A Multiscale Theory of Swelling Porous Media II. Dual Porosity Models For Consolidation of Clays

Transport in Porous Media 28: 69108, 1997.
c 1997 Kluwer Academic Publishers. Printed in the Netherlands.
69
A Multiscale Theory of Swelling Porous Media:

II. Dual Porosity Models for Consolidation of
Clays Incorporating Physicochemical Effects
MARCIO
A. MURAD1 and JOHN H. CUSHMAN2
1
Laboratorio Nacional de Computaca o Cientifica, LNCC/CNPq, Rua Lauro Muller 455, 22290
Rio de Janeiro, Brazil
2
Center for Applied Math, Math Sciences Building, Purdue University, W. Lafayette, IN 47907
U.S.A. e-mail: jcushman@math.purdue.edu
(Received: 13 August 1996; in final form: 21 February 1997)
Abstract. A three-scale theory of swelling clay soils is developed which incorporates physicochemical effects and delayed adsorbed water flow during secondary consolidation. Following earlier
work, at the microscale the clay platelets and adsorbed water (water between the platelets) are
considered as distinct nonoverlaying continua. At the intermediate (meso) scale the clay platelets and
the adsorbed water are homogenized in the spirit of hybrid mixture theory, so that, at the mesoscale
they may be thought of as two overlaying continua, each having a well defined mass density. Within
this framework the swelling pressure is defined thermodynamically and it is shown to govern the
effect of physico-chemical forces in a modified Terzaghis effective stress principle. A homogenization
procedure is used to upscale the mesoscale mixture of clay particles and bulk water (water next to
the swelling mesoscale particles) to the macroscale. The resultant model is of dual porosity type
where the clay particles act as sources/sinks of water to the macroscale bulk phase flow. The dual
porosity model can be reduced to a single porosity model with long term memory by using Greens
functions. The resultant theory provides a rational basis for some viscoelastic models of secondary
consolidation.
Key words: swelling clay soil, mixture theory, homogenization, consolidation, swelling pressure,
disjoining pressure, dual porosity.
1. Introduction
Swelling clay soils consisting of an assemblage of clusters of hydrous alluminium
and magnesium silicates with an expanding layer lattice are widely distributed in
the earths crust. Their behavior is of paramount importance in almost all aspects
of life, where they are responsible for many reactions and processes. For example,
compacted clays such as bentonite have been extensively used to impede the movement of water through cracks and fissures. They play a critical role in various waste
isolation scenerios such as barriers for commercial land fills. In the context of oil
and gas production, drilling muds play a critical role ([45, 75]). Swelling clays also
play a critical role in the consolidation and failure of foundations, highways and
runways. For example, the foundation engineers face problems when a foundation
S.A. INTERPRINT: (K.B. 6) PIPS Nr.: 135627 MATHKAP

tipm1248.tex; 29/07/1997; 6:53; v.7; p.1
70
MARCIO
A. MURAD AND JOHN H. CUSHMAN
is built on expansive soils, since as the moisture content increases, the soil swells
and heaves upward, and as the moisture decreases, the soil compacts and the ground
surface recedes and pulls away from the foundation walls [76]. An accurate model
capable of capturing the swelling of clays will have important consequences in
civil and petroleum engineering, hydrology, geology and soil science.
Since Terzaghi [72] and Biot [16,17] first proposed linear poroelastic models
for deformable media, the criterion for rupture and failure of soils has been based
on the concept of effective stress. Effective stresses are defined as the difference
between total applied stresses and bulk water pressure. Classically these stresses are
interpreted as contact stresses, i.e. transmitted between points of the intergranular
contact. It has been experimentally verified that the classical Terzaghi effective
stress principle describes accurately coarse-grained soils such as sands, silts and
low and medium plastic clays such as kaolinite or illite. However, its classical
form has been found to be inadequate for explaining deformation of swelling
clays, particularly active plastic clays such as bentonite and montmorillonite. The
reason is that the classical effective stress principle assumes that no other forces
except the effective stress and pore pressure are present. The existence of physicochemical forces within and between the clay particles are excluded. Interparticle
forces arising from physico-chemical mechanisms have been demonstrated to be
of paramount importance for active clays. Researchers have heuristically modified
Terzaghis effective stress principle to account for physico-chemical forces and
as a consequence different mechanistic pictures of the various stresses have been
derived (see Sridharan and Rao [70], Sridharan [68], Lambe [48], Morgensten and
Balasubramonian [60]). A comparison between these different mechanistic pictures
can be found in, e.g., Hueckel [39, 40] or Graham et al. [31].
