and vary the volume
of the solution applied to the soil (ECI isotherm
Environmentally Conservative Isotherms) (USEPA,
1992).
The best known isotherm models applied to interpretate the batch equilibrium test results are linear isotherm, Freundlich isotherm and Langmuir
isotherm. The first two do not present a limit for the
amount of solute to be adsorbed by the soil, whereas the
Langmuir isotherm considers that there is a maximum
of particles which can be adsorbed to a certain amount
of soil (Fetter, 1999; Tindall et al, 1999; Sparks, 2003).
For HCH isotherms found in other studies, Freundlich and linear isotherms were the most commonly
used models (Rood et al, 1994; Sprynskyy et al, 2008;
Wadaskar et al 2010).
For the present case, ECI isotherms were attempted,
because of the difficulty in obtaining the pollutant
solution with the highest concentrations possible for
the HCH isomers.
The soil retardation factor (Rd ) for linear isotherms
can be obtained by the following equation:
where n is the soil porosity and d is the soil specific

dried mass.
PRELIMINARY RESULTS
Figure 8 presents the results of the equilibrium time

determination step, through a logarithm scale chart of
the degree of sorption versus time. The logarithm scale
facilitates the visualization of the S values in this case,
since they have different orders of quantity.
Figure 8 shows that -HCH, followed by the
-HCH, have higher degree of sorption by the soil than
the -HCH and -HCH, independently of the contact
time between soil and solution. This fact could explain
614
Table 6. Kd values calculated in the equilibrium time

determination step.
for the risk analyses and remediation studies is 3.0, for

this site, for all HCH isomers.
Kd (L/Kg)
Equilibrium
time (h)
-HCH
-HCH
-HCH
-HCH
2
4
8
24
48
72
168
1.58
0.50
0.18
0.40
0.21
0.62
0.15
1.28
0.45
0.10
0.39
0.17
0.48
0.23
0.67
0.43
0.13
0.55
0.16
0.57
0.13
0.47
0.39
0.16
0.76
0.10
0.61
0.07
This paper described the challenges to carry out the

batch equilibrium test when the substance of interest is
organic and has low solubility in water. A brief review
of the literature was presented and also the details of
the methodology developed for acquiring the pollutant solution, equilibrium time and sorption isotherms
steps.
The preliminary results shown that the HCH isomers for the study site is less adsorbed by the soil
than expected, presenting low Kd and Rd values. Also,
the interpretation of these results could help to understand the behavior and the different phases in which
HCH appears in the environmental media: solid state
adsorbed by the soil particles and dissolved phase in
very low concentrations in the groundwater.
Table 7. Rd values estimated in the equilibrium time determination step data.

Rd
Equilibrium
time (h)
-HCH
-HCH
-HCH
-HCH
2
4
8
24
48
72
168
8.91
3.50
1.90
3.01
2.07
4.08
1.77
7.39
3.23
1.48
2.97
1.87
3.39
2.17
4.37
3.17
1.63
3.73
1.79
3.84
1.64
3.35
2.94
1.80
4.78
1.49
4.03
1.35
CONCLUSION
ACKNOWLEDGMENTS
why -HCH and -HCH were the isomers mostly

found during the groundwater sampling investigation.
In Figure 8, it is not clear for what time the S turns
constant (equilibrium time). But the greatest degrees
of adsorption were observed to occur at 2, 24 and
72 h. The adsorption capacity of the -HCH isomer
was described by Sprynskyy et al (2008) as 5.6 g/g
in 48 h equilibrium time and the Kd value calculated
using linear isotherm was 50 L/Kg. It is worth noting that the soil and solution conditions had different
conditions in this study.
Based on the results presented in Figure 8, the sorption step was carried out using 24 h of shaking. The
analytical results, however, did not confirm the same
adsorption behavior observed during the equilibrium
time determination step and the duplicates results did
not present repetitively analytical results. Hence, it was
not possible to construct the sorption isotherms along
this study, and the possible problems those occurred
are been investigated.
Table 6 presents the punctual Kd calculated through
Equation 2, using the equilibrium time determination
step data. It is important to note that the majority of
Kd values are higher for equilibrium time of 2 h.
Using Equation 3 and the soil properties shown in
Table 5, it was possible do estimate the retardation factors (Rd ) for each point investigated. If 24 h of shaking
will be adopted as equilibrium time, as recommended
by ASTM D4646-03 (2008), it can be concluded that
an average and conservative Rd value to be adopted
The authors acknowledge the Brazilian Development

Bank (BNDES), the Water and Electricity Department
of the State of So Paulo (DAEE) and the Chemical
Analysis Laboratory (LAQ) of the IPT, for supporting
this research and its continuation.
REFERENCES
ABNT ISO (2004). Resduos Slidos Classificao, NBR
10004.
ABNT ISO (2004). Procedimentos para obteno de extrato
lixiviado de resduos slidos, NBR 10005.
ABNT ISO (2004). Procedimento para obteno de extrato
solubilizado de resduos slidos, NBR 10006.
ASTM C1733 (2010). Standard Test Method for Distribution
Coefficients of Inorganic Species by the Batch Method.
ASTM D4646 (2003 Reapproved in 2008). Standard Test
Method for 24-h Batch-Type Measurement of Contaminant Sorption by Soils and Sediments.
ATSDR. (2005). Disease Registry. Toxicological Profile for
Alpha-, Beta-, Gamma- and Delta-Hexaclorocyclohexane.
Atlanta, Gergia.
Bear, J. (1972). Dynamics of fluids in porous media. U.S.A:
American Elsevier Publishing Company, 764p.
Boscov, M.E.G. (2008). Geotecnia Ambiental. Oficina de
textos, So Paulo, Brasil. 248p.
Fetter, C. W. (1999). Contaminant Hydrogeology. Prentice
Hall, 2nd Ed., 500p.
IPT (2000). Diagnstico de contaminao de solo e gua
subterrnea de Bota-fora situado em Santo Andr, SP.
Parecer Tcnico IPT n 7.775/00, So Paulo, SP.
IPT (2009). Desenvolvimento de alternativas tecnolgicas
para tratamento de solo e gua subterrnea, contaminados com HCH (Hexaclorociclohexano), na rea do
Bota-fora situado em Santo Andr, SP. Relatrio tcnico
IPT n 115053-205, So Paulo, SP.
615
Mackay D.; Shiu, W.Y.; Ma, K.C.; Lee, S.C. (2006). Ilustrated
handbook of physical-chemical properties and environmental fate for organic chemicals. Second Edition. Volume IV Nitrogen and Sulfur Containing Compounds and
Pesticides.
RAIS. (2011). The Risk Assessment Information System.
The University of Tennessee. Pgina: http://rais.ornl.gov/.
Consultado em 14/12/2011.
Rood, G.A.; Broekman, M.H.; Aalbers, Th.G. (1994). Investigating a Leaching Test for PCBs and Organochlorine
Pesticides in Water and Building Materials. Environmental
Aspects of Construction with Waste Materials. Holanda.
Sparks, D. L. (2003) Environmental Soil Chemistry, San
Diego: Academic Press, 2a edio, p. 111112.
Sprynskyy M.; Ligor, T; Buszewski, B. (2008). Clinoptilolite in Study of Lindane and Aldrin Sorption Processes
From Water Solution. Jornal of Hazardous Materials, 151,
pgs. 570577. Polnia.
Tindall, J.A.; Kunkel, J.R.; Anderson, D.E. (1999): Unsaturated Zone Hydrology for Scientists and Engineers.
Prentice Hall, New Jersey, USA.
United States Environmental Protection Agency (USEPA,
1992). Batch-type procedures for estimating soil adsorption of chemicals Technical Resource Document,
EPA/530/SW-87/006-F, Washington, April.

2007). SW-846, Method 3540C. Soxhlet Extraction.
December, 1996.
2007). SW-846, Method 3562. Supercritical Fluid
Extraction of Polychlorinated Biphenyls (PCBs) and
Organochlorine Pesticides. February, 2007.
Wadaskar, J.V., Ekhe, J. D. & Kale, S. P. (2006): Adsorptiondesorption of HCH and Endosulfan on a Soil. Environmental Technology, 27:9, 1011-1017.
Weil L, Dure G, Quentin K-E (1974) Wasserlslichkeit von
insektiziden chlorierten Kohlenwasserstoffen und polychlorierten Biphenylen im Hinblick auf eine Gewsserbelastung mit diesen Stoffen. Wasser Abwasser Forsch
6:169175.
World Health Organization (WHO) (1992). Alpha- and BetaHexachlorocyclohexanes; Geneva, EUA, 1992.
World Health Organization (WHO) (2004). Lindane in
Drinking Water; Geneva, EUA, 2004.
World Health Organization (WHO) (2011). International
Programme on Chemical Safety (INCHEM). Pgina:
http://www.inchem.org/. Consultado em 13/12/2012.
Yong, R. N.; Nakano, M.; Pusch, R. (2012). Environmental
soil properties and behavior. CRC Press. 435 p.
616

The resistivity piezocone for high resolution geo-environmental site

investigation: A Brazilian example
M.T. Riyis
ECD Sondagens Ambientais Ltda, Sorocaba-SP, Brazil
H.L. Giacheti
Universidade Estadual Paulista, Bauru-SP, Brazil
ABSTRACT: A suitable decision-making on managing a contaminated site characterization program is strongly

dependent of the diagnosis process. A detailed diagnosis can be done based on a Conceptual Site Model (CSM)
elaboration using high resolution site characterization tools. The piezocone (CPTu) test is a high resolution
tool which allows attaching several specific sensors, like the resistivity probe. This hybrid device is called the
resistivity piezocone (RCPTu). A simulated geo-environmental site characterization program was performed
on an erosion site using different tools (direct push tools soil samplers, hollow stem auger (HSA) drilling
and RCPTu tests) to develop the CSM for a site similar to the Brazilian conditions. It was observed a good
agreement between the site profiles interpreted by the different methods. The resistivity sensor attached to the
piezocone improved the interpretation and the decision-making process on site was significantly better for the
CSM elaboration. The RCPTu test data also allowed identifying the hydrogeological heterogeneities. The present
study shows that the RCPTu test is also a useful and powerful tool to development an accurate CSM in a Brazilian
condition, especially in an approach that prioritizes high resolution geo-environmental investigation.
INTRODUCTION
The study of contaminated sites is relatively new

in Brazil. The two major specific legal instruments
were published just in 2009: the So Paulo State
Law 13577/2009 and the Federal Resolution 420/2009
(CONAMA National Environmental Board) that
applies to all Brazilian territory. In So Paulo State,
resolutions, regulations and procedures are implemented and monitored by CETESB the Environmental Company of the State of So Paulo.
In Brazil, the legal liability for remediation of a contaminated site is responsibility of who have caused the
contamination, of who got financial gains with it, or of
the owner of the area. It is his or her obligation to investigate and to recover the site. The Legally Responsible
Person currently hires a Technical Manager, which
is called a Consultancy. Usually, this Consultancy
subcontracts or supervise some outsourced services,
like chemical analysis laboratories, drilling companies, and other contractors which carry out the field
work (soil, groundwater and vapor sampling, installation of monitoring wells, among other). The Consultancies do their best to reduce the cost of their project,
so, they just strictly complies the requirements of the
Brazilian environmental agency. In most cases, the
area is not properly investigated; therefore, it is not
suitably remedied.
The goal of the Legally Responsible Person, in
the U.S. model (US EPA, 2004), is to rehabilitate the
area more effectively as possible, faster and cheaper,

considering the project globally. In this approach, a lot
of funds are invested in research and development of
methodologies, tools and equipment for site assessment, investigation and diagnosis of contaminated
areas.
To achieve these goals, some North American companies implemented the Triad Approach and Expedite
Site Assessment which have three basic principles:
a) Decision-making must be rapid and on site;
b) The Conceptual Site Model (CSM) must be continuously reviewed and, at the end of the diagnostic
work in the field, it should be with the lowest possible degree of uncertainties, and these uncertainties
must be manageable (Aquino Neto, 2009; Quinam,
2012; Killinbeck, 2012);
c) The data collection equipment must be of high resolution and allow a dense collection of data in a
short time (Clearly, 2010; Linch, 2004, Martin &
St Germain, 2008).
Considering these principles, the cone (CPT),
the piezocone (CPTu) and the resistivity piezocone
(RCPTu) tests are very useful site investigation tools,
because:
a) It collects a high density data very fast (Lee et al,
2008, Welty, 2012; Killinbeck, 2012);
b) It gives high resolution profile with relatively low
cost (Lee et al, 2008; Shin III et al, 2004; Giacheti
et al, 2006; Quinam, 2012);
617
c) It allows decision-making on site; it provide a very

good understanding of the geology, a most critical
variable on the CSM elaboration (Quinam, 2012;
Welty, 2012; Killinbeck, 2012; Clearly, 2009);
d) It provides the development of a very good
hydrostratigraphical profile (Quinam et al, 2010;
Clearly, 2009; Vienken et al, 2012);
e) It allows to gather soil and groundwater data which
would be very difficult (or impossible) to obtain.
All these tests (CPT, CPTuand RCPTu) are frequently used in the investigation of contaminated sites
in the U.S. (Vienken et al, 2012).
Unfortunately, an equivocated approach is currently
used in Brazil. The site investigation campaigns to
assess contaminated sites end up being much slower
and with a lot of uncertainties because the decisions
are not taken on site, and traditional site investigation
methods and low resolution tools are used. Besides
the choice of an inappropriate approach, the traditional
methods are not usually adequately used, resulting in
inefficient remediation projects, with extremely long
periods, and very high costs for the Legally Responsible Person and to society, often making impossible
to properly revitalize the area.
This paper presents a case study which uses the
RCPTu test as a tool to develop a solid CSM
with a low degree of uncertainty in a simulated
geo-environmental site investigation, conducted in
accordance with the high resolution investigation
approach. The RCPTu tests are a powerful tool
to complement hydrogeological studies, providing a
high density and high definition data to delimit with
accuracy the hydrostratigraphical profiles of the area
with potential to be used also in Brazil.
2
STUDIED SITE
The studied site is an erosion process installed in

Bauru city, central part of So Paulo State, Brazil.
The city is settled in a natural amphitheater of about 5
km radius, modeled by several streams at the headwaters of Bauru river. The regional relief presents
wide and gentle hills, and the rocks are sand-stones
of Marlia and Adamantina Formations. The typical
soil is residual from sandstone of Marlia, Adamantina
and Cenozoic Formations which sand texture with little clay. The soil formation in tropical climate, marked
by the alternation of rainy seasons and droughts, intensify the leaching of thin surface horizon, producing
a porous and permeable structure, usually with deep
groundwater table.
This erosion process is described by Ide et al (2010),
and its location is presented in Figure 1, along the gua
Comprida Creak. It is originated due to the destruction of rainwater dissipation system, poor design, poor
construction and lack of maintenance. The process
reached a huge dimension with the installation of a
land subdivision for houses (residential condominiums) in areas of high erosion susceptibility, combined
with poor design and lack of infrastructure.
Figure 1. An aerial view of the studied site.
Figure 2. The erosion process in the major erosion branch

(Ide et al, 2010).
The rapid evolution was due to several heavy rainfalls, typical in this region, which hit and destroyed
the water dissipating system, contributing to an intense
erosion progress on its main branch. Figure 2 shows
the erosion process in the studied site. According to
Ide et al (2010) a revitalization project was develop to
rectify the bed of the stream since the rapid evolution
of erosion process. However, the lack of maintenance
is contributing to slope instability and minor erosions
are taking place again at the site.
Several site characterization campaigns, including
field observation, laboratory and in situ testing have
been carried out at the site to understand and explain
the erosion process (Ide et al, 2010). Recently, a new
campaign including resistivity piezocone tests were
also carried out in order to help identifying details in
the site stratigraphy.
3.1 Direct push soil samplers single tube

This technique consists in pushing, by means of a
percussion device, a tubular steel tool known as soil
618
Figure 4. Hollow stem auger drilling in used at the studied

site.
Figure 3. Single Tube Sampling Diagram (AMS, 2012).
sampler, which measures about 1.40 m long by 55 mm

outer diameter and 46 mm internal diameter. This sampler contains a transparent device of polymer material
(HDPE or PVC) called liner. The liner is where
the retrieved soil samples are kept after spiking and
removal of the set of Direct Push Tools (DPT). The
liners are 1.20 m long with outer diameter of 44 mm.
The DPT is pushed with some percussion equipment. In this study, the used equipment was an SB-50
Atlas Copco hydraulic hammer coupled to a truckmounted hydraulic rig (ECD-SMA-01). Besides the
sampler, the DPT has extending rods and adapters.
After pushing, the DPT is retrieved, the liner is
removed from the sampler, the lower end of the liner is
identified with a red HDPE plastic cap, and the DPT
is reassembled and replaced in the borehole, with a
extending rod, to sample 1.0 m deeper (Figure 3).
The choice for a Single Tube Direct Push soil sampler rather than another, as the Dual Tube or Piston
Sampler, was done because it is faster and more widely
used in Brazil. The other two modalities mentioned are
used in Brazil only in special cases.
In cases where the borehole collapses, casing is necessary. In such cases, the soil sampling is carried out
along with the Hollow Stem Auger (HSA) drilling.
In the present work, the rig used with the HSA tools
is the truck-mounted ECD-SMA-01, owned by ECD
Sondagens Ambientais, a Brazilian company for geoenvironmental investigation, with a maximum torque
of 5000 Nm mounted on a VW 9150 truck.
Initially, the DPT set is pushed into the ground.Then
it is taken out, dismantled and the liner is removed from
inside of the sampler for soil tactile-visual identification. Then, HSA drilling is carried out to the depth
already sampled. When this depth is reached, the DPT
set inside of HSA is taken out and the DPT is placed
back by Direct Push soil sampler inside the hollow
borers. The DPT is hammered one more meter, and

the sampling procedure is repeated (Figure 4).
The lack of casing allows the collapse of the
borehole wall, and consequently non representative
sampling. So, in collapsible soils (especially sandy
soils), Direct Push sampling should be performed with
concomitant HSA drilling or using Dual Tube DPT.
3.2 RCPTu tests
The piezocone is an instrumented probe which is vertically pushed into the ground at a standard rate of
20 mm/s. Measurements of tip resistance (qc ), sleeve
friction (fs ) and pore pressure response at up to three
locations are typically recorded every 25 or 50 mm.
Changing in the friction ratio (Rf , equal to fs /qc , in
percent) are often used to identify soil behavior type
based on classification charts. Pore pressure records
gives information about the response of the ground to
the probe during the pushing and consequent migration
of fluids. The pore pressure measurement is done by a
pressure transducer inside the piezocone. The conventional procedure to measure pore pressure is saturating
the porous element with water or glycerin. Larsson
(1995) and Elmgren (1995) have suggested the use of
a slot filter filled with grease to replace this procedure.
The slot filter filled with grease is much easier to
prepare and handle than the porous piezo-element saturated with glycerin (Mondelli et al, 2009). The pore
pressure measurement is done by a pressure transducer inside the piezocone, which is brought in-to stiff
contact with the pore water in the soil by fill-ing the
inner cavity with water and pressing the grease into
the cavities inside the cone tip.
The RCPTu test is similar to any other CPT test.
The additional procedure is to add a signal generator
to the data acquisition system to control the current
level and frequency for the electrical conductivity
(or resistivity) measurements.
Figure 5 presents a Wenner-type resistivity piezocone with an array of four electrodes. Resistivity
measurements are taken with the inner electrodes
and the current is applied through outer electrodes.
619
Figure 5. Schematic representation of a resistivity piezocone probe with a four-electrode array (Mondelli et al.
2007).
carried at a constant velocity of 20 mm/s, interrupted

only to connect new rods, every 1 m interval.
Measurements were recorded at regular intervals
of 20 cmm, and displayed in real time on a computer
connected to a sonic data transmission system. Thus,
the following parameters were measured: tip resistance (qc ), lateral friction (fs ); pore pressure (u2 ) and
electrical conductivity (EC).
4 TESTING DATA AND DISCUSSION
4.1 Direct push soil sampling
Figure 6. RCPTu Push Rig ready for the test.
These measurements are also digitally recorded at

25 or 50 mm intervals, providing essentially continuous field data in addition to all the other standard
piezocone measurements.
The probe used in the tests is the RCPTu system
produced by Geotech AB (Sweden), NOVA Acoustic
model (wireless). The position to measure pore pressure in this cone is the standard (u2 ), according to
international practice. The slot filter filled with grease
was used to measure pore-pressure. The equipment has
a relationship of unequal areas (a) equal to 0.84. This
hybrid probe, owned by ECD Sondagens Ambientais,
has a resistivity (or conductivity) sensor, also produced
by Geotech AB. This sensor provides the electrical
conductivity data, also by the NOVA Acoustic wireless system. The wireless data transfer and acquisition
were performed with a microphone, which is part of
the NOVA Acoustic system.
The anchoring of the pushing equipment was done
using double helicoids of 180 mm in diameter and
extended rods with square coupling. The RCPTu tests
were carried out using a hydraulic system attached to
a Tractor to perform the test.
The test procedure followed the Brazilian Standards
(ABNT-MB 3406/1990), quite similar to the ASTM
D3441 standard. The hydraulic system was positioned,
leveled and anchored (Figure 6). After that, the baseline was taken and the resistivity piezocone test was
The new site investigation campaign started by soil

sampling with DPT, a technique recommended by US
EPA and, in So Paulo State, by CETESB. It is recognized as the best technique for soil sampling for
investigation of contaminated sites. The Single Tube
was chosen as the type of DPT for an initial sampling
strategy. A parallel section to the creek and perpendicular to the direction of the studied erosion was
established by Ide et al (2010) to identify the site stratigraphical profile. In this section, a distance of 20 m
was established between one point and another as the
default for the initial data collection.
Several soil samples were taken in that row with
Single Tube DPT. The samples were collected in liners
every meter depth and they were described on site by
visual-tactile identification method.
During the completion of the samplings, there were
some factors that demonstrate the limitation of this
method to establish an accurate CSM, with the resolution required for a proper investigation of contaminated sites. They are: wall collapse of the borehole;
hydraulic pressure loading material into the casing;
difficulty in recovering saturated sandy soil samples.
These limitations are discussed as follows.
4.1.1 Wall collapse
In the first sample point (DP-01) it was possible to
experience this kind of difficulty. After retrieving the
fifth sample (from 4.0 to 5.0 m depth), when the DPT
was relocated into the borehole, it was noticed that
the bottom of the sampling pit was about 4.7 m depth,
instead of 5.0 m, as expected. This means that at least
0.3 m of material fell to the bottom of the borehole
and the following sample, which would be 5.0 to
6.0 m depth was contaminated with material which
does not belong to this layer. Then, the following two
samples (5.0 to 6.0 and 6.0 to 7.0 m depth) were collected this way, and the soil from the bottom part of
the liner was discharged.
In the sample point DP-05, the same procedure
was adopted; however, due to the need for more
information on this sampling point, the sample was
recollected, i.e., after collecting the sample of 56 m
depth, the DPT was taken out, and the probe was
inserted once again, 1 m away from the previous one.
The HSA rotary drilling was used up to 5 m depth
and then, the soil sampling from 5.0 to 6.0 m was performed inside the hollow augers. After sampling, the
620
HSA drilling was continued up to 6.0 m depth, and

the soil sample was collected inside the hollow auger
for 6.0 to 7.0 m depth. When the sample from 5.0 to
6.0 m depth was recollected, it was noticed that there
was a layer of about 0.2 m of plastic gray clay with
sand grains, which had not been detected in previous
sampling (with no casing), certainly due to contamination of the sample material from the upper layers.
The procedure was repeated up to 8 m depth, but the
samples below 6 m depth were not representative due
to another factor: the hydraulic pressure bearing material into the hollow augers. This factor is going to be
discussed in the next section.
4.1.2 Hydraulic pressure loading material into the

casing
As previously described, some samples were not representative due to sample contamination with material
beneath the casing caused by the hollow augers. The
material inlet occurs because of the hydraulic pressure associated with a friable and low consolidated soil
(fine sand characteristics). If the hydrostatic pressure
is high enough (estimated at 30 KPa), the soil tends
to enter in the hollow augers and block the retrieving of representative soil samples. There exist some
tools (like the Piston Sampler, manufactured by AMS
Inc, for example) that minimize this problem; however,
they are not widely available in Brazil yet.
Likewise, the technique used in SPT tests for
geotechnical site investigation of washing inside
the casings with water cannot be used for geoenvironmental site investigations because it interferes
in the aquifer being investigated.
Figure 7. Profile interpreted using Direct Push soil sampler

for the studied site.
Figure 8. Typical resistivity piezocone test data (RCPT-02).
4.2 Resistivity piezocone tests (RCPTu)

4.1.3 Difficulty in recovering saturated sandy soil
samples
In the sample point DP-06, another problem occurred:
excessive amount of water in the sandy soil sample.
Consequently, the sampler penetrated the ground with
the casing (HSA), but when it was removed from
the borehole, there was almost no recovered sample
because it escaped from the DPT sampler due to the
large amount of water coming out of it.There is another
tool (core catcher, by Geoprobe) that minimizes this
effect; however, it is not commercially available in
Brazil yet.
The soil profile interpreted by the samples collected
on a DPT sample point is shown on Figure 7.
It is clear that the CSM elaborated using the DPT
has several uncertainties arising, mainly from the limitation of the data collection as well as the limitations
of how to interpret the data, since the tactile-visual soil
identification depends on the professional who is performing it variations of two orders of magnitude are
common (Ahlers, 2012) and the scale with which
samples can be described is inappropriate for variations in hydraulic conductivity that it is necessary to
be identified.
The next step on the CSM elaboration was to perform

RCPTu tests. The goals of using this tool were to obtain
better detailed profiles at key points, refine the CSM,
check the accuracy and limitations of this tool compared to others previously used in the site (SPT and
DPT), and check in detail the position of any minimal
lenses of clay that can act as an aquitard and confine
a deeper aquifer.
Ten RCPTu tests were carried out in the same section where the soil samples were collected, most of
them right besides DPT soil samplings. With these
data, it was possible to refine the CSM on site considering the hydrostratigraphic profiles interpreted from
the data obtained by the RCPTu tests.
A typical RCPTu test and the interpreted profile are
presented in Figure 8 for the RCPT-02. The soil type
was identified using Robertson et al (1986) classification chart as a reference to be confirmed and adjusted
based on the samples retrieved with the DPTs. This
interpretation is presented in the laptop screen on site,
allowing identify the soils, the storage and the flux
zones and refining the CSM right after every new test.
It was possible to detect some clayey (or fine
materials) lenses on RCPTu profiles, and identify
the storage zones and the flux zones (Quinan et al,
621
Figure 9. Hydrostratigraphic profile interpreted on site for

the same piezocone test data previously presented (RCPT-02).
2010; Quinan, 2012; Welty, 2012). In the profile

shown on Figure 8, it is possible see at least four
clayey lenses (aquitards or storage zones): Between
5.405.60 m; between 6.957.05 m; between 7.10
7.20 m; and between 8.809.00 m. All these lenses
were detected by 3 or 4 sensors.
The Figure 9 shows the interpretation of the hydrostratigraphic profile elaborated on the field for the same
test (RCPT-02). A closer look at the hydrogeological
heterogeneities on a detailed scale allowed inferring
the present of eight storage zones as shown in Figure
9. These heterogeneities are very important to CSM
elaboration and to future remediation projects on a
site, as discussed by Welty (2012) and Clearly (2009).
It can be observed that the resistivity sensor complements nicely the piezocone data (qc , fs and u)
interpreted by using a classification chart and it provides a more suited information to the scale that the
investigation of contaminated area needs.
The CSM was significantly improved in the field
based on the RCPTu test data. It was also possible to
determine the exact position of the sampling screens
of the monitoring wells that a geo-environmental site
investigation requires.
It is important to point out that the installation of
monitoring wells, although it is considered in Brazil as
the best approach to conduct a hydrogeological study,
must be preceded by a consistent CSM with the best
resolution and detail as possible. If this previous model
does not exist, and the only information is the drilling
itself for the well installation, the uncertainties associated with the installation process and the region where
this well is effectively monitoring are so questionable
that can derail any accurate decision.
CONCLUSIONS
It was observed that the interpretation of RCPTu

tests data are consistent with the results obtained by
traditional methods for geo-environmental site investigation (Direct Push sampling and Hollow Stem Auger
drilling), indicating that this test could be used in the
similar way as the traditional ones, since the achieved
results are at least equivalents.
RCPTu tests are faster and give a greater level of

detail than the traditional method for the CSM elaboration. Geological and hydrogeological heterogeneities
are accurately detected in centimeters, which does not
occur with DPT sampling and monitoring wells.
In the studied site, the RCPTu tests presented less
limitations than the traditional methods, mainly when
compared to the Direct Push soil sampling. They
reached greater depths and there were no unreliable
or non-recovered samples. The results proved to be a
more solid and more efficient tool to support a proper
site investigation and diagnosis.
It was also observed that the site investigation program achieves a better result when decision-making
is taken in real time, in the field, based on a highresolution data.
Even for the Brazilian conditions, the RCPTu tests
are a very useful tool for developing a suitable hydrogeological conceptual site model, especially in an
approach that prioritizes collection of high density
data in high-resolution site investigation, to take the
decision in real time on site.
REFERENCES
ABNT-MB 3406/1990:- Soil Penetration Cone Test In Situ.
Ahlers, R. 2012. NAPL Theory. In: North American Environmental Field Conference. Tampa, FL. March/2012.
AMS Inc. 2012. Piston Sampling System. Disponvel em
http://www.ams-samplers.com/category.cfm?CNum=151.
Acessed in 01/02/2012.
Aquino Neto, V. 2009. Roteiro para Execuo de
Investigao Detalhada e Plano de Interveno em
Postos e Sistemas Retalhistas de Combustveis. in
http://www.cetesb.sp.gov.br/tecnologia/camaras/eventos/
20_out_2009/4.pdf. Acessed in 10/01/2011
ASTM D3441/1998: Standard Test Method for Deep, Quasistatic, cone and friction-cone penetration tests of soil
Clearly, R. 2009. Por que a maioria dos sistemas de remediao falha? Anais do I Congresso Internacional do
Meio Ambiente Subterrneo (CIMAS). So Paulo-SP.
September, 2009.
CONAMA. 2009. Conselho Nacional do Meio Ambiente,
Resoluo 420/09, de 28 de dezembro 2009. Dispe
sobre critrios e valores orientadores de qualidade do
solo quanto presena de substncias qumicas e estabelece diretrizes para o gerenciamento ambiental de reas
contaminadas por essas substncias em decorrncia de
atividades antrpicas. Legislao Federal. DOU n 249,
de 30/12/2009, pg. 8184.
Einarson, M. D.; Schrimer M.; Pezeshkpour, P.; Mackay,
D.M.; Wilson, R.D. 1999. Comparison of eight innovative
site characterization tools used to investigate an MTBE
plume at Site 60, Vandenberg Air Force Base, California.
1999 Petroleum Hydrocarbons and Organic Chemicals in
Ground Water: Prevention, Detection, and Remediation
Seminar. NGWA/API. Houston, Texas. Novembro, 1999.
Elmgren, K. 1995. Slot-type pore pressure CPT-u filters.
Behaviour of different filling media. Proceedings of
CPT95 2:912.
Giacheti, H. L.; Peixoto, A. S. P.; Mondelli, G.; De Mio,
G. 2006. Resistivity Piezocone for Geo-environmental
Site Characterization of Soils: Two Case Histories from
Brazil. In: 5th International Congress on Environmental
622
Geotechnics, Cadiff. International Congress on Environmental Geotechnics, 2006.

Ide, D. M. 2009. Investigao geotcnica e estudo de um processo erosivo causado por ocupao urbana. Dissertao
(Mestrado em Geotecnia) EESC-USP, So Carlos.
Ide, D.M.; Silva, R.A.; Corghi, F.N. & Giacheti, H.L. 2010.
Erosion Process caused by an Urban Setting in Brazil. 6th
Proc.New Delhi, India. 812 November 2010. 12731278.
Killenbeck, E. 2012. Depositional Systems: Why is Geology Important? In: North American Environmental Field
Conference. Tampa, FL. Maro/2012.
Larsson, R. 1995. Use of a thin slot as filter in piezocone
tests. Proceedings of CPT95 2:3540.
Lee, D. S.; Elsworth, D.; Hryciw, R. 2008. Hydraulic Conductivity Measurement from On-the-Fly uCPT Sounding
and from VisCPT. Journal of Geotechnical and Geoenvironmental Engineering, Vol. 134, No. 12, December,
2008. 17201729.
Martin, T; St Germain, R. 2008. Direct Push Site Characterization Direct Push of NAPL with Laser-Induced
Fluorescence (LIF). In: North American Environmental
Field Conference. Tampa, FL. Janeiro/2008.
Mondelli, G., Giacheti, H. L., Boscov, M. E. G., Elis, V. R.,
and Hamada, J., 2007, Geoenvironmental Site Investigation Using Different Techniques in a Municipal Solid
Waste Disposal Site in Brazil, Environ. Geol., Vol. 52,
pp. 871887.
Mondelli, G.; De Mio, G.; Giacheti, H. L.; Howie, J. A.
(2009). The use of slot filter in piezocone tests for
site characterization of tropical soil. In: 17 th International Conference on Soil Mechanics and Geotechnical
Engineering, Alexandria, 2009. v. 2. pp. 973976.
Quinnan, J. A. 2012. Real-Time High Resolution LNAPL

and DNAPL Characterization Part 1: Philosophy and
Background. In: North American Environmental Field
Conference. Tampa, FL. March/2012.
Quinnan, J. A.; Welty, N. R. H.; Killinbeck, E. 2010. Hydrostratigraphic and permeability profiling for groundwater
remediation projects. In Proceedings of the 2nd International Symposium on Cone Penetration Testing (CPT10),
Huntington Beach, CA., 911 May 2010. Edited by
P.K. Robertson and P.W. Mayne. CPT10 Organizing
Committee. Paper 333.
So Paulo. 2009. Lei Estadual 13577/09, de 08 de julho
de 2009. Dispe sobre diretrizes e procedimentos para
a proteo da qualidade do solo e gerenciamento de reas
contaminadas, e d outras providncias correlatas. So
Paulo. 2009. Legislao Estadual.
Shinn III, J. D. 2004. Enhanced Access Penetration System Draft Final Technical Report. Applied Research
Associates, Inc. 2004.
United States Environmental Protection Agency USEPA.
2004. EPA Removes Love Canal for Superfund List. in
http://yosemite.epa.gov/opa/admpress.nsf/e34cbb5867df82a085257359003d480c/40a7e32b0bbe56018525712a00
6fca02!OpenDocument. Assessed in 01/07/2012.
Vienken,T; Leven, C.; Dietrich, P. 2006. Use of CPT and other
direct push methods for (hydro-)stratigraphic aquifer
characterization a field study. Canadian Geotechnical Journal, Volume 49, Number 2, February 2012,
pp. 197206(10).
Welty, N. R. H. 2012. Permeability Mapping: Strategies
and Methods. In: North American Environmental Field
Conference. Tampa, FL. March/2012.
623

A case study of contaminant transport and fate at a landfill

site in China
M. Sedighi & H.R. Thomas
Geoenvironmental Research Centre, School of Engineering, Cardiff University, UK
H.J. Xie, Y.M. Chen & C. Guan

College of Civil Engineering and Architecture, Zhejiang University, Hangzhou, China
ABSTRACT: This paper presents a numerical study of contaminant transport beneath a landfill site. The case
study is a landfill site located in Huainan city in Anhui province in China. The numerical study presented
includes the transport and reaction of calcium and sodium ions in the soil beneath the landfill site. The numerical
study has been carried in conjunction with a field scale experimental programme. The numerical modelling
has been based on a two-dimensional finite element analysis with an overall domain size of approximately
300 m. The results of the simulations are compared with those obtained from a field investigation reported
for the landfill site. Major chemical transport parameters are established via a calibration exercise. Numerical
simulations were then performed to predict the long term behaviour of the landfill in relation to theses two
chemicals.
INTRODUCTION
This paper presents a study of transport and fate of

chemicals in a landfill site in China which has been
constructed over a clay deposit with no major liner
system beneath the waste. The research presented
includes a numerical investigation of transport and fate
of major inorganic chemicals in the mentioned landfill
site. The results reported of field-scale experimental
investigation in this landfill site, carried out by the coauthors from Zhejian University (Xie et al.) have been
used in conjunction with the numerical modeling work
presented here.
The numerical simulations have been conducted
using the coupled thermal, hydraulic, chemical and
mechanical model; COMPASS, developed at the
Geoenvironmental Research Centre at Cardiff University (e.g. Thomas and He, 1995; Thomas et al.,
2012). The chemical reactive transport module of this
numerical model has been utilized to simulate the
transport and adsorption processes of inorganic chemicals. The results of two series of calibration tests
are reported where the numerical results are compared with the field data. These results include i) the
transport of chloride ions (Cl ), providing the major
advective and diffusive transport parameters and ii) the
transport and adsorption of sodium ions (Na+ ), which
provides the soil retardation properties for sodium
ions. By adopting the transport and adsorption parameters obtained from the calibration exercise, the long
term fate of these two ionic species was numerically
simulated.
2.1
CASE STUDY DESCRIPTION AND SITE

INVESTIGATION
Huainan landfill site
Huainan landfill is located approximately 5 km Southeast of Anhui province in China. This landfill has
been in operation for approximately 17 years. Figure
1 presents the location of the landfill site. A comprehensive site investigation on this landfill site has been
carried out by the co-authors from Zhejian University
in 2007 (Xie et al.). The study included a series of laboratory tests on samples from seven boreholes drilled
to the bedrock of the landfill site. The locations of the
boreholes are shown in figure 1 and 2.
Boreholes No.1 and No.2 are located beneath the
waste disposal area. Boreholes No.3 to No.6 are
located towards the north of the landfill site which
represents the downstream conditions. Borehole No.7
is located upstream of the landfill site and relatively
far from the disposal area.
2.2
Soil characteristics and properties
The physical properties of the underlying soil layers

have been reported from the analysis of the samples
obtained during the site investigation. The experimental study included standard soil characterisation and
determination of major geotechnical properties such
as density, water content, porosity and permeability
(NSPRC, 1995).
Figure 2 presents the soil profile and major underlying layers at the landfill site. As shown in this figure,
625
Figure 1. Huainan landfill site location and plan view of the boreholes.
Figure 2. Schematic cross-section of the Huainan landfill site and the underlying soil profile.
three layers of soil were identified which included i)

a layer of ploughed soil with a thickness of approximately 12 m, ii) a 24 m silty clay layer and iii) a clay
layer which extends to the bedrock of approximately
58 m thickness.
The average porosity of the soil layers was reported
to be about 0.4. Shallow ploughed earth has a high
water content of about 30%. The water content was
reported to decrease below 25% with depth. The
hydraulic conductivity of the ploughed earth, the
clayey silt layer, and the old clay were reported to
be 3.0 107 , 5.08.0 108 , and 3.0 109 m/s,
respectively (Xie et al.). Regional ground-water flow

underneath the landfill site was reported to be in a
south to north direction.
2.3 Distribution of chemicals
The distribution of the major chemicals present in the
underlying soils and the pore-water in landfill site has
also been determined from the soil samples collected
and the associated laboratory analysis. The average
background pore water concentration of chloride and
sodium in the soil was found to be 189 and 1155 mg/l,
626
respectively which were obtained from BH7, located

far from the waste disposal site.
The concentrations of chemicals in waste leachate
were reported from the analysis of the samples
obtained from boreholes 1 and 2. At the time of the
site investigation, the leachate level of the landfill was
found to be around 8 meters above ground. The concentrations of sodium and chloride were found to be
higher than 3000 mg/l. Samples from the boreholes
were found to have high concentrations of sodium and
chloride.
3
NUMERICAL SIMULATIONS
Two series of numerical simulations were carried out i)

to investigate the major transport mechanisms of chloride and sodium ions in the underlying soil system and
ii) to predict the long term migration of ionic species
in the landfill site. These simulations were as part of
a wider scope of work to study the fate of other contaminants present in this landfill site including heavy
metal ions and organic compounds.
The numerical simulations were carried out using
the reactive transport module of the coupled THCM
model COMPASS. This model has been extensively
verified and used to simulate the reactive transport of
chemicals in unsaturated soils (e.g. Seetharam et al.
2007; Thomas et al., 2012; Sedighi et al., 2012).
3.1 Theoretical formulation of chemical
transport
where, vl is the water velocity calculated based on

Darcys law. Dm represents the dispersion coefficient
matrix. i and Di0 are toruosity factor and the tracer
diffusion coefficient of the ith chemical component.
3.2 Numerical modeling
The governing equations describing the transport of
chemicals are numerically solved via finite element
and finite difference approaches for spatial and temporal descritization, respectively. The Galerkin weighted
residual method is adopted to discretize the flow
variables.An implicit mid-interval time-stepping algorithm is used to achieve temporal discretization. The
simulations were carried out under two-dimensional
analysis.
The domain analysed is a quadrilateral with the size
of 300 m (length) and 10 to 13 m (depth). The soil system was considered to be homogeneous and saturated.
The domain was descritized into 840 unequally sized
quadrilateral elements.
The bottom boundary of the domain was considered
to be impermeable to chemicals. The pore water pressures at upstream and downstream boundaries were
assumed to be fixed to establish a constant pore water
velocity in the domain. At surface boundary, the chemical concentrations in the area of waste disposal were
considered to be fixed and equal to the average values
of each chemical reported for BH1 and BH2 at surface.
3.3 Material properties and parameters
The theoretical formulation of the chemical module

of the model incorporates the mass conservation law
and major transport mechanisms of chemicals including advection, dispersion and diffusion. The effects
of chemical reactions have been included in a general form of multiple homogeneous and heterogeneous
geochemical reactions via linking the transport model
with a geochemical model (Sedighi et al., 2010).
Under the conditions in which the geochemical
reactions are limited to adsorption of chemicals, a
retardation factor has also been introduced to the mass
conservation equation which simplifies the analysis.
The theoretical formulation of chemical reactive
transport, considering the adsorption by retardation
factor, can be given as:
Average values of the properties of the soil layers have

been used including an average hydraulic conductivity
of the soil layers, i.e. 1.5 107 m/s and porosity equal
to 0.4. An average value for tracer diffusion coefficient
of ions in pure water at ambient temperature was also
used, i.e. Di0 = 1.0 1010 m2 /s (Lasaga, 1997).
The turtuosity factor for chemical diffusion was
adopted as a function of water content and porosity
(Millington and Quirk, 1961), given as:
where n represents the porosity of the soil.

4
where Rf represents the retardation factor, l is the volumetric water content, ci is the chemical concentration
of the ith component in pore water and t represents the
time. Ji represents the total chemical flux of the ith
chemical component, which comprises the advetive,
dispersive and diffusive flows.
The components of the chemical flux can be
given as:
MODEL CALIBRATION FOR CHEMICAL

MIGRATIONS
4.1 Transport of chloride ions

The effective transport parameters of chemicals
including the water velocity and dispersion parameters were obtained by a series of calibration exercise
of chloride migration. The results obtained from these
simulations were compared with the experimental data
reported on chloride distribution after 17 years of landfill operation reported for boreholes No.1 to No.6.
Since chloride is a non-reactive species, the migration behaviour can be interpreted as pure transport,
627
Figure 3. Distribution profile of chloride ions at the locations of different boreholes after 17 years of landfill operation
(Symbols represent the experimental data reported from the site investigation).
Figure 4. Distribution profile of sodium ions at the locations of different boreholes after 17 years of landfill operation
(Symbols represent the data reported from the site investigation).
i.e. non-reactive transport. The advective and diffusive/dispersive flow regimes can therefore be obtained
by the best fitted results.
Figure 3 presents the results of the numerical simulations of chloride migration after 17 years at the
locations of boreholes No.1 to No.6. The experimental
data for the chloride profile in each borehole reported
from the site investigation is also shown in this
figure.
The results indicate that chloride migration through
advection occurs mainly in the first 3 meters of the
underlying soil. In other words, a non-uniform water
flow regime was observed from the numerical results,
which was considerably higher in the first few meters
of the underlying soil. The higher permeability of the
first soil layer compared to the silty clay and clay layers
obviously contributes to such a non-uniform advective
regime.
As a result of the numerical simulations, the horizontal and vertical water velocity in the soil layer
was obtained to be 3.0 107 and 3.0 109 m/s,
respectively. The values of longitudinal and vertical
dispersion coefficients were found to be 15 and 0.15,
respectively.
4.2 Reactive transport of sodium ions
A similar calibration procedure migration was adopted
for the migration of the sodium ions. The water
velocity and dispersion coefficients calculated from
628
Figure 5. Prediction of distribution of chloride ions (a) and sodium ions (b) with time at ground surface at the locations of
boreholes No.3 to No.5.
model calibrations of chloride ions were utilised

to simulate the transport of sodium ions. However, since the migration of sodium also involves
adsorption effects, the retardation factor associated
with interaction of sodium with soil layers was also
considered.
Figure 4 shows the results of the simulations. The
results of the field investigation of sodium distribution in each borehole are also presented in this figure.
The retardation factor associated with sodium interaction with the soil was found to be equal to 1.5.
The retardation value obtained indicates a relative low
interaction between the soil and sodium ions. This
could be related to the fact that the first few meters
of the underlying soil, found as the dominant pathway of ions, has been reported as ploughed earth layer
which is expected to be less reactive compared with
clay soils.
PREDICTION OF LONG TERM FATE OF

CHEMICALS
Numerical simulations were then performed to predict

the long term fate of chloride and sodium ions. The
parameters associated with the water velocity, dispersion coefficients and retardation factor obtained from
the calibration tests, were utilised.
Some results achieved from the simulations are presented in figure 5. The results of chloride migration
from simulations up to 100 years are presented in figure 5a (soil surface of boreholes No.3 to No.5). The
results indicate that the chemical profiles reach steady
state after approximately 50 years.
The results of simulations for the case of sodium
migration are presented in figure 5b. Similar distribution patterns are observed. However, due to the
adsorption reaction of sodium ions with the soil, steady
629
state conditions are achieved relatively later, compared

with chloride ions, i.e. after approximately 60 years.
6
CONCLUSIONS
A numerical investigation of transport of calcium and

sodium ions in a landfill site, located in south-east
China, has been presented in this paper. The results
of a series of 2D calibration exercise of chloride
and sodium migration for the first 17 years of landfill operations were discussed and compared with the
results of the field investigation reported on this site.
Flow and adsorption parameters were established via
the calibration exercises. The results obtained from
an analysis of chloride and sodium migration indicated a non-uniform advective flow regime of ions
with depth localised in the first few meters of the soil
below ground. The results were found to be consistent
with the higher permeability of the first layer of the
soil beneath the landfill.
Simulations of prediction of the long term migration of the ions are therefore presented. Further numerical investigations are being carried out to investigate
the transport and fate of other contaminants in the
landfill case study presented.
Millington, R.J. & Quirk, J.M. 1961. Permeability of porous

solids. Transaction Faraday Society, 57:12001207.
NSPRC (National Specifications of the Peoples Republic of
China). 1995. Environmental Quality Specifications for
Soils, GB 15618-1995 (in Chinese).
Sedighi, M., Thomas, H.R. & Vardon, P.J. 2012. On the reactive transport of chemicals in unsaturated expansive soils,
The 4th International Conference on Problematic Soils,
China, 211218.
Sedighi, M., Thomas, H.R., Vardon, P.J. & Seetharam,
S.C. 2010. Advances in modelling the non-isothermal
behaviour of multicomponent chemicals in unsaturated
soils, The 6th International Congress of Environmental
Geotechnics (6ICEG), Delhi, 16191622.
Seetharam, S.C., Thomas, H.R., & Cleall, P.J. 2007. Coupled thermo-hydro-chemical-mechanical model for unsaturated soils-Numerical algorithm. International Journal
of Numerical Methods in Engineering, 70: 14801511.
Thomas, H.R. & He,Y. 1995. Analysis of coupled heat, moisture and air transfer in a deformable unsaturated soil,
Thomas, H.R.Sedighi, M. &Vardon, P.J. 2012. Diffusive reactive transport of multicomponent chemicals under coupled
thermal, hydraulic, chemical and mechanical conditions,
Journal of Geotechnical and Geological Engineering, 30:
841857.
Xie, H.J. Chen, Y.M. & Guan, C. Leachate migration in soils
beneath an uncontrolled landfill site at Huainan, Anhui,
China: a field Investigation, (under preparation).
REFERENCES
Lasaga, A.C. 1998. Kinetic Theory in the Earth Sciences.
Princeton Series in Geochemistry, Princeton University
Press.
630

Effect of geometrical sand drains on consolidation of soft contaminated

marine clays using Oedometer
M.V. Shah
Applied Mechanics, L.D. College of Engineering, Ahmedabad, Gujarat, India
A.V. Shroff
Applied Mechanics, The M.S University of Baroda, Vadodara, Gujarat, India
ABSTRACT: The object of the present study is to determine consolidation parameters of contaminated marine
kaolinitic clay with inward radial flow using geometrical shape vertical sand drains under constant magnitude
of constant pressure. Efficacy of sand drains of different diameters and shapes like circular, band, tripod and
plus shape keeping same surface area are modeled in laboratory using Oedometer and its effect to accelerate
compressibility by draining lechate is found using isochrones. The optimum diameter and geometry of sand drain
is investigated using n values (ratio of zone of influence to the diameter of drain) equal to 11.04, 16.93 and
21.71 using 254 mm size modified hydraulically pressurized Oedometer. Basic aim of this paper is to evaluate
mechanics of pore water pressure of marine clay using isochrones drawn for vertical drains. Authors theoretical
solution based on isochrones is fitted with experimental results to compute degree of consolidation and results
show a fair agreement between measured and predicted values. Circular shape sand drain can be predominately
utilized for consolidation of contaminated marine clays with dual advantage.
INTRODUCTION
Consolidation using vertical sand drains is the oldest ground improvement technique for soft soils. But
its use for waste management is very limited.India is
a country having a rich resource of natural materials. Use of this natural materials as filler material for
vertical drains need a long tern study on discharge
capacity (longitudinal and transverse permeability),
mechanical properties (tensile strength, strain characteristics),ecological properties( chemical and biological degradation, contamination) and finally durability
under required time. Now a days numbers of prefabricated vertical geodrains of band shape are available
in the market to accelerate rate of consolidation for
soft soils. But due to their corrugated shape and use
of plastic sheet its use becomes limitation for consolidation of contaminated clays with drainage as lechate.
Use of sand drains to drain out lechate and accelerate
compressibility of such marine clays can be made in
India because of easy availability of sand at cheaper
rate. Keeping in view authors decided to investigate
sand drains of different geometry and size by keeping
same surface area. This paper presents the complete
experimental model study of prefabricated vertical
sand drains of differentn values of circular, plus,
band and tripod shape indigenously developed at the
geotechnical engineering laboratory of The Maharaja
Sayajirao University of Baroda, India. For exact simulation, large size modified oedometer was used in the
study. The authors also recommend new theoretical

solution to compute degree of consolidation for any
drain material, diameter and shape of drain.
2
OBJECTIVE OF THE PAPER
The aim of the present work is to study consolidation

due to radial drainage of marine Kaolinite clay using
sand drains and to study the effect of n value (ratio of
drain dia to sample dia) and geometry on consolidation
parameters. The hydraulically pressurized Oedometer
modified by Shroff and Shah (2006) for pore pressure
measurement during radial flow is employed in the
present investigation.
Paper presents experimental model of sand drains
reinforced soft soil mass to expedite the rate of consolidation due to radial drainage. The effects of sand drains
of different n values (ratio of drain diameter to sample diameter) on consolidation characteristics (coefficient of consolidation Cr due to radial drainage) of
marine kaolinitic clay are undertaken to investigate the
settlement characteristics.
The variation of ratio of zone of influence (re ) to
the diameter of drain has been varied to access the
optimization of drain diameter for better performance
with respect to increase of quick strength on the basis
of pore pressure dissipation characteristics. Keeping
the same surface area optimum drain geometry (circular, band, plus & tripod shape) is investigated for all
drains.
631
template and guide frame connection according to

the shape of the drain. The four mandrels of circular shape, plus shape, band shape & tripod shape were
used to form the borehole. The drain was filled with
de-aired saturated sand (coefficient of permeabiliy
k= 8.67 103 cm/sec) with the aid of small diameter flexible tubing by siphoning action to cause nearly
no smear. During the installation of mandrel it was seen
that the centre of sample should match with centre of
drain. Marine kaolinitic clay of Bhavnagar city (India)
with index properties (G = 2.58, LL = 70%, PI = 34.0,
CH group) was mixed to form slurry with twice the
liquid limit using a de-aired distilled water and was
placed in Oedometer.
Figure 1. Advanced laboratory set-up of Consolidation
using modified Oedometer.
Comparison of theoretical isochrones and experimental isochrones obtained from pore pressure measurements at various radial distances using Authors
solution for the case of equal vertical strain condition.
To compare the average degree of consolidation
(Ur) for different n values and drains geometry with
authors theoretical solution.
Comparison of post shear strength achieved at the
end of consolidation for various drains.
3
EXPERIMENTAL INVESTIGATIONS
The large size oedometer of diameter 254 mm were

used for testing the remolded samples. A uniform pressure is applied by means of conventional hydraulic
pressure system on the convoluted rubber jack which
transfers uniform pressure on the soil sample placed
in the cell. Pore pressure measuring points at the three
radial distances located at 120 each with r/re distances
as r/4, r/2 and 3r/4 respectively.
3.1
Modified oedometer and pore water pressure

measurements
Modified hydraulically pressurized oedometer (Fig.1)

is machined from brass metal with three plates i.e.top,
middle and bottom. Three radial points are provided
on the bottom plate, while drainage facility is provided
at the centre of the top plate. The vertical settlement is
measured at the center of the sample by means of brass
hollow rod attached to the jack and passing out the cell
cover to suitable dial gauge. Excess pore water pressures are measured at three radial points using both
conventional Bishops system and data logger system
along with displacement and pore pressure transducers interfaced with PC. For all the oedometer tests
diameter to height ratio was kept constant.
3.2
Preparation and installation of sand drains and

clay sample
The axial hole was formed with a thin walled mandrel, having area ratio of 0.8 to 1.6 attached with
3.3 Testing program

After installation of sample and fixing top platen
along with dial gauge, displacement transducers, pore
pressure transducers, bishops apparatus are connected to their respective locations. Load increment
is applied from constant pressure system to sample
through water filed jacket, with closed position of the
drainage valve at the top plate. Pressures are applied
in the range of 20 kPa, 40 kPa, 80 kPa, 160 kPa and
320 kPa with p/p = 1.0. Each load increment is kept
constant for more then 96 hrs and secondary compression is also recorded. After completion of the
test, the sample is taken for final moisture content
measurement.
4 THEORETICAL SOLUTION, RESULTS AND
ANALYSIS
Shroff & Shah (2009, 2010) proposed a new
mathematical theory for one dimensional consolidation through radial flow which incorporates
non-homogeneity, time effects intrinsic to the soil
skeleton along with physico-chemical changes, compressibility of pore fluid and solid, variation of compressibility and permeability during consolidation,
type of the drain material, tortousity effect, kh /kv ratio
under load variation, effect of n value (drain diameter) and drainage path. In the theoretical derivation of
above theory it is reflected that the rate of outflow
of fluid, that is rate of change of fluid in the element is dependant on excess fluid pressure gradient
which will be function of the type of drain material
used in fabricating vertical geodrain being k will be
function of both porosity and void ratio e which ultimately reflects in coefficient of consolidation (Cr ) in
the derivation. Further lumped parameter () is the
ratio of (Ce /Cr ) re , where Cr is coefficient of consolidation due to radial drainage and Ce is coefficient due
to permeability and porosity which is to some extent
due to small magnitude of radial strain which helps
in redistributing loads to the surface at interface of
drain giving same settlement supporting equal strain
condition. In turn Cr will vary with the drain material reflecting in the values of (). In the present case,
the optimum lump parameter () is deduced for all the
632
Figure 3. Dissipation of excess pore water pressure vs. logt

in min for three radial points at 160 kPa (n = 11.04, psd).

in min for three radial points at 160 kPa (n = 11.04, csd).
drains on the basis of average degree of consolidation

(UR ).

in min for three radial points at 160 kPa (n = 11.04, bsd).

in min for three radial points at 160 kPa (n = 11.04, tsd).
4.1 Experimental plots and review of results

For analysis engineering stress of 160 kPa is selected
here for presentation of results. The notations used for
circular shaped sand drain is csd, band shape sand
drain as bsd, plus shape sand drain as psd and tripod shape sand drain as tsd. n = 11.04 was found to
be most efficient one whose results are shown below.
Also volume of excess pore water collected was compared with rate of settlement of sample and degree
ofsaturation.
a) Fig. 2, 3, 4 & 5 clearly indicates that csd is more
appropriate in accelerating the rate of compression
as the flow path of water particle is in radial direction giving a direct way for pore water to dissipate.
In case of psd the corners of the drain which are at
right angles to each other is hindering the flow path
because at corners the particle rearrangement at low
pressure is not comparatively fast while bulging
phenomenon is observed in middle portion of csd
but not found in psd and tsd.At the end of test it
is observed in upper portion of all sand drains.
b) Comparing settlement analysis, for 40 kPa and for

50% consolidation csd takes 70%, 96%, 50% less
time in compare to psd,tsd,bsd respectively, while
for 80% consolidation cssd takes 73.5%, 73.5%,
53% less time in compare to psd, tsd, bsd respectively, while for 160 kPa and for 50% consolidation
csd takes 87%, 145%, 61% less time in compare to
psd,tsd,bsd respectively, while for 80% consolidation csd takes 51%, 77%, 44% less time in compare
to psd,tsd,bsd respectively. (for same n value and
for same pressure)
c) Comparing pore pressure analysis, for 40 kPa and
for 50% consolidation cssd takes 34.7%, 47.3%,
26.5% less time in compare to psd,tsd,bsd for
radial point r1respectively, while csd takes 14.3%,
22.4%, 10.2% less time in compare to psd, tsd,
bsd for radial point r2 respectively, while csd
takes 15.8%, 25%, 5.3% less time in compare to
psd,tsd,bsd for radial point r3 respectively, Similarly for 80% consolidation cssd takes 20%, 33.3%,
633
Figure 6. Variation of void ratio vs. applied stress for all

geometrical drains.
Figure 8. Coefficient of horizontal permeability (Kh ) at

middle radial point r2 vs. applied stress for bsd (n = 11.04).
Figure 7. Coefficient of consolidation due to radial drainage

(Cr) vs. applied stress for bsd (n = 11.04).
8.3% less time in compare to psd,tsd,bsd for radial

point r1 respectively, while csd takes 16.7%,
29.2%, 8.3% less time in compare to psd,tsd,bsd for
radial point r2respectively, while csd takes 13.9%,
25%, 8.3% less time in compare to psd,tsd,bsd for
radial point r3 respectively.
d) Similarly for 160 kPa and for 50% consolidation
cssd takes 25%, 45%, 12.5% less time in compare
to psd, tsd, bsd for radial point r1 respectively,
while csd takes 25%, 40.6%, 10.9% less time in
compare to psd,tsd,bsd for radial point r2 respectively, while csd takes 26.3%, 31.6%, 15.8% less
time in compare to psd,tsd,bsd for radial point r3
respectively, Similarly for 80% consolidation csd
takes 26.7%, 26.7%, 17.3% less time in compare to
psd,tsd,bsd for radial point r1 respectively, while
cssd takes 20%, 33.3%, 13.3% less time in compare
to psd,tsd,bsd for radial point r2respectively,while
csd takes 19%, 28.6%, 9.5% less time in compare
to psd,tsd,bsd for radial point r3 respectively. (for
same n value and for same pressure)
e) Fig.6 shows straight-line relationship between pressure versus void ratio defining the characteristic
behavior of normally consolidated kaolinitic clays.
In case of csdvalue of Cr increased at lighter loads
thereafter remaining constant while in case of psd
(fig.7) value of Cr increased at top while declined
at 40 kPa & 80 kPa and then increased at 320 kPa.
The value of compression index varied in the range
of 0.79 to 0.35 for n = 11.04. Also the variation in
horizontal permeability at every stress shows that
coefficient of permeability remains nearly constant
for all drains.
f) From settlement analysis, n = 11.04, for 50% consolidation csd shows 33.33%, 40.71%, 48.17%
higher Cr value in compare to bsd, psd, tsd for
40kPa pressure, while for 80% consolidation csd
shows 34.67%, 41.9%, 41.44% higher Cr value in
g)
h)
634
i)
j)
compare to bsd, psd, tsd for 40kPa pressure, while

bsd shows 11% & 22% higher Cr value in compare
to psd, tsd and for 80% consolidation bsd shows
11% & 10% higher Cr value in compare to psd &
tsd respectively, while for 50% consolidation csd
shows 38%, 46%, 58% higher Cr value in compare
to bsd, psd, tsd for 160kPa pressure, while for 80%
consolidation csd shows 31%, 33%, 42% higher
Cr value in compare to bsd, psd, tsd for 160kPa
pressure, while bsd shows 13% & 33% higher Cr
value in compare to psd, tsd and for 80% consolidation bsd shows 3% & 17% higher Cr value in
compare to psd & tsd respectively..
From pore pressure analysis, n = 11.04, for 50%
consolidation, for radial point r1, csd shows 22%,
20%, 28% higher Cr value in compare to psd,
bsd and tsd for 40 kPa pressure, while n = 11.04
for 80% consolidation csd shows 13%, 7%, 21%
higher Cr value in compare to psd, bsd and tsd
for 40 kPa pressure. Similarly for radial point r2,
csd shows 8%,9%, 14% higher Cr value in compare to psd, bsd and tsd for 40 kPa pressure, while
n = 11.04 for 80% consolidation csd shows 10%,
7.5%, 18% higher Cr value in compare to psd, bsd
and tsd for 40kPa pressure. Similarly for radial
point r3, csd shows 9%, 5%, 15% higher Cr value
in compare to psd, bsd and tsd for 40kPa pressure,
while n = 11.04 for 80% consolidation csd shows
8%, 7.5%, 15% higher Cr value in compare to psd,
bsd and tsd for 40kPa pressure.
From fig. 8 using settlement analysis, it is observed
that for csd, the Kh50 of n = 11.04 is 11%, 24%,
39% higher in compared to bsd, psd,tsd at 160kPa
pressure, while Kh50 of bsd is 15% and 31%
higher in compared to psd and tsd at 160 kPa
pressure.
From pore pressure analysis, comparing csd values of various shapes of sand drains it is observed
that n equal to 11.04 for 160 kPa for Kh1 50 is
17.6%, 27.8%, 6.8% higher in compare to
bsd, psd, tsd while Kh1 80 is 16.1%, 22.5%,
22.3% higher. Similarly Kh2 50 is 17.6%,
29.6%, 10.1% higher in compare to bsd, psd, tsd
while Kh2 80 is 22.5%, 26.9%, 16.2% higher.
While Kh3 50 is 16.4%, 24.2%, 17.7% higher
in compare to bsd, psd, tsd while Kh3 80 is 23.5%,
31.3%, 20.5% higher respectively.
There is a considerable gain in shear strength of soft
mass at the end of consolidation. The shear strength
increased averagely from 18 kPa to 165 kPa for all
Figure 11. Comparison of Theo & Exp average degree of

consolidation (Ur) at 160 kPa for csd & bsd (n = 11.04).
Figure 9. Experimental isochrones with respect to time for
csd at 160 kPa (n = 11.04).
Figure 10. Comparison of Theoretical & Experimental

isochrones of tsd at 160 kPa (n = 11.04).
the four shapes of drains. The highest gain was

observed in circular shape and plus shape sand
drain.
k) Fig. 9 &10 shows the nature of dissipation and rate
of dissipation of excess pore pressure at three radial
points in form of isochrones with respect to time.
Also is observed that experimental isochrones w.r.t
time factor (Tr) fits well with theoretical isochrones
for all pressures and for all n values.
l) Fig. 11 represents the comparison of average degree
of consolidation (Ur) versus time factor (Tr) for
csd & bsd at applied pressure of 160kPa.The
lumped parameter which is best fitted with experimental results for csd is equal to +0.1 while for
psd it is +0.25, for bsd it is +0.3, for tsd it is
0.4. The above results clearly indicate that authors
theory fits best with experimental results for any
pressures for any n value and for any shape of
drain.
CONCLUSIONS
From the results and analysis it can be concluded that

dissipation and development of pore water pressure
is influenced by geometrical shape of sand drain. It
can strongly be concluded that sand drains can be
used to consolidate contaminated clays as it drain outs
lechate without much clogging sand drain. Shape of
sand drain helps to judge permittivity of marine clays
under various pressures as excess pore water drains
out in form of lechate there is change in fabric structure (orientation of particle) due to addition of cations
and anions like sulfate, chloride, iron, aluminum, zinc,
ammonia, etc which either help in compressibility or
may be responsible for hardening the clay. Moreover
contaminated marine clays like kaolin mixed with oil
or heavy metal toxic compounds, sand drains having
good filter capacity will reduce further contamination
of such organic compounds and if necessary certain
sand drain can also acts a semi-rigid pile to increase
bearing capacity of such soils covering large area.
Highest time is taken by tsd while lowest time is
taken by csd.Out of three n values, n value equal
to 11.04 is most optimum one. The time required to
achieve 80% consolidation increased with increase in
pressure at middle radial point r2 for all four sand
drains. More the complexity in geometry of drains less
the advantage of surface area and more the effect of
tortousity. This can be very well seen for plus drain
which shows same magnitude of pore water dissipation in compare to circular drain by taking more
time. Horizontal permeability decrease with increase
of pressure under lighter load and then it is more or
less constant for higher intensity loading for radial
drainage. Less decrement in horizontal permeability
is observed in case of csd while in case of psd
and bsd permeability is decreasing with increase of
pressure.
The four shapes of sand drains investigated so far,
circular shape remains to be most economical as it
accelerates dissipation rate of excess pore water pressure at lesser time. Also it is concluded that band shape
is more efficient than plus shape as it takes lesser time
and also less disturbance is observed during its installation. Sand is a best filler material whose maximum
635
advantage can be taken where it is easily and cheaply

available for fabricating sand drains. Further analysis of above results and particularly the deformation
pattern obtained for various geometrical sand drains,
the three major inferences were noted, they are what
type of particle arrangement is found in case of csd
under each loading and what type of strain it exerts
on clay sample. Simultaneously what is the position
of pore and what is the effect of strain on pore and
is there any diffusion of pores because of gases produce due to contamination of liquids. Second noting
is regarding effect of lateral confinement to clay sample as such conditions rarely occur in field and due to
lateral confinement of soil sample what type of geometrical changes occur in pore (volumetric form) and
is there any impact of it on vertical compressibility,
third noting is regarding inward slope(gradient) created under final load on soil sample and almost the
verticality of drain did not remain but there due to
radial flow of pore water( turbid water) the corners of
psd and tsd were rounded to much extent, which
initially might have created hindrance for free dissipation of pore water pressure. Overall sand drains can
efficiently be used for dissipating contaminated pore
water and accelerating rate of compressibility and thus
increasing bearing capacity or shear strength of soil.
By performing long duration tests on marine clays,
authors have observed that coagulation of lechate and
microbial activity of such clays can be well studied and
modeled by using large size oedometer under steady
state condition. It was observed that degree of consolidation computed using authors theoretical solution
lies below the Barrons equal strain theory(1948) for
higher degrees of consolidation but overall experimental results obtained from pore pressure measurements
fits with theoretical results with appropriate lump
parameter . Authors equal strain theory fits best with
experimental results for all the shape of drains and
at any pressures. The lump parameter which fits
appropriately with geometry of drain, material of drain
and diameter of drain gives clear clue to design engineer for selection of most optimum drain as per site
conditions.
ACKNOWLEDGEMENTS
The authors are highly thankful to the Prof.(Dr)
Ambika Misra, Dean, Faculty of Technology & Engineering, Prof. (Dr) I.I. Pandya Head, AMD, Prof.
(Dr) D.L. Shah, P.G.Incharge, AMD, The Maharaja
Sayajirao University of Baroda, Vadodara and Prof.
M.N. Patel, Dean, L.D. College of Engineering
and Prof. (Dr) V.S. Purani, Head, AMD, LDCE,
Ahmedabad, India for providing all necessary research
facilities.
REFERENCES
Barron, R. 1948, Consolidation of fine grained soils by drain
wells, Trans A.S.C.E., Vol. 113, pp. 718754.
Hansbo, S. 1960, Consolidation of clay, with special reference
to influence of vertical sand drains, Proc. Swedish Geotech Instt. No. 18, pp. 160.
Hansbo, S. 1981, Consolidation of Fine-grained soils by
prefabricated drains, Proc. of 10th Int. Conf. on Soil
Mechanics and Foundations Engineering, Stockholm,
Vol-3, pp. 677682.
Rowe P.W and Barden. L. (1966), A new consolidation cell,
Geotechnique, 16:2:162.
Shroff, A.V. and Shah, M.V. 2006, Effect of ratio of influence
zone and type of vertical drain on consolidation of soft clay
due to radial flow, Proc. Fourth International conference
on Soft- soil Engineering, Canadian Geotechnical Society,
University of Alberta, Canada, Vol-I, pp. 765773.
Shah, M.V. and Shroff, A.V. (2010), Soil-Structure interaction of soft clay using prefabricated Vertical geodrains
under seismic stresses, Proc. Fifth International conference on Recent Advances in Geotechnical Earthquake
Engineering and Soil Dynamics, Missouri University of
Science and Technology, International Society of Soil
Mechanics and Geotechnical Engineering, San Diego,
CA, USA, Vol-II, paper no. 5.53a.
Shah, M.V. and Shroff, A.V. (2010), Non-Linear Theory
of Consolidation through Radial Drainage for Soft Soils
using Vertical Drains, Proc. of14th Danube-European
Conf. on Geotechnical Engg, Bratislava, Slovakia, Vol-I.
Taylor, D, W, (1948), Fundamentals of Soil Mechanics, John
Wiley and Sons, New York.
636

Geostatistical analysis of groundwater nitrates distribution

in the Plaine dAlsace
R.-L. Spacagna
University of Cassino and Southern Lazio, Department of Civil and Mechanical Engineering, Cassino, Italy
C. de Fouquet
Ecole Nationale Suprieure des Mines de Paris, Goscience-Gostatistique, Fontainebleau, France
G. Russo
University of Cassino and Southern Lazio, Department of Civil and Mechanical Engineering, Cassino, Italy
ABSTRACT: The groundwater of the Plaine dAlsace (France) is one of the largest water reservoir in Europe.
It is highly vulnerable to pollutants coming from anthropical activities (both industrial and agricultural). The
monitoring activities carried on by the local national agencies on the groundwater pollution concurred to form a
very large public database on different aspects of water quality. In this study the nitrate concentration is examined
in detail for two years for which dense sampling was available, namely 1997 and 2003. A bivariate variographical
analysis is performed, and monovariate and bivariate estimations are carried out. The evolution of the spatial
distribution is estimated by means of co-kriging, allowing the variance evaluation of the estimation error. The
risk of groundwater quality degradation over the whole area is finally analysed by evaluating the probability of
exceeding a given threshold of pollutant concentration at a given time.
2 THE GROUNDWATER OF THE PLAINE

DALSACE
INTRODUCTION
The cartographic representation of a pollutant distribution in a large area is a key tool in the safeguard
of the hydraulic resources. From the monitoring data,
a correct description of the pollution can be reached
taking account of the regional character of the relevant variables (Matheron 1972) in the geostatistical
framework (de Fouquet 2006). The groundwater of the
Plaine dAlsace in the North East of France is highly
vulnerable to pollutants, due to the intense anthropic
activities (agricultural, industrial, etc.) on the whole
area (BRGM 2006). The use of the hydraulic resource
for domestic and industrial purposes makes its safeguard essential for the sustainable development of the
region. The present study examines the distribution of
nitrates and its evolution for the years 1997 and 2003.
DAgostino et al. (1998) analysed the distribution of
nitrates in the groundwater of Lucca plain (Italy) with
reference to three different periods of the same year.
In this study the concentration of nitrates measured in
the 1997 and 2003 were analyzed with reference to
the same time interval, namely August and September
of each year. A multivariate approach allowed the representation of the evolution of nitrates concentration
in the reference time interval. The risk of groundwater quality degradation over the whole area is finally
analysed by evaluating the conditional probability of
exceeding a given threshold of pollutant concentration
at a given time.
The Alsace region, in the north-east of France, is characterized by a narrow depression, the Rhine rift, of
40 km in width and about 300 km in length, separating
the Black Forest and the Vosges, whose lower part is
the plain dAlsace.
The geological formation present in the plain
(Fig. 1) is mainly composed of recent alluvial soils
mainly due to the action of the Rhine and its tributaries.
These soils have a well-graded size distributions, ranging from gravels down to clays. The average thickness
of the alluvial layer is approximately 70 m, with local
higher depths (down to 200 m) in some areas, and a
relevant reduction nearby the lateral sides of the plain.
The shallow groundwater belongs to the hydrogeological basin of Rhine-Meuse, one of the largest in
France, and is characterized by an high vulnerability mainly due to the intense use of the soil. For this
reason, an extended monitoring activity guarantees
the safeguard of the water resources. The Association
pour la protection de la nappe phratique de la Plaine
dAlsace (APRONA) was established on 1995 under
the initiative of the Regional Council of Alsace, the
Water Agency Rhine-Meuse and the Prefecture of the
Alsace region, with the aim of managing the regional
monitoring network with reference to the piezometric level and the water quality. Actually more than
ten years of qualitative and quantitative observations
637
are available to the public through the national portal of access to information on groundwater resources
ADES.
On the basis of piezometric levels collected over
a network of about 200 piezometers installed in
the whole area, APRONA studied the hydrodynamic
behaviour of the groundwater on a regional scale
(APRONA 1999). In Figure 2 the main flow directions
are represented. The north and north-east directions
are congruent with the main development of the river
bed; alteration of those directions have been evidenced
along the west side of the area corresponding to water
flow amount coming from the Vosges formation.
The Plain dAlsace is particularly exposed to risks
of water contamination by nitrates, primarily due to
the use of nitrogen fertilizers for the agricultural activities carried on in the region. The Direction Regionale
de lEnvironnement (DIREN) Alsace follows closely
the phenomenon of diffuse pollution of groundwater
of Plain dAlsace collecting data on nitrates concentrations. The data, available through the web portal
of ADES, have been considered in this study for two
years, namely 1997 and 2003, in order to analyse
Figure 1. Alsace main geological units: 1) Rhine rift, 2)

Kaiserstuhl Volcan, 3) sub-Vosges and sub-Schwarzwald
hills, 4) Mesozoic sedimentary formation, 5) Paleozoic
basement (Sell et al., 1998).
and represent the evolution of the nitrates pollution

phenomena in the whole area.
3
EVOLUTION OF NITRATES POLLUTION
3.1 Data sets

Nitrate concentration was detected at 702 measuring
points in the year 1997, and 605 in 2003. The measuring points common to both data sets are 585. As well
as in terms of spatial location, the data can also differ
in terms of the reference period of the measurements.
The observation period was therefore reduced to a time
interval in which concentration measurements were
made both for the 1997 data set and the 2003 data
set. This period lasts 7 weeks, namely from the 18th
of August to the 3rd of October. After the selection,
the measuring points for the two sets are respectively
696 for 1997 year and 601 for 2003, while data having the same spatial location and reference period are
574. During the reference time interval the piezometric
level of the groundwater was not subjected to relevant
variations. In Table 1 the relevant statistical parameters of the 1997 data set and 2003 data set have been
reported with reference to the entire sets and to common data, while in the Table 2 the same data have
been specified for the selected 7 weeks measurements.
No relevant differences were found between statistical
parameters of sets referred to different time intervals.
Figure 2. Groundwater flow directions (APRONA, 1999).
638
Nitrates concentration varied in a wide interval,

with values higher than the limit value of 50 mg/l set
by the French regulations; nevertheless average concentrations were lesser than the reference threshold.
3.2
Explorative data analysis
The measurement points are not uniformly distributed

over the studied area, as evidenced in the Figure 3,
Table 1. Relevant statistical parameters of 1997 and 2003
data set.
1997
N of point
Min. [mg/l]
Max. [mg/l]
Mean [mg/l]
Dev. std. [mg/l]
Coeff. of var.
2003
Full
Common
Full
Common
702
0.17
295
28.15
27.50
0.98
585
0.17
295
28.36
27.63
0.97
605
0.50
217
27.28
24.60
0.90
585
0.50
217
27.30
24.39
0.89
Table 2. Relevant statistical parameters of 1997 and 2003

data set (7 weeks observations).
and the histograms markedly asymmetric, with high

frequencies in the low concentrations range and low
frequencies in the high range. Frequencies higher
than the threshold value have been represented with
black; the maps show that points characterized by high
levels of concentrations are located along the west
side of the study area, corresponding to the boundary between the Vosges mountains and the Plaine
dAlsace, where the groundwater is very shallow and
the viticulture activity is particularly intensive.
The good correlation ( = 0.86) between data of
1997 and 2003 sets is shown by Figure 4. The very high
and high values (greater than 100 mg/l and between
50 mg and 100 mg/l) are mainly located under the
bisector line, as well as the curve representing the
mean values for each class of observation. It could suggest a decrease in the average concentration of nitrates
moving from 1997 data to 2003 data. The same observations apply for the 1997 and 2003 histograms, with a
evident reduction of 2003 frequencies in the high values range. In the low concentrations range, the 2003
frequencies seems to be lower than 1997 indicating a
slight increase, that locally could have even exceeded
the threshold value.
3.3 Variografic analysis
1997
N of point
Min. [mg/l]
Max. [mg/l]
Mean [mg/l]
Dev. std. [mg/l]
Coeff. of var.
2003
Full
Common
Full
Common
696
0,17
295
28,06
23,00
0,98
574
0,17
217
27,28
24,60
0,90
601
0,50
217
27,36
24,67
0,90
574
0,50
217
27,30
24,39
0,89
The simple variogram is calculated as half the average,

for distances h, of the standard deviation between two
values z at distance of h:
where N (h) is the number of distinct pairs of values at

distance h.
Figure 3. Explorative analysis for 1997 and 2003 data sets.
639
Figure 4. Joint analysis of the two data sets. a) data location map, b) correlation between nitrate concentrations in 1997 and
in 2003, histograms of nitrate concentrations in c) 1997 and d) 2003.
Directional variographic analysis on different

observation scale highlighted anisotropy of the concentration of nitrates at great distances with the
preference direction 15 (to the North), whereas no
anisotropy was detected at small distances (up to
5 km). The experimental variogram was then calculated with reference to the directions 15 , 60 , 105
and 150 setting a 4 km step.
According to the results of the explorative analysis,
the 1997 and 2003 data sets show a good correlation.
In order to investigate their spatial correlation a multivariate analysis has been carried on by calculating the
cross variogram following the subsequent definition:
Table 3. Model parameters of 1997 and 2003 simple

variograms.
Model
Sill
Range
97
Nugget effect
Isotropic spherical
Anisotropic spherical
166
360
280
03
Nugget effect
Isotropic spherical
255
140
248
5 km
N15: 65 km
N105: 15 km
5 km
N15: 65 km
N105: 15 km
Table 4.
97,03
The simple variograms for 1997 data (97 ) and 2003

data (03 ) and the cross variogram (97,03 ) were interpreted by means of a co-regionalisation linear model,
considering the phenomenon as a sum or superposition
of independent processes occurring at different spatial
scales. The parameters of the theoretical simple variogram models and the cross variogram (97,03 ) model
parameters have been reported respectively in Table 3
and in Table 4. In Figure 5 the variograms 97 , 03 and
the cross variogram 97,03 have been represented.
Model parameters of cross variogram.