The nature of physico-chemical forces remains controversial. In contrast to the
effective stress, net attractive(A)-repulsive(R) forces between the clay particles do
not depend upon direct contact. They have at least three components: the Van der
Wals attraction, electrostatic (or osmotic) repulsion and surface hydration (a structural component). This latter component arises due to the hydrophilic structure of
the platelets which manifest short range bonding forces between the minerals and
water. These forces are usually referred to as hydration forces (Israelachvili [41]).
The complex mechanisms underlying the constitutive behavior of a hydrophilic clay
soil are a consequence of its complicated microstructure. Due to their tremendous
specific surface area and their charged character, clusters of clay platelets when
hydrated form particles consisting of an assemblage of platelets and adsorbed
water. These particles swell under hydration and shrink under desication. The
platelet-water hydration forces cause the macroscale behavior of clays to significantly differ from granular nonswelling media. The hydration forces modify the
thermodynamical properties of the water in the interlamellar spaces and consequently its properties vary with the proximity to the solid surface (Low [5456],
Grim [32]). The interlamellar water is termed adsorbed water to distinguish it
from its bulk or free-phase counterpart (i.e. water free of any adsorptive force).
tipm1248.tex; 29/07/1997; 6:53; v.7; p.2
A MULTISCALE THEORY OF SWELLING POROUS MEDIA
71
It has been argued by Low and co-workers [57, 58], Derjaguin and co-workers
[23, 24, 25] and Israelachvili and co-workers [29, 41, 42], that for intersticies
, the structural hydration forces play a crucial role in swelling.
smaller than 30 A
More specifically, these authors have advocated that for small interlamellar spaces
electrostatic forces play a negligible role in swelling and are too weak to explain the
anomalous behavior of the adsorbed water. Low [58] has advocated that hydration
forces are the main cause of the in-depth perturbation of the adsorbed water by
the clay mineral surfaces and consist of the dominant component of the swelling
pressure, i.e. the overburden pressure excess that must be applied to the saturated
mixture of clay and adsorbed water to keep the layers from moving apart. Similar
arguments have been claimed for the disjoining pressure (the microscopic counterpart of the swelling pressure) at the finest scale (microscale), where the adsorbed
water is viewed as a thin film coating the mineral surfaces. Derjaguin and Churaev
[24] have favored the claim that hydration forces play a decisive role in the stability
of colloids and thin films.
A proper description of the microscopic behavior of the adsorbed water is of
utmost importance when modeling consolidation of a swelling clay soil. It is well
known that the consolidation process usually involves two different compression
stages, primary and secondary. The primary stage is due to the drainage of the bulk
phase water, where bulk phase pressure is gradually transferred to the effective
stress. Secondary compression is characterized as continued deformation after the
bulk water has been substantially drained. Both mechanisms can occur simultaneously with the application of load, or it can be assumed that the secondary stage
begins after the primary stage is complete. The secondary stage is characterized
by the creep viscous behavior of the clay structure. Attempts have been made to
characterize creep and stress relaxation using phenomenological rheological models at the macroscale. Although viscoelastic constitutive equations, consisting of
linear or nonlinear springs in combination with linear or nonlinear dashpots, are
the most widely used in modeling of secondary compression (see, e.g., [10, 19,
30, 44, 46, 50, 74, 77]) there is not yet a complete understanding of the mechanisms underlying this phenomenon. Sridharan and Rao [71] have suggested that
secondary compression is related to the delayed deformation of the clay particle
after the bulk water has been drained from the larger pores in the primary structure
(see also Hueckel [40]). This secondary deformation appears somewhat related to
the delayed drainage of the adsorbed water relative to the bulk phase flow. Since
the adsorbed water is strongly attracted to the clay surfaces, it is more difficult to
drain than the bulk water. In addition, the interaction between the adsorbed water
and clay surfaces produces an excess in viscosity relative to the bulk water and is
manifest in a viscous stress relaxation (Zeevart [77]). The delayed particle deformation due to the adsorbed water flow has also been discussed by Barbour and
Fredlung [9] in the context of osmotically induced consolidation resulting from
osmotic concentration gradients.