Model
Sill
Range
Nugget effect
Isotropic spherical
162
220
242
5 km
N15: 65 km
N105: 15 km
3.4 Estimations
The estimate of concentration values over not sampled
areas was performed by ordinary kriging (Chils &
Delfiner 1999), defined as the best linear estimator
without bias. Weights assigned depend on the spatial relationship between the measured values, directly
640
Figure 5. Simple variograms 97 , 03 and the cross variogram 97,03 .
Figure 7. Co-kriging estimation results on 2003 data set: a)

nitrate concentration map (mg/l); b) standard deviation of the
estimation error map (mg/l).
Figure 8. Comparison between ordinary kriging and cokriging standard deviations.

Figure 6. Ordinary kriging estimation results on 1997 data
set: a) nitrate concentration map (mg/l); b) standard deviation
of the estimation error map (mg/l).
relating to the model of the variogram assumed in the

analysis.
The 1997 nitrates concentrations have been calculated by ordinary kriging on the 97 variogram model
and represented in Figure 6a. The standard deviation
of the estimation error, shown in Figure 6b, shows
systematically a small value around the monitoring
point and an increase with increasing distance, depending on the model of spatial variability assumed. The
2003 nitrates concentrations represented in Figure 7a
have been calculated by cokriging based on the simple
variogram models (97 , 03 ) and the cross variogram
(97,03 ), as defined before. Figure 7b shows the standard deviation of the estimation error. If compared
with the estimate obtained by ordinary kriging on the
03 variogram model, the cokriging estimate of 2003

nitrate concentrations seems to be more precise. In
order to highlight this point, the estimation error standard deviations of both ordinary kriging and cokriging,
evaluated at nodes of the geographical grid, have been
represented in Figure 8, showing systematically the
better performance of the cokriging method.
In order to estimate the evolution of nitrates concentrations during the selected time interval, the difference D between 1997 and 2003 concentrations at
the same measure points was introduced. The simple
and cross variograms (namely D and 97,D ) of the difference D and the 1997 concentration are derived from
the concentration bivariate model :
641
4 THRESHOLD-EXCEEDING PROBABILITY
MAPS
and
The models parameters are summarized in Table 5.

As the cokriging ensure the consistency between the
estimations of different variables, it is equivalent to
cokrige the two concentrations and to calculate their
difference or directly to cokrige the difference from
the two concentrations data (Rivoirard, 2003).
Figures 9a and 9b show respectively the nitrate concentrations in 1997 estimated by means of cokriging
and those estimated by cokriging of D and 1997; in Figure 9c the standard deviation of cokriging estimation
error is presented.
Table 5. Model parameters of simple and cross variograms
D and 97,D .
Model
Sill
Range
Nugget effect
Isotropic spherical
97
60
44
5 km
N15: 65 km
N105: 15 km
97,D
Nugget effect
Isotropic spherical
4
140
38
5 km
N15: 65 km
N105: 15 km
The comparison of a relevant parameter with a reference value or threshold is a recurring task in the
analysis of environmental problems. In this study, the
evaluation of underground water pollution has been
performed with respect to the reference value set by
French regulations.
Starting from a finite number of nitrate concentration measurements, the probability that the concentration exceeds a threshold value can be evaluated
also where no measures are available. To this end, the
probability distribution of the variable under examination should be known. As shown by exploratory
data analysis, the considered regionalized variable Z
(nitrate concentration) does not follow a specific probability law. It is therefore necessary to transform the
variable Z in another random function Y following a
Gaussian probability law. This transformation is called
anamorphosis:
If the nitrate concentration is known in x , then the

transformed variable is obtained as:
The probability of exceeding threshold has been

calculated following non-linear methods (Rivoirard,
1991). The probability that the variable Z exceeds the
Figure 9. Estimation of nitrate concentrations and their evolution between 1997 and 2003: a) 1997 cokriging; b) cokriging
of the difference 20031997; c) standard deviation of estimation errors of the cokriged difference.
642
Figure 10. Variograms of the Y 03 .
limit value for nitrate concentration is given by the

following relationship (conditional probability):
where s is the threshold value and 1 (s) is the

reference value after anamorphosis.
The nitrate concentration Z is deduced directly
from the transformed random variable Y. In the case
of ordinary kriging Z is obtained according to the
relation:
whereYKO is the estimate variable by ordinary kriging,

KO is the standard deviation of the ordinary kriging
estimation error, N is a gaussian random function not
dependent on YKO . The correction factor w is linked
to the Lagrange parameter m, in turn depending on
YKO (de Fouquet, 2006). The nitrate concentration Z
becomes:
Figure 11. Threshold-exceeding probability map of 2003

nitrate concentrations.
of the experimental variograms showed the absence

of anisotropy at small scale (h < 5 km), whereas an
anisotropy along the main direction N10 was detected
03
at great scale. The estimation of the YKO
was then performed by means of ordinary kriging, obtaining the
variance of the estimation error 2KO and the Lagrange
parameter .
The threshold-exceeding probability map has been
represented in Figure 11 for the area under study with
reference to 2003. As expected, the higher probability of high nitrate concentrations is distributed along
the western boundary of the plain dAlsace, while the
areas around the river Rhine are potentially far from
the risk of pollution.
5
and the exceeding threshold probability follows:
From the anamorphosis of the variable Z 03 (nitrate

concentration for the year 2003) the random variable
Y 03 was obtained, and the Gaussian behaviour of the
transformed variable was assessed. The experimental variograms of Y 03 were then calculated at small
scale and large scale (Fig. 10). Theoretical modelling
CONCLUDING REMARKS
The distribution of nitrates concentration in the Plane

dAlsace groundwater has been analyzed by geostatistical analysis with reference to the 1997 and 2003 years
monitoring data. Based on the correlation between
the data collected in the same monitoring points, a
reduction of the average high concentrations of nitrate
was evidenced for 2003 data. The spatial structure of
the nitrates concentrations was investigated by means
of multivariate analysis variography, highlighting the
anisotropy closely related to the prevailing direction
of groundwater flow. The estimation of nitrates concentrations allowed the representation of pollutants
643
distribution in the studied area. The accuracy of the

estimate was improved by using jointly two independent data set (namely 1997 and 2003 nitrates concentration measurements). Potentials of the followed
approach in the pollution analysis have been evidenced
by the threshold-exceeding probability maps, which
represent a relevant provisional tools in the safeguard
of groundwater resources from pollution phenomena.
REFERENCES
APRONA, 1999. Carte des directions dcoulement de
la nappe de la Plaine dAlsace, septembre 1999,
www.aprona.net.
BRGM. 2006. Aquifres et eaux souterraine de France.
Collection Scientifique et Technique.
Chils, J.P., Delfiner, P., 1999. Geostatistics: Modeling
Spatial Uncertainty. Wiley, New York.
DAgostino V., Greene E.A., Passarella G., Vurro M., 1998.

Spatial and temporal study of nitrate concentration in
groundwater by means of coregionalization. Environmental Geology 36 (34) December 1998. Springer-Verlag.
de Fouquet C., 2006. La modlisation gostatistique des
milieux anthropiss. Memoire des Sciences de la Terre N
200613. Academie de Paris, Universit Pierre et Marie
Curie.
Matheron G., 1972. The theory of regionalized variable and
its applications. Les Cahiers du Centre de Morphologie
Mathmatique. Ecole des Mines de Paris, Fontainebleau.
Rivoirard, J. 1991. Introduction au krigeage disjonctif et la
gostatistique non linaire. Centre de Gostatistique de
Fontainebleau Ecole des Mines de Paris.
Rivoirard J., 2003. Course on multivariate geostatitics.
C-173. Centre de Gostatistique de Fontainebleau. Ecole
des Mines de Paris.
Sell Y., Berchtold JP., Callot H., Hoff M., Gall J.C., Walter
J.M.. 1998. LAlsace et les Vosges Gologie, milieux
naturels, flore et faune. Edition Delachaux et Niestl.
644
Degradation extraction and inerting systems

for the reclamation of polluted sites

The use of lime in the stabilization and solidification of lateritic soil

contaminated with tannery effluent
A.O. Eberemu
Department of Civil Engineering, Ahmadu Bello University, Zaria
ABSTRACT: An assessment of the use of lime in immobilizing and stabilizing some ions contained in tannery
effluent contaminated lateritic soil was carried out. The contaminated soil was treated with up to 10% lime.
Batch equilibrium studies was used to check its efficacy. Anions such as Cl and SO2
4 and cations such as
Na+ , Cr3+ , and Fe2+ were observed. The cations monitored in the lime treated contaminated soils were found to
decrease with increasing lime treatment, which was due to the immobilization of the metal ions by adsorption
and encapsulation in the Calcium Silicate Hydrate gel formed. The pH was alkaline and increased with lime
treatment which favors the immobilization process. Results also reveal that there was reduced SO2
4 content
and desorption of Cl ions with increased lime treatment. The index properties of the soils also improved with
increased lime treatment. Optimum results were obtained at 10% lime treatment suggesting its suitability.
INTRODUCTION
The degradation of our environment due to discharge

of tannery effluent is a serious problem in several countries around the world. The situation is even worse
in developing countries like Nigeria where little or
no treatment is carried out before discharge into the
environment particularly with the small artisan tanneries. Inspite of many steps taken to improve the
situation waste waters generated by tannery industries
continue to increase. Municipalities and industries are
presently confronted with the problem of remediating
soils contaminated with tannery effluent particularly
with the present awareness and increased monitoring
by regulatory environmental agencies.
Stabilization/Solidification (S/S) is a widely used
method for remediation of contaminated sites particularly those contaminated with hazardous waste.
(Yukselem et al., 2001). In S/S application, the treatment immobilizes the contaminant in the media by
mixing it with a binding reagent which prevents its
migration to the receptor. Solidification and stabilization is an effective method for treating soils containing asbestos, metals, radioactive materials, corrosive
and cyanide compounds, inorganics, and semi-volatile
organics etc. Solidification removes free liquids,
reduces hazardous constituent mobility by lowering
waste permeability, minimizing constituent leachability, and providing stability for handling, transport, and
disposal.
The chrome tanning process results in the release of
toxic metals especially chromium III passing through
the waste water which are not easily eliminated by
ordinary treatment processes (Franco et al., 2005).
Tannery effluent are highly characterized by high
salinity, high organic loading and specific pollutants

such as chromium (Colak et al., 2007).
Some of the chemicals used in tanning process
include lime, sodium bi-carbonate, common salt,
sodium sulphate, chromium sulphate, fat liquors, vegetable oils and dyes (Tuduwada et al., 2007). Tannery
effluent collected from industrial outlets has also been
found to contain high concentration of Total dissolved
solid (TDS), chromium, chloride, ammonium nitrates,
and sulphate (Calheiros et al., 2008). Besides these,
chemicals such as zinc chloride, mercury chloride, and
formaldehyde are also used as disinfectants; sodium
chloride (NaCl) in curing and as bleaching powder.
Sodium sulphate, chromium and sodium fluoride are
normally used to prevent putrefaction; sodium sulphate, ammonium chloride, borax and hydrochloric
acid are used in de-liming. Furthermore, basic and
acidic dyes are also used in leather finishing (Amita
et al., 2005). Hence the tannery waste is always characterized by a strong colour reddish dull brown, high
BOD, high pH and high dissolved solids. The other
major chemical in the effluent from the waste of the
tannery industry are sulphide and chromium.
These chemicals pollute the soil affecting both its
geotechnical and geo-chemical properties and ultimately leach to contaminate ground water. The need
to reuse contaminated soils and reduce the harmful
effects of the contaminants present in the effluents
from contaminating ground water through leaching by
immobilizing the chemical species and stabilizing the
soil is the justification for this work; ultimately making the environment safer. The aim of this work is to
investigate the potency of up to 10% lime in the immobilization and stabilization of contaminants present in
a lateritic soil that is contaminated with tanery effluent.
647
Table 1.
Physical properties of natural soil.
Table 2.
Oxide composition of hydrated lime used.
Property
Description/Values
Oxide
Content (%)
Liquid limit (%)

Plastic limit (%)
Plasticity Index
Natural Moisture Content (%)
Standard Proctor
OMC (%)
MDD (Mg/m3 )
Standard Proctor Compacted
State Cohesion (kN/m2)
Angle of internal friction ()
Specific gravity (Gs)
USCS
AASHTO
Group index
Percent pass. BS sieve No. 200
pH
Colour
42.8
18.8
24.0
13.8
AL2 O3
SiO2
SO3
CaO
TiO2
V2 O5
Fe2 O3
NiO2
CuO
Y2 O3
BaO
MoO3
CdO
4.4
8.56
0.03
53.4
1.32
0.03
1.12
0.008
0.13
0.1
0.06
0.0
1.7
2
2.1
20.0
1.66
57.0
17.0
2.57
CL
A-7-6
11
58.75
6.1
Reddish brown
Table 3.

Soil
The soil used for this study is a naturally reddish brown

lateritic soil with inclusion of white mottles obtained
by method of disturbing sampling from a borrow pit
in Shika Area of Zaria (Latitude 11 15 N and Longitude 7 45 E) Nigeria. A study of the geological and
soil maps of Nigeria after Akintola (1982) and Areola
(1982) respectively show that the sample taken belong
to a group of ferruginous tropical soil derived from
acid, igneous and metamorphic rock. Previous studies
of soil in this area have been shown to contain koalinite
as the dominant clay mineral (Osinubi 1998). The soil
is classified as A-7-6 (10) according to AASHTO soil
classification system (AASHTO, 1986) and low plasticity clay (CL) according to the unified soil classification system (ASTM 1992). The in-situ properties of
the soil from the borrow show that the natural moisture
content of the soil as at time of sampling is 13.8%, with
water table at about 12.0 m below ground level but rises
to the ground surface at the peak of the rainy season.
The physical properties of the soil are shown in Table 1.
Soil samples were transported to the laboratory in
bags, air dried and passed through 4.76 mm aperture
sieve to reduce the amount of clods before usage.
2.2
Lime
The hydrated lime used was obtained from theAhmadu

Bello University Water treatment plant. The oxide
composition of the hydrated lime is shown in Table 2.
2.3
Properties of effluent used.
Property
Unit (mg/l)
Sodium ion, Na+

Chromium, Cr+
Iron, Fe2+
Calcium, Ca2+
pH
Total dissolved solid, TDS
Total suspended solid, TSS
Sulphide
Sulphate
Chloride
Chemical Oxygen Demand
Biological Oxygen Demand
Electrical Conductivity, EC
9.0178 l
0.2440
1.2052
1.4353
8.03*
780
220
0.05
53
2410
448
0.314
0.15**
**units = S m1 ; *no unit.
Spectrometer in National Research Institute for Chemical Technology (NARICT) in Zaria, Kaduna State
Nigeria.
2.4 Sample preparation
The contaminated soil used was simulated in the laboratory. The soil was thoroughly mixed with 50%
moisture content tannery effluent by dry weight of soil
and stored in polythene bags. 50% moisture content
was used since this was just above the liquid limit. The
soil was thoroughly mixed with the effluent to ensure
good reaction and contact between the particles of the
soil and individual molecules of the effluent. Thoroughly mixed soils were then placed in a polyethene
bag to ensure no loss or gain of moisture and particle
within the enclosure. The mixture was left to cure for
5days in the polyethene bag after which air drying was
carried out before usage.
Effluent
Tannery waste effluent was collected from the National

Institute of Leather Research Technology, Zaria,
Nigeria. The chemical composition of the effluent used
is given in Table 3. Test to determine the property of the
effluent was carried out using the Atomic Absorption
2.5 Methods
2.5.1 Soil index properties
All tests were carried out on the contaminated soil and
samples of the contaminated soil treated with 2, 4,
6, 8, and 10% lime content, respectively. The index
648
properties of the soils were characterized using sieve

analysis and Atterberg limits; which were carried out
in accordance with specifications outlined in BS 1377
(BSI, 1990).
2.5.2 pH
The pH of the contaminated and treated contaminated
soils was carried out at the Institute of Agricultural
Research (IAR) of Ahmadu Bello University Zaria
using a table bench pH machine.
2.5.3 Batch equilibrium test
Batch equilibrium test was used to monitor the leached
amount of the various contaminant species in the solution as against the Toxicity Characteristic Leaching
Procedure (TCLP) machine for sake of availability. After the test some of the chemical species are
adsorbed and what is not adsorbed is leached out into
the solution. The batch equilibrium test was carried out
in the Institute of Agricultural Research (IAR) A.B.U,
Zaria.
Batch equilibrium test, was carried out on the
contaminated soil lime mixtures using stepped lime
treatment (0%, 2%, 4%, 6%, 8% and 10%) dry weight
of soil. A 1:4 soil-solution ratio by weight of dry soil
was used (Batch-type, 1987). A known mass of dry soil
(25 g) for each of the contaminated soil-lime combinations was placed in a 500 ml plastic container. A
known volume (100 cm3 ) of distilled water/deionized
water was introduced into the soil in the container and
sealed. The contaminated soil lime-distilled water
mixture was placed in a table shaker (HS 500 Janke &
Kunker; Ika-Werk) clamped down and allowed to mix
for a period of 48 hours at a speed of 30 hub/mins.
At the end of the experiment, the solid and solution
phases were separated using a filter paper where the
filtrate, otherwise known as leachate was collected in
a conical flask in which the filter paper was wrapped
inside. The liquid content in the conical flask were then
decanted off into a test tube bottle of 70ml in capacity.
Each soil sample leached about 50-60ml leachate out
of the soil sample. After which the leachate was taken
for Atomic Absorption Spectrophometer analysis and
titrimetric analysis to determine the concentration of
the cations and anions, respectively that were leached
out; and hence to ascertain if the lime had any potential to immobilize some of the ions present in the
contaminated soil.
3
3.1
Figure 1. Particles size plot of the natural Soil.
Figure 2. Variation of Atterberg limits with lime content.
associated with soils having a greater quantity of clay

particles or particles having higher surface activity.
The Atterberg limits of the effluent contaminated
soil lime mixtures are shown in Fig. 2.
The Atterberg limit results show improved geotechnical properties of the soil with a decrease in liquid
limit (LL), an increase in plastic limit (PL) and a resulting decrease in plasticity index (PI). The liquid limit
decreased from 4329%, the plastic limit increased
from 2126%, thus resulting in a decrease in plasticity index (PI) values from 183%. Similar results were
obtained by (Ferguson 1993; Nicholson and Kashyap
1993; Cokca 2001; Eberemu et al., 2011) when fly ash
and rice husk ash was used, respectively. This trend
is also consistent with the findings of Muntohar and
Hantoro (2000) where the liquid limit reduces with
increasing lime treatment.
These are due to physio-chemical changes on the
clay sized particles caused by cation exchange between
the lime and clay sized fractions of the effluent contaminated soil; forming compounds possessing cementations properties calcium silicate hydrates with soil
particles leading to a more stable material.
Index properties
The grain size distribution of the natural uncontaminated soil show that 58.75% passed through sieve with
aperture size 0.075 mm. Particle size plot is shown in
Fig. 1.
Atterberg limits are indices of the quantity of claysized particles and their mineralogical composition.
Typically, higher liquid limits and plasticity indices are
3.2 pH
pH affects the behaviour of clay in suspensions; a
low pH enhances a positive edge to negative surface interaction, often leading to flocculation from
suspension. Stable suspensions or dispersions of clay
particles often require pH conditions. The type of
adsorbed cation is of great importance in influencing
649
Figure 4. Variation of measured chloride ion leached with

lime treatment.
Figure 3. Variation of pH with increasing lime treatment.
the behaviour of the material in suspension and the

nature of the fabric in sediments. Monovalent cations
promote deflocculation, whereas clay suspension ordinarily flocculates in the presence of di and trivalent
cations.A high pH environment of the soil lime-water
mixtures normally promotes the dissociation of calcium cations supplied by the hydrated lime to replace
the cations normally present on the surface of the clay
mineral (NLA, 2001). The role of pH in influencing
inter-particle repulsions and the existence of positive
edge charges in low pH environment is the single most
important factor controlling the adsorption and desorption of cations during the batch equilibrium test
(Mitchell, 1976).
The pH of the natural soil was found to increase
after contamination; it was initially 6.1 before contamination with tannery effluent and increased to 9.2
after contamination. The variation of the pH values
with lime treatment is shown in Fig 3.
With increased lime treatment the pH of the contaminated soil increased, for up to 6% lime treatment before decreasing slightly between 8% and 10%
treatment. This could probably be due to hydroxides
which were created by the dissolution of oxides in
the contaminated soil lime mixtures used, also most
importantly it should be noted that lime contains large
amounts of calcium carbonate in it which chemically
is a base that assists in neutralizing acidic soils and
further increasing its alkalinity and thus its pH value.
Osinubi and Amadi (2003) suggested that increased
pH value made the surface to become more negatively
charged, and an increase in pH will create a condition
for the establishment of attractive electrostatic forces
that can enhance the adsorption of cationic species.
3.3 Anions
Anions monitored include chloride and sulphate ions.
3.3.1 Chloride (Cl )
The plot of variation of leached chloride concentration
in the contaminated soil with lime treatment is shown
in Fig. 4. The chloride ion concentration increased
from 1.7 to 28.1 mg/l for up to 10% lime treatment.
It was observed that for the varying lime treatments,
the leached chloride ion (Cl ) in the solution was
Figure 5. Variation of leached sulphate ions with lime

treatment.
found to increase with increasing percentages of lime

treatment.
Elevated concentration of Cl ion in solution is due
to the desorption of Cl within the contaminated soil
with increased lime treatment. This could probably be
due to the higher pH environment with increased lime
treatment. Desorption of chlorine was also observed
by Barone et al. (1990) in their studies on chlorine diffusion in intact shale. Desorption (leaching of chlorine
ions) was also recorded by Tukka (2012) in her work
on stabilization/solidification potential of rice husk
ash treated lateritic soil contaminated with tannery
effluent.
3.3.2 Sulphate ions (SO2
4 )
The variation of leached sulphate ions (SO2
4 ) with
lime treatment is shown in Fig. 5. Sulphate ions
decreased from 7.73 to 0.97 mg/l with increased lime
treatment. The result showed reduced leached SO2
4
ions with increased lime treatment. This could probably be due to the formation of calcium silicate hydrate
gels (CSH) due to ion exchange between the lime and
contaminated soil particles. When the pore water of
the soil comes in contact with lime, hydration occurs,
leading to the formation of cementitious products,
CSH gel. These cementitious products bind the sulphate ions within the soil grains thus hardening and
forming a hardened skeleton matrix, which encapsulates the ions; hence immobilizing it and leading to
650
Figure 6. Variation of leached chromium ions with lime

treatment.
Figure 7. Variation of leached sodium ions with lime

treatment.
reduced amounts of sulphate being leached into solution. Furthermore, the hydration of lime subsequently
enhances the rise of pH value of pore water, which
is caused by dissociation of the bivalent calcium ion
(Ca++ ) from the hydrated lime during hydrolysis. The
dissociation of Ca++ eventually causes flocculation
and agglomeration of soil particles into larger sized
grains.
3.4
Cations
The cations monitored include chromium (Cr3+ ),

sodium (Na+ ) and Iron (Fe2+ ). The variation of
chromium ion is shown in Fig. 6. Chromium ion
decreased from 0.2 to 0.045 mg/l with increased lime
treatment. The variation of sodium ions is shown
in Fig. 7. Sodium ions decreased from 1.303 to
0.752 mg/l with increased lime treatment. The variation of Iron ions is shown in Fig. 8. Iron ions decreased
from 1.39 to 0.544 mg/l with increased lime treatment.
Generally, the cations monitored in the lime treated
contaminated soils were found to decrease with
increasing percentage of lime treatment; this phenomenon can be accorded to various reasons. The
formation of CSH gels which is the imminent substance formed in soil lime reaction (Eades and Grim,
1960); a general breakdown of clay particles due to
chemical attack by lime, whereby calcium ions initially adsorbed on to clay particle surfaces by cation
exchange, slowly diffuse into the clay structure by further exchange gradually breaking apart the structural
Figure 8. Variation of leached Iron ions with lime treatment.
layers of the clay, and reacting to produce fine foils

and filaments of calcium (aluminium) silicate hydrate
gel (Wild et al., 1986).
Several researchers (Taylor, 1981; Stewart and
Bailey, 1983; Zhao-Qi and Young, 1984; Wild et al.,
1986) have shown that the hydration of tri-calcium
silicate which is present in soil-lime mixes can incorporate cations such as Na+ , Fe2+ , Cr3+ into the CSH
structure. This gel helps in immobilizing cations by
absorbing the ions into its gel. Ivey et al. (1990) proposed that chromium would substitute for silicon and
chemically incorporate into CSH gel. Bone and Saker
(1994) also indicated that metals are more likely to be
physically encapsulated in CSH gel. Lin et al. (1997)
showed that in addition to substitution of chromate ion
for silicon there is also hydration of Tri calcium silicate
(C3 S).
Another reason for this occurrence could be owed
to the incorporation of cations into a crystalline matrix
system, this is possible by adsorption into the binder
soil matrix; this occurs at a very high alkaline medium
that brings about enhanced surface adsorption in mineral edge sites. Furthermore, it could also be due to the
pH of the system; the fraction of clay sized particles
in the system and the redox precipitation of insoluble
substances (Osinubi and Amadi, 2003). It should also
be noted that the fraction of clay sized particles in the
soil have a pronounced effect on the stabilization and
solidification potential of the binder (lime).
CONCLUSION
The effect of up to 10% lime content on the stabilization/solidification potential of lateritic soil contaminated with tannery effluent was investigated. Atterberg
limit results show improved geotechnical properties
of the soil with a decrease in liquid limit, an increase
in plastic limit and a resulting decrease in plasticity
index; producing a more stable material.
The tannery effluent contained a variety of cations
and anions, of which Fe2+ , Na+ , Cr3+ , Cl and
SO2
were investigated. Batch equilibrium studies
4
were used to monitor the leached ions. The pH of the
leached solution was found to generally increase with
651
lime treatment; it increased for up to 6% lime treatment before decreasing slightly between 8% and 10%
treatment.
The chloride ion concentration increased from 1.7
to 28.1 mg/l for up to 10% lime treatment. Increased
concentration of Cl ion in solution is due to the
desorption of Cl ions within the contaminated soil
with increased lime treatment. Sulphate ions decreased
The result showed reduced leached SO2
4 ions with
increased lime treatment. This probably could be due
to the formation of calcium silicate hydrate gels (CSH)
due to ion exchange between the lime and contaminated soil particles. These cementitious products bind
the sulphate ions within the soil grains thus hardening and forming a matrix, which encapsulates the ions;
hence immobilizing it and leading to reduced amounts
of sulphate being leached into solution.
For the cations monitored, chromium ion decreased
from 0.2 to 0.045 mg/l with increased lime treatment; sodium ions decreased from 1.303 to 0.752 mg/l
with increased lime treatment; iron ions decreased
Generally, the leached cations in the lime treated contaminated soils were found to decrease with increasing
percentage of lime treatment. This phenomenon can
be accorded to the formation of CSH gels which is
the imminent substance formed in soil lime reaction;
that helps in immobilizing the cations by absorbing
the ions into its gel, hence reducing the amounts of
leached cations into the environment.
These findings show that up to 10% lime can be
used to remediate lateritic soil contaminated with tannery effluent by stabilizing and immobilizing some
of the ionic species present in the soil. This can be a
first step in treatment before disposing of such soils in
designated landfills.
ACKNOWLEDGEMENT
I wish to acknowledge the contribution of Mr Victor
Udale Okpanachi of Civil Engineering Department of
Ahmadu Bello University Zaria, in making this work
a success.
REFERENCES
AASHTO (1986) Standard specification for transportation
materials and methods of sampling and testing, 14th Ed.,
Washington, D.C.
Akintola FA (1982) Geology and Geomorphology, Nigeria
in Maps, R. M Barbour, Editor, Hodder and Stoughton,
London.
Amita DA, Verma S, Tare V, Bose D. (2005) In: Oxidation of
Cr (III) in tannery sludge for Cr (VI): Field observations
and theoretical assessment. J Hazd. Mat. B 121: 215222.
Areola O (1982) Soils. Nigeria in Maps, R. M, Barbour, Editor
Hodder and Stoughton, London U.K.
ASTM (1992) Annual book of ASTM standards, Vol. 04.08,
Philadephia.
Barone FS, Rowe RK, Quigley RM (1990) Laboratory determination of chloride coefficient in an intact shale. Can.
Geotech Journal. Vol. 27, 177184.
Batch-type (1987) Batch-type adsorption procedures for
estimating soil attenuation of chemicals, EPA/530-SW006: Draft Tech. Resources Document for Public Comment (NTIS-146-155), US Environmental Protection
Agency, Cincinnati, Ohio.
Bone D. and Sarkar, S.L. (1994) The Present State of the Art
of Immobilisation of Hazardous Heavy Metals in CementBased Materials. In: Advances in Cement and Concrete
(Eds. M.W.Grutzeck and S.L. Sarkar). ASCE Durham,
New Hampshire, July 2429, 1994, ASCE, New York,
merican Society of Civil Engineers, pp 481495.
British Standard Institute(BSI). (1990). Methods of testing
soils for civil engineering purposes. BS 1377, London.
Calheiros, C.S.C., Rangel, A.O.S.S., Castro, P.M.L., 2008.
The effects of tannery wastewater on the development
of different plant species and chromium accumulation in
hragmitesaustralis. Arch. Environ. Con. Toxicol. 55 (3),
404414.
Cokca, E. (2001). Use of class C fly ashes for the stabilization
of an expansive soil. J. Geotech. Geoenviron Engr., 127(7),
568573.
Colak H, Soylak M, Turkoglu O (2005). Determination of
trace metal content of herbal and fruit teas produced and
marketed from Turkey. Trace Elem. Elec. 22: 192195.
Eades, J. L., and Grim, R. E. (1960). Reaction of hydrated
lime with pure clay minerals in soil stabilization: Highway
Research Board Bulletin 262 (N.A.S.-N.R.C. Publ. 771),
pp. 5163.
Eberemu,A. O.,Amadi,A.A. and Sule, J. (2011) Desiccation
effect on compacted tropical clay treated with rice husk
ash ASCE Geotechnical Special Publication (GSP) No.
211. ASCE. In Advances in Geotechnical Engineering.
Volume 2 of 6. Editors: Jie Han Daniel A. Alzamora ISBN:
978-1-61782-594-1. pp. 11921201.
Ferguson, G. (1993). Use of self-cementing fly ashes as a
soil stabilization agent. Fly ash for soil improvement,
Geotechnical special publication No. 36, ASCE, New
York, 114.
Franco, A.r., Calheiros, C.S.C., Pacheco, C.C., Marco, P.,
Manaia, C.M. and Castro, P.M.L. (2005), Isolation and
characterization of polymeric Galloyl- Ester- degrading bacteria from a tannery discharge place, Microbial
Ecology, 50: 550556.
Ivey, D.B., Heinmann, R.B., Neuwirth, M., Shumborski, D.,
Conrad, Mikula, R.J. and Lam, W.W. (1990) Electron
Microscopy of Heavy Metal Waste in Cement Matrices.
Journal Materials Science Letters 25, pp. 50555062.
Lin, T.T., Lin, C.F., Wei, W.C. and Lo, S-L. (1993) Mechanisms of Metal Stabilisation in Cementitious Matrices: Interactions of Tricalcium Aluminate and Copper
Oxide/Hydroxide. Environmental Science andTechnology
28, pp. 13121318.
Mitchell J.K. (1976) Fundamental of Soil Behaviour. John
Wiley and Sons, Inc. pp. 422.
Muntohar A.S. and Hantoro, G. (2000), Influence of the rice
husk ash and lime on engineering properties of clayey subgrade, Electronic Journal of Geotechnical Engineering
5, Paper#019. www.ejge.com
Nicholson, P.G, Kashyap,V. (1993). Flyash stabilization of
tropical Hawaiian soils. Fly ash for soil improvement,
Geotechnical special publication No. 36, ASCE, New
York, 1529.
NLA (2001). Using lime for soil stabilization and modification; a proven solution. National Lime Association.
http://www.lime.org
652
Osinubi, K.J., (1998b). Permeability of lime-treated lateritic

soil. J. Transp. Engrg. ASCE, 124 (5), pp. 465469.
Osinubi, K. J. andAmadi,A.A. (2003)Attenuation of cationic
contaminants in municipal solid waste leachate by fly ash
stabilized laterite. Nigerian Journal of Engineering, Vol.
11. No.2, pp. 8287.
Stewart, H. R. and Bailey, J. E. (1983). Microstructural studies of the hydration products of three tricalcium silicate
polymorphs. J. Mat. Sci. 18, 36863694.
Taylor, H. F. W. (1981) Modern chemistry of cements.
Chemistry and Industry 9, 620625.
Tudunwada, I. Y. Essiet, E.U. and Mohammed, S.G. (2007).
The effects of Tannery sludge on heavy metals concentration in cereals on small-holder farms in Kano, Nigeria
Not Bot Hort Cluj 35(2), 5560.
Tukka, D. D., (2012). The use of rice husk ash in the stabilization and solidification of a lateritic soil contaminated
with tannery effluent. Unpublished B. Eng. Project Dept.
of Civil Engineering. Ahmadu Bello University, Zaria.
Nigeria. Pp.53.
Wild, S. Arabi, M. and Leng-Ward, G. (1986). Soil-lime
reaction and microstructural development at elevated
temperatures. Clay Minerals. 21, 279292.
Yukselen, M. A. and Alpaslan B., (2001). Leaching of metals
from soil contaminated by mining activities. J. Hazard
Mater 87 (13) pp. 289300.
Zhao-Qi Wu and Young, J. F. (1984). The hydration of tricalcium silicate in the presence of colloidal silica. J. Mat.
Sci. 19, 34773486.
653

Preliminary results on the stabilization of dredged sediments

from the Port of Taranto
A. Federico, A. Murianni, E. Miccoli, C. Vitone & M. Nobile
Technical University of Bari, Bari, Italy
G. Intern
Port Authority, Taranto, Italy
ABSTRACT: This paper presents the preliminary results of an experimental research into the stabilization
with lime and/or cement of clayey sediments dredged from the Port of Taranto (Italy). The testing program is
still on-going as part of a cooperation with the Port Authority of Taranto. The aim of the research is to study the
effect of artificial stabilization on physical properties and behavior of the soil in order to choose the best option
for re-use. In this paper, the physical and mechanical properties of the natural sediments are briefly described
and their plasticity properties are compared with those of the treated clays. For each stabilizing agent, the effect
of treatment is analyzed and the experimental data interpreted by means of a general plasticity model. The model
was developed to provide a method for evaluating the effect on clay plasticity of the treatment with a given
additive after a given curing time.
INTRODUCTION
Research into the behavior of artificially stabilized

soils has become increasingly important and is developing along various lines in order to find the most
effective ways to improve specific aspects of the
behavior of natural soils depending on the particular
engineering problem to be resolved. Since the early
1960s, soils have been treated with cementing agents
to improve their mechanical behavior (e.g. Eades &
Grim 1966, Croft 1967, Bell 1976, Broms & Broman
1977, Mitchell 1981, Terashi & Tenaka 1981, Bell
1989, Locat et al. 1990, 1996, Fratalocchi et al. 1996,
Smith 2005, Dalla Rosa et al. 2008, Cecconi et al.
2011, Russo & Croce 2011). However, there has been
little research into the prediction of changes in the plasticity properties and mechanical behavior of dredged
materials (Terashi & Tanaka 1981, Rajasekaran & Rao
1997, Tang et al. 2001, Tremblay et al. 2001, Burgos
et al. 2007, Chiu et al. 2008, Grubb et al. 2010, Huang
et al. 2011). Although more problematic than clayey
soils with low water content (i.e. close to the plastic
limit), stabilization-solidification of dredged slurries
is an interesting method of recovery, enabling them to
be used for either embankments or earthworks.
This is the case of clayey sediments from the Port
of Taranto, in the south of Italy, where a new urban
development plan aims to improve and expand the harbor and update the infra-structure of the Port itself.
This would make the Port, which is located in a
strategic area in the Mediterranean sea (Fig. 1), more
internationally competitive.
Figure 1. The Port of Taranto in the Mediterranean Sea.
The plan involves extensive deepening of the seabed

in the Port, which will require the disposal and/or
re-use of about 19 million cubic meters of dredged
submarine sediments. The many potential options for
their re-use need to be assessed with reference to their
cost and to the quality and real volume of the sediments themselves. It follows that the assessment must
also take account of the specific uses to which the
recovered sediments will be put, varying from land
reclamation in the area of the Port of Taranto to the capping of contaminated submarine sediments in order to
mitigate the environmental emergency affecting that
area.
This paper presents the preliminary results of ongoing experimental research jointly carried out by the
Port of Taranto Authority and the Technical University
of Bari.
655
The research aims to explore the effects of artificial

stabilization on the physical properties and behavior
of dredged natural clayey sediments in order to choose
the best option for their re-use.
To this purpose, ongoing laboratory testing at the
Technical University of Bari is seeking to determine the effect of their stabilization with lime and/or
cement.
The paper first reports some of the data concerning
the geotechnical characterization of the natural submarine clayey sediments, acquired from both laboratory
and in-situ testing carried out before the present investigation (Port Authority of Taranto 2004, Cotecchia
2005, SGS SELC 2009).
The physical properties of the natural clay are
then compared with those of treated clay. Specifically, the effect of treatment with lime, cement and
a cement/lime mixture on the high plasticity (CH)
clay from the Port of Taranto is analyzed and the
experimental data interpreted by means of a general
plasticity model. The model was developed to provide
a method for evaluating the effect on clay plasticity
of the treatment with a given additive after a given
curing time.
Figure 2. The area of the Port of Taranto. Site 1: First in-situ

campaign; Site 2: Sampling site for the present study.
NATURAL SUBMARINE CLAYEY

SEDIMENTS
Figure 2 shows the area of the Port of Taranto and

the location of the two sampling sites. During the first
in-situ investigation (Site 1 in Figs. 23), 14 continuous boreholes were drilled to depths of up to 30 m
below the sea-bed and 74 undisturbed samples were
taken from depths varying from the sea-bed up to
30 m below. In addition, 14 cone penetration tests
(CPTUs) were carried out at depths of up to about
15 m below the seabed. The submarine sediments in
the area are mainly made up of so-called Subapennine
Clays, which, being grey-blue to grey-green, are commonly known as blue clays. This Lower Pleistocene
formation is in transgression over either the Gravina
Calcarenite or the Altamura Limestone and its thickness, in the area under study, may be more than 100 m
(Mastronuzzi et al. 1999)
The data from the clay samples are fairly homogeneous in terms of both physical and mechanical
properties so that the material may be considered as
belonging to a single lithotype. The clay is characterized by high carbonate content (mainly calcite and,
to a lesser degree, dolomite), and therefore it can be
considered a marly clay.
The clay minerals are mainly represented by illite
followed by chlorite, montmorillonite and kaolinite.
The organic content of the clay is low and was found
to decrease with depth from 2.2% for the shallow samples (i.e. up to 6 m below the seabed) to 1% for the
deeper ones.
According to the literature (Keller 1982, Booth &
Dahl 1986, Huang et al. 2012), below a minimum content threshold of about 34% organic matter does not
Figure 3. Site 1 in Figure 2. Location of the boreholes and

the CPTU tests carried out by the Port Authority of Taranto
(Cotecchia 2005, SGS SELC 2009).
affect soil properties. It follows that the clay under

study can be considered inorganic clay, since the quantity of organic matter is too small to interfere during the
cementing process when a cementing agent is added
to the soil.
Figures 4a and b show the vertical profiles of the
clay and sand fractions respectively. The vertical profile of the clay fraction (CF) with respect to the sea
level is S-shaped, ranging between 25 and 45%, with
the minimum recorded at about 10 meters below sea
level. The sand fraction (SF) varies from 5% to about
42% (Fig. 4b) and, in particular, the SF of samples
taken from boreholes S4-S8 in the south-eastern part
of the area in Figure 3 is about or below 20%, whereas
in samples taken from the inner part (S10-S14) it is
higher than 20%. Figure 5 shows the granulometric
spindle of soils from the Port of Taranto: the upper
part of the spindle is represented by the deeper samples
656
Figure 6. Left: Vertical profile of the void ratio (e); right:

consistency index (CI) of submarine sediments from the Port
of Taranto (Cotecchia 2005, SGS SELC 2009).
Figure 4. Vertical profiles of a) the clay fraction and b) the

sand fraction of marine sediments from the Port of Taranto
(Cotecchia 2005, SGS SELC 2009).
Figure 7. Casagrande plasticity chart of marine sediments

from the Port ofTaranto (Cotecchia 2005, SGS SELC 2009).
Figure 5. Granulometric spindle of submarine sediments
from the port of Taranto (from Cotecchia 2005, SGS SELC
2009, modified).
S13-C5 and S4-C6, which appear to be characterized

by higher clay and lower silt content (about CF = 53%
and MF = 45%) than the shallower samples (about
CF = 27% and MF = 57%).
The void ratio (e) and the consistency index (CI) in
Figure 6 do not show a clear trend with depth.
Apart from some scattering in the data recorded in
the first few meters, the void ratio appears to vary
between 0.60.8 and the CI of most of the samples
is either below or about one. The Casagrande plasticity chart (Fig. 7) shows that the clay samples can
be classified as being of medium to high plasticity (CL-CH). Specifically, for almost all the samples,
the plasticity index (PI) ranges between 20 and 30%
and the liquid limit (wL ) between 40% and 60%.
The most plastic samples (PI = 3540%) are in general
the shallowest ones, taken from boreholes S4-S6 in the
south-eastern part of the area in Figure 3 (SF about or
below 20%).The high carbonate content (2035%) of
these clays may be ascribed to carbonatic cementation.
A significant number of in-situ CPTUs were performed on the submarine sediments to determine
undrained shear strength. The vertical profiles of both
the tip resistance (qc ) and the pore pressure (u) are

shown in Figure 8. Analysis of the results of the CPTUs
seems to suggest the presence of a crust in the top
four meters of sediment, within which the soil is harder
than at higher depths.
At higher depths (from four meters below the
sea-bed), the CPTUs show a fairly uniform vertical
profile, as is typically observed for overconsolidated
clays, and an average undrained strength (cu ) of about
200 kPa can be derived.
Direct shear tests carried out on undisturbed specimens show that the failure envelopes are characterized
by the following shear strength parameter values:
c = 045 kPa and = 28 10 (Port Authority of
Taranto 2004, Cotecchia 2005, SGS SELC 2009).
STABILISED SEDIMENTS
3.1 Materials and methods

The soil used for the present study was taken from
two submarine boreholes, G02 and G03, drilled in Site
2 (see Fig. 2) within ten meters below the sea-bed.
Table 1 shows the average soil composition and index
properties. The soil is a clayey silt of medium-high
plasticity that can be classified as CH and considered
657
Figure 9. Specimens prepared after mixing the slurry with

the additives and placed in a temperature-controlled room to
be cured in sea water.
Figure 8. Vertical profiles of the tip resistance (qc ) and the

pore pressure (u) from the CPTU tests. The exact position
of the tests within the area of the Port of Taranto is shown in
Figure 3 (Cotecchia 2005, SGS SELC 2009).
Table 1.
clay.
w
%
Composition and index properties of untreated
CF MF SF wL PI e
% % % % %
18.8 33 52
s
kN/m3 kN/m3
15 55 29 0.566 0.88 19.86
26.19
representative of the shallower submarine sediments

in the external part of Site 1.
The soil was cut into small pieces, dried at room
temperature, and then finely pulverized into a powder
(100% passing through Sieve No. 40).
The clay powder was mixed with sea water
(w = 10.05 kN/m3 ) to form a slurry with a water
content (w0 ) of 83%, i.e. about 1.5 times the liquid
limit (wL ).
The samples were prepared by using different additives, namely CEM II/A-S 42.5 Portland cement (C),
quicklime (L) and a mixture of 75% cement and 25%
quicklime (C/L). Each agent was added to the slurry in
varying percentages (2%, 4%, 8%) with respect to the
dry weight of the soil. Several specimens were prepared after thorough mixing of the slurry with the
additives in a mixer. Thereafter the specimens were
placed in a temperature-controlled room to be cured
in sea water (see Fig. 9). The physical properties of the
stabilized soils were determined after different curing
times, i.e. 2, 7 and 28 days.
Figure 10a. Plasticity paths of clay treated with cement.

The symbol size is proportional to the curing time (i.e. small
symbols AX: 2 days, medium symbols BX: 7 days and large
symbols CX: 28 days).
3.2 Physical characteristics of the clays

Figures 10a, b and c show the positions on the
Casagrande plasticity chart (Casagrande 1948) of
clays treated with the various additives, i.e. lime,
cement, and cement/lime. In all the Figures, the plasticity of the untreated clay is also shown (symbol +).
Moreover, for each percentage of additive, the plasticity points are shown in the Figures by a symbol
whose size is proportional to the curing time. If the
percentage of additive used is X, then points AX, BX,
CX on the chart represent the plasticity of the soil
after curing times of 2, 7 and 28 days respectively. The
Figures also show line A and the bisecting line
(dashed line).
658
Figure 10b. Plasticity paths of clay treated with lime. The

symbol size is proportional to the curing time (i.e. small
symbols AX: 2 days, medium symbols BX: 7 days and large
symbols CX: 28 days).
in the Figure, they are numbered I to VIIII following

a counter-clockwise sequence.
For example, path S-EI , which is directed upwards
and to the right with an inclination of less than 45 to
the horizontal, is included in Sector I and is therefore
characterized by an increase in PI which is due to an
increase in both the liquid limit (wL ) and the plastic
limit (wP ). In contrast, a path lying in Sector II is also
characterized by an increase in wL and PI but, in this
case, the increase in plasticity is due to both an increase
in wL and a reduction in wP .
Moreover, the model shows that, for each path
in Sector I, wL increases more rapidly than wP (i.e.
wL > wP ). The Figure also clearly shows that:
i) horizontal paths are characterized by PI = 0 so
that wL = wP ; ii) along vertical paths wL = 0 and
PI is equal and opposite to wP ; iii) wP is equal to
zero 0 along the (dashed) bisecting lines, is less than
0 in Sectors II to V and is greater than 0 in Sectors VI
to VIII and Sector I; iv) it is only in Sectors II and VI
that wL and wP can have opposite sign. For further
details about the plasticity model see also Federico
et al. (in prep.).
3.4 Plasticity paths of clay treated with cement
Figure 10c. Plasticity paths of clay treated with the

cement/lime mixture. The symbol size is proportional to the
curing time (i.e. small symbols AX: 2 days, medium symbols
BX: 7 days and large symbols CX: 28 days).
3.3 A general plasticity model

Figure 11 shows an original plasticity model that was
used to interpret the plasticity paths followed by the
treated clays shown in Figures 10ac.
In the model in Figure 11 points S and E represent the starting and end points respectively of each
possible path within the plasticity chart. The dashed
lines are bisecting lines that divide each quadrant into
two sectors, so that eight sectors may be identified;
Figure 10a shows the plasticity paths followed by the

clay when treated with 2, 4 and 8% cement content.
The first general consideration that can be made is
that, regardless of the percentage of cement used, the
three points Ai , Bi and Ci appear to be almost aligned.
The Figure also clearly shows that more than 2%
cement has to be added to bring the plasticity of the
clay below line A. Two days of curing produced an
increase in wL in all cases. However, for 2% and 4%
cement content, paths U-A2 and U-A4 lie in Sector
I of the plasticity model, whereas with 8% cement
content path U-A8 lies in Sector VIIII. This means
that, unlike what happens in the other two cases, the
plastic limit increases more than the liquid limit (i.e.
wP > wL ), causing a reduction in the plasticity
index that is observed only when the clay is treated
with 8% cement.
Curing for more than two days causes all the plasticity points to move downwards and to the left so that
regardless of cement content, a reduction in both wL
and PI is recorded. However, for cement content of 2
and 4%, the plasticity follows paths that lie in Sector
V (i.e. PI falls because wL decreases more rapidly than
wP ), whereas for 8% cement content, the reduction in
PI is due to both falling wL and a slight increase in wP
(Sector VI).
It follows that the most substantial effect on clay
plasticity of adding more than 4% cement is to induce
a greater permanent increase in wP . After 28 days
of curing, plasticity paths U-C2, U-C4 and U-C8
show that:
659
the effects on the clays physical properties induced

by adding 2% cement have been almost entirely
cancelled out so that point U is almost coincident
with point C2;
Figure 11. A plasticity model for the interpretation of the plasticity paths.
the wL of clay with 4% cement content has fallen

back to a level slightly below that of untreated clay,
but a permanent reduction in wP means the clay has
lower plasticity (path U-C4);
point C8 shows that, in the case of clay with 8%
cement content, neither wL nor wP (which exhibited
an even greater increase) have returned to the values
of untreated clay, so that the treated clay has both
higher wL and lower PI (path U-C8).
3.5 Plasticity paths of clay treated with lime
Figure 10b shows the plasticity paths followed by clay
with 2, 4 and 8% lime content. As in the case of clay
treated with cement, points AX, BX and CX appear to
be aligned and plasticity clearly moves below line A
only with an additive content of more than 2%. Two
days of curing does not produce significant changes
in the PI of clay treated with either 2 or 4% lime
(paths U-A2 and U-A4), due to similar increases in
wL and wP (i.e. a horizontal path between sectors VIII
and I in Fig. 11). With 8% lime, path U-A8 shows
that wP increases as in the previous cases, but wL
does not change significantly, meaning that an overall reduction of PI is recorded, corresponding to a
path within Sector VII in Figure 11. Thanks to the
pattern established in the first two days of curing
(i.e. the inhibited increase in wL accompanied by a

more substantial increase in wP ), path U-C8 lies within
Sector VIII.
Comparison of the three U-CX paths in Figure 10b
also shows that the highest clay plasticity is obtained
within 28 days with just 4% of lime added. Higher
quantities of additive both delay the increase in wL
and promote that of wP , leading to a net reduction in
clay plasticity.
What is particularly interesting is that the effect
of curing time on the results of treatment with lime
appears to be the opposite of what is observed when
adding cement. When lime is used, the longer the curing time the further the process develops, so that the
plasticity points continuously shift further away from
point U as time passes. In terms of plasticity paths, this
means that, since both wL and wP are found to increase
from AX to CX irrespective of lime content, the plasticity points move upwards and to the right following
paths that are all included in Sector I (paths A2-C2,
A4-C4 and A8-C8 in Fig. 10b).
3.6
Plasticity paths of clay treated with

cement/lime mixture
Figure 10c shows the plasticity paths followed by clay

with 2, 4 and 8% cement/lime mixture content.
660
As in the previous cases, for all three additive percentages, the plasticity points AX, BX and CX are
aligned. However, in this case, the plasticity points
representative of a clay with 2% cement/lime mixture
lie below line A after 28 days of curing.
Clay treated with 2% follows a path (A2-C2) mainly
characterized by falling plasticity (a downward path
in the model in Fig. 11) and an invariance in wL .
In contrast, clay with 4% additive follows a path
(A4-C4) which is characterized by the invariance
of the plasticity index (a horizontal path in the model
in Fig. 11) and a significant increase in wL .
As observed in the previous cases, with 8% additive
content a large increase in wP , even higher than the
increase in wL , is recorded after two days of curing
(path U-A8: Sector VIII).
This reduces the plasticity of the clay to well below
that of the untreated clay. With longer curing times, wL
increases more than wP , so that path A8-C8 lies within
Sector I; the final plasticity point (C8) has slightly
lower PI and higher wL than C2 and C4.
4
CONCLUDING REMARKS AND FUTURE

RESEARCH
This paper presents the preliminary results of experimental research which aims to study the effect of
artificial stabilization on the physical properties and
behavior of dredged natural clayey sediments. In particular, the research focused on stabilization with lime
and/or cement of clayey sediments dredged from the
Port of Taranto in the South of Italy.
The effect of the treatment on high plasticity (CH)
clay from the Port ofTaranto is analyzed and the experimental data are interpreted by means of a new general
plasticity model. In this respect, it appears that the
model can be helpful in order to develop a method to
estimate the effect on clay plasticity of treatment with
a given additive for a given curing time.
From this research, the following conclusions can
be drawn: i) regardless of the type and percentage
of cementing agent, clays treated for different curing
times (up to 28 days) follow aligned paths on the plasticity chart; ii) if either lime or cement is used, more
than 2% additive must be used in order to bring the
plasticity of the treated clay below line A; iii) in general, in order to reduce PI in just two days of curing,
at least 8% additive has to be used; iv) the effect of
curing time on treatment with lime appears to be the
opposite of what is observed when adding cement.
Focusing on point iv), it should be noted that when
lime is used, the curing time makes the process develop
further, so that the points on the plasticity chart move
further away from untreated clay (i.e. both wL and wP
increase); in contrast, if 2% cement is added, extending
the curing time brings the plasticity point back towards
that of untreated clay, so that the increases in both wP
and wL during the first two days of curing are cancelled
out after 28 days.
Only when 8% cement is added does the treated
clay recover neither its initial wL nor its initial wP , so
that the final plasticity is characterized by both higher

wL and lower PI.
Determinations of clay plasticity properties are
ongoing in order to explore the effect of stabilization
with lime, cement and cement/lime mixture after about
four years of curing. Moreover, a laboratory testing
program aimed at studying the effect of soil stabilization on both compression and shearing behavior is still
in progress.
ACKNOWLEDGEMENTS
The authors wish to thank the Port Authority of Taranto
for its financial support, without which this research
project would not have been possible, and UNICALCE
and CEMENTIR for providing the additives used for
the research.
REFERENCES
Bell, F.G. 1976. The influence of the mineral content of clays
on their stabilization by cement. Bulletin of theAssociation
of Engineering Geologists 13(4): 267278.
Bell, F.G. 1989. The enhancement of the properties of clay
soils by the addition of cement or lime. In Taylor & Francis
Group (eds.), Soil Mechanics and Foundation Engineering; Proceedings 12th International Conference, Rio de
Janeiro, 1318 August 1989. Rotterdam: Balkema.
Booth, J.S. & Dahl, A.G. 1986. A note on the relationships
between organic matter and some geotechnical properties of marine sediment. Marine Geotechnology 6(3):
281297.
Broms, B.B. & Boman, P. 1977. Stabilization of soil with lime
column: Design handbook. Kungsbacka, Sweden: Lime
Column AB.
Burgos, M., Samper, F., Alonso, J. J. 2007. Improvements
carried out in very soft dredged mud soil in the port of
Valencia (Spain). In Cullar et al. (eds.), Soil Mechanics and Geotechnical Engineering; Proceedings 14th
European Conference, Madrid, 24-27 September 2007.
The Netherlands: IOS Press.
Casagrande, A. 1948. Classification and identification of
soils. ASCE-Transactions 113: 901930.
Cecconi, M., Ferretti, A., Russo, G., Capotosto, A. 2011.
Mechanical properties of two lime stabilised pyroclastic
soils. In Choong-Ki Chung et al. (eds.), Deformation
Characteristics of Geomaterials; Proceedings International Symposium, Seoul, 13 September 2011. The
Netherlands: IOS Press.
Chiu, C. F., Zhu, W., Zhang, C. L. 2008. Yielding and
shear behaviour of cement treated dredged materials.
Engineering Geology 103(12): 112.
Croft, J.B. 1967. The influence of soil mineralogical composition on cement stabilization. Gotechnique 17(2):
119135.
Cotecchia, V. 2005. Realizzazione di una vasca di contenimento per fondali di dragaggio in una zona compresa tra
il molo Ovest e Punta Rondinella nel Porto di Taranto.
Geotechnical and Geological Report, Port Authority of
Taranto.
Dalla Rosa, F., Consoli, N. C., Baudet, B. A. 2008. An
experimental investigation of the behaviour of artificially
cemented soil cured under stress. Gotechnique 58(8):
675679.
661
Eades, J.L. & Grimm, R.E. 1966. A quick test to determine lime requirements for lime stabilization. Highway
Research Record 139: 6172.
Federico, A., Murianni, A., Miccoli, E., Vitone, C., Intern,
G. 2013. Physical properties and compression properties
of submarine clayey soils from the Port of Taranto treated
with lime or cement (in prep.).
Fratalocchi, E., Manassero, M., Pasqualini, E., Roccato,
M., Spanna, C. 1996. Predicting hydraulic conductivity of cement-bentonite slurries. In Bouazza et al. (eds),
Environmental Geotechnics, Proceedings 2nd International Congress, Osaka , 58 November 1996. Rotterdam:
Balkema.
Grubb, D., Chrysochoou, M., Smith, C. & Malasavage, N.
2010. Stabilized Dredged Material. I: Parametric Study.
Huang, Y, Zhu, W., Qian, X., Zhang, N., Zhou, X. 2011.
Change of mechanical behaviour between solidified
and remolded solidified dredged materials. Engineering
Geology 119: 112119.
Huang, P., Bobet, A., Santagata, M. 2012. Identification
of low Organic-Content Soils: an engineering approach.
Geotechnical Testing Journal 35(4): 596606.
Keller, G.H. 1982. Organic matter and the geotechnical
properties of submarine sediments. Geo-Marine Letters
2(3-4): 191198.
Locat, J., Brub, M.A., Choquette, M. 1990. Laboratory investigations on the lime stabilization of sensitive
clays: shear strength development. Canadian Geotechnical Journal 27(3): 294304.
Locat, J., Tremblay, H., Leroueil, S. 1996. Mechanical and
hydraulic behaviour of a soft inorganic clay treated with
lime. Canadian Geotechnical Journal 33(4): 654669.
Mastronuzzi, G., Palmentola, G., Sans, P. 1999. La storia geologica. In: Mastronuzzi G. & Marzo P. (eds.) Le
Isole Chradi fra natura, leggenda e storia. Mottola:
Stampasud.
Mitchell, J. K. 1981. Soil improvement State of the art

report. In Nils Flodin (ed.), Soil Mechanics and Foundation Engineering; Proceedings 10th International Conference, Stockolm, 1519 June 1981. Rotterdam: Balkema.
Port Authority of Taranto 2004. Prove di laboratorio ai fini
della caratterizzazione geologica e geotecnica dei terreni
pertinenti alla vasca di contenimento e alle opera previste.
Final Report.
Rajasekaran, G. & Narasimha Rao, S. 1997. Lime stabilization technique for the improvement of marine clay. Soils
and Foundations 37(2): 97104.
Russo, G. & Croce, G. 2011. Experimental investigation
on durability of a lime stabilised soil. In Andreas Anagnostopolous et al. (eds.), Soil Mechanics and Geotechnical Engineering; Proceedings XV European Conference,
Athens, 1215 September 2011. The Netherlands: IOS
Press.
SGS SELC 2009. Caratterizzazione geotecnica dei sedimenti della darsena Capitaneria di Porto a Taranto. Final
report G148/RT3/Finale.
Smith, J. 2005. Soil stabilization. History lesson. Ground
Engineering Ground Improvement Supplement IXXI.
Tang, Y.X., Miyazaki, Y., Tsuchida, T. 2001. Practices of
reused dredgings by cement treatment. Soils and Foundations 41(5): 129143.
Terashi, M. & Tanaka, H. 1981. Ground improved by deep
mixing method. In Nils Flodin (ed.), Soil Mechanics and
Foundation Engineering; Proceedings 10th International
Conference, Stockolm, 1519 June 1981. Rotterdam:
Balkema.
Tremblay, H., Leroueil, S., Locat, J. 2001. Mechanical
improvement and vertical yield stress prediction of clayey
soils from eastern Canada treated with lime or cement.
662

Sustainable remediation of contaminated sites

C.N. Mulligan
Concordia University, Montreal, Canada
S. Dumais & R. Noel-de-Tilly

Golder Associates, Montreal, Quebec, Canada
ABSTRACT: Selection of the most appropriate remediation technology must coincide with the environmental
characteristics of the site and the ongoing fate and transport processes. In this paper, we will examine the means to
select the most appropriate technique for site remediation, evaluate the progress of the remediation and determine
the long term restoration of the site using the GOLDSET software (www.goldset.com). The two main approaches,
in situ and ex situ treatment, are examined further. Removal of the contaminated soil requires evaluation of the
risks, determination of disposal methods and/or potential beneficial use. To work towards sustainability, waste
must be minimized, natural resources must be conserved, landfill deposition should be minimized, biodiversity
must not be lost and must be protected. Innovative integrated decontamination technologies must be utilized. The
evaluation of three case studies of contaminated sites using sustainability indicators and analyses are presented.
The adaptability of the software was shown for the different scenarios.
INTRODUCTION
Proper management of the geoenvironment is essential

for future generations. The principles of sustainability
require recognition that geoenvironmental natural and
cultivated resources are renewable and non-renewable.
It is necessary to recognize that renewable geoenvironmental natural resources can be easily threatened and
can become unusable as a resource. For example, pollution of receiving waters will render such waters unacceptable for human consumption, therefore rendering
this renewable geoenvironmental resource useless.
Figure 1 summarizes some of the impacts of
geoenvironmental contamination. The impacts to the
Figure 1. Impacts of geoenvironmental contamination

(adapted from Yong et al. 2006).
geoenvironment must be evaluated and procedures for

avoidance and mitigation of these impacts need to be
established. Public health must be protected, natural
resources and assets of the site maintained.
Although many have accepted the principles of sustainable development, there are still significant challenges in applying them to project and process design
or operation. The nature of sustainability concepts
often do not translate easily into workable activities,
or technological evaluations. Therefore, it is difficult for engineers and decision makers to understand
whether their actions contribute to sustainable development, especially, in the long term. While there have
been developments regarding sustainability assessment tools, there is still a lack of tools for the
necessary thorough and comprehensive assessment of
sustainability.
During the evaluation of technologies to treat soil,
it is necessary to factor in the targets, exposure routes,
future land use, acceptable risks, legislation, and resultant emissions to determine the criteria and tools for
evaluating technologies and protocols for environmental management of contaminated soil. Other factors
that need to be considered to evaluate site remediation
technologies include: (a) disturbance to the environment, (b) energy use and consumption, (c) solid wastes
generated, (d) emissions of contaminants and greenhouse gases into the air, and (e) water and materials
used (Yong et al. 2006).
In evaluating the technologies, contamination of
the subsurface material (subsurface geologic material)
and the underlying aquifers from pollutant sources
must be examined before and during remediation
663
Figure 3. Methodology used for decision making process

(Ouyed & Vincent-Guimond 2010).
Figure 2. Procedures required to determine indicators for

soil remediation processes (adapated from Yong et al. 2006).
processes. The transport of pollutants in soils must

be determined through laboratory tests and field studies to support remediation projects. Procedures and
analytical-computer models have been developed to
determine and/or predict the movement, distribution
and concentration of pollutants in the subsurface soil.
These essential elements required for assessment or
evaluation of the impacts from surface discharge phenomena such as spills, dumping, etc. are shown in
Figure 2.
There is a growing need to incorporate sustainability into projects due to pressure from all stakeholders.
To integrate this into projects, decision support tools
are needed. According to the EPA (2008), minimization of energy use, air emissions, water impacts, material and waste use, land and ecosystems is required.
The focus should be more on sustainable remediation such as phytoremediation and biobarriers, and
use of renewable energies. In Canada, some sustainable remediation initiatives are led by the Interstate Technology & Regulatory Council (www.itrcweb
.org/gd.asp), Government of Canada (Public Works,
Health Canada) and more recently by Surf Canada
(www.surfcanada.org).
GoldSET (Golder Sustainability Evaluation
Tool), a sustainability decision support tool for engineering projects was created by Golder Associates
(www.gold-set.com). It is a robust and transparent framework to embed sustainable development
practices into design, construction and operational
decision-making phases of any engineering projects.
The sustainability tool has been applied for various
applications such as site remediation and mining tailings across Canada, the U.S., South America, Europe
and Australia. The tool includes a number of quantitative and qualitative indicators for the three dimensions
of sustainability: environment, society, and economy.
Indicators provide a way of describing the situation
surrounding the project with a weighting scheme
allowing the relative importance of each indicator to
be reflected.
The first step in an evaluation involves the description of the site and identification of key issues of
concern to all stakeholders as shown in Figure 3. Site
conditions, project objectives, and key stakeholders
and their issues must be identified.
The second step is to identify and elaborate on
various remediation options and/or alternatives that
are thought to be suitable for the site specificities
and project restrictions. Those options will then be
assessed from an economic, social and environmental
viewpoint.
Together, the entire set of indicators should be representative of a projects performance, impacts and
cost. The tool provides several modules, which already
include a set of pre-selected indicators, chosen based
on international and industry-specific standards, as
well as legal requirements. Indicators can be added
or removed, depending on needs and the context of the
project.
For each indicator, every project option must be
ranked according to a scoring scheme previously
determined by the user. The software automatically
calculates the aggregate score for each project option
and compares their performance across the three
dimensions (environmental, social, and economic) of
sustainable development.
Results are presented as a visual diagram that
clearly illustrates the strength and weaknesses of each
option in regards to sustainability. The evaluation process is iterative by nature, and further refining can
be accomplished if additional information is available
or if a new option is proposed. Sensitivity analysis
can also be performed on the results to improve the
reliability of the findings.
The right method to communicate your choices and
results is often crucial. The methodology entails a
collaborative approach to the table by engaging all
stakeholders into the decision making process through
every step of the assessment, allowing them to make
decisions with broad consensus that can be easily
defended.
Various technical options are then identified, followed by the tailoring of the specificities for each
option. The scoring scheme attached to each indicator provides a mechanism to assess the performance
of each option with respect to the indicator, producing
a comparative graphical result of each options sustainability performance. The data is both quantitative and
qualitative. The most sustainable option is portrayed
by the largest, most balanced triangle with respect to
the three axes of environmental, social, and economic
performance.
664
for comparison of remediation options for benzene

contaminated groundwater at the given site.
Identified options included:
Pump and treat and activated carbon

Pump and treat and air stripping
Biosparging and soil vapour extraction (SVE)
Natural attenuation.
2.2
Figure 4. Assignment of weights of importance (Ouyed &
Vincent-Guimond 2010).
The module allows for both detailed designengineering phase option assessments and design
selections for soil and/or groundwater treatment. It
operates through the assessment of different project
alternatives against a number of quantitative and qualitative sustainability indicators for each of the four
dimensions of sustainability: environment, society,
economy and technology. The indicators provide a
comprehensive assessment of a project, with a weighting scheme allowing the relative importance of each
indicator to be reflected (Figure 4). The newly developed tool provides a framework to compare different
project alternatives relative to the three sustainability
dimensions on an iterative basis. All qualitative indicators have scoring schemes consisting of 3 levels.
Quantitative indicators have both relative and absolute
scoring schemes. For the specific quantitative indicator like greenhouse gas emissions and Net Present
Value (NPV), the framework is adopted to a level of
detailed calculators. Relative scoring schemes assign
a score of zero to the lowest performing option, while
assigning 100 to the best performing option. Absolute
scoring schemes have a fixed scoring scale independent of the options, and score the options relative to
this fixed scale. These fixed values were adopted from
accredited organizations (UNEP, WHO, etc.) as benchmarking values for consumption of natural resources
or concentration of pollutants in the media.
2
CASE STUDIES FOR THE EVALUATION
Three different case studies were used to evaluate the

software according to the methodology in Figure 3.
They included the following scenarios.
2.1
Case study for a benzene- contaminated site
A benzene spill occurred along the highway near a

small town of 1600 residents following a train derailment. A certain amount of benzene then leached into
the ground water aquifer used by the residents for
their domestic well water. The area of the affected
site was 6 hectares. The concentration of benzene
in ground water was 55 g/L which needed to be
reduced to its maximum contaminant level (MCL)
of 5 g/L for drinking water. The software was used
Case study for an agricultural site
A 2 hectare agricultural site was contaminated with

elevated levels of copper, nickel and zinc. Metal
drums, electrical transformer casings and laboratory
wastes were the primary sources of the metal pollution.
The levels of each metal were found to be 205 mg/kg
higher than the level B according to the Ministre
du development durable, de lenvironnement et des
parcs (MDDEP). Contaminants from the soil reached
surface waters via two agricultural drainage ditches
adjacent to the contaminated site. Due to the proximity to food crops, animals, and residential communities, the contaminants must be removed from
the soil to prevent further transport of contaminants
to sensitive receptors. Three options for remediation
were considered including soil washing, solidification/
stabilization, and phytoremediation. Each method was
evaluated on the basis of economic, environmental,
and social impact. The software was employed to
compare the technological options.
2.3
Case study for a fuel-contaminated site
A retail fuel sales and service station has six underground storage tanks (USTs). In 1992, three of the
USTs were excavated and removed following failure
of the line tightness testing. In 1995, the remaining three USTs were excavated and removed along
with associated piping, dispenser pumps, and island.
Approximately 1000 m3 of petroleum contaminated
soil were excavated and stockpiled on- site during the 1992 and 1995 tank removal activities. The
most representative compounds inside the petroleum
composition considered were benzene, toluene, ethylene dibromide, ethyl benzene, xylenes, naphthalene
and tetraethyl lead. The soil concentrations were for
benzene, 0.03 mg/kg; toluene, 0.37 mg/kg; ethyl benzene, 0.082 mg/kg; xylene, 11 mg/kg and tetraethyl
lead, 0.036 mg/kg. Two main options were evaluated.
Biopiles which is a bioremediation treatment through
the use of aerated biopiles was selected as a remedial
action. The alternative was low temperature thermal
treatment. The software was then used to compare the
alternatives.
3
CASE STUDY FOR A

BENZENE-CONTAMINATED SITE
The multi-criteria analytical tool was used to evaluate the strengths and weaknesses of engineering
projects with respect to the environmental, social, and
economic dimensions of sustainable development. It
665
Table 1.
List of indicators considered for the analysis.
Environmental aspect
Type of Indicator
Soil quality
Sediment quality
Contaminated soil erosion
Groundwater quality
Free product
Surface water quality
Waterborne contaminant migration
Water usage
Impact on fauna during project
Impact on fauna after project
Soil vapor intrusion
Greenhouse gas emissions
Energy consumption
Quantity of wastes
Hazardous wastes
Residual impact of technology
Economic aspect
Social aspect
NPV of option costs