tipm1248.tex; 29/07/1997; 6:53; v.7; p.3
72
MARCIO
Both the heuristic modifications of Terzaghis effective stress principle and

the ad-hoc viscoelastic models for secondary consolidation need to be rigorously
justified by upscaling the microscale behavior. Theoretical approaches which have
been used to develop models of porous media include mixture theory and other
methods which propagate microscopic governing equations to the larger scale
(e.g. homogenization, volume averaging, etc.). Hybrid mixture theory, HMT (see
Hassanizadeh and Gray [33, 34]) consists of classical mixture theory in the sense
of Bowen [18] applied to a multiphase system with volume averaged balance
equations. HMT is applicable to a multi-phase mixture in which the characteristic
length of each phase is small relative to the extent of the mixture. An average
value for each phase property is established at every point in the mixture, forming
M coexisting continua at each point. Macroscale dependent variables are defined to
be as consistent as possible with their microscale counterparts, so that an analogue
of classical Gibbsian thermodynamics can be developed. Variables such as swelling
pressure must often be defined in a somewhat nonintuitive fashion. Constitutive
equations are developed on the averaged scale and are subject to constraints placed
by the entropy inequality (Coleman and Noll [21]). HMT has been used extensively
to improve our understanding of flow and deformation in nonswelling porous media
(see Hassanizadeh and Gray [3537]). More recently, the theory has been clarified
and extended to derive remarkable results for two scale single porosity swelling
systems such as smectitic clay pastes (Achanta et al. [1, 2]). This framework has
since been extended to three-scale swelling systems (i.e. porous systems composed
of swelling porous particles and bulk fluid filled voids or cracks), by Murad et al.
[61], Bennethum and Cushman [12], and Murad and Cushman [62].
A three-scale model (micro, meso and macro) of a porous matrix consisting of
porous swelling particles is depicted in Figure 1. The particles are in contact with
one another and bulk water. Each particle consists of clay colloids and adsorbed
water. In Murad et al. [61], Bennethum and Cushman [12, 13] and Murad and
Cushman [62] the adsorbed water is treated as a separate phase from the bulk
water. At the microscale the model has two phases, the disjoint clay platelets and
the adsorbed water. At the mesoscale (the homogenized microscale) the model
consists of the clay particles and the bulk water. The macroscale consists of the
bulk water homogenized with the mesoscale particles. To propagate information
between scales, several types of upscaling methods can be used. For example, one
can upscale the microscale to the mesoscale using methods such as, e.g., homogenization or volume averaging. Since the microscopic adsorbed water is viewed
as a thin film coating the clay minerals, this method of upscaling would be very
complex as it would involve averaging the thin film governing equations coupled
with those governing the large deformations of the solid phase (e.g. elasticity,
viscoelasticity). Alernatively, one can perform such upscaling by adopting the HMT
of Hassanizadeh and Gray [34, 35] together with a proper theory of constitution
including appropriate internal variables needed to capture the swelling character of
the system. The mesoscopic internal constitutive variable which captures particle
tipm1248.tex; 29/07/1997; 6:53; v.7; p.4
73
swelling is the volume fraction. Hence one can perform a simpler upscaling by
pursuing the framework of Achanta et al. [2] and adopting a volume fraction
hybrid mixture theory in the sense of Bedford and Drumheller [11].
This type of upscaling was adopted in Murad et al. [61] who used the hybrid
mixture theory of Achanta et al. [2] to upscale the microscale to the mesoscale. Then
assuming Stokesian slow bulk-water movement, the homogenization procedure
(Bensoussan et al. [15], Sanchez-Palencia [64]) is used to upscale the mesoscopic
governing equations to the macroscale. The extension of this approach to coupled
water flow and solid deformation with disconnected (entrapped) bulk phase fluid
and clay particles was developed in Murad and Cushman [62]. In their framework,
macroscopic governing equations for flow and deformation were rigorously derived
by upscaling the microstructure. A different approach was adopted in Bennethum
and Cushman [12, 13]. Here information was propagated to the mesoscale by
averaging the microscale balance laws, but a constitutive theory was not developed
on the mesoscale. Rather, the mesoscopic equations were again averaged directly to
the macroscale and a constitutive theory developed at this latter scale by exploiting
the entropy inequality in the sense of Coleman and Noll [21]. The fundamental
difference between these two approaches is the development of a constitutive theory
on the mesoscale in the former and not in the latter. Each approach has advantages
and disadvantages which are governed by the type of experiments one wishes to
run. The upscaling technique from the meso to the macroscale pursued in Murad
and Cushman [62] was a straightforward homogenization of the entire hydrophilic
swelling clay soil. It yielded a macroscopic Darcys law in which the velocity is
a superposition of the adsorbed and bulk water velocities. This type of model has
been referred to as a parallel flow type model (Showalter [66, 67]) because the
secondary mesoscopic adsorbed/bulk water flow is neglected and the geometry of
the cells is suppressed in the upscaling procedure. One of the disadvantages of this
approach is that both the particles and bulk phase fluid have the same time scale and
therefore the model cannot incorporate delayed adsorbed water flow. This latter
type of flow is crucial for explaining secondary consolidation.