Potential litigation
Financial recoveries
Environmental reserve
Standards, laws and regulations
Service Reliability and performance
Reuse of the property
Corporate image
Reliability
Technological uncertainty
Logistics
Community health and safety

Workers health and safety
Drinking water supply
Direct local employment
Opportunities for local business generation
Public disruption
Quality of life
Public use
Cultural heritage
Impact on the landscape
Management practices
allowed for the comparison of different options on a

balanced, impartial, and comprehensive basis. The following section describes the overall conclusion of the
analysis. Table 1 provides a list of the indicators used
for the analysis.
monitoring costs. However, the selection of natural attenuation may require additional institutional
controls and local regulations.
3.3 Social evaluation of selected remediation

options
3.1
Environmental evaluation of selected

remediation options
Using the indicators in Table 1 scores (0 to 100) and

weights (1 to 3) were assigned to each indicator for
each option. 0 represents the worst scenario whereas
100 represents the best practice or best scenario.
Although a life cycle assessment is not performed by
the software, the indicators should be designed to the
changing technological and environmental conditions
and at different stages in the process. Results showed
that, with respect to the environment, natural attenuation, and biosparging and SVE appear to be more
appropriate for this site, since they have minimum
impact on soil quality, on fauna and flora resulting
during the project, soil vapour intrusion, greenhouse
gas emissions, energy consumption, and quantity of
wastes. On the other hand, activated carbon and air
stripping seem to be less appropriate, because of the
disturbance to the natural condition of the site due to
excavation and groundwater extraction, higher energy
consumption due to pumping of groundwater, and
higher production of greenhouse gases due to the
heavy trucks usage.
3.2
Economical evaluation of selected remediation

options
With respect to the economic aspect, activated carbon and air stripping have the highest initial costs
while natural attenuation and biosparging have higher
With respect to the community health and safety,

workers health and safety, public disruption, and
impact on the landscape, natural attenuation and
biosparging would have less impact on community,
while activated carbon and air stripping have more
impact on community due to excavation of wells, truck
traffic and potential accidents. Furthermore, there is
no limitation in terms of remediation time, natural
attenuation and biosparging appears to be best options
among the selected options for remediation of the
contaminated aquifer.
On the other hand, if the contaminated aquifer is the
only drinking water supply for the society, other remediation options such as activated carbon or air stripping
should be considered. The required costs for site remediation are higher than bioremediation options due to
the urgency for an acceptable source of drinking water.
3.4 Overall summary of the analysis

The overall comparison can be shown in Table 2 and
Figure 5. It is not surprising that natural attenuation
is the most sustainable option due to the low impact
on the site and costs. It shows the most balanced
approach. However the duration of the remediation
(approximately 20 years) would be a major issue if
the site is needed in the near future. If this is the case,
then biosparging would be the next most sustainable
option which could be accomplished in a reasonable
time frame.
666
Table 2.
Chosen indicator scoring and their weights for the benzene contaminated site.
Biosparging
Natural
attenuation
Pump and
treat
Pump and treat

with AC
Weight
Environmental
Soil quality
Sediment quality
Soil erosion
Groundwater quality
Free product
Surface water quality
Contaminant migration
Water usage
Impacts on biota after project
Impacts on biota during project
Soil vapour intrusion
Energy consumption
Quantity of wastes
Hazardous wastes
Residual impact
90
0
100
100
0
0
100
50
90
66
90
92
54
63
50
100
100
0
0
100
90
100
90
50
100
100
100
100
100
100
50
100
0
0
0
100
90
0
90
50
45
66
0
75
0
52
50
66
100
0
0
100
100
100
90
50
0
0
0
0
54
0
50
100
1
1
1
3
1
1
3
1
2
2
2
1
3
1
1
3
Social
Community health and safety
Water health and safety
Water supply
Direct employment
Opportunities for business
Public disruption
Quality of life
Public use
Cultural heritage
Impact on landscape
100
66
50
66
45
100
66
0
100
66
50
100
100
50
33
90
0
66
0
100
100
75
100
66
50
66
45
73
66
0
100
66
50
100
66
50
66
100
100
66
0
100
66
75
3
2
2
Economic
Net present value
Financial reserve
Service reliability and performance
Reuse of property
Corporate image
Reliability
0
50
25
0
100
66
100
90
90
90
100
50
25
0
90
66
100
100
90
90
91
50
25
0
90
66
100
90
90
90
1
50
50
0
90
100
100
90
90
100
3
2
2
1
2
2
2
1
3
2
4 AGRICULTURAL SITE CASE STUDY

Twenty-four qualitative and quantitative indicators
were chosen to assess the best technology available
in the decision making software for the agricultural
site. Three options for remediation were considered
including soil washing, solidification/ stabilization,
and phytoremediation. A wide spectrum of indicators
was chosen. There were five quantitative indicators
selected. Indicators were given weights as seen in
Table 3. They incorporated the concerns of the client
with those of the public. Since this project was carried
out on private land, concerns for the client were given
higher scores.
Eleven environmental indicators were selected due
to the wide range of issues to be addressed. Soil
1
1
1
2
2
1
3
quality, effects on local habitats, water quality, and air

emissions were addressed in particular.
Six social indicators were addressed including
impact on landscape, local job creation and diversity,
re-use of the property by the corporation, quality of
life of execution, worker safety and public safety.
Five economic indicators were included in the program: economic advantages for the local community,
service reliability and performance, potential litigation
costs, net present value of costs and reliability.
Analysis of the final report of GoldSET illustrates the following results (Table 3 and Figure 5).
Phytoremediation has the lowest impact on the environment compared to the other methods, stabilization/
solidification (in-situ), and soil washing (ex-situ).
Excavation, transportation, and the wastes that are
667
produced during these two last methods dramatically

change the environment conditions.
Although all three technologies have some effect on
the society, phytoremediation has the lowest impact.
The societal effect of soil washing and S/S processes
is almost the same at 65%, and 68%.
Comparison of the economic results shows that
S/S is the best option, and that soil washing due to
substantially expensive excavation and transportation
costs.
Another important factor is time. Since the site
is close to a residential area, then it is necessary to
treat the soil quickly. The treatment period for soil
washing and S/S performance is appropriate, while

phytoremediation takes extensive.
The overall results in Figure 6 show that phytoremediation is the best choice compared to the two other
methods. It has the lowest cost, and the lowest environmental impacts. Although the process period is
extensive, this problem can be reduced by planting
willows around the site to prevent the contaminants
from migration to the surface water. The site has the
potential for re-use after the remediation enhancing
the sustainability of the remediation process.
Based on the the GoldSET output, phytoremediation was selected as the most favourable solution. This
is due primarily to the low environmental and social
impacts of the project, in addition to its moderate cost.
5
PETROLEUM CONTAMINATED SITE

CASE STUDY
Another site for this case study is fundamentally a

contaminated soil with petroleum, based on various
assumptions that were made. The remediation was
considered to be ex situ since a large amount of the contaminated soil was already removed. In addition it was
an industrial application, with no residential and commercial places nearby. Therefore, the main impact of
the remediation would be to the project workers.Therefore based on this case, the list of indicators that were
quantitative and qualitative was generated as shown in
Table 4. The energy consumption and the greenhouse
gas emissions are significantly higher for the LTTD
Figure 5. Comparison of the site remediation options for

the benzene contaminated site showing the environmental,
1social and economic indicator scoring and their weights.
Figure 6. Quantitative indicators and comparison of the

results for the agriculture site.
Table 3.
Quantitative indicators for the results of the agricultural site.
Indicator
Units
Stabilization
Soil washing
Phytoremediation
Environmental indicator
Water usage
Wastes
Energy consumption
Liters
Tonnes
Tonnes CO2 equiv
GJ
10,000
7.9
129.2
1835
713,150
2007.9
671.9
9662
7,200,000
600
82.6
0.16
Social
Duration of work
Years
0.38
1.16
15
Economic
Net present value
4,000,000
28,354,000
184,700
668
Table 4.
site.
List of qualitative and quantitative indicators considered for the analysis of options for a petroleum contaminated
Environmental aspect
Type of indicator
Quantitative
Energy consumption
Quantity of wastes
Hazardous wastes
Qualitative
Soil quality
Soil vapor intrusion
Economic aspect
Social aspect
NPV of option costs
Duration of work
Financial recoveries
Service Reliability and performance
Reuse of the property
Corporate image
Reliability
Logistics
Workers health and safety

Quality of life
Impact on the landscape
Figure 7. Comparison of the technologies considered for

remediation of the petroleum-contaminated site.
process. However, many of the other indicators were

not significantly different for the two processes.
The overall results are presented in Figure 7. As can
be seen, the first option of biopiles is more sustainable than the second one according to the economic,
environmental and social elements. The biopile technique is the more balanced triangle, in relation to the
three factors.The LTTD technique has a higher score in
terms of societal aspects (88%) and the project duration is only 0.19 years. The biopile technique is the
better method option to treat the soil than LTTD technique as the LTTD requires more soil movement and
in consequence is more risky than the other.
CONCLUSIONS
The module can be customized for a specific or new

application to optimize design decisions. Through this
multi-criteria analysis framework, alternatives can be
compared regarding environmental, economic, and
societal aspects that are relevant to treatment decisions. It provides evaluations for different issues such
as the lifecycle costs, regulatory risks, energy and
greenhouse gas emissions, reuse opportunities, and
social acceptability. A thorough understanding of the
soil environment is needed through soil testing and
modeling to provide accurate input information on soil
and water indicators and risks to the environment. A
fourth dimension, technical, could be added to enhance
the evaluation process as was recently done for the
development of a wastewater module (Alimahmoodi
et al. 2012).
The module can be used to provide credible decisions and assess tradeoffs as seen in the three case
studies and is easy to understand and use. It showed
that it is dynamic and adaptable for a variety of case
studies. The sustainability of the various wastewater
treatment options can be compared when criteria are
identified and weighted and performance measures
selected to fit the specific conditions. As new and
improved treatment technologies are developed, more
management options will be available to offer greater
sustainability along with increased reliability and flexibility. Thus, site treatment systems can offer a higher
level of sustainability to users, the community, and the
environment.
The developed module is applicable for assessment
of sustainability of existing processes as well as for
detailed assessments of different designs and technology selections. Its outcome also helps the decision
makers for general project planning decisions, process
revamping, or the future upgrades and prioritization.
Further refinement of the scoring and weighting will
allow an enhanced evaluation process. More efforts in
the future should be made to green the remediation
technologies through the use of solar or wind energy,
reducing GHG emissions, improving air, soil, sediment and water quality, reducing waste generation and
material use (Hers 2012).
669
ACKNOWLEDGEMENTS
The authors would like to acknowledge the technical
and financial contribution of Concordia University for
the realization of the project.
REFERENCES
Alimahmoodi, M., Mulligan, C.N., Chalise, A., Grey, V. &
Noel-de-Tilly, R. 2012. Tool for evaluating the sustainability of wastewater treatment systems, 1st International
Specialty Conference on Sustaining Public Infrastructure,
Edmonton, Alberta, 69 June 2012.
EPA. 2008. Green remediation: Incorporating sustainable

environmental practices into remediation of contaminated
sites USEPA , April.
Hers, I, 2012. Where are we headingInnovation in Green
and Sustainable Remediation, GeoEnvirologic,Vancouver,
29 Feb., 2012.
Ouyed, M. & Guimont-Hbert, V. 2010. GoldSET-A sustainability Evaluation Tool for Engineering Projects. CSVA
2010 Conference, Montreal, 1516 Nov. 2010.
Yong, R.N., Mulligan, C.N. & Fukue, M 2006. Geoenvironmental Sustainability, Boca Raton, FL USA: CRC
Press.
670

Prediction of time to reduce tsunami sediment salinity by rainfall

after the Great East Japan Earthquake
K. Tada, H. Komine & S. Murakami
Ibaraki University, Department of Urban and Civil Engineering, Nakanarusawa, Hitachi, Ibaraki, Japan
ABSTRACT: On March 11, 2011, the Great East Japan Earthquake disaster, comprising a gigantic earthquake
motion and a giant tsunami, caused huge geotechnical damage. The desalination of tsunami sediment and of
salt-affected of soils has become necessary, but the enormous scale of materials over such a vast area presents
severe difficulties. Nevertheless, tsunami sediment can be used as materials for renewal. Therefore, quantitative
evaluation of desalinization of tsunami sediment by rainfall was conducted to assess the long-term influence of
salinity. The necessary time for desalinization of tsunami sediment in affected areas can be predicted based on
experimentally obtained results.
INTRODUCTION
On March 11, 2011, the Great East Japan Earthquake

disaster, comprising a gigantic earthquake motion
and a giant tsunami, caused widespread geotechnical
damage and shifted enormous quantities of tsunami
sediment (Fig. 1). An estimated 1328 million tonnes
must be treated (Ministry of environment, 2011).
The material is anticipated for use in renewed construction under foundations, in embankments, or as
a raw material in cement (Ministry of environment,
2011). However, tsunami sediment is highly saline.
Desalination of tsunami sediment and salt-affected
soil are necessary before it can be used. The quantitative results of tsunami sediment salinity reduction
by rainfall were been analyzed to examine the longterm influence of salinity. The necessary time for
desalinization of tsunami sediment in an area can
be predicted based on the experimentally obtained
results.
FUNDAMENTAL PROPERTIES OF TSUNAMI

SEDIMENT AND ARTIFICIAL TSUNAMI
SEDIMENTS
Tsuami sediment samples were collected from affected

areas in Yamada City, Iwate (TS-A) and Kesennuma
City, Miyagi (TS-B) (Fig. 2). These materials were
sieved using a sieving machine. TS-A was less than
20 (mm). TS-B was 220 (mm). Furthermore, artificial tsunami sediments including materials A, B, C
and D (ATS-A, B, C and D, respectively) (Fig. 3) were
Figure 2. Tsunami sediments (TS-A and TS-B).
Figure 1. Iwate area after the 2011 Great East Japan

Earthquake (2011).
Figure 3. Artificial tsunami sediments (ATS-A, B, C and D).
671
produced to evaluate salinity reduction. The artificial

samples were used to simulate salt-affected soil by
soaking in artificial seawater for 7 days. Then, these
samples were air-dried. The grain size distributions of
artificial materials are shown in Figure 4. Their fundamental properties are presented in Table 1. Tsunami
sediment shows a difference in the value of particle density: it is 2.482.58 (g/cm3 ) in TS-A and 2.73
(g/cm3 ) in TS-B. Results show that the wood chip contents influenced the particle density. The ignition loss
of TS-A was 11.4%; that of TS-B was 4.4%.
EXPERIMENT PROCEDURES
Leaching tests were used for evaluating environmental

impact attributable to waste. This study used column leaching tests to investigate EC characteristics of
tsunami sediment, which are important to assess the
tsunami sediment salinity reduction. Figure 5 depicts
the test apparatus used for column leaching tests. An
overview is presented in Figure 6. Distilled water percolated from the top to the bottom of the tsunami
sediment specimen at a constant rate of discharge
of 100 (mL/d). To assess the influence of the grain
size distribution, the flow rates in artificial tsunami
sediment were 300, 3000, 8000 (mL/d). Specimens
were produced to simulate natural sedimentation by
free fall. As presented in Table 2. The acrylic column
heights are 30 mm, 50 mm and the inside diameter is

75 mm. The resulting water that passed through the test
specimen was collected at a liquidsolid ratio (L/S) of
0.210 (L/kg) with a sampling bottle. It was analyzed
at EC.
EXPERIMENT RESULTS AND DISCUSSION
EC leaching characteristics of artificial tsunami sediment using column leaching tests are presented in
Figure 7. EC concentrations decreased gradually with
increased L/S. The salinity reduction is less than 200
(mS/m) at the Ministry of Land Infrastructure, Transport and Tourism in Japan Tourism, 2012). The EC
concentrations in ATS-A, B, C and D were constant
as L/S = 0.61.2 (L/kg). The result shows that reducing EC is influenced by the amount of passing water,
but it is unaffected by the grain size distribution. EC
leaching characteristics of tsunami sediment in column
leaching tests are depicted in Fig. 8. In TS-B, EC leaching is constant at L/S = 0.7 (L/kg), and it is constant at
Figure 5. Column leaching test apparatus.
Figure 4. Grain size distributions of artificial tsunami

sediment.
Table 1.
Figure 6. Overview of column leaching test apparatus.
Fundamental properties of tsunami sediments and artificial tsunami sediments.
Sample
Sampling Area
Grain Size
Particle Density
Ignition Loss
Natural water content
(mm)
(Mg/m3 )
(%)
(%)
TS-A
TS-B
Yamada City, Iwate

Less than 20
2.482.58
13.2
10.7
Kesennuma City, Miyagi

220
2.72
1.7
41.2
672
ATS-A
ATS-B
ATS-C
2.65
Simulated soil
Fig. 4
2.66
2.67
ATS-D
2.65
Table 2.
Experimental conditions of column leaching tests.
Sample
Height
Rate of discharge
Dry density
Initial water content
Void ratio
EC
TS-A TS-B
(mm)
(mL/d)
(Mg/m3 )
(%)
()
(S/m)
50
100
1.25
10.4
1.18
0.08
50
100
0.596
40.8
3.24
0.1
ATS-A
300
1.30
0.1
1.04
30
3000
1.30
0.1
1.04
0.17
ATS-B
8000
1.30
0.1
1.04
300
1.20
0
1.21
30
3000
1.20
0
1.20
0.53
ATS-C
8000
1.19
1.2
1.24
300
1.21
2.8
1.22
30
3000
1.21
5.1
1.22
1.0
ATS-D
8000
1.20
5.6
1.22
300
1.21
0.2
1.22
30
3000
1.20
0.2
1.23
0.4
8000
1.21
0.2
1.21
Figure 8. EC leaching characteristics of tsunami sediment.
Figure 9. Mechanism of reducing salinity in tsunami

sediment.
L/S = 2.5(L/kg) for TS-A. The result was influenced

by the amount of wood chips. The pattern of water
soaking in wood shows a two-step process: more than
half of the final absorbed water occurred at the first
contact of liquid water with wood. This was followed
by a period of very slow water uptake (J. Khazaei,
2008). Figure 9 shows the mechanism of tsunami sediment salinity reduction when the specimen includes
many wood chips. First, saline soil is removed. The
salinity concentration decreases concomitantly with
increased passage of water in wood.
Figure 7. EC leaching characteristics of artificial tsunami

sediment.
PREDICTION OF SALINITY REDUCTION

TIME BY RAINFALL
Tsunami sediment is anticipated for using as a material in compacted embankments, and as a material for
renewal such as cement raw material. We predicted
673
Table 3.
Required time for desalinization of tsunami sediment.
Sample
Occurring area
Average precipitation
Rate of discharge
Dry mass
(mm/d)
(mL/d)
(g)
Use Case
Standard of the EC
Desalted day
(mS/m)
(d)
TS-A
TS-B
Yamada City, Iwate

4.54
153
131.4
Kesennuma City, Miyagi

4.02
136
275.6
Embankment
Tree planting
maintenance
Incineration
Embankment
Tree planting
maintenance
Incineration
200
467
610*
250*
176
200
321
610*
224
250*
309
*The standard of tree planting maintenance and incineration is determined by the chlorine concentration (Cl (mg/100 g)). Tree
planting maintenance is 1000 mg/100 g (1.0%); incineration is 400 mg/100 g (0.4%). Therefore, the Standards of the EC are
calculated from Cl (mg/100 g) = 170 EC (ms/cm) 30 (Ministry of Agriculture, Forestry, and Fisheries, 2011).
for 224 days. For incineration 250 (mS/m), TS-A

required desalination for 76 days, TS-B was sufficiently desalted in 309 days. The result underscores
the need to examine the time necessary for desalting
of tsunami sediment using respective materials.
6
CONCLUSION
This study investigated EC leaching characteristics

of tsunami sediment using column leaching tests
and assessed the necessary time for tsunami sediment desalting in a particular area. The following
conclusions were obtained.
Figure 10. Time necessary to reduce tsunami sediment

salinity.
the time necessary to reduce salinity concentration in

tsunami sediment by rainfall using Eq. (1) and Eq. (2).
In these equations, L/S denotes salinity reduction

[L/kg], where A denotes the cross-section area of specimen [m2 ], Ms is the dry mass of the specimen [g], Q
denotes the rate of discharge [mL/d], and T represents
the salinity reduction time [d]. p denotes the average
precipitation levels of 10 years [mm/d], which were
4.54 mm/d and 4.02 mm/d in Yamada City, Iwate, and
Kesennuma city, Miyagi in Japan during 20012011.
The time necessary for reducing tsunami sediment
salinity is depicted in Fig. 10. The standards of EC
difference using respective materials were used. The
embankment case was 200 [mS/m]: TS-A was desalted
in 467 days, and TS-B was desalted in 321 days. The
tree planting maintenance case was 610 (mS/m): TSA required no desalting, and TS-B required desalting
(1) Tsunami sediment shows different behavior of EC

according to whether wood chips are included or
not.
(2) Before using tsunami sediment for embankment
and incineration purposes in Yamada city, Iwate,
467 and 176 days are necessary for desalting,
respectively, although tree planting maintenance
cases require no desalting.
(3) Before using tsunami sediment for embankments,
tree planting maintenance, and incineration in
Kesennuma city, Miyagi desalting for 321, 224
and 309 days are necessary, respectively.
(4) Results underscore the need to examine the necessary time for tsunami sediment desalination before
its use as various materials.
REFERENCES
Khazaei, J. 2008. Water absorption in three wood varieties,
Cercetari Agronomice n Moldova, Vol. XLI, No. 2 (134).
Ministry ofAgriculture, Forestry, and Fisheries. 2011. Desalination manual for farmland.
Ministry of environment. 2011. The 2011 Great East Japan
Earthquake tsunami sediment processing guidance.
Ministry of Land, Infrastructure, Transport and Tourism.
2012. Basic mode of thinking for the good use of to
residential and development laying earth on the ground
reproduction material contributing to quick restoration,
renewal.
674

Green and sustainable remediation of contaminated Indian Ridge

Marsh site in Chicago, USA
E.N. Yargicoglu & K.R. Reddy
University of Illinois at Chicago, USA
ABSTRACT: The remediation and restoration of heavily industrialized former wetlands and mesic prairies in
the Great Lakes region pose several special challenges due to: 1) widespread and heterogeneous distribution of
contaminants; 2) the variety of contaminant classes present; 3) complex hydrogeologic regimes due to extensive
and variable industrial filling and dredging; and 4) the presence of sensitive ecological receptors and habitats,
including nesting areas for several threatened bird species. Indian Ridge Marsh (IRM) is one of several degraded
wetlands in the Calumet region that are slated for remediation and redevelopment as part of the Calumet Open
Space Reserve (COSR). The goals of this work were to: 1) assess historically documented contamination based
on previous Phase I & II ESAs; 2) identify Areas Of Concern (AOCs) that present the greatest risks to human
and ecological receptors to determine the extent/intensity of remedial treatments required to meet established
cleanup requirements; 3) evaluate the remedial options available based on applicability, cleanup efficiency and
sustainability metrics; and 4) recommend an appropriate remedial strategy to the Chicago Park District.
INTRODUCTION
Recent efforts by the City of Chicago and the Illinois Department of Natural Resources to restore historically industrialized wetlands and prairies in the
Calumet region (southeast Chicago) have prompted
the evaluation of potential remedial options for several tracts of land slated for redevelopment as part
of the Great Lakes Restoration Initiative (GLRI), a
multi-agency effort to increase funding for remediation and protection of Great Lakes ecosystems. This
work sought to evaluate appropriate remedial actions
to reduce contaminant concentrations in impacted
media to acceptable levels and recommend a feasible
remedial strategy for one of these sites Indian Ridge
Marsh (IRM) to the Chicago Park District (CPD)
for consideration. New tools (SiteWise, Sustainable Remediation Tool) for green and sustainable
remediation (GSR) were employed to estimate the
environmental impacts of potential remedial options
and determine the most sustainable and cost-effective
remedy.
1.1
Identified Contamination at
Indian Ridge Marsh
Indian Ridge Marsh (IRM) has significant and

widespread historic contamination, including documented impacts to soil, sediments, surface water
and groundwater. Restoration of wetland and prairie
habitats at IRM holds significant ecological value,
especially for several endangered birds (e.g. black
crowned night heron) that nest seasonally in these

areas (Kamins et al., 2002). Multiple contaminant
classes are present on-site including: heavy metals,
pesticides, volatile organic compounds (VOCs), polycyclic aromatic hydrocarbons (PAHs), pesticides, and
one observed instance of an LNAPL plume containing
petroleum hydrocarbons. The contaminated areas that
posed the greatest risk to human and ecological health
were identified through comparison of measured sample concentrations to risk based screening levels
(RBSLs) established in the State of Illinois Administrative Code, Tiered Approach to Corrective Action
Objectives (TACO) and the Calumet Area Ecotoxicological Protocol (CAEP). Areas of Concern (AOCs)
were established based on the geographic distribution
of samples with contaminant levels exceeding established RBSLs (refer to map in Fig. 1). The AOCs
were targeted for direct remediation, and data regarding contaminant distribution in the subsurface, depth
to the water table, and area of impacted media from
each AOC were used to estimate overall energy use and
emissions associated with remediation of these areas.
The results from these analyses, coupled with sitespecific considerations, allowed the remedial design
to be tailored towards more sustainable alternatives by
informing the planning process on the source and magnitude of environmental impacts throughout different
phases of the remediation process (in SiteWise
only: 1. Remedial investigation; 2. Remedial action
construction; 3. Operations & maintenance; and 4.
Long-term monitoring) and across different remedial
alternatives (both SiteWise and SRT).
675
Figure 1. Area map showing three wetlands slated for restoration as part of the Millennium Reserve, proposed as part of the
Great Lakes Restoration Initiative. Inset map shows Areas of Concern (AOCs) identified at IRM.
1.2
Prior environmental site assessments
Since the late 1990s, six Environmental Site Assessments (ESAs) were conducted documenting the presence of VOCs, SVOCs, pesticides, and heavy metals
variably distributed throughout the soil, sediment,
groundwater, and surface waters. The source of the
contamination originates from both onsite and offsite
activities, including historic legal and illegal dumping
of waste and slag. Sources of offsite contamination include the Lake Calumet Cluster Sites (LCCS),
located directly adjacent and topographically upgradient from IRM to the west, which is believed to have a
direct impact on the IRM sediments and surface waters
through discharge of overland flow from LCCS by way
of culvert control structures passing beneath the N&S
rail lines. The LCCS, formerly used for both regulated and unregulated industrial facilities and waste
disposal, is flanked on the north and west by landfills,
and was placed on the National Priorities List (NPL) in
2010 receiving Superfund designation. LCCS is currently undergoing remedial actions which will impact
potential future contaminant transport into IRM.
METHODS
Qualitative and quantitative analyses were conducted

to evaluate potential environmental impacts associated
with each remedial option using Green and Sustainable

Remediation (GSR) tools such as: the Green Remediation Evaluation Matrix (GREM), created by the
California Department of Toxic Substances Control;
SiteWise sustainability evaluation tool, developed
jointly by US Navy, US Army Corps of Engineers
(USACE) and Battelle (Bhargava & Sirabian, 2011),
and the Sustainable Remediation Tool (SRT), developed by the Air Force Center for Engineering and the
Environment (AFCEE). Following a qualitative evaluation of sustainability metrics following GREM (i.e.
noise; worker safety; aesthetics), quantitative evaluation of energy/resource consumption was conducted
using both SRT and SiteWise.
Estimates of material and labor needs, treatment
time, volume of affected soil or groundwater to be
treated (based on the surface area and depth of contamination in each AOC), and assumptions specific to
certain treatments were made for each remedial alternative and input into SiteWise and SRT. The output
of these calculations is dependent on user-input of
project specifications, requiring estimates of labor and
material needs for each remedy that may vary in latter
stages of project design and implementation.
Output from these models includes estimates of:
project energy and water consumption, greenhouse gas
emissions (GHG) (CO2 , N2 O, NOX , SOX ), and accident/injury risk to workers. Refer to the SiteWise
and Sustainable Remediation Tool User Manuals for
676
Figure 2. Select output from SRT analyses among active remedial alternatives for groundwater treatment at Area F. Shown
are estimated emissions of CO2 and other criteria air pollutants (NOx , SOx , PM10 ).
specific equations and conversion factors employed by

the software to generate the reported estimates (Bhargava and Russell, 2011 and AFCEE, 2010). Figure 2
shows an example of output provided by the SRT tool,
comparing air pollutant emissions for several remedial alternatives considered for Area F. Because SRT
does not include phytoremediation as a remedial alternative, results from SRT only provide comparisons
among active remedies that can be employed if treatment time is a constraint. Given the end-use of the site
involves habitat restoration and preservation however,
overall project cost and environmental impact remain
more important in remedy selection than treatment
time. As such, a passive, in situ treatment with minimal
site disturbance (e.g. phytoremediation) is ideal. These
initial estimates, coupled with the continued use of
SiteWise during remedy implementation will allow
detailed accounting of the environmental impacts of
the project without excessive (and costly) sampling
and analyses of affected media and emissions.
Modeling estimates are applied initially to Areas
C and F, which have the highest contaminant concentrations and most complex contaminant mixtures;
under a cost-limiting scenario, these AOCs represent
the minimum area necessitating active remediation.
The remaining areas (A, B, D and E) may be monitored
for natural attenuation (MNA) of on-site contaminants until sufficient funding is available for full-site
remediation. Ideally, all AOCs will be remediated;
however, treatment of the entire contaminated area
(16.7 acres) may be cost-prohibitive.
A remedial strategy was chosen from the results

of both quantitative and qualitative sustainability
assessments. The criteria for selecting applicable
remedial technologies are based on site-specific conditions, including: geologic setting, local hydrology
(i.e. hydraulic connection of groundwater to surface
wetlands and the nearby Calumet River), the nature
of topsoil and surficial sediments (low permeability
clay-rich glacial till and silty sands; heterogeneous distribution of fill materials) the nature and distribution
of identified contaminants, and the end-use of the site.
3
RESULTS
Due to the high environmental impact rating for active

remedial technologies (e.g. pump-and-treat) observed
from qualitative comparisons and the incompatibility
of certain remedial technologies with long-term habitat restoration goals (i.e. irreversible negative impacts
to soils that may prohibit future plant growth, as can
occur with solidification and stabilization (S/S) and
electrokinetic treatment; disruptions to site hydrology
that will impact wetland habitats from the installation of permanent caps or vertical barriers for
in situ containment), detailed estimates of project
metrics using SiteWise were limited to two remedial alternatives: phytoremediation and excavation.
Both options are compatible with native prairie and
wetland rehabilitation of the site, though excavation
is anticipated to remove contaminants more quickly
677
Table 1. Summary of SiteWise comparison of sustainability metrics between phytoremediation with enhanced biostimulation (Phyto-EB) and Excavation at Area C.
Remedial
alternatives
GHG
emissions
Energy
usage
Water
usage
NOx
emissions
SOx
emissions
PM10
emissions
Accident risk
fatality
Accident riskinjury
Phyto-EB
Excavate
Medium
High
Medium
High
High
Low
Medium
High
Low
High
Low
High
High
High
High
Medium
than phytoremediation, which generally require 3 to

5 growing seasons for complete contaminant removal
(ITRC, 2009).
Table 1 summarizes the results generated by
SiteWise for these two options at Area C. In general,
phytoremediation has a lower environmental impact
than excavation, though water consumption can be
high if an irrigation system is utilized. The use of
fertilizers also increased the environmental impact
of phytoremediation significantly, and measures to
reduce fertilizer use (i.e. addition of organic wastes
or compost to enhance growth) should be employed to
mitigate these impacts.
3.1
Selection of remedy based on sustainability

metrics and site-specific considerations
Several treatment types were deemed inappropriate

for the site conditions and/or contaminant chemistries
and excluded from extensive sustainability assessments. Several site-specific considerations narrowed
the range of feasible remedies, including:
The shallow water table (315 ft bgs), presence of

numerous surface ponds, and extensive wetlands
limited the use of technologies that were restricted
for use in the vadose zone (e.g. soil vapor extraction; air sparging) or those that required extensive
dewatering (e.g. thermal desorption) of the soils.
The wide-spread distribution of shallow subsurface
contamination (comprising a total area of 16.7
acres) poses logistical difficulties for treating or
removing large volumes of soil. In situ treatments
are thus preferable.
The presence of mixed contaminant types (heavy
metals, PAHs, VOCs, SVOCs, etc.)
necessitated a technology that can be applied to
a variety of chemical compounds (e.g. solidification & stabilization (S/S), excavation and disposal,
phytoremediation).
The heterogeneous nature and low hydraulic conductivity of the surficial sediments (fill material,
silty sands interbedded with clay lenses; K = 105
to 103 cm/s) limits the effectiveness of technologies that require pumping large amounts of liquids
through contaminated sediments (e.g. soil flushing) or rely on high groundwater flow rates (e.g.
permeable reactive barriers).
Future site use as part of the Calumet Open Space
Reserve (COSR) includes habitat and ecological
restoration goals, so remediation should minimize
the degree of permanent or irreversible site disturbance (i.e. excessive removal of native vegetation
that may not otherwise be recovered; severe chemical or electrochemical treatment of soils that may
permanently alter soil pH or otherwise prohibit
wetland restoration).
Qualitative evaluation of four remedial options
revealed the highest environment impact ratings are
assigned to excavation and capping, associated with
high site disturbance and use of heavy machinery, as
well as final disposal and monitoring considerations.
Phytoremediation is a more passive treatment, with
several advantages over traditional soil and groundwater remedies. Water uptake and transpiration by plants
is powered by solar energy (photosynthesis), allowing
for a low-energy, low-cost plant-mediated filtration of
contaminated groundwater that is effective for several
contaminant classes. Assuming minimal additional
topsoil is required to initiate rooting of tree stands and
prairies grasses, phytoremediation will require fewer
trips with large trucks to excavate and infill large
volumes of soil than S/S and excavation, thus minimizing transportation costs. The use of an irrigation
system, which would significantly increase water use
and energy costs, can be eliminated to encourage root
growth toward contaminated aquifers as well as reduce
overall project costs. The primary tradeoff is increased
treatment time, which can be vary widely depending
on how quickly the plants can grow once established.
Growth rates will slow if conditions are suboptimal
and soils may require amendments to improve plant
health and maintain high removal rates.
3.2 Proposed remedial design for
indian ridge marsh
The recommended strategy for remediation of IRM
will use phytoremediation in conjunction with biostimulation of existing soil microorganisms to enhance
degradation of organic contaminants at all identified AOCs. Native tree species with high growth
and transpiration rates, deep rooting depths, and the
ability to accumulate and/or sequester contaminants
of concern will be employed. Trees will be planted
in stands and spaced 10 ft apart to achieve maximum growth density and remedial efficiency. In areas
with both groundwater and soil contamination (B,
C, E and F), 50% of trees will be placed in lined
trenches to encourage root growth towards the contaminated aquifer. The liners will be modeled after
678
the proprietary ANS TTTS TreeWell system used

successfully at Argonne National Laboratory with the
same tree species (willows, cottonwoods, and poplars).
This technique also allows for greater tree densities
in the stands, as root systems will not grow as wide,
reducing the lateral extent of each tree in the root zone.
All treated areas will receive soil amendments in the
form of organic compost and an initial application of
balanced NPK (10-10-10) fertilizer to stimulate new
root growth. Oxygen reactive compounds (ORCs) will
be mixed into tilled soils during planting. This form of
oxygen additive is preferred over direct O2 injection
because it is less energy-intensive, less costly, does not
require the installation of injection wells, and releases
O2 in the soil over time rather than in pulses, improving
long-term performance of the plants. One drawback of
ORCs is the potential to raise the local soil pH, which
will be counteracted by the addition of acidifying soil
amendments (e.g. granular S, gypsum or Al2 (SO4 )3 ;
leaf litter) (Rentz et al., 2003). The addition of O2
to the soil is intended to stimulate microbial activity
in the rhizosphere, enhancing rhizodegradation processes associated with the plants as well as microbial
degradation processes that occur in natural soils when
sufficient nutrients and O2 are available (Rentz et al.,
2003). Regular applications (2-3 times per growing
season) of organic compost will provide ample nutrients for biostimulation processes and maintain overall
soil quality and pH.
A vegetative cover of grasses (Lotus corniculatus)
and legumes (Lolium perenne and Phalaris arundinacea) will be put in place in between treated areas
to help stabilize soils, maximize total water use, minimize erosion, and keep shallow soils dry to promote
deeper rooting depths of the phreatophytic trees (EPA,
2003; ITRC, 2009). The vegetative cover also serves
to reduce the flow of contaminated surface waters
to the nearby Calumet River or other off-site waterways by increasing infiltration into shallow soils. This
will also serve to minimize the production of leachate
as precipitation flows through contaminated soils and
groundwater. Additionally, the grasses and legumes
will help remediate contaminants in shallow subsurface soils that have less contact with the deeper root
systems of the Willows, Poplars, and Cottonwoods.
To minimize cross-contamination of surface waters
with contaminated sediments and soils, a riparian
buffer zone (510 ft wide) will be installed around
surface water reservoirs in close proximity to AOCs.
The riparian buffer will slow water transport between
surface and groundwater, while limiting erosion of
surficial sediments, helping to contain existing contamination within the site boundaries. The buffer zone
will consist of Cattails, Small Duckweed and Common
Reed already present on-site; additional plants will be
added in areas that lack sufficient native vegetation to
serve this purpose.
The remedial progress of each AOC will need to be
evaluated after every 5 years, the approximate length
of one growth cycle for the selected trees. This cycle
refers to the 46 years that the trees require to grow
from saplings to mature trees, at which point growth

rates and phytoremedial efficiency decrease. At the
end of each cycle, mature trees will be replaced in
order for new saplings to be planted. It is projected
that a minimum of 3 growing cycles (up to 15 years)
will be required to reduce the contamination levels
to an acceptable amount (ITRC, 2009). Areas with
higher contaminant concentrations (i.e. Areas C and
F) will require more growth cycles than areas which
have lower levels of contamination, which may be
remediated within the first growth cycle. The number of cycles needed at each AOC will be determined
as remedial progress is monitored and overall uptake
and degradation rates can be quantified at the site.
3.3 Selected plants: Site-specific requirements for
remedial efficacy
The plants that will be installed, in addition to any
existing vegetation found suitable for phytoremediation, will include trees of the Salicaceae family,
e.g. Willows (Salix spp.) Cottonwoods (Populus deltoides) and Poplars (Populus spp.). These trees are
phreatophytes, meaning they have long root systems
that extend to the water table and can grow quickly,
consuming large amounts of water (Zalesny et al.,
2007; EPA, 2003; ITRC, 2009). Their high growth
and transpiration rates translate to improved remedial efficiency, and their tendency to seek out the
saturated zones allows growth to be targeted to contaminated aquifers (EPA, 2003). Willows and Poplars
have demonstrated remedial efficacy for a variety
of contaminants present at IRM at similar locations
(e.g. Argonne National Laboratory-East; EPA, 2003),
including the ability to degrade TCE and PCE into
non-toxic byproducts (Newman et al., 1997b; Westphal & Isebrands, 2006). Though high transpiration
rates are ideal for phytoremedial applications, the rapid
of growth of these trees may impact the hydrogeology of the site and excessive draining from surface
waters may negatively impact wetland communities.
As such, monitoring of the water table and adjacent
surface waters is recommended to ensure the wetland
habitats are maintained for nesting bird populations
throughout the treatment.
As the end use of the site involves habitat rehabilitation and restoration of native flora and fauna, care will
be taken during the site investigation phase to identify existing vegetation that can remain in place during
the course of phytoremediation to minimize site disturbance. Table 2 lists several species of vegetation
identified at IRM that have demonstrated phytoremedial efficacy for one or more COC on-site and
our recommendation for use of those species in the
phytoremedial system.
These plants including the common reed (Phragmites spp.), small duckweed (Lemna minor), and
eastern cottonwood (Populus deltoides) should be
analyzed to determine if contaminant uptake and/or
accumulation has or is occurring and whether or not
they should be considered hazardous materials to both
679
Table 2.
Existing vegetation at IRM & potential phytoremedial applicability. An asterisk indicates a native species.
Plant Name (Species)
Targeted Contaminants
Recommendation for Use?
Common water plantain (Alisma subcordatum)

Path rush (Juncus tenuis)*
Small duckweed (Lemna minor)*
Switchgrass (Panicum virgatum)
Common reed (Phragmities spp.)
TBD or N/A
TBD or N/A
Pb, Cr(VI), certain pesticides
Anthracene, PAHs, Pyrene
Benzene, Trichloroethane, Toulene,
PCE, TCE, Cu, Fe, Mn
TCE, PCE
TBD or N/A
TBD or N/A
TBD or N/A
TBD after analysis

TBD after analysis
Yes
Yes
Yes
Eastern cottonwood (Populus deltoides)

Box elder (Acer negundo)
Hackberry (Celtis occidentalis)
Green ash (Fraxinus penn.)
ecological receptors and human workers. If analysis

of contaminant concentrations in the branches, leaves
or stems of the existing plants indicate that these
species may have phytoremedial efficacy for one or
more COCs on-site (e.g. hyperaccumulation of inorganics detected; root zones of plants correspond to
local areas of contaminant removal), they may be left
in place to assist installed plants in remediation of
soils and groundwater to minimize clearing of existing
vegetation and resultant habitat disturbance.
Other native plants at IRM include various wetland
and prairie species (e.g. cattails and reeds [Typha and
Phragmites spp.]; switchgrass [Panicum virgatum]),
although much of the marsh is dominated by plants
(primarily the Common Reed) with higher salt tolerances (20,000 mg/L) than native vegetation (Bosko,
1998a). Upland trees at IRM are also found to be
resistant to high soil salinity and include the eastern cottonwood (Populus deltoides), box elder (Acer
negundo), hackberry (Celtis occidentalis), and green
ash (Fraxinus penn.) (Roadcap et al., 1999; Bosko,
1998b). The cattails and reeds have demonstrated
remedial efficacy and will be left in place to aid
in remediation and stabilization of surface waters.
Because the common reed is currently dominating
much of the wetland habitat, it may be more sustainable to leave it in place during treatment so removal
can be timed with the planting of native species during the final cleanup and restoration stages. Doing so
would reduce the number of initial plantings required,
saving on transportation and energy costs.
3.4
Incorporating best management practices

into remedial actions
The environmental impacts of the proposed remedy

can be minimized significantly by adhering to best
management practices (BMPs) throughout remedy
design, construction, and implementation. Table 3 outlines key BMPs for the four major stages of the project:
remedial investigation; construction and installation;
operations and maintenance; and long-term monitoring. Many of these practices are consistent with goals
regarding minimal site disturbance and will aid in
the transition from remediation to habitat restoration
following site cleanup.
Yes
TBD after analysis
TBD after analysis
TBD after analysis
The use of on-site vegetation that demonstrates

remedial efficacy can significantly reduce overall land
disturbance and project cost by reducing the number
of new plantings, while also aiding habitat restoration
efforts after cleanup by minimizing the replacement
of existing vegetation with tree stands planted for
phytoremediation that may later be removed.
Moreoever, existing vegetation should require less
fertilizer than new plants without established root systems. Fertilizers require large amounts of energy to
be produced, and ideally should only be applied in the
initial phases to promote new growth. To minimize fertilizer use, vegetation and leaf litter removed to clear
space for tree stands should be incorporated into surface soils to enhance soil fertility and organic content.
Another major source of GHG in phytoremediation is tilling of the land prior to planting tree stands.
The use of ORCs can reduce the depth and frequency
of tilling required for sufficient soil aeration, though
some tilling will be required initially to incorporate
the ORCs with the soil. Proper management of the
phytoremedial system will require regular monitoring
of plant health to assess the need for additional soil
amendments. This will ensure that only the necessary
amount of fertilizers is applied to ensure ready plant
growth.
CONCLUSIONS
Based on the site conditions and history of widespread,

low-level contamination on and off-site, a passive
remedial strategy with minimal site disturbance is recommended. Due to the mixed contaminant chemistries
present, the shallow water table and heterogeneous
subsurface hydrology, other remedial technologies
were disqualified as appropriate treatments for all contaminants of concern at IRM. In terms of compatibility
with future site use and sustainability metrics, phytoremediation is the ideal technology for remediation
of Indian Ridge Marsh. This technology is in line with
future site use goals as part of the Calumet Open Space
Reserve (COSR) that include: preservation of wetland habitats; improvement of existing habitat, which
will be addressed as overall soil quality and vegetative
health is improved over the course of treatment; and
680
Table 3.
Best management practices (BMPs) for optimal performance during phytoremediation.
Phase
Best management practices
Remedial
Investigation
Construction &
Installation
Native vegetation with demonstrated phytoremedial applicability (or applicable for use as riparian
buffer) will be identified & left in place to minimize site disturbance and capital costs
Planting will occur in early Spring to maximize the length of the growing season
No irrigation system will be employed to allow tree roots to grow to depth of water table, reducing
water consumption and increasing remedial effectiveness (EPA, 2003)
Saplings will be trucked in from local nurseries
Existing vegetation to be cleared that is not hazardous waste will be chipped and amended to soils,
providing additional organic carbon & nitrogen, reducing fertilizer needs & transportation/disposal
costs of cleared materials
Soil amendments (ORCs, organic fertilizer/compost) will be mixed in with soil during tilling & planting
phases, rather than injected via injection wells
Well-defined work areas & use of compost over work areas to reduce soil compaction
Soil should be damp during installation to minimize dust production & potential exposure of
contaminated soils/sediments to workers
Periodic evaluation of treatment efficiency & appropriate modifications for optimal performance
(e.g. addition/removal of trees; watering during severe drought)
Organic compost (e.g. leaf litter, organic wastes, woodchips) will be added to soils yearly to fertilize
soils and reduce soil pH if above pH 8 (optimal pH range for poplar growth is 5.58.0)
Granular sulfur or aluminum sulfate will be added for soil acidification as needed to areas with soil
pH > 9 (1.5 lbs/100 ft2 in early spring and fall)
Lost plants will be replaced throughout treatment period (estimated 5% annual loss) to maintain high
removal rates
Reuse of existing wells for monitoring; use of direct-push methods over rotary drilling for construction
of new wells
Operations &
Maintenance
Long-term
Monitoring
creation of new habitats, which can be incorporated

into planting schemes after high levels of contamination are reduced in the early cycles of tree growth and
replacement.
It is recommended that an initial survey of existing
vegetation on-site be conducted to determine applicability to phytoremedial processes. Further sampling
of affected media in under-represented areas will be
necessary to better constrain the spatial extent of areas
of high-level contamination. This will allow the proposed design to be tailored to current conditions and
optimized to utilize existing vegetation with minimal
site disturbance.
Further benefits from this remedial alternative
extend from educational and public outreach opportunities that can be incorporated into the remediation and
habitat rehabilitation process. Information on native
vegetation and wildlife at IRM can be disseminated
throughout community bulletins and through posted
signs on-site that inform the public of ongoing remedial activities and what steps are being taken to ensure
that sensitive habitats are being protected. This will
improve public acceptance of the remedial activities
at IRM and garner support for habitat restoration
goals and improvement of degraded sites and wetlands
throughout the Calumet region.
REFERENCES
Air Force Center for Environmental Excellence, 2010
Sustainable Remediation Tool User Guide. United States
Air Force.
Bhargava, Mohit, and Sirabian, Russell. 2011. SiteWise

Version 2 User Guide. Battelle Memorial Institute, Columbus, Ohio.
Bosko, M.T. 1998a. Phase 1 Ecological Reclamation Study,
Lake Calumet Cluster Site, Chicago, Illinois. Roy F.
Weston, Inc., Chicago, IL.
Bosko, M.T. 1998b. Lake Calumet Area Ecological Analysis.
Roy F. Weston, Inc., Chicago, IL.
Hinchman, R. R., M. C. Negri, and E. G. Gatliff. 1997.
Phytoremediation: Using Green Plants to Clean Up Contaminated Soil, Groundwater, and Wastewater. Submitted
to the U.S. Department of Energy, Assistant Secretary for
Energy Efficient and Renewable Energy under Contract
W-31-109-Eng-38.
ITRC (Interstate Technology & Regulatory Council). 2009.
Phytotechnology Technical and Regulatory Guidance and
Decision Trees, Revised. Washington, D.C.: Interstate
Technology & Regulatory Council, Phytotechnologies
Team, Tech Reg Update. www.itrcweb.org
Kamins, N., S. Malec, L. Westphal, C. LeBlanc, et al. 2002.
Calumet Ecological Management Strategy (EMS). City
of Chicago, Dept. of Environment, Natural Resources
Division.
Newman, L. A., S. E. Strand, N. Choe, J. Duffy, G. Ekuan,
M. Ruszaj, B. B. Shurtleff, J. Wilmoth, P. Heilman, and
M. P. Gordon. 1997b. Uptake and Biotransformation
of Trichloroethylene by Hybrid Poplars. Environmental
Science and Technology 31: 106267.
Rentz, J. A., B. Chapman, P. J. J. Alvarez and J. L. Schnoor.
2003. Stimulation of Hybrid Poplar Growth in PetroleumContaminated Soils through Oxygen Addition and Soil
Nutrient Amendments. International Journal of Phytoremediation. 5(1):5772.
Roadcap, G.S., M.B. Wentzel, S.D. Lin, E.E. Herricks, R.K.
Raman, R.L. Locke, and D.L. Hullinger. 1999. An Assessment of the Hydrology and Water Quality of Indian Ridge
681
Marsh and the Potential Effects of Wetland Rehabilitation on the Diversity of Wetland Plant Communities.
University of Illinois. Prepared for EPA. November.
U.S. Environmental Protection Agency, National Risk Management Research Laboratory, Office of Research and
Development. 2003. Deployment of Phytotechnology in
the 317/319 Area at Argonne National Laboratory-East:
Innovative Technology Evaluation Report. Cincinnati,
Ohio.
Westphal L. M. and J. G. Isebrands. 2006. Phytoremediation

of Chicagos Brownfields: Consideration of Ecological
Approaches and Social Issues USDA Forest Service,
North Central Research Station.
Zalesny Jr., Ronald S. and Bauer, Edmund O. 2007. Selecting and Utilizing Populus and Salix for Landfill Covers: Implications for Leachate Irrigation. International
Journal of Phytoremediation 9:6, 497511.
682
Active and passive barriers for polluted sites

Critical aspects related to Fe0 and Fe0 /pumice PRB design

S. Bilardi, P.S. Calabr & N. Moraci
Universit Mediterranea di Reggio Calabria, Dipartimento di Ingegneria Civile, dellEnergia,
dellAmbiente e dei Materiali (DICEAM), Reggio Calabria, Italy
ABSTRACT: A Permeable Reactive Barrier (PRB) is an in situ remediation technology. The design of a PRB
is an interdisciplinary problem which involves mainly geotechnical, chemicals and hydraulic issues. This paper
presents considerations on the most critical aspects related to the design of PRB in particular from a geotechnical
perspective. The considerations originate from the results of column tests carried out on zero valent iron (ZVI)
and granular mixtures of ZVI and pumice using nickel as contaminant. In particular, the following topics are
discussed: (1) long term performance of PRB, (2) critical aspects related to the PRB design and (3) investigations
on long and short term behavior of PRB through column tests.
LONG TERM PERFORMANCE OF PRB
A Permeable Reactive Barrier (PRB) is an in situ remediation technology. It consists of a diaphragm wall,
filled with a permeable reactive material, and installed
perpendicularly to the groundwater flow. When contaminated groundwater flows through it treatment
processes occur. A PRB is a passive treatment system
since the water flows under its natural gradient.
PRB systems have been installed worldwide with
more than 200 applications (ITRC 2011) located
mainly in the USA. The most common reactive
medium used is the zero valent iron (Fe0 or ZVI).
Despite numerous installations of PRBs their longterm performance is not well known, therefore they
are currently not fully accepted in Europe. The lifespan of the PRB can be defined as the duration of
time when the barrier is able to intercept the contaminated groundwater and trap the contaminants within.
It is mainly controlled by reactivity and permeability reduction. In particular, PRB clogging may result
in preferential flow, through more permeable zone, or
blockage of flow.
Several literature studies have indicated different
possible causes of ZVI barrier hydraulic efficiency
reduction. The iron corrosion process by the water
or dissolved oxygen leads to the formation of Fe2+
and OH which cause an increase of pH. In these
conditions minerals precipitation occurs causing a
possible decrease of porosity (Jeen et al. 2011). Corrosion process induced by water in anaerobic conditions leads also to the formation of hydrogen gas
which, if entrapped in the reactive medium, can reduce
the porosity of the reactive material (Henderson &
Demond 2011). Formation of a biofilm due to microbial activity (Gu et al. 1999) and the retention of
fine particles coming from upstream soil in the PRB
pores can also contribute to the permeability reduction.
Finally, according to recent research (Car et al. 2008,

Zhao et al. 2011), another possible cause of porosity reduction is the expansive nature of iron corrosion
products being their volume (between 2 and 6 times)
higher than that of the original metal.
One of the removal mechanisms activated by ZVI
is the oxidation-reduction reaction, efficient only for
redox-sensitive contaminants. Furthermore the presence of iron corrosion products can be beneficial in
terms of contaminant immobilization through adsorption and co-precipitation processes. The reasons for
iron reactivity decrease could be formation of biofilms
or oxide layers on the ZVI surface which could cover
the active redox sites and thus reduce the reactive surface area or cause a declining rate of electron transfer
(Gu et al. 2002, Komnitsas et al. 2007, Bilardi et al.
2012b).
In order to extend the lifespan of the ZVI, especially
in term of hydraulic conductivity, it can be mixed with
an inert and/or porous material like sand or pumice
(Komnitsas et al. 2007, Moraci & Calabr 2010, Ruhl
et al. 2012).
2
CRITICAL ASPECTS RELATED TO THE

DESIGN OF PRB
The choice of the most suitable reactive medium is

strictly linked to the type of contamination. On the
contrary, the main geotechnical characteristics of the
medium (i.e. grain size curve, porosity, permeability)
exclusively result from the proprieties of the aquifer.
The factors influencing the choice of the grain size
distribution of the reactive medium are the permeability, porosity and grain size distribution of the soil that
constitutes the aquifer. In fact, the barrier should not
modify the natural groundwater flow.
Therefore, the important geotechnical consideration regarding the choice of the grain size distribution
685
of the medium, which was sometime not properly

considered, is the necessity for the reactive medium to
satisfy the filter design criteria towards the surrounding soil.
In one-dimensional flow conditions a filter must
satisfy three main design criteria: the internal stability, the retention and the permeability criteria (Moraci
2010, Moraci et al. 2012b).
When water flows thorough a soil, fine particles
can be washed out leading to internal erosion (i.e. piping). A filter should be fine enough to retain loose
soil particles (retention criteria). On the other hand it
should be coarse (or permeable) enough to allow seepage flow and avoiding the development of high internal
pore pressure (permeability criteria). In order to satisfy
the two criteria, Terzaghi (1922) proposed for granular
filters respectively the two following conditions:
where D15 (or d15 ) and d85 are the diameters corresponding to 15 and 85% finer in the grain size
distribution. The subscripts D and d refer to filter and
soil respectively.
Another important characteristic of the granular filter is the internal stability which is the ability to prevent
loss of its own small particles due to disturbing forces
such as seepage and vibration (Kenney & Lau 1985).
It is important that the filter does not undergo, due
to the action of dragging exerted by the fluid, appreciable variations of its particle size distribution and
permeability. In particular, Kzdi (1969) proposed a
method where the particle size distribution of the filter is split into coarser and finer portions at an arbitrary
grain diameter and the traditional retention criteria is
applied using the following expression:
v
where D15
is the particle size at which 15% is finer by
mass taken from the coarse particle size distribution,
k
k
D85
and D15
are the particle size at which 85 and 15%
are respectively finer by mass taken from the finer
particle size distribution.
Recently, a new theoretical method was developed
to evaluate the internal stability of granular soils
(Moraci et al. 2012a).
The design of a PRB requires the determination
of the thickness of the barrier. Usually two different approaches are used for organic compounds or
metals. In the first case it can be determined as
(v)*(t), where v is the groundwater flow velocity
and t is the residence time or contaminant contact
time with the reactive media. The residence time for
volatile organic compounds (VOCs) may be estimated
from reasonable first-order rate which depends on
the contaminants constituent degradation rate, maximum contaminant concentrations, and groundwater
flow rate (ITRC 2011).
The solution to a first-order decay rate is as follows:
where Ct = the concentration (mass per unit volume)

at time t; C0 = the initial concentration; k = the firstorder degradation coefficient.
Otherwise regarding metals, observed breakthrough curves can be used to quantify the removal
capacity of the reactive material or the ratio between
the removed contaminant mass and the mass of
exhausted reactive medium (ITRC 2011).
3
COLUMN EXPERIMENTS
3.1 Materials and methods

The used ZVI is of the type FERBLAST RI 850/3.5,
distributed by Pometon S.p.A., Mestre Italy. The
material is characterized by uniform grain size distribution (Fig. 1). The coefficient of uniformity
U = d60 /d10 is 2. The mean grain size (d50 ) is about
0.5 mm.
The used pumice originates from Lipari (Aeolian
Islands, Sicily Italy). The material is characterized by
uniform grain size distribution (Fig. 1). The coefficient
of uniformity U is 1.4. The mean grain size (d50 ) is
about 0.3 mm. This type of pumice has been chosen
since it was the available fraction closest to ZVI in
dimension.
The mixture of ZVI/pumice investigated was mixed
with a weight ratio respectively of 30:70.
The contaminated aqueous solutions were prepared
by dissolving nickel(II) nitrate hexahydrate (purity
99.999) in distilled water.
The aqueous concentrations of Ni and total dissolved iron, derived by iron corrosion, were measured
by Atomic Absorption Spectrophotometry (AAS
Shimadzu AA 6701F) using conventional Standard
Methods (APHA 2005).
The pH and Eh values were directly measured on
the samples (WTW GmbH, inolab pH/Cond 720).
During column tests hydraulic conductivity was
determined, using the falling-head or constant-head
permeability methods (Head & Keeton 2008).
3.2 Column tests apparatus
Columns experiments were carried out by using polymethyl methacrylate (PMMAPlexiglas) columns
with an internal diameter of 5 cm and a height of 50 cm,
equipped with sampling ports located at different distances from the inlet. A peristaltic pump (Ismatec ISM
Figure 1. Grain size distribution of Fe0 and pumice.
686
930) was used to feed the columns under constant

upward flow.
In order to ensure that the column tests are sufficiently representative to be used to determine PRB
design parameters, the physical conditions such as
composition and grain size distribution of the reactive material and hydraulic conditions should be kept
as similar as possible to those in situ. In particular,
according the latter aspect, the two main issues are
possible channeling and wall effects, associated with
the use of column tests, and the accurate reproduction
of in situ hydrodynamic conditions.
The prevention of channeling and wall effects
ensures that a small column of reactive medium
behaves similarly to an indefinitely large actual PRB
and in particular that the behavior of the reactive medium is not influenced by the proximity of
the column surface. According to Badruzzaman &
Westerhoff (2005), these effects may be avoided if the
ratio between the column diameter (D) and the mean
particle size (d) is greater than 50. This ratio, for the
material used, was about 130 for ZVI alone and about
150 for the ZVI/pumice mixture.
Reproduction of hydrodynamic conditions is somewhat more complex in practice.The best way to comply
with this criterion (Crittenden et al. 1991), is to keep
the experimental flow rates and Peclet numbers as similar as possible to the field values. Peclet number (Pe )
quantifies the relative importance of advection, molecular diffusion, and dispersion for solute transport and
is therefore important for evaluating the hydrodynamic
conditions of the test (Fetter 1999, Chun Kit 2004):
where the parameter v is the groundwater velocity in

the longitudinal direction (m2 sec1 ), d is the mean
or effective grain diameter (m), and D is the diffusion
coefficient of the contaminant in an aqueous environment (m2 sec1 ). A Peclet number smaller than 0.4
indicates that solute transport is diffusion controlled,
a number between 0.4 and 6 suggests a transition
between diffusion control and dispersion/convection
mechanisms and Pe greater than 6 indicates dominance
of dispersion and convection over molecular diffusion
(Fetter 1999).
column (P) and the volume of the column (V), s is

the particle density (g cm3 ) of the column medium.
In the case of mixtures the porosity was calculated as:
where Vv is the void volume, Vt is the volume of the

column occupied by the reactive medium, Ppom and
PZVI are the weight of Pumice and ZVI respectively
present in the column, ZVI is the ZVI particle density
(s ), pom can be referred to the pumice particle density
(s ) or to the apparent specific weight as . In the first
and latter case the inter-particular porosity (n) and the
total porosity (nt ) (included intra-particular porosity
of the pumice) can be respectively calculated.
The initial porosity of the granular mixture
ZVI/pumice has been estimated to be about 45%,
neglecting the internal porosity of the pumice. The initial porosity of used ZVI medium has been estimated
to be about 48%.
Three different values of flow rate (i.e. 0.1, 0.5
and 2.5 ml/min) and nickel concentrations (8, 40 and
95 mg/l), where used in the column experiments.
The flow rates 0.1, 0.5 and 2.5 ml/ml correspond
to a Darcy velocity of 0.07, 0.38 and 1.9 m/day
respectively.
In order to obtain information about hydrodynamic
conditions in the column tests the Peclet number
(Equation 5) was calculated (Table 1). For D, the diffusion coefficient of the contaminant, a literature value
of 1.59 109 m2 sec1 was used in the calculation
(O et al. 2009).
The analysis of the Peclet number reveals that diffusion plays a major role in tests carried out at flow rates
representative of field conditions, but that dispersion
and convection are not negligible. For tests employing
a moderately accelerated flow rate the Peclet number is significantly higher and the role of diffusion
is greatly reduced relative to field conditions. For
tests at the highest hydraulic load, the Peclet number is considerably greater than 6 and the effect of
diffusion is negligible. This indicates that the hydrodynamic conditions in accelerated column tests could
differ significantly from field conditions.
3.3 Column tests parameters

Column tests parameters can be referred to the test
conditions (reactive medium, contaminant solution
and flow rate) and to the reactive medium (volume,
mass and porosity of the reactive medium).
Porosity of the columns was calculated through
Equation 6:
Table 1.
where b is the bulk density (g cm3 ) which is the

ratio between the weight of the medium present in the
687
Determination of Peclet number.
Reactive medium
Q
(ml/min)
Pe
()
Mix. ZVI/Pum.
Mix. ZVI/Pum.
Mix. ZVI/Pum.
ZVI
ZVI
ZVI
0.1
0.5
2.5
0.1
0.5
2.5
0.40
2.14
11.14
0.42
2.29
11.92
The removal capacity of the reactive medium can

be expressed by the following Equation:
where Mrem = mass of removed contaminant, MZVI =

mass of exhausted material in a given column length,
referred to ZVI only.
The mass of removed contaminant was calculated
through a mass balance.
4
4.1

Long term behaviour of ZVI and granular
mixture ZVI/pumice
duration of the test (T). Efficiency reduction can be due

to reactivity or hydraulic conductivity reduction. The
first cause was considered when the breakthrough was
observed at the outlet of the column. Hydraulic conductivity reduction was considered when a reduction
of about two order of magnitude was observed. In Figures 23 hydraulic conductivity is plotted as function
of cumulative contaminant mass in input (Fig. 2) and as
function of time (Fig. 3), for the mixture ZVI/pumice
(Figs 2a, 3a) and for the ZVI (Figs 2b, 3b).
The results indicate that the hydraulic conductivity
behaviour is very sensitive to high flow rates and/or
concentration increase. In fact acceleration of column
tests through flow rate (keeping constant initial nickel
concentration at 40 mg/l) or contaminant concentration (keeping constant initial nickel concentration at
8 mg/l) leads to a change of the causes of efficiency
reduction of the reactive media. In particular, hydraulic
Generally, in column tests should be used values of

flow rate that approximate the in situ conditions, but
these circumstances imply a longer period of time until
the exhaustion (breakthrough) of the reactive medium
can be reached, since normally in situ velocities are
quite low. When the duration of the tests has to be
reduced, column tests can be accelerated by increasing,
proportionally to the in situ conditions and considering the desired PRB lifespan, the mass of contaminant
in input into the column (Bilardi et al. 2012a). Acceleration of column tests can be achieved using higher
values of flow rate and/or initial contaminant concentration respect to field conditions. In a previous study,
having the object to evaluate if accelerated column
tests could be representative of field conditions, column tests were carried out under different conditions
of flow rate (Q) and initial contaminant concentration
(C0 ) (Bilardi et al. 2012a, Bilardi 2012).
In Table 2 are specified the mass of ZVI and Pumice
used in the columns, the mass of Ni treated before
efficiency reduction (MNi ) and the corresponding
Table 2.
Column tests details.
Reactive medium
(gr)
Q
C0
MNi
(ml/min) (mg/l) (gr)
ZVI(240)/Pum(560)**
ZVI(240)/Pum(560)***
ZVI(240)/Pum(560)***
ZVI(240)/Pum(560)*
ZVI(240)/Pum(560)**
ZVI(240)/Pum(560)***
ZVI(240)/Pum(560)***
ZVI (1680)**
ZVI (240)***
ZVI (1680)***
ZVI (1680)***
ZVI (240)**
ZVI (1680)**
0.1
0.5
2.5
0.1
2.5
0.1
2.5
0.1
0.5
2.5
2.5
0.1
2.5
40
40
40
8
8
95
95
40
40
40
95
8
8
T
(days)
1.92 334
0.14
5
0.24
1.6
*
*
2.10 73
0.42 30
0.05
0.16
1.44 250
0.06
2.3
1.2
8.3
2.39
7
0.19 170
0.51 18
*The reactive medium is still maintaining its efficiency.

**Efficiency reduction due to hydraulic conductivity
reduction.
***Efficiency reduction due to reactivity reduction.
Figure 2. Hydraulic conductivity vs. cumulative contaminant mass input for a) the mixture ZVI/pumice and b)
the ZVI.
688
conductivity reduction is more frequent in two situation: i) when a low value of flow rate is used (i.e.
0.1 ml/min) or in case of low initial nickel concentration (i.e. 8 mg/l) and higher values of flow rate (i.e.
2.5 ml/min). Probably in these conditions iron corrosion by water is more favoured. As a consequence
production of bubbles gas and iron corrosion products
probably contribute to porosity reduction.
Therefore the long term efficiency of the reactive medium is strictly dependent on the groundwater
chemistry, contaminant concentration and groundwater flow rate. Therefore accelerated column tests
carried out using higher contaminant concentration or
flow rate respect to in situ conditions do not allow to
investigate the longevity of a reactive medium.
Results showed in Table 2 reveal that pumice is able
to extend the lifespan of the barrier. In fact the mass
of nickel treated before efficiency reduction (MNi ) is
for the granular mixture greater than the ZVI, unless

for column tests carried out at high flow rate (i.e.
Q = 2.5 ml/min) and using high values of nickel concentrations (i.e. 40 and 95 mg/l). In the latter cases it
should be considered that the mass of reactive medium
contained in the ZVI column is seven time greater than
that contained in the mixture.
The column tests carried out at nickel concentration of 8 mg/l and a flow rate of 0.1 ml/min, using
the mixture ZVI/pumice, is still maintaining its efficiency after a contaminant mass in input equal to 0.3 gr
(260 days).
4.2
Removal capacity of ZVI and granular

mixture ZVI/pumice
In Table 3 the removal capacity (RC) was calculated

using Equation 8. The RC is referred to the length
of column containing completed exhausted material
(L). In the first column of Table 3 in parenthesis
the mass of ZVI or of pumice referred to the length
L of the column is specified. The highest removal
capacity was calculated for column tests carried out
at low flow rate (0.1 ml/min) and initial contaminant concentration equal to 40 mg/l. The presence
of pumice allows to reduce the amount of ZVI to
be used in the barrier but increase its thickness and
thus the residence time which is beneficial for nickel
removal.
During a column test measures of pH are important
in order to have a fast information regarding the reactivity of the material. The pH measurements (data not
shown), carried out during column tests, revealed an
increase of pH along the length of the column up to
values of 910. The observed pH increase is due to iron
corrosion which releases OH ions into the solution.
pH is restored to the initial value of the influent solution (i.e. 5.56) when breakthrough occurs (Bilardi
et al. 2012a).
Table 3.
Removal capacity values.
Reactive medium
(gr)
Q
ml/min
C0
mg/l
MNi
gr
L
cm
RC
mg/gr
ZVI(240)
ZVI(134)/Pum(300)
ZVI(240)
ZVI(122)/Pum(86)
ZVI(240)/Pum(560)
ZVI(240)
ZVI(275)/Pum(193)
ZVI(1680)
ZVI(240)/Pum(560)
ZVI(979)
ZVI(240)/Pum(560)
ZVI(240)
ZVI(240)/Pum(560)
0.1
0.1
0.1
0.1
0.1
2.5
2.5
2.5
2.5
2.5
2.5
0.5
0.5
40
40
8
8
95
8
8
40
40
95
95
40
40
1.48
1.97
0.19
0.3
0.42
0.51
2.33
1.2
0.24
2.39
0.05
0.06
0.14
3
28
3
8
50
3
18
22.5
50
13
50
3
50
5.51
14.48
0.8
1.98
1.71
2.12
8.20
0.70
0.95
2.43
0.21
0.1
0.55
Note: MNi is the mass of Nickel flowed (gr), L is the length of

column contained exhausted material (cm), RC is the removal
capacity (mg Ni/gr ZVI).
Figure 3. Hydraulic conductivity vs. time for a) the mixture

ZVI/pumice and b) the ZVI.
689
tests are showed for two column tests. The first one
contained the mixture ZVI/pumice and using a flow
rate of 2.5 ml/min and an initial nickel concentration
of 40 mg/l (Fig. 4a).
The second one is the column contained the mixture
ZVI/pumice and using a flow rate of 0.1 ml/min and
an initial nickel concentration of 95 mg/l (Fig. 4b).
The general trend shows that the iron released by
the reactive medium is always below the limit concentration allowed in groundwater by Italian regulation
and equal to 0.2 mg/l (Gazzetta Ufficiale della Repubblica Italiana 2006) at the begging of the tests or when
the reactivity of the medium is maximum and the contaminant is easily removed. Iron concentration tends
to increase with the time in particular when the reactive medium starts to lose its reactivity and the pH is
restored to the initial value (Bilardi et al. 2012a).
Regarding release tests, the values of nickel concentrations revealed at the beginning of the test are
probably due to a plug flow effect of the contaminants
originally present in the column. After the contaminant contained into the void is completely expelled the
reactive medium does not release nickel or iron showing a good retention capacity.
Figure 4. Nickel and iron concentrations measured at the

outlet of a) ZVI/pumice granular mixture (Q = 2.5 ml/min
and C0 = 40 mg/l) and b) ZVI/pumice granular mixture
(Q = 0.1 ml/min and C0 = 95 mg/l).
4.3
Iron or contaminant release
It is important to evaluate during a column test the possible release of secondary reaction products derived by
the reaction between the contaminated solution and the
reactive medium. In this case the total dissolved iron
due to anaerobic corrosion of ZVI and by the abiotic
degradation of NO
3 was measured at the outlet of the
columns (Bilardi et al. 2012a).
The possible release of contaminants after exhaustion of the medium or when uncontaminated water
flow through was also evaluated. The medium after
reactivity exhaustion was permeated with distilled
water and the concentrations of nickel and iron were
measured at the outlet of the column (release test).
Measurements of the concentration of total dissolved iron at the outlet of the columns and release
CONCLUSIONS
In this paper were summarized the issues on the long

term behaviour of ZVI permeable reactive barriers and
the main geotechnical characteristics requested to the
reactive medium for a correct design of the barrier. In
particular, it was underlined the necessity for the reactive medium to satisfy the filter design criteria towards
the surrounding soil, thus the necessity to evaluate with
attention the particle size distribution and porosity of
the reactive medium.
Column tests results showed that the long term
behaviour of the reactive media cannot be predicted
through column tests accelerated respect to in situ
conditions. In fact, in these cases the main causes of
efficiency loss of the reactive media can be different.
The possible release of iron deserves attention during column tests, results revealed an increase of iron
above limit concentration when the reactive medium
loses its reactivity. For this reason it is necessary to
consider the possible replace of the reactive medium
or the placement of a medium downstream the barrier
able to intercept iron release.
In order to predict the long term behaviour of a PRB
it is necessary to develop numerical models able to take
in account the possible clogging and/or reactivity loss
of the medium.
REFERENCES
APHA, AWWA, WEF 2005. Standard Methods for the examination of water and wastewater, 21st ed. Washington D.C.
(USA): American Public Health Association.
Badruzzaman, M. & Westerhoff, P. 2005. The application of
rapid small-scale column tests in iron-based packed bed
690
arsenic treatment systems. Advances in arsenic research:

integration of experimental and observational studies and
implications for mitigation. New York: American Society
of Chemistry.
Bilardi, S. 2012. Short and long term behaviour of Fe0
and Fe0 /pumice granular mixtures to be used in PRB
for groundwater remediation. Ph.D. thesis in geotechnical engineering and materials chemistry (XXIV Cycle),
Universit Mediterranea di Reggio Calabria.
Bilardi, S., Calabr, P.S. & Moraci, N. 2012a. Are accelerated column tests used in permeable reactive barriers
design sufficiently reliable?. Proceeding of the third international conference Hazardous and industrial waste
management. 1214 September 2012. Crete: Greece.
Bilardi, S., Amos, R.T., Blowes, D.W., Calabr, P.S. & Moraci
N. 2012b. Reactive Transport Modeling of ZVI Column Experiments for Nickel Remediation. Ground Water
Monitoring & Remediation 33(1): 97104.
Car, S., Nguyen, Q.T., LHostis, V. & Berthaud, Y. 2008.
Mechanical properties of the rust layer induced by
impressed current method in reinforced mortar. Cement
and Concrete Research 38(89): 10791091.
Chun Kit, L. 2000. Laboratory and full-scale studies of a
permeable reactive barrier on the dechlorination of chlorinated aliphatic hydrocarbons. Ph.D. thesis, Hong Kong
University of Science and Technology.
Crittenden, J.C., Reddy, P.S., Arora, H. & Hand, J.T.D.W.
1991. Predicting GAC performance with Rapid SmallScale Column Tests. American Water Works Association
83(1).
Fetter, C.W. 1999. Contaminant Hydrogeology, 2nd edition.
New Jersey, USA: Prentice-Hall.
Gazzetta Ufficiale della Repubblica Italiana, 2006. Norme
in materia ambientale. Poligrafico dello Stato, Roma:
Gazzetta Ufficiale n. 88 del 14 aprile 2006 (in Italian).
Gu, B., Phelps, T.J., Liang, L., Dickey, M.J., Roh, Y., Kinsall,
B.L., Palumbo, A.V. & Jacobs, G.K. 1999. Biogeochemical dynamics in zero-valent iron columns: implications
for permeable reactive barriers. Environmental Science
Technology 33(13): 21702177.
Gu, B., Watson, D.B., Wu, L., Phillips, D.H., White, D.C. &
Zhou, J. 2002. Microbiological characteristics in a zerovalent iron reactive barrier. Environmental monitoring and
assessment 77(3): 293309.
Head, K.H. & Keeton, G.P. 2008. Permeability, shear
strength & compressibility tests. Manual of Soil Laboratory Testing vol. 2. United Kingdom: Whittles Publishing.
Henderson, A.D. & Demond, A.H. 2011. Impact of Solids
Formation and Gas Production on the Permeability of
ZVI PRBs. Journal of Environmental Engineering 137(8):
689696.
ITRC (Interstate Technology & Regulatory Council)