A notable consequence of the HMT framework of Murad and Cushman [62]
is the appearance of a new stress component, the hydration stress tensor, which
accounts for physico-chemical effects. However, experimental validation for these
stresses and their relation with other measurable physicochemical quantities such
as swelling pressure still needs to be clarified. Our first goal is to provide insight into
the physical interpretation of this physicochemical quantity. We then compare the
constitutive equations of Murad and Cushman [62] with Lows swelling pressure
concept and then we provide an alternative way of measuring hydration forces in
active clays. In addition, we show that hydration forces lead to the appearance of
a new thermodynamic quantity which governs the excess in pressure of the clay
particles relative to the classical pore pressure of Biot [16, 17] for nonswelling
granular media. We shall illustrate that, although Lows swelling pressure has
tipm1248.tex; 29/07/1997; 6:53; v.7; p.5
74
MARCIO
Derjaguins disjoining pressure as a microscale counterpart, this new excess in

pore pressure does not.
Furthermore, to derive three-scale macroscopic governing equations we adopt
the homogenization procedure to upscale the mesoscale governing equations of
the clay particles together with the Stokesian slow bulk-water movement. We aim
at deriving macroscopic equations that incorporate the delayed flow of adsorbed
water during secondary consolidation. Within the framework of the homogenization
procedure this can be achieved naturally by simply adopting a different scaling
law for the mesoscopic conductivity for the clay particles than that of Murad
and Cushman [62]. This scaling captures the secondary mesoscopic adsorbedbulk water flow. This yields a macroscopic picture which exhibits an additional
capacity term to account for the momentum interchange between the particles and
surrounding bulk fluid and leads to a dual porosity or distributed microstructure
model for swelling clay soils. This technique has been successfully used to model
naturally fractured reservoirs in which the system of fractures plays the role of
the bulk system (where the macroscopic flow takes place) and the matrix blocks
behave analogous to the clay particles and are treated as sources/sinks to the bulk
phase (see, e.g., Arbogast, Douglas and Hornung [4, 5, 27] and references therein).
The macroscopic bulk phase flow is influenced at the mesoscale through distributed
source/sinks of momentum which govern the creep constitutive equations for the
macroscale effective stress tensor. In the linear case the dual porosity model can
be solved by using a Greens function and then reducing it to a single integrodifferential equation of Volterra type in which the kernel appears related to the
geometry of the clay particles. The integro-differential equation can then be related
to some viscoelastic models proposed for secondary consolidation. We remark that
the derivation of the memory effects (due to the delayed adsorbed water secondary
flow) within the HMT three-scale framework of Bennethum and Cushman [12]
would require a more general exploitation of the entropy inequality involving
history dependent constitutive variables. Hence, we adopt the homogenization
procedure.
In the theory developed herein we will assume that the exchangeable cations
are concentrated on the clay surface, such that the platelets negative surface charge
is effectively screened. In other words, as in Low [57, 58], we will consider that
surface hydration is the dominant component of the swelling pressure.
2. Constitutive Equations for the Mesoscale Swelling Clay Particles
We begin by reviewing the main mesoscale results of the constitutive theory developed by Murad and Cushman [62] for a system composed of clay-platelets and
adsorbed-water (the clay particle). Then we re-examine the constitutive theory to
obtain a better feel for the hydration and swelling stresses.
tipm1248.tex; 29/07/1997; 6:53; v.7; p.6
75
Figure 1. Three-scale model for clay (from Murad et al. [61]).
2.1. NONEQUILIBRIUM RESULTS

Consider the clay particles as a mixture of two phases (the solid clay platelets and
liquid adsorbed water) viewed as coexisting continua, which undergo independent
motions x = x (X ; t); = l; s with respect to each reference configuration (here
x denotes the spatial position of the particle of the -phase at time t with respect
to a reference position X ). Let the subscript = l; s denote the adsorbed liquid
and solid phase respectively and let ,t , and A denote the averaged density,
symmetric particle stress tensor, volume fraction and intensive Helmholtz potential
of phase . Further, let T denote temperature (assumed equal