2011. Permeable Reactive Barrier: Technology Update.
Washington, D.C.: Interstate Technology & Regulatory
Council, PRB: Technology Update Team. www.
itrcweb.org.
Jeen, S-W., Gillham, R.W. & Przepiora, A. 2011.
Predictions of Long-term Performance of Granular Iron
Permeable Reactive Barriers: Field-scale Evaluation.
Journal of Contaminant Hydrology 123(1-2): 5064.
Kedzi, A. 1969. Increase of protective capacity of flood
control dikes. Department of Geotechnics, Technical University, Budapest, Hungary, Report n.1. [In Hungarian].
Kenney, T. C. & Lau, D. 1985. Internal stability of granular
filters. Canadian Geotechnical Journal 22(2): 215225.
Komnitsas, K., Bartzas, G., Fytas, K. & Paspaliaris, I. 2007.
Long-term efficiency and kinetic evaluation of ZVI barriers during clean-up of copper containing solutions.
Minerals Engineering 20(13): 12001209.
Moraci, N. & Calabr, P.S. 2010. Heavy Metals Removal
and Hydraulic Performance in Zero-Valen Iron/Pumice
Permeable Reactive Barriers. Journal of Environmental
Management 91(11): 23362341.
Moraci, N. 2010. Geotextile filter: design, characterization and factors affecting clogging and blinding limit
states. Theme lecture. Proceeding of the 9th International
Conference on Geosynthetics, 2327 May 2010, vol. 1:
413435. Guaruj: Brazil.
Moraci, N., Mandaglio, M.C. & Ielo D. 2012a. A New
Theoretical Method to Evaluate the Internal Stability of
Granular Soils. Canadian Geotechnical Journal 49(1):
4558.
Moraci, N., Ielo D. & Mandaglio M.C. 2012b. A new theoretical method to evaluate the upper limit of the retention
ratio for the design of geotextile filters in contact with
broadly granular soils. Geotextiles and Geomembranes
35: 5060.
O, J., Jeen, S., Gillham, R.W. & Giu, L. 2009. Effect of initial
corrosion rate on long-term performance of iron reactive
barriers: Column experiments and numerical simulation.
Journal of Contaminant Hydrology 103(34): 145156.
Ruhl, A.S., nal, N. & Jekel, M. 2012. Evaluation of
two-component Fe(0) fixed bed filters with porous materials for reductive dechlorination. Chemical Engineering
Journal 209: 401406.
Terzaghi, K. 1922. Der Grundbruch and Stauwerken und
Seine Verhuntung Forcheimer-Nummer Wasserkr, 17,
445449, quoted by Vafai (1996).
Zhao, Y., Haiyang, R., Hong, D. & Weiliang, J. 2011. Composition and expansion coefficient of rust based on X-ray
diffraction and thermal analysis. Corrosion Science 53(5):
16461658.
691

Flow numerical modeling for efficiency assessment of vertical

barriers in landfills
E. Koda & P. Osinski
Department of Geotechnical Engineering, Warsaw University of Life Sciences, SGGW, Warsaw, Poland
T. Koanka
GEOTEKO Projekty i Konsultacje Geotechniczne Sp. z o.o., Warsaw, Poland
ABSTRACT: There are several engineering disciplines, where construction of cut-off walls is commonly
used, these are land reclamation, hydroengineering structures, landfill containment systems or deep dewatered
excavation. The construction of vertical barriers is mainly based on cut-off wall mono or diphase technology
with the use of bentonite-cement mixture as a filling material. The paper mainly considers groundwater flow
numerical modeling conducted on landfill areas, where such solutions as vertical bentonite barriers were applied.
The remedial works on site consisted of several, commonly known technical solutions, however the present paper
focuses mainly on remediation effects of constructing cement-bentonite vertical cut-off walls surrounding the
landfills body. The example is provided to prove that the efficiency of proposed remedial solution is a method
which could be widely utilized on such contaminated sites as old MSW landfills. Moreover, to prove the reliability
of conducted research on quality of construction works and quality of obtained results, a presentation of outcomes,
confirmed by numerical modeling, and based on groundwater monitoring on site is provided.
INTRODUCTION
Nowadays, the vertical barriers are constructed with

the use of technologies most commonly based on
deep foundation and ground reinforcement techniques
(Brand 1994). The application of cut-off wall appears
to be a great solution as a part of the landfill site
reclamation process. The fundamental idea of cellular cut-offs is an isolating by using screen with
linking cross wall at certain longitudinal distances
(Koda & Stepien 2001). It protects the surrounding
environment against the pollutants transport and significantly decreases the level of soil and groundwater
contamination. The selection process of appropriate
construction technique depends on geological structure of the ground, durability and chemical resistance
of the material used, also the ability of heavy machinery entrance at the site and economical factor have
to be considered. Migration of the pollutants depends
on geological conditions, which can be favourable to
the transport of water and the substances it carries
(large thickness of permeable layers, high hydraulic
gradients) or it can completely stop and keep the transport away from the landfill (impermeable layers, low
gradients).
Since early 1962 to 1991 Radiowo landfill has
been used as MSW storage area for capital city of
Poland, Warsaw. Actually, it covers ca. 15 ha area, and
it is higher than 55 m. Since 90s only non-composted
wastes from Radiowo compostory are stored on the
landfill. The organic matter content of the waste is

ca. 4%. Central and southen parts of the landfill was
filled with old municipal wastes (1030 years), while
the upper layers in the north part was filled with fresh
non-composted products.
2
BENTONITE VERTICAL BARRIERS ON

SELECTED LANDFILL
The examples of introducing solutions described

above is the reclamation process on Radiowo landfill
(located near Warsaw) which consisted of bentonite
cut-off wall construction on boundaries of the landfill
body, leachate collecting drainage, degassing system,
ditches restoration, and landfill body engineering and
bioengineering of landfill slopes. The scheme of the
landfill where remedial works were introduced is
presented on figure 1.
The vertical barrier on the landfill was constructed
in mono phase diaphragm wall technique with the
use of bentonite self hardening suspension known
as Solidur. The purpose of the cut-off walls was to
decrease to a minimum the contaminants migration to
the aquifer of a usable groundwater level. The exact
content of the suspension was determined according
to the physical and chemical laboratory test results of
groundwater, lecheate and soil samples.
Moreover, during the full construction process the
geological supervision of works had been provided
693
Figure 1. Elements of remedial works applied on landfill (Koda et al. 2009).
monitoring and modeling of changed hydrogeological

conditions had to be performed.
3
Figure 2. In situ hydraulic conductivity test of bentonite

material at Radiowo landfill site (Koda & Skutnik 2003).
(control of geological structure and depths of impermeable layers during excavation), and suspension samples for laboratory tests were collected consistently.
The main purpose of the laboratory tests was to determine density and permeability of the suspension, as it
appears to be a crucial parameter for structure in such
conditions (Fratalocchi & Pasqualini 1998, Lipinski
et al. 2007). For further measurements of hydraulic
conductivity there were also in situ tests conducted.
They were performed with the use of BAT system as
outflow test at the different depths of the barrier
(Koda & Skutnik 2003).
The graph presented in figure 2 indicates that
the measured coefficient of permeability of bentonite
barrier makes it practically impermeable (presented
results consider hydraulic conductivity measured after
28 days). The examples of tests presented in the paper
are only small amount of all performed, however
all of them proved that permeability characteristics
are sufficiently promising to approve it as a reliable
containment system of the landfill (Lipinski et al.
2007, Soga & Yoshi 2010). However, to make all the
tests more reliable a further research, consisting of
FLOW AND TRANSPORT MODELING

BASED ON FEMWATER
The most reliable method of predicting the groundwater flow and transport is a numerical modeling.
Such simulation could be conducted with the use
of FEMWATER model (included in GMS package)
(Lin et al. 1997). To present the possibilities that the
software provides, the overall idea attached to it and
some examples of utilising it in engineering reality are
presented below.
The basis of FEMWATER model is the solution of the problem of groundwater flow described
by Richards equation, using 3-D finite elements.
The presented model is integrated with preprocesor/
postprocessor GMS (Groundwater Modelling
System) (EMRL 2003). The main equation for the
groundwater flow modeling is a partial differential
formula proposed by Richards (EMRL 2003):
where: F(h) = differential water capacity (d/dh);

= volume moisture content; h = height of hydrostatic pressure; t = time; kr = relative hydraulic conductivity; ks = hydraulic conductivity tensor in saturation zone; z = location height; q = source element.
In the saturation zone; F = 0 (search to zero in
FEMWATER), and kr as-sumed maximum values
( is equal to porosity, while kr = 1). In FEMWATER model, relations: F(h), (h) and kr (h) can be
defined as functions described by van Genuchten
(1980) or/and as linear functions. The parameter values of unsaturated zone, demanded in the numerical
model, for sands were accepted on the basis of the
literature data (Carsel & Parrish 1988, Rowe et al.
694
1995, Shackelford & Rowe 1998), while for wastes

permeability parameters were determined from site
test.
The assumed initial conditions for the flow
equation:
where: R = dewatering area; hi = defined initial condition, which can be received from area investigation
or from solution of flow equation (1) for established
state.
Boundary conditions for flow equation:
Dirichlets conditions:
Figure 3. The 3-D GMS numerical mesh for Radiowo
landfill (Koda 2005).
Neumanns conditions:
Cauchys conditions:
where: xb , yb , zb = coordinates on bounds; n

unit vector is normal to bound; hd = determined
value of Dirichlets functional; qn = Neumanns flow;
qc = Cauchys flow; Bd , Bn and Bc = bounds, respectively by Dirichlet, Neumann and Cauchy.
There are possible more sophisticated linked conditions, which enable to consider variable conditions
in precipitation period or conditions in period without
rainfall. The flow equation (1) subordinates both initial
and boundary conditions described by equations (3),
(4) and (5). This is possible to be solved by Galerkin
finite elements scheme. The model can be applied both
for steady flow and for the flow unstable with time.
The following sets of information can be calculated
and printed:
1. Information on the flow on the bound (changes
velocity, increasing and the sum of flow through
bounds).
2. Hydraulic height, considered as the sum of the
location level and hydraulic pressure height
(defined in every node).
3. The level of hydraulic pressure (defined in every
node).
4. Discharge capacity (defined in every node).
5. Volume moisture content (defined in Gauss point
at every element).
6. Darcy velocity (volume moisture content).
Figure 3 presents the generated model mesh, while
assigned boundary conditions for the task is shown
in figure 4. The analysed technical solution for the
model simulation consists of the protection system
concerning the vertical barrier to stop leachate outflow from the landfill. The model mesh consists of
8903 elements and 5289 nodes. The total area covered by model is approximately 88 ha, including 16
Figure 4. Boundary conditions for the numerical model of

Radiowo landfill (Koda 2005).
ha of the landfill area. In the numerical model, taking

into account geological deposits in subsoil and hydrogeological conditions in landfill surroundings (Koda
2004), four landfill subsoil materials were distinguish.
695
Beginning from the older (deep deposited) to

younger (occurred directly below surface level),
these are:
Tertiary clays, with coefficient of hydraulic conductivity ks = 5 109 m/s,
sandy clays, with coefficient of hydraulic conductivity ks = 1 107 m/s,
variety grained sands, with coefficient of hydraulic
conductivity ks = 5 104 m/s,
dense sands, with coefficient of hydraulic conductivity ks = 5 105 m/s.
The parameter values of unsaturated zone,
demanded in the numerical model, for sands were
accepted on the basis of the literature data (Carsel and
Parrish, 1988). They were, as follows:
differential water capacity, F = 0.0725,
volume water capacity , reached from = 0.045
for h = 4 m to = 0.36 for h = 0,
relative hydraulic conductivity, reached from kr = 0
for h = 4 m to kr = 1 for h = 0.
Municipal waste, with hydraulic conductivity coefficient ks = 1 105 m/s (based on laboratory tests),
are placed in the model centre as well as in the upper
part of model. In is assumed, that parameter values
of wastes in the unsaturated zone can be the same as
for sandy soils. These parameters were assumed as
follows:
F = 0.00588,
, reached from = 0.057 for h = 60 m to
= 0.41 for h = 0,
kr , reached from kr = 0 for h = 60 m to kr = 1 for
h = 0.
Along the landfill border, there is the vertical
bentonite barrier (cut-off wall) with hydraulic conductivity coefficient of ks = 5 1010 m/s (value obtained
from field tests, Koda 2012).
It was assumed, that parameter values of bentonite
material in the unsaturated zone are similar as for clays.
These parameters as follows:
F = 0.00725,
, reached from = 0.07 for h = 4 m to = 0.36
for h = 0,
kr , reached from kr = 0.0885 for h = 4 m to kr = 1
for h = 0.
On the situation map, the model outside borders
is overlapped with surface streams on the landfill surroundings.These streams were shaped in the numerical
model as assigned in Dirichlets boundary conditions.
The constant hydraulic gradient equals to the water
table level in these streams, reached from 17 (in the
northern part of area) to 23.45 m a.s.l. (in the southern
part of area).
Neumans boundary conditions (qN = 3.17
108 m/s) were assigned in the model for the part of
the landfill area where the leachate and rain water
from the compostory plant are pumped (re-circulation
system). In the case of the intensive precipitation,
Figure 5. Groundwater contour map for subsoil without the

vertical barrier (position of groundwater table in m above 0
of Vistula River).
Neumans boundary condition was also taken into consideration in the model, i.e. on the landfill slopes of
qN = 3.17 1010 m/s, while on the surroundings area
it was qN = 1.58 109 m/s.
Initial conditions were assumed from preliminary
calculations for the exemplary projection, in which
polluted water pumping on the landfill surface is not
provided.
Numerical calculations were carried out for the two
following examples:
inflow of polluted water (re-circulation of leachate
and rain water from compostory area) on the landfill surface, where the vertical barrier has not been
constructed yet,
inflow of polluted water on the landfill surface,
where vertical barrier has already been constructed.
4
MODELING RESULTS
The results of the groundwater flow numerical simulation were worked out as the groundwater contour
map (figures 5 and 6) and the maps of groundwater
table level changes caused by the vertical barrier (figure 6). The shape of groundwater level and its changes,
696
Figure 6. The groundwater contour map for subsoil with the

vertical barrier, after 5 years of performance (in m above 0
of Vistula River).
presented on the maps, are suitable for stable situation

of groundwater flow conditions.
Due to the vertical barrier construction, the groundwater table level in the landfill area was also changed.
On the landfill, the groundwater level will increase
from approximately 0.25 m in the southern part to
more than 0.75 m in the north-western part of the landfill. On the surrounding area, the groundwater level
may increase or decrease, depending on the zone of
area (figure 7).
The largest decrease of the groundwater level (ca.
1 m) will take place in the northern part of the area,
directly close to the vertical barrier. However, small
increase of the groundwater level (ca. 0.2 m), will
occur in the part close to the south part of the landfill. The vertical barrier composes the obstacle for
groundwater flow from the southern to the northern
direction.
5
CONCLUSIONS
The influence of the vertical bentonite barrier on

the groundwater flow on the landfill surroundings
is observed in the frame of the local water quality
Figure 7. The contour map of the groundwater level changes

(in meters) caused by the vertical barrier after 5 year of
performance.
monitoring. For the final assessment of the vertical

barrier efficiency, long-term monitoring observations
are necessary. The numerical modelling method is useful for investigation of the vertical barrier influence
on the groundwater flow. The results of the numerical modeling for Radiowo landfill, presented in the
paper, proved the isolation role of the vertical bentonite
barrier When the vertical barrier wall was closed,
the quality of the groundwater (the first layer) on
Radiowo landfill surroundings has been essentially
improved. Such statement was proved by analyzing
the monitoring results presented in table 1.
Additionally, on figure 8 a significant reduce of
chloride e concentration is presented. The results are
obtained from a long term monitoring provided at the
site (please see the location of piezometers on figures
5, 6 and 7).
As presented above the numerical modeling could
also be used for the assessment of the efficiency of
engineering solutions on site. Moreover, the modeling
allows assessing the cut-off wall effectiveness according to its preventing character. The example provided
proved that such solution is a reliable containment
system providing the sufficient limitation of the pollutants migration.
697
Table 1. The influence of the vertical bentonite barrier

surrounding landfill on groundwater quality at Radiowo
landfill.
Pollution
indicator
Unit
Electrical
S/cm
conductivity
11.07.00
10830
BOD5
mgO2 /dm3
650
CODCr
mgO2 /dm3
1758
Ammonium
Nitrogen
mgNNH4 /dm3 97,7
Chlorides
mgCl /dm3
Sulphates
3 690
mgSO2
4 /dm
Copper
mgCu/dm3
2,1
Lead
mgPb/dm3
0,2
2374
Closureof the vertical barrier, Nov. 2000
Date/Concentration of
pollutants
7.11.05 02.03.10
3170
921
49
8,5
193
124
0,29
<0,04
274
31,9
805
131
0,228
0,05
<0,004 0,005
Figure 8. Moving of the maximum chlorides concentration

(in mg Cl/dm3 ) towards flow in the first groundwater layer
between two piezomenters (Koda 2012).
REFERENCES
Brandl H. 1998. Vertical barriers for municipal and hazardous waste containment. Proceedings Symposium on
Developments in geotechnical Engineering, Bangkok:
301334.
Carsel R.F., Parrish R.S. 1988, Developing joint probability
distribution of soilwater retention characteristics. Water
Resources Research, Vol. 24, No. 5:755760.
EMRL. 2003. Groundwater modeling system tutorial volume
I-IV, UT: Brigham Young University.
Fratalocchi E. Pasqualini E. 1998. Permeability over time

of cement-bentonite slurry walls. Proceedings of the 3rd
International Congress on Environmental Geotechnics.
Lisbona Vol. 2: 509514.
Koda E. 2011. Stability and pollutant transport from remediated landfills with the use of Observational Method,
SGGW, Warsaw [In Polish].
Koda E., Golimowski J., Wiencaw E. 2009. Assessment of
efficiency of the old landfill protection system based on
transport modelling and monitoring research. 17th International Conference on Soil Mechanics and Geotechnical
Engineering Alexandria. Vol. 3: 19771980.
Koda E., Skutnik Z. 2003. Quality control tests of vertical bentonite barriers for old sanitary landfill containment. Proceedings of the 13th European Conference on
Soil Mechanics and Geotechnical Engineering, Prague.
Vol. 1: 409414.
Koda E., Stepien M. 2001. Application of Vertical Bentonite
Barriers Surrounding Sanitary Landfills. 4th International Waste Forum Systems of Waste Management.
PoznanPia: 112
Koda E., Wiencaw E. 2005. Flow and transport modeling in
old landfill subsoil with vertical barrier. Proceedings of
the 16th International Conference on Soil Mechanics and
Geotechnical Engineering. Osaka. Vol. 2: 921924.
Koda, E. 2004. Local water quality monitoring on surroundings of the sanitary landfill.Annals ofWarsawAgricultural
University. SGGW, Land Reclamation. 35a: 131144.
Lin H.C., Richards D.R., Yeh G.T., Cheng J.R., Cheng H.P.,
Jonem N.L. 1997. FEMWATER: a three-dimensional
finite element computer model for simulating densitydependent flow and transport in variably saturated
media. Technical Report CHL-97-12. US: Army Corps of
Engineer
Lipinski M. J., Koda E., Wdowska M.K. 2007. Assessment of key geotechnical characteristics of a groundwater
protective vertical barrier, Proceedings of the 14th European Conference on Soil Mechanics and Geotechnical
Engineering, Madrid. Vol. 2: 767772.
Rowe R.K., Quigley R.M., Broker J.R. 1995. Clayey Barrier Systems for Waste Disposal Facilities. London:
Chapman & Hall.
Shackelford C.D., Rowe R.K. 1998. Contaminant Transport
Modeling. Proceedings of the 3rd International Congress
on Environmental Geotechnics, Lisboa. Vol. 3: 939956.
Soga K., Joshi K.. 2010. Long-term Performance of cementbentonite cut-off walls: A case study. Proceedings of the
6th International Congress on Environmental Geotechnics, New Delhi. Vol. 1: 151164.
Van Genuchten M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soil. Soil
Science Society Journal. 44: 892898.
698

Design, installation and performance of containment system

at a petroleum-release site
A. Sanzeni
University of Brescia, DICATAM, Brescia, Italy
M. Morando, E. Crescini & A. Guerini

NCE srl, Brescia, Italy
ABSTRACT: A sealed sheet pile barrier was designed and installed, coupled with a hydraulic system for
groundwater level control, to contain the migration of contaminants from a petroleum-release site in Italy. The
environmental data regarding the nature, concentration and distribution of contaminants have been integrated
with a comprehensive geotechnical investigation campaign designed and carried out to define the physical
and mechanical properties of a heterogeneous subsoil deposit. Such a thorough investigation was essential for
assessing the feasibility of sheet piles installation and to investigate the possibility of creating a vertical cut-off
wall driven into a low-permeability clayey soil layer. The barrier was installed in problematic conditions due
to the presence of superficial coarse grain soil layers, the large number of above and underground services,
and the proximity to a riverbed slope. Performance pumping tests and chemical analyses on water samples are
demonstrating the effectiveness of these systems in containing contaminants migration.
INTRODUCTION
In 2003 at the Rome Refinery site, the finished products loading bay area was involved in an accidental
release of petroleum hydrocarbons from a pipeline,
causing an impairment of soil and groundwater.
During the following site investigation, concentrations above legal threshold values (Italian Legislative
Decree no. 152/06) of dissolved hydrocarbons, MTBE
and BTEX compounds were detected in a 300 m wide
area, adjacent to a seasonal creek (Rio Galeria), located
downstream of the expanding plume. The release accident was followed by immediate implementation of
emergency safety measures, consisting of a hydraulic
barrier equipped with total fluid pneumatic pumps and
skimmers, in order to contain the contaminated plume
and to remove free product floating on the water table.
Based on detailed soil and groundwater investigations, in 2007 Rome Refinery started the design
process for a comprehensive remediation project (on
the entire refinery site), that included a permanent containment system for controlling the off-site migration
of contaminants in groundwater. The system consisted
of a sealed sheet pile barrier coupled with groundwater level control system, plus a Soil Vapor Extraction
system (SVE) for the remediation of volatile organic
compounds detected in the vadose zone. In 2009,
a more extensive investigation campaign, including
geotechnical field and laboratory testing, was performed to obtain data for the design of the permanent
containment system; environmental data regarding the

nature, concentration and distribution of contaminants
have been therefore integrated with a comprehensive
geotechnical characterization for the design of the
containment system. Due to the complex geological
setting of the region, the soil profile at the realease
site is highly heterogeneous and comprises a superficial layer of fill material, over a number of layers
of dense sandy to soft clayey soils, with water table
interacting with the nearby seasonal river.
The sheet piles containment system was completed
in 2011 and a number of permanent monitoring wells
were added to the existing network, in order to allow
extensive periodical measurements of groundwater
level and contaminants concentration, upstream and
dowstream of the barrier.
The barrier was installed in challening conditions
due to the presence of superficial coarse grain soil
layers and to the limitations originated by the presence
of the refinery main access road, adjacent to the barrier route. In addition, the presence of aboveground
technological plants and underground services and
pipelines, as well as the proximity to the riverbed slope,
generated supplementary obstacles to the construction
of the physical barrier.
This paper describes the investigation activities,
the design and installation of the contaiment system, and presents a selection of data obtained by the
implemented monitoring system.
699
ENVIRONMENTAL AND GEOTECHNICAL

INVESTIGATIONS
3.1 Previous studies

Soil and groundwater characterization carried out
by Rome Refinery between 2003 and 2007 allowed
to identify the main geological and hydrogeological properties and the nature and extent of the
contamination.
Soil borings were driven down to 15 m below
ground level (b.g.l.) at 79 different locations, all
equipped as observation wells, some of them completed with larger diameter casing in order to function
as extraction wells.
Groundwater table depth varies from 7 m b.g.l. in
the eastern part of the area, nearby the riverbed, to
4 m b.g.l. on its west side, with an hydraulic gradient
varying between 4% in the south-eastern portion, next
to the river, and 1%, at the North-west side.
Other hydraulic properties of the aquifer were
estimated by means of pumping tests conducted at
different locations, allowing to identify a significant
variation in transmissivity, between 5 106 m2 /s in
the northern sites portion (where aquifer thickness
is 7 m) and 1 103 m2 /s at its south-eastern side
(aquifer thickness of 12 m).
Dissolved petroleum hydrocarbons were measured
in groundwater up to values in the range of 1000
to 2000 g/l. Total petroleum hydrocarbons were
detected up to concentrations in the range of 5000 to
10000 mg/kg in the soil matrix.
The first emergency solutions (taken immediately
after the 2003 accident) consisted in the installation
of a hydraulic barrier to prevent the plume migration
toward the river. This operation made it possible to
recover most of the floating product and to initiate the
mitigation of the dissolved contaminants.
Figure 1. Plan view of the loading bay area.
3.2 Environmental investigation and studies
Figure 2. Cross-section of the refinery area and riverbed

slope.
SITE DESCRIPTION
Figure 1 shows the site layout: the finished products

Loading Bay covers 70 000 m2 and is located to the
West of Rio Galeria, at an elevation of approximately
20 m above sea level (a.s.l.). Figure 2 shows an illustrative cross-section of the refinery area and the riverbed
slope.
During the design process (started in 2007) supplementary environmental investigations and studies
were carried out by Rome Refinery and its consultants (see Figure 3). As a result, in 2010 the owner
obtained the Authorities approval of a comprehensive
remedial project for the entire site, including activities
on other areas that are not discussed in this paper.
In the Loading Bay area, the result of monitoring
activities carried out during the emergency operations
(hydraulic containment) suggested the opportunity to
install a physical barrier coupled with a water level
control system. In addition, the results of monitoring
activities carried out on soil gas and data obtained from
a pilot test indicated the opportunity of running a SVE
system up and downstream of the cut-off wall.
3.3 Geotechnical investigation
In 2009 a geotechnical investigation, including field
and laboratory tests, was performed in order to
700
Figure 3. Plan view of the petroleum release site, showing contaminant concentrations before sheet pile installation
(2007 campaign).
characterize subsoil conditions for the design of the

sheet pile cut-off barrier.
The geotechnical campaign consisted of several
bore-holes, with execution of Standard Penetration
Tests (SPT) and Lefranc in-situ permeability tests,
as well as the installation of water table observation wells. A considerable number of disturbed and
undisturbed soil samples were obtained for subsequent
laboratory testing. The site investigation campaign
was completed with the execution of 7 piezocone
static penetration tests (CPTU) performed along the
sheetpile route, down to 1819 m below ground level.
The laboratory investigation included classification
tests (particle size distribution and Atterberg limits), 1-D consolidation and uniaxial compression tests
on undisturbed specimens and oedometer hydraulic
conductivity tests.
The results of the geotechnical investigation along
the barrier layout indicated the presence of a highly
heterogeneous soil deposit that included: 1) a superficial layer of loose-medium fill (made ground, approximately 2 m thick, NSPT = 1020, qc = 25 MPa); 2) a
layer of medium-dense sand (25 m thick, NSPT = 20
25, qc = 510 MPa); 3) a heterogeneous 6 m thick layer
of silt and silty sand (NSPT = 210, qc = 25 MPa);
4) a 23 m thick deposit of low-medium plasticity
clay (5060% of silt and clay), and finally; 5) another
layer of silty and clean sand. Figure 4 shows an interpreted cross-section of the soil stratigraphy beneath
the refinery area along the sheet pile wall layout.
The results of the field investigation campaign were
interpreted on the basis of the authors experience with
local soil deposits, with empirical correlations with
results of SPT and CPTU tests (Colombo & Colleselli,
2004), and were compared with results of laboratory tests. Table 1 summarizes the main geotechnical
parameters of each soil layer, Figure 5 shows typical
SPT results of the 2009 campaign.
Figure 4. Interpreted geotechnical cross-section along the

refinery access road.
701
Table 1.
Main geotechnical parameters.
Layer Type
#
TopBottom c
(m b.g.l.)
( ) (kPa)
cu
k
(kPa) m/s
1
2
3
4
5
0.02.0
2.06.0
6.014.0
14.016.0
16.0
75
Fill, R
SC-SM
SM
CL
SM
34
30
30
27
30
0
510
0
1015
0
5 106
4 108
Figure 6. Typical result of the geo-radar investigation,

interpreted cross-section.
Figure 5. Typical SPT results along the physical barrier

layout.
3.4
Geophysical subsurface utilities

investigation
A geo-radar investigation was carried out to identify

the presence of underground utilities along the sheet
pile route.
Figure 6 shows an interpreted cross-section along
the barrier layout in the loading bay area with
identification of possible submerged services and
pipelines.
The length of sheet piles (15 m) was defined in

order to create a vertical cut-off wall driven into the
low-permeability clayey soil layer no. 4 (Table 1 and
Figure 4), able to reduce seepage below the barrier.
In order to guarantee an overall target permeability in the order of 108 m/s, polyurethane sealants
(relative permeability 0.3 1010 m/s) were also
implemented. The design of the sheet piles barrier
included calculations on required durability of the
steel structure as well as stability analyses to assess
safety of the nearby riverbed slope (ArcelorMittal,
2008).
4.2
4 THE DESIGN OF THE CONTAINMENT
SYSTEM
4.1
Geotechnical aspects of the sheet piles design
Based on the results of the geotechnical investigation,

Larssen 604 profile (S355GP steel) was considered
feasible for the subject site.
Hydraulic containment system
The physical containment system was coupled with 7

extraction wells (127 mm diameter PVC casing, driven
down to 15 m b.g.l.) to control the upstream hydraulic
head/groundwater level (Figs. 7 and 8).
An additional hydraulic containment system, consisting of 5 extraction wells, was installed in the
northern part of the loading bay area, where the physical barrier was unfeasible due to the presence of
702
Figure 8. Cross-section of physical barrier with SVE system

and hydraulic containment system.
Figure 7. Physical barrier, equipped with groundwater

extraction wells for up-gradient level control and monitoring
wells (inside/outside physical barrier).
Figure 9. Plan view of the area with the SVE system layout.
underground services. All the wells were equipped

with submersible pumps with variable-speed controllers, level and flow-rate monitoring sensors and
data loggers.
4.3 Soil Vapor Extraction system

In the northern portion of the physical barrier, a Soil
Vapor Extraction (SVE) system was designed and
installed for the remediation of volatile organic compounds in the vadose zone. As presented in Figure 9, 14
SVE wells (9 located outside and 5 inside the physical
barrier) were installed in the area. The wells were connected to a vacuum unit, which operates at a flow-rate
between 200 and 450 m3 /h, and to a catalytic oxidizer
for the abatement of the volatile contaminants, stripped
out from the soil.
CONSTRUCTION AND MONITORING

SYSTEM
The systems in the Loading Bay area (i.e. physical barrier, hydraulic containment and SVE) were installed
between September 2010 and August 2011. Before
installing the sheet pile barrier, a 1.0 m deep excavation was executed to remove the road pavement and
the most superficial-dense soil layer (Fig. 10a, 10b).
Based on the results of the geotechnical investigation, the sheet pile elements were driven (by means
of low amplitude, high frequency vibrodriving equipment) to a depth of 15.5 m b.g.l. into the clayey soil
layer #4 (Table 1). Due to the presence of the mediumdense sandy soil layer #2, between 2.0 and 6.0 m b.g.l.,
the installation was occasionally carried out using
703
Figure 11. Photographic image of the sheet pile barrier

during installation near the Canteen building (December
2010).
Figure 12. Result of a pumping test performed to evaluate

hydraulic performance of the barrier.
vibrodriving load was reduced because of the proximity to the Canteen building of the production plant
(Fig. 11).
Figure 10. Photographic images taken during installation.
water jetting or pre-augering. This procedure however

did not prove sufficient for a 30 m long segment of the
barrier (between BH 6 and BH 7 in Figure 5), where a
number of sheet piles could be driven only to a depth
of 12.5 m b.g.l. due to the presence of an unexpected
consistent clay layer between 11.0 and 15.0 m b.g.l.
Also, during the installation of this wall segment, the
RESULTS AND CONCLUSIONS
At the end of the construction, a set of tests was performed to verify the effectiveness of the hydraulic
confinement provided by the cut-off barrier. The
tests were executed with the couples of monitoring
wells installed, respectively, up and downstream of the
barrier (Fig. 7). During each test, a constant 0.1 l/s
flow-rate was extracted from the well located upstream
of the barrier and groundwater level was measured in
the corresponding downstream well. Under these operating conditions, the water level in the upstream wells
was lowered by 2.02.5 m, while the downstream wells
(located 23 m downstream of the barrier) remained
substantially unaffected (Figure 12).
The results of the pumping tests, were confirmed
by groundwater level monitoring sessions, periodically carried out after the installation of the system.
As presented in Figure 13, the physical barrier and
704
The positive effects of the systems (including

the SVE) are also confirmed by quality monitoring
sessions periodically performed on soil-gas and
groundwater. After almost one year from the
installation/start-up of the system, a significant reduction of contaminants was detected in almost all the
monitoring points located in the area downstream of
the physical barrier: the concentration of petroleum
hydrocarbons in soil-gas and ground water decreased
by one order of magnitude from the initial (2007)
measured values; Benzene and MTBE concentrations in groundwater were permanently lower than the
maximum allowable limits.
ACKNOWLEDGEMENTS
The authors wish to thank the HSE team and the Rome
Refinery management for the productive collaboration
and for giving permission of publication.
REFERENCES
Figure 13. Groundwater level contour lines after installation

of the physical barrier and the hydraulic containment system.
ArcelorMittal Commercial RPS. 2008. Piling Handbook.

8th edition.
Cestari, F. 2005. Prove geotecniche in situ. 3rd edition.
Segrate: Geo-graph.
Colombo, P. & Colleselli, F. 2004. Elementi di Geotecnica.
3rd edition. Bologna: Zanichelli.
the hydraulic containment system are able to control the groundwater flow towards the seasonal creek
(Rio Galeria) and, therefore, the off-site migration of
groundwater contaminants.
705

Hydraulic barrier performance of soil bentonite mixture cutoff wall

A. Takai, T. Inui & T. Katsumi
M. Kamon
Kagawa National College of Technology, Kagawa, Japan
S. Araki
Raito Kogyo Co., Ltd., Tokyo, Japan
ABSTRACT: Soil-bentonite mixture (SBM) is one of the materials with a hydraulic conductivity low enough to
be used as a containment barrier for contaminated soil. The main concern regarding in-situ containment barriers is
the assurance of their high hydraulic barrier performance which is the result of the coupled phenomena of physical
and chemical factors. This paper addresses the results of laboratory tests that focused on the hydraulic barrier
performance of SBM. Based on the results, hydraulic conductivity decreased when the bentonite powder content
is reduced and the chemical concentration increased. The cation in pore water led to a more significant increase
in the hydraulic conductivity of SBM than that in the permeant. Piezocone tests (CPTU) were employed to a
large-scale soil tank to evaluate the vertically homogeneity. CPTU has the potential to be used for the evaluation
of vertical homogeneity of constructed wall, since the profiles of physical properties change with the bentonite
powder content and the hydraulic conductivity could be estimated briefly by pore pressure dissipation test.
1
INTRODUCTION
Vertical cut-off walls are often constructed at

contaminated sites to contain the contaminants and
prevent their migration in the aquifer (Evans 1994,
Yeo et al. 2009). Soil-bentonite mixture (SBM) is one
of the most widely used materials since it provides
extremely low hydraulic conductivity as well as sufficient deformability (e.g. Grube 1992). Besides, high
homogeneity of the constructed SBM cut-off wall can
be achieved by employing the trench cutting and remixing deep wall (TRD) method as shown in Figure 1
(e.g. Kamon et al. 1998).
The main concerns of containment barriers are the
assurance of their hydraulic barrier performance and
homogeneity. Homogeneity plays a fundamental role
Figure 1. Construction procedures of SBM cut-off wall with

TRD method (Katsumi et al. 2008b).
on the quality of containment barriers since a larger

variability in the hydraulic conductivity leads to a
higher flux of contaminant out of the barrier system
even though the average hydraulic conductivity values
are equal (Britton & Filz 2007, Yesiller & Shackelford
2010).
This paper addresses the experimental results of
hydraulic conductivity tests for the SBM specimens,
which were made with different bentonite powder
contents and with different chemical concentrations
in permeants as well as in soil pore water. Piezocone test (CPTU) was also employed to a largescale soil tank in order to assess its potential as a
quality control/quality assurance (QC/QA) method
of SBM cut-off wall. CPTU is commonly used to
estimate profiles of soil characteristics (e.g. Chai
et al. 2011, Jeffries et al. 1993, Manassero 1994,
Robertson 2009). Since the profiles of the physical properties obtained from CPTU, which are corrected cone resistance (qt ), sleeve friction ( fs ), and
pore pressure (u), change by the soil characteristics, it is possible to classify the soil type by the
penetration.
Hydraulic conductivity of clay-based materials is
obtained as the results of physical and chemical
aspects, such as viscous resistance of fluid and interactions of chemicals with the clay mineral surface
(Katsumi et al. 2008a, Takai et al. 2012). Also about
cone penetration, both friction of soil and dissipation of pore pressure must be considered as a
coupled-phenomena on the interpretation of testing
707
results. On-site and in-situ evaluation of hydraulic

barrier performance should be carried out, considering both factors affecting the hydraulic conductivity
and the behavior of surrounding soils during the cone
penetration.
2
composite soil was prepared by mixing sandy gravel

and volcanic cohesive soil at their natural water contents with a ratio of 25:4 by dry weight. This ratio was
determined based on a boring log obtained at the pilot
test site. Table 1 lists the physical properties of these
soils.The chemical concentration in the pore water was
adjusted to the target value by using calcium chloride
(CaCl2 ) solution, artificial sea water and heavy fuel oil
A. Ca2+ was selected since it is commonly found in
natural aqueous system as well as in water discharged
from industrial processes or leached from waste. Sea
water and heavy fuel oil A were used to evaluate the
chemical compatibility of SBM constructed near the
seashore and at an oil contaminated site, respectively.
In practical construction procedures of SBM cutoff wall using TRD method, first, the trench cutting is
conducted supplying bentonite slurry to maintain the
workability of the soil inside the trench. Then, bentonite powder is re-mixed in the trench after the base
machine goes back to the initial position. To simulate
these construction processes in the laboratory, 10%concentration hydrated bentonite slurry was blended
with the soil. The additive content of the bentonite
slurry was determined based on the flowability of the
soil-slurry mixture (approximately 150 mm flow value
according to JIS R 5201). Once a mixture of suitable
EXPERIMENTAL PROCEDURE
2.1
SBM preparation
Two different soils were used for these studies as base

soil of SBM: 1) composite soil which is a mixture of
volcanic cohesive soil and sandy gravel collected at a
pilot scale test site and 2) silica sand #7. These soils
were sieved through a 4.75 mm-opening screen. The
Table 1.
Physical properties of two soils used.

Composite Silica
soil
sand
Grain size distribution (%)

Gravel fraction [2 mm ]
Sand fraction [75 m2mm]
Silt & Clay fractions [75 m]
Hydraulic conductivity (m/s)
5.6
70.8
23.6
1.5 107
0
99.6
0.4
3.9 105
Table 2. Testing conditions for hydraulic conductivity test.
Bentonite powder
content
kg/m3
Hydraulic
conductivity
m/s
Test No.
Soil type
Chemical type in pore water
Type of permeant
P-1
P-2
P-3
P-4
P-5
Composite soil
Distilled water
0.25 M-CaCl2
1.0 M-CaCl2
Seawater
50%-ethanol
100
5.0 1011
2.2 1010
1.4 1010
1.2 1010
4.9 1011
N-1
N-2
Composite soil
0.1 M-CaCl2
50
100
2.3 1010
1.2 1010
N-3
0.01 M-CaCl2
0.1 M-CaCl2
100
2.2 1010
N-4
N-5
N-6
0.025 M-CaCl2
0.1 M-CaCl2
50
75
100
1.5 109
6.8 1010
5.6 1010
N-7
N-8
N-9
0.05 M-CaCl2
0.1 M-CaCl2
75
100
125
1.1 109
1.0 109
7.4 1010
N-10
N-11
N-12
0.1 M-CaCl2
0.1 M-CaCl2
100
125
150
1.3 109
1.1 109
7.0 1010
N-13
N-14
N-15
Sea water
5,000 mg/kg-heavy fuel oil A
10.000 mg/kg-heavy fuel oil A
Sea water
0.1 M-CaCl2
0.1 M-CaCl2
100
100
100
9.8 1010
1.0 1010
8.3 1011
Distilled water
25
50
100
2.7 109
8.7 1010
1.1 1010
S-1
S-2
S-3
Silica sand
Composite soil: a mixture of volcanic cohesive soil and sandy gravel (collected at a pilot test site).
CaCl2 : concentration in pore water of parent material, Heavy fuel oil A: concentration in parent material (dry weight basis),
Seawater: pore water of the parent material was replaced by the seawater.
2
708
flowability was established, 25 to 150 kg/m3 of bentonite powder was added and re-mixed using the soil
mixer for 2 minutes.
2.2 Hydraulic conductivity test
All testing conditions for the hydraulic conductivity
test are summarized in Table 2. After pre-consolidation
at 30 kPa in the oedometer, SBM specimens, having
2030 mm in height and 60 mm in diameter, were
set on the hydraulic conductivity test apparatuses. A
flexible-wall permeameter with a fallen-head system
according to ASTM D5084 was employed. A confining pressure of 30 kPa and a hydraulic gradient of
approximately 3050 were applied during permeation.
Permeation was continued until the following requirements were confirmed: 1) the volume of the effluent
and the influent were almost equal, 2) the change in
hydraulic conductivity values with time was negligible, 3) pore volumes of flow were greater than 2,
and 4) the electrical conductivity of the effluent was
almost equal to that of the influent in the P and Nseries (e.g. Shackelford et al. 1999). The electrical
conductivity was of the collected effluent was measured in the bottle after certain amount of the solution
was accumulated. P-series was designed to assess the
chemical compatibility of SBM attacked by the solution containing chemicals, such as CaCl2 , seawater and
50%-ethanol. In this series, bentonite in SBM has been
firstly hydrated by original pore water. In the N-series,
the effect of the content of bentonite powder (CBP ) and
the chemicals in original pore water on the hydraulic
conductivity was verified. In this series, swelling of
bentonite is impeded due to the chemicals in the soil
pore water at a site. In the S-series, SBMs made with
two soils were tested to evaluate the effects of the type
of base soil.
Figure 2. Detail of a large-scale soil tank used for CPTU.
in JGS 1435-2003. During the test, the profile of qt , fs ,

and u were logged continuously. At 0.75 m depth, the
penetration was suspended to evaluate the hydraulic
conductivity with the pore pressure dissipation test.
3
2.3
Piezocone test (CPTU)
Figure 2 shows the detail of a cylindrical soil tank

used for CPTU. The tank has 1.0 m in inner diameter
and 0.8 m in height. A self-standing steel mesh with
0.5 m diameter was set inside the tank as a boundary for the SBM layer and surrounding layer. For this
test, silica sand #7, which is the same soil used for
the hydraulic conductivity test, was used as the base
soil of SBM. The lower half of the cylindrical steel
mesh was filled up with SBM of CBP = 100 kg/m3
with a density of 1.87 g/cm3 , and the upper half
was filled up with SBM of CBP = 50 kg/m3 with that
of 1.91 g/cm3 . These densities were determined by
the specimens after the hydraulic conductivity tests.
Silica sand was filled around SBM layer to allow the
dissipation from the SBM. During the preparation,
water level in the soil tank was kept higher than the
soil surface to assure that no air exists in the soil.
The top of the tank has a circular hole for the cone
penetration. The penetration rate was 1.0 cm/s; the airpressure of 30 kPa was applied on the soil surface via
an air-bag to prevent volume changes of the soil layers. The piezocone probe in this research is specified
3.1
Factors affecting the hydraulic barrier

performance
3.1.1 Effect of chemicals

Figure 3 illustrates the hydraulic conductivity values
for CBP = 100 kg/m3 , affected by the chemicals in the
permeant (Kamon et al. 2006). In the case when SBM
is permeated with 0.1 M CaCl2 solution, the hydraulic
conductivity value is 3.5 times higher than the case of
distilled water. Another finding is that the difference in
hydraulic conductivity is negligible for higher CaCl2
concentration than 0.1 M. SBM can achieve its low
hydraulic conductivity of 1.2 1010 m/s even against
the sea water. Hydraulic conductivity value against
50%-ethanol permeation is almost equal with that
against distilled water permeation. Intrinsic permeability values, which take into account the viscosity of
the permeant, are 5.0 1020 m2 against 50%-ethanol
permeation and 3.0 1020 m2 against distilled water
permeation. Therefore, the effect of organic solvent on
hydraulic barrier performance is small. These observations confirm that the hydraulic conductivity of SBM
does not increase significantly with any permeant if
709
Figure 5. Effects of content of bentonite powder and CaCl2

concentrations in the pore water.
Figure 3. Effects of chemicals in the permeant.
Figure 4. Effects of chemicals in the pore water.
the bentonite in SBM is initially well hydrated with

soil pore water.
Figure 4 shows the hydraulic conductivity values
of SBM (CBP = 100 kg/m3 ) for different chemicals in
pore water of the original soil. The higher hydraulic
conductivity values were obtained with the higher
CaCl2 concentration. Although the pore water containing calcium ions is diluted by the water fraction of the
bentonite slurry, there was a significant increase in
hydraulic conductivity. Hydraulic conductivity values
increased linearly in association with CaCl2 concentration in the range of 00.05 M. However, for concentrations of CaCl2 higher than 0.05 M, it still increased
but at a smaller rate. For 0.1 M, the hydraulic conductivity reached a value higher than 1.0 109 m/s.
When the pore water was replaced by seawater, the
hydraulic conductivity of SBM reached the same values as in the case of 0.05 M CaCl2 . In the cases that
heavy fuel oil is contained in the parent material, SBM
displays very low hydraulic conductivity values due to
both of the hydrophobicity of oil and the swelling of
bentonite.
Figure 5 illustrates the electrical conductivity (EC)
of the permeant and the soil pore water versus
hydraulic conductivity values. EC has a positive
correlation with electrolyte concentration, therefore,

is one of the important factors to evaluate chemical
compatibility. As shown in this figure, the increase
of hydraulic conductivity with increasing EC of soil
pore water is significant compared with that of the
permeant.
By comparing the effects of divalent cations on the
hydraulic conductivity when they exist in the permeant and in pore water of the parent material, it can be
concluded that the cations in the pore water have a
more significant effect on the barrier performance of
SBM. The hydraulic conductivity for 0.1 M in permeant was 3.5 times higher than that for 0 M in permeant.
By contrast, the increase in the CaCl2 concentration in
pore water from 0 to 0.1 M resulted in an increase in
the hydraulic conductivity of more than one order of
magnitude. This difference indicates that the prehydration of bentonite is a crucial factor for the chemical
compatibility of the SBM. Its effect is relevant even
in the cases that the divalent cation concentrations in
pore water are lower than 0.05 M as shown in Figure 4.
Thus, the concentration of the cations and their variation in groundwater at the site should be considered in
evaluating the hydraulic barrier performance of SBM.
3.1.2 Effect of the bentonite powder content
Figure 6 shows the hydraulic conductivity values in
N-series for various CBP . The higher content of bentonite powder results in obtaining the lower hydraulic
conductivity, even when the CaCl2 concentrations in
the pore water were relatively high (0.05 or 0.1 M).
Even for the highest CaCl2 concentration of 0.1 M, the
hydraulic conductivity was lower than 1.0 109 m/s
in the case of CBP = 150 kg/m3 . This can be attributed
to the shorter distance of montmorillonite layer by
adding the higher amount of bentonite powder. Montmorillonite is a mineral and is main component of
bentonite. Shorter distance of montmorillonite leads
the higher chemical potential of adsorbed water and the
thicker diffused double layer. From these observations,
when the SBM vertical cut-off wall is installed at a
site where the inorganic chemical concentration in the
soil and/or groundwater is high, its hydraulic barrier
710
Figure 6. Effects of content of bentonite powder and CaCl2

concentrations in the pore water.
performance can be enhanced by adding a higher

amount of bentonite. In the case that the SBM was prehydrated (CaCl2 concentration in pore water = 0 M), a
hydraulic conductivity lower than 1.0 109 m/s was
achieved by adding 50 kg/m3 of powder bentonite for
the composite soil used in this study.
This relationship between CBP and the hydraulic
conductivity can also be observed for SBM made with
different type of base soil as shown in Table 2. The
hydraulic conductivity value of SBM made with composite soil was 5.0 1011 m/s and that of SBM made
with silica sand was 1.1 1010 m/s with adding bentonite powder of 100 kg/m3 against the distilled water
permeation. In the cases of CBP = 50 kg/m3 , although
permeants used in N-1 and S-2 are 0.1 M CaCl2 solution and distilled water respectively, SBM made with
composite soil can achieve lower hydraulic conductivity than SBM made with silica sand. Hence, SBM
made with composite soil shows a better hydraulic barrier performance than SBM made with silica sand even
when the same amount of bentonite powder was added
because of the particle size distribution of original soil
shown in Table 2. From these observations, fine particle contents in the base soil also contribute to the
low hydraulic conductivity in terms of soil skeletons
of SBM.
3.2 Post-construction verification using CPTU
3.2.1 Quality assessment of constructed cut-off wall
Figure 7 shows continuous changes of corrected cone
resistance (qt ), sleeve friction ( fs ), and pore pressure
(u) with depth obtained from CPTU. This figure indicates that these three values changed notably around
0.5 m depth, which is the boundary of the two different compositions of SBM. The qt values in upper
layer (CBP = 50 kg/m3 ) became larger than those in
lower layer (CBP = 100 kg/m3 ). In the upper layer, the
qt values attained approximately 1.3 MPa, and in the
lower layer it was approximately 0.1 MPa regardless
of the depth. Similarly for fs values, higher values
were attained in the upper layer than in the lower
layer. Although larger variations were observed on fs
values, the same trend as qt can be confirmed. Pore
Figure 7. Results of CPTU; [Left] Corrected cone resistance

(qt ), [Center] Sleeve friction ( fs ), [Right] Pore pressure (u).
water pressure values in the upper layer became so

reduced that reached negative values, and those in the
lower layer recovered until they reached values near
the hydrostatic pressure.
Generally, cone resistance values and sleeve friction values from sand stratum are higher than those
values from silty soil and/or clay stratum due to their
particle sizes (e.g. Pradhan 1998, Mayne et al. 2009).
In this research, the steel mesh in the soil tank was
filled with SBMs made with different content of bentonite powder. In the case of CBP = 50 kg/m3 (in upper
layer), fine particle content of the SBM is approximately only 4% and this material is classified as sandy
gravel (SG) according to JGS 0051-2009. Considering
cut-off walls are mainly constructed in sand stratum,
partial lean-mix of bentonite in constructed SBM cutoff wall also acts as sand stratum when CPTU is carried
out as an in-situ QC/QA method. These results indicate
that CPTU can detect hydraulic defects in SBM cutoff wall when the physical properties in the lean-mix
part are obviously different. Hence, CPTU has some
possibilities to be applied as a QC/QA method at the
post-construction.
3.2.2 Estimation of hydraulic conductivity by pore
pressure dissipation test
In this research, the pore pressures were collected at 1
second intervals at 0.75 m depth from the soil surface
after cone penetration was suspended. Figure 8 shows
a result of pore pressure dissipation test. Here, degree
of excess pore pressure dissipation (U ) defined by the
following equation (1) are plotted on the vertical axis.
The horizontal axis is elapsed time with a logarithmic
scale.
where ut = pore pressure at time t; u0 = hydrostatic

pore pressure in situ; and ui = initial pore pressure at
start of dissipation test. Since dissipation rate varies by
711
Table 4. Coefficient of consolidation and hydraulic conductivity at each degree of excess pore pressure dissipation.
Degree of
dissipation
%
Horizontal coefficient
of consolidation, ch
m2 /kN
Horizontal hydraulic
conductivity, kh
m/s
20
30
50
2.5 103
2.4 103
2.9 103
1.5 1010
1.4 1010
1.8 1010
Figure 8. Pore pressure dissipation test result at 0.75 m

depth.
Table 3. Time factors T (from Houlsby and Teh, 1988).
Degree of dissipation
%
Time factor
20
30
50
0.038
0.078
0.245
compressibility and permeability of surrounding soil

around the cone probe, hydraulic conductivity of SBM
should be estimated by the dissipation process. In the
calculation of hydraulic conductivity, it is frequently
recommended to use the time for 50% dissipation, t50 .
In this research, both of the time for 30% dissipation,
t30 , and the time for 20% dissipation, t20 , were also
used for the prediction in order to shorten the time for
quality evaluation at a site.
After each time (t20 , t30 , and t50 ) were determined
from the experimental results, horizontal coefficient
of consolidation, ch (m2 /day), was calculated from the
following equation (2).
where T = time factor; R = penetrometer radius

(=17.8 mm); t = time (s); and Ir = rigidity index. Time
factor T is a dimensionless factor that depends on
the location of the pore pressure element, and analytically calculated as shown in Table 3 by Houlsby and
Teh (1988) for the cone probe which has the element
above cone base as used in this study. Rigidity index
Ir is assumed 25 for SBM. Based on the calculated
ch , horizontal hydraulic conductivity, kh (m/s), can be
estimated by the following equation (3).
where ch = horizontal coefficient of consolidation

(m2 /day); mv = coefficient of volume compressibility (m2 /kN); and w = unit weight of water
Figure 9. Comparison of hydraulic conductivity from

CPTU and from hydraulic conductivity test.
(=9.81 kN/m3 ). In this research, mv = 5.3 104 m2 /

kN measured by a consolidation test was applied on the
assumption that the compression of the surrounding of
the probe is one-dimensional; the same as specimen in
the consolidation test. From the above equations, it
can be seen that the hydraulic conductivity estimated
by the pore pressure dissipation test has an inverse
correlation with time for the dissipation as shown in
the following equation (4).
Calculated horizontal coefficient of consolidation

values and horizontal hydraulic conductivity values
at each degree of excess pore water dissipation are
summarized in Table 4. As shown here, the variation
of values was negligible regardless of the degree of
excess pore water dissipation. Moreover, the hydraulic
conductivity values calculated by CPTU were in the
range of 1.41.6 times of those values calculated
by the hydraulic conductivity test as shown in Figure 9. Therefore, hydraulic conductivity of SBM can
be briefly estimated by the dissipation test. These
observations suggest that the operation of dissipation test is suitable as evaluation of hydraulic barrier
performance, which is to say quality assurance, at
post-construction of SBM cut-off wall.
3.2.3 Self-sealing of bore hole after CPTU
Bore hole after CPTU must be sealed or refilled with
another material with low hydraulic conductivity to
712
Figure 10. Brief scheme of design and evaluation.

Photo 1. Appearance of specimen before CPTU.
Photo 2. Appearance of SBM surface after CPTU.
ensure adequate wall thickness. As described in the

introduction, high deformability even after the construction is one of the most significant characteristics
of SBM, unlike other materials with solidification
materials. Therefore, the bore holes after CPTU can
be expected to be sealed by itself with time.
Photo 1 and 2 show the appearance of SBM layer
in a soil tank before and after CPTU, respectively. As
seen in these photos, although an approximately 2.0
cm depth residual deformation due to the cone penetration was observed; the deeper area of the bore hole
was sealed by SBM itself with time. This result indicates that SBM cut-off wall can maintain the designed
hydraulic barrier performance with its self- sealing
capacity even after CPTU has been conducted to check
in-situ hydraulic conductivity.
3.3 Design and evaluation of the hydraulic barrier
performance
In order to assure high hydraulic barrier performance
and to determine adequate mixing ratio of SBM,
implementation of the hydraulic conductivity tests
is a crucial step doing the designing process at the
pre-construction stage. As discussed above, physicochemical properties of in-situ soil and bentonite
powder content will affect the subsequent hydraulic

barrier performance of SBM. Bentonite powder content should be determined according to physicochemical properties of in-situ soil, such as particle size
distribution and chemical concentration in the soil pore
water.
After SBM cut-off wall is constructed with adequate mixing ratio based on the experimental results
of the hydraulic conductivity tests, CPTU should be
employed to the constructed SBM cut-off wall in
order to ensure its homogeneity and in-situ hydraulic
barrier performance without excavation. The homogeneity of the constructed wall can be checked by the
trends of three parameters obtained during cone penetration. The homogeneity assurance using the CPTU
test should be operated at certain intervals along the
wall because it is not economically feasible to evaluate the wall overall. If the profile of parameters during
CPTU has any unstable trends, there are possibilities
that lean-mix part due to the heterogeneity exists inside
the wall. In such case, re-mix in the cut-off wall should
be done to improve the homogeneity until homogeneous profiles are achieved. After the homogeneity is
ensured, excess pore pres sure dissipation test should
be conducted to estimate hydraulic conductivity values
at certain depth intervals. Since the hydraulic conductivity will become lower with higher earth pressure in
the deeper area of the cut-off wall, it is more important
to determine the hydraulic conductivity values in the
shallow area. When the estimated hydraulic conductivity values exceed the criteria, re-addition of bentonite
powder will be required.
A brief scheme of the experimental work in this
study is summarized in Figure 10 in terms of the
pre-/post-construction. Currently, in-situ hydraulic
conductivity is mainly evaluated by conducting the
hydraulic conductivity tests in laboratories after the
samples are collected and delivered from the sites.
However, this evaluation contains some problems such
as: 1) it requires a lot of time to measure the values,
2) the number of samples which can be tested is limited and 3) the measured hydraulic conductivity values
have to represent the wall overall because a linearly
continuous evaluation is not feasible with this method.
Hence, although requirements about quality depend
713
on the specifications of each site, in-situ and on-site

quality assurance method should be established to
ensure the designed hydraulic barrier performance.
A linearly continuous evaluation can enhance the
reliability of in-situ containment techniques.
CONCLUSIONS
This paper presents the experimental results of

hydraulic conductivity tests and of CPTU for the SBM
focusing on its chemical compatibility and on the quality assurance at post-construction. The conclusions are
summarized as follows:
1) When no divalent cation exists in soil pore water,
the hydraulic conductivity of the SBM permeated
with 1.0 M CaCl2 solution is only 3.5 times as
high as that with the distilled water. Sufficient
hydration of the bentonite will maintain the high
hydraulic barrier performance permeated with the
CaCl2 solution, seawater, and 50%-ethanol.
2) Calcium ions in soil pore water have more significant effect on the subsequent hydraulic conductivity than those in permeant. However, hydraulic
barrier performance can be enhanced by mixing a
higher content of bentonite powder.
3) Fine particle contents of in-situ soil also affect the
hydraulic conductivity in terms of soil skeletons
of SBM.
4) Profiles of three parameters obtained during cone
penetration changed notably at a boundary of two
different composition of SBM by the difference
of clay fraction. CPTU offers the possibility of
detecting lean-mix part inside SBM cut-off wall.
5) Hydraulic conductivity can be estimated by excess
pore pressure dissipation test with temporal stop
of cone penetration. Estimated valued at different
degree of excess pore pressure dissipation were
almost equal and in the range of 1.41.6 times
of those values calculated by the hydraulic conductivity test. In-situ hydraulic conductivity can be
measured on-site in a shorter time compared with
other methods.
6) SBM cut-off wall can maintain hydraulic barrier
performance with its self-sealing capability even
after CPTU is conducted to check in-situ hydraulic
conductivity.
ACKNOWLEDGEMENT
The hydraulic conductivity tests were conducted with
the cooperation of Mr. Ogawa and Mr. Matsuhashi;
both former students of Kyoto University. The authors
would like to express their appreciation to Mr.
Yoshimura and Mr. Terao at Soil and Rock Engineering Co., Ltd. as well as to Mr. Koga, former student
at Kyoto University, for their great support conducting
CPTU.
REFERENCES
Britton, J.P. & Filz, G.M. 2007. High uniformity versus low
hydraulic conductivity for vertical barriers in contaminant containment applications. Burns, S.E., Culligan, P.J.,
Evans, J.C., Fox, P.J., Reddy, K.R., and Yesiller, N. (eds.),
Geoenvironmental Engineering 163. ASCE: on CD.
Chai, J.C., Agung, P.M.A., Hino, T., Igaya, Y. and Carter, J.P.
2011. Estimating hydraulic conductivity from piezocone
soundings. Geotechnique 61(8): 699708.
Evans, J.C. 1994. Hydraulic conductivity of vertical cutoff
walls. Daniel, D.E. and Trautwein, S.J. (eds.), Hydraulic
Conductivity and Waste Contaminant Transport in Soils.
ASTM STP 1142. ASTM. Philadelphia: 7994.
Grube, W.E. 1992. Slurry trench cut-off walls for environmental pollution control. Paul, D.B., Davidson, R.R., and
Cavalli, N.J. (eds.), Slurry Walls: Design, Construction,
and Quality Control, ASTM STP 1129. ASTM: 6977.
Philadelphia.
Houlsby, G.T. & Teh, C.I. 1991. An analytical study of the
cone penetration test in clay. Geotechnique 41(1): 1734.
Jeffries, M.G. & Davies, M.P. 1993. Use of CPTu to estimate
equivalent SPT N60. Geotechnical Testing Journal 16(4):
458468.
Kamon, M., Ohta, H., Aoi, M. & Asada, S. 1998. Development of new river protection method by continuous
inclined diaphragm wall. P.S. Sco e Pinto (ed.), Environmental Geotechnics: 3538. Balkema, Rotterdam.
Katsumi, T., Ishimori, H., Onikata, M., & Fukagawa, R.
2008a. Long-term barrier performance of modified bentonite materials against sodium and calcium permeant
solutions. Geotextiles and Geomembranes 26(1): 1430.
Katsumi,T., Kamon, M., Inui,T. &Araki, S. 2008b. Hydraulic
barrier performance of SBM cut-off wall constructed
by the trench cutting and re-mixing deep wall method.
Khire, M.V., Alshawabkeh, A.N. and Reddy, K.R. (eds.),
GeoCongress 2008: Geotechnics of Waste Management
and Remediation, Geotechnical Special Publication 177.
ASCE: 628635.
Manassero, M. 1994. Hydraulic conductivity assessment of
slurry wall using piezocone test. Journal of Geotechnical
& Geoenvironmental Engineering 120(1): 17251746.
Mayne, P.W., Matthew, R.C., Sarah, M.S., An-Bin, H. &
Jorge, G.Z. 2009. Geomaterial behavior and testing.
Hamza, M., Shahien, M., and El-Mossallamy, Y. (eds.),
Proceedings of the 17th International Conference on Soil
Mechanics and Geotechnical Engineering. The Academia
and Practice of Geotechnical Engineering 4: 27772872.
Pradhan, T.B.S. 1998. Soil identification using piezocone
data by fuzzy method. Soils and Foundations 38(1):
255262.
Robertson, P.K. 2009. Interpretation of cone penetration
tests a unified approach. Canadian Geotechnical
Journal 46: 13371355.
Shackelford, C. D., Malusis, M. A., Majeski, M. J. &
Stern, R. T. 1999. Electrical conductivity breakthrough
curves. Journal of Geotechnical and Geoenvironmental
Takai, A., Inui, T., Katsumi, T., Kamon, M. & Araki, S. 2012.
Factors affecting the hydraulic barrier performance of
soil-bentonite mixture cut-off wall. Journal of Japan
Society of Civil Engineers, Ser. C (Geosphere Engineering) 68(1): 114. (in Japanese)
Yeo, S.S., Shackelford, C.D. & Evans, J.C. 2009. Hydraulic
conductivity and compressibility of soil-bentonite backfill
amended with activated carbon. Journal of Geotechnical
Yesiller, N. & Shackelford, C. D. 2010. Chapter 13: Geoenvironmental Engineering. Das, B.M. (ed.), Geotechnical
Engineering Handbook: 1352.
714

Author index
Abdelaziz, S.L. 511

Abou-Bekr, N. 299
Andronic, A. 383
Arairo, W. 357
Araki, S. 707
Baehler, M. 483
Barast, G. 275
Beaven, R.P. 61
Benson, C.H. 79
Bezuijen, A. 255
Bilardi, S. 685
Boscov, M.E.G. 579
Bouazza, A. 211, 435
Bourabah, M.A. 299
Brianzoni, V. 243
Caki,
L. 347
Calabr, P.S. 685
Carrubba, P. 313
Casini, D. 501
Casini, F. 467
Castellanza, R. 541
Castelli, F. 307
Castiglia, C. 387
Cecconi, M. 533
Charlier, R. 443
Chatterjea, K. 419
Chen, Y.M. 365, 625
Chiu, A.C.F. 371
Ciantia, M.O. 541
Clauss, F. 501
Conte, M. 313
c, S. 347
Cori
Costa, S. 377
Crescini, E. 699
Croissant, D. 291
de Fouquet, C. 637
DeJong, J.D. 103
Delage, P. 119
Della Vecchia, G. 199
Dewaele, P. 189
Di Donna, A. 475
Di Emidio, G. 255
di Prisco, C. 541
Di Sante, M. 243
Djeran-Maigre, I. 275
Djran-Maigre, I. 357
Dominijanni, A. 169
Dumais, S. 663
Dupray, F. 85, 475, 483
Eberemu, A.O. 647
Endo, K. 335
Era, N. 401
Erol, S. 491
Evans, J.C. 149
Federico, A. 655
Fei, X. 321
Feng, S.J. 329
Ferrari, A. 549
Fleming, I.R. 189
Flores, G. 573
Franois, B. 491
Fratalocchi, E. 243
Frchtenicht, H. 501
Gabas, S.G. 579
Gastone, F. 223
Gens, A. 21, 413, 467
Giacheti, H.L. 617
Ginn, T.R. 103
Gourc, J.P. 181, 313
Guan, C. 625
Guerini, A. 699
Guo, R.Y. 365
Harnas, F.R. 425
Hartog, N. 35
Hassanizadeh, S.M. 35
Hepburn, B.D.P. 525
Holmes, D.S. 61
Hosney, M.S. 283
Hu, X.P. 371
Huang, M. 393
Hudson, A.P. 61
Hussien, H.M. 599
Imanishi, H. 335
Intern, G. 655
Inui, T. 335, 573, 587, 707
Iyer, K.R. 593
Jefferis, S.A. 557
Joshi, K. 149
Kamon, M. 587, 707
Kang, J. 267
Karkush, M.O. 599
715
Katsumi, T. 335, 573, 587, 707

Katzenbach, R. 501
Kelly, D. 189
Khire, M.V. 407
Kim, J.K. 261
Koanka, T. 693
Koda, E. 693
Kodikara, J. 377
Komine, H. 671
Kwak, C.W. 261
Laloui, L. 85, 475, 483, 549
Lee, S.R. 451
Lei, G.H. 371
Leong, E.C. 419, 425
Levasseur, S. 443
Li, X.L. 413
Li, Y.C. 365
Lima, A. 413
Macedo, L.S. 609
Mahmoodlu, M.G. 35
Mainero, D. 387
Malusis, M. 267
Manassero, M. 169
Manoli, D.M. 383
Mano, F.P. 609
Martin, J.R. 511
Martinez, B.C. 103
Maugeri, M. 307
Mazzieri, F. 243
McDougall, J.R. 189
Merritt, A.S. 557
Miccoli, E. 655
Millard, A. 357
Mimouni, T. 85
Minder, P. 565
Mondelli, G. 609
Moraci, N. 685
Morando, M. 699
Mulligan, C.N. 663
Murakami, S. 671
Murianni, A. 655
Musso, G. 199
Napoli, P. 533
Nelson, D.C. 103
Ng, C.W.W. 371
Nobile, M. 655
Noel-de-Tilly, R. 663
Okawara, M. 335
Olgun, C.G. 511
Olivella, S. 231, 467
Osinski, P. 693
Otsuka, Y. 335
Oxarango, L. 181, 313
Pane, V. 533
Park, I.J. 261
Park, J.B. 261
Parker, D. 189
Pasqualini, E. 243
Pathak, P. 593
Powrie, W. 61
Priceputu, A. 383
Priono, 419
Prunier, F. 357
Pulat, H.F. 341
Puma, S. 169
Puzrin, A.M. 221, 565
Rahardjo, H. 419, 425
Rakic, D. 347
Ramli, H. 573
Rao, B.H. 519
Raoof, A. 35
Raskin, L. 321
Razakamanantsoa, A. 275
Reddy, K.R. 131, 675
Riyis, M.T. 617
Romero, E. 199, 413
Rouf, M.A. 435

Rowe, R.K. 3, 283, 435
Russo, G. 637
Sakamoto, H. 335
Sakanakura, H. 335
Salehnia, F. 443
Sanzeni, A. 699
Saravanathiiban, D.S. 407
Satyanaga, A. 425
Sedighi, M. 525, 625
Serbah, B. 299
Serbulea,
M.S. 383
Sethi, R. 223
Shackelford, C. 45, 267
Shah, M.V. 631
Shin, H. 451
Shroff, A.V. 631
Singh, D.N. 593
Singh, R.M. 211, 435
Soga, K. 149
Spacagna, R.-L. 637
Stocco, S. 387
Stoltz, G. 181, 291
Suzuki, H. 335
Tada, K. 671
Taha, M.R. 459
Taha, O.M. 459
Taibi, S. 299
716
Takai, A. 335, 587, 707

Thiel, R. 189
Thomas, H.R. 525, 625
Tinet, A.J. 181
Tosco, T. 223
Touze-Foltz, N. 291
Vaunat, J. 413
Verastegui, R.D.F. 255
Verrascina, T. 401
Viggiani, G.M.B. 467
Vitone, C. 655
Wang, B. 211
Wang, H. 393
Wang, J.Y. 419, 425
Witteveen, P. 549
Xie, H.J. 625
Xue, J. 377
Yargicoglu, E.N. 675
Yasutaka, T. 335
Yu, Y. 283
Yukselen-Aksoy, Y. 341
Zaboon, A.T. 599
Zarotti, C. 533
Zekkos, D. 69, 321
Zhan, L.T. 365
Zhang, X. 329

Coupled Phenomena in Environmental Geotechnics

Enviado por

Dados do documento

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

Coupled Phenomena in Environmental Geotechnics

Enviado por

Direitos autorais:

Formatos disponíveis

COUPLED PHENOMENA IN ENVIRONMENTAL GEOTECHNICS

PROCEEDINGS OF THE INTERNATIONAL SYMPOSIUM, ISSMGE TC 215, TORINO,

Cover photo credits:

ISBN: 978 1 138 00060 5 (Hardback + CD Rom)

Coupled Phenomena in Environmental Geotechnics Manassero et al (Eds)

Coupled modelling of barriers for radioactive waste disposal

Liquid and gas flow in municipal solid waste

Experimental evidence of anisotropy in municipal solid waste

Impact of subgrade water content on cation exchange and hydraulic conductivity of

Effect of mechanical settlement on biogas transfer in MSW landfills

Mass loss and volume change: From sand-salt analogues to MSW

Modeling the coupled chemo-hydro-mechanical behavior of compacted active clays

Soil effective thermal conductivity from energy pile thermal tests

Modification of clay consolidation properties using microbial gas production

Modelling gas and water flow in environmental problems

Migration of aggressive solutions through cement-bentonite slurry walls

Some geotextiles properties useful for HDPE geomembrane puncture protection

Landfill waste characterization

Dynamic characterization of municipal solid waste by SDMT

Correlation between compressibility and degradation for a PTMB waste

2011 East Japan earthquake and tsunami Geoenvironmental challenges

Shear strength of municipal waste materials from two landfills in Serbia

Stability and settlement analysis of landfills and natural slopes

A model for anaerobic degradation of municipal solid waste

Water retention and shrinkage properties of a cemented dredged material

Soil mass stability analysis using Euler-Lagrange coupling

Piggyback liner detection by means of spatially constrained inversion of SWM data

Three-dimensional rotational-translational mechanism for the stability analysis of landfill

Coupled thermo-hydraulic pulse tests on two reference Belgian clay formations

Improved correction factor for indirect determination of permeability function

Capillary barrier system for landfill capping

Modeling of strain localization around the radioactive waste disposal galleries

Constitutive model for unsaturated soils based on the effective stress

Underground energy issues

Effect of groundwater flow on the THM behavior of an energy pile

Thermal, hydraulic and mechanical performances of enhanced grouting materials

Long term performance of heat exchanger piles

Soil thermal behavior of a horizontal ground source heat system

Natural and anthropogenic bio-chemical processes within soils and rocks

Chemo-mechanical weathering of calcarenites: Experiments & theory

Chemo-mechanical behaviour of a low activity clay

Exploiting physico-chemical modification of soils in closed face tunnelling

Combining biomineralization with chemically enhanced drainage in soils

Characterization of polluted sites and related aquifers

Interaction of metal adsorption and solubilization in a lateritic clay

Novel techniques for simulating and monitoring impact of contaminants on geoenvironment

Studying the effects of contamination on the geotechnical properties of clayey soil

The resistivity piezocone for high resolution geo-environmental site investigation:

A case study of contaminant transport and fate at a landfill site in China

Geostatistical analysis of groundwater nitrates distribution in the Plaine dAlsace

Sustainable remediation of contaminated sites

Prediction of time to reduce tsunami sediment salinity by rainfall after the

Active and passive barriers for polluted sites

Flow numerical modeling for efficiency assessment of vertical barriers in landfills

Design, installation and performance of containment system at a petroleum-release site

Hydraulic barrier performance of soil bentonite mixture cutoff wall

Coupled Phenomena in Environmental Geotechnics Manassero et al (Eds)

UNDER THE AUSPICES OF

INTERNATIONAL GEOSYNTHETICS SOCIETY

Coupled Phenomena in Environmental Geotechnics Manassero et al (Eds)

Coupled Phenomena in Environmental Geotechnics Manassero et al (Eds)

Italy, President of Italian Geotechnical Society

ISSMGE TC 215 Members

Coupled Phenomena in Environmental Geotechnics Manassero et al (Eds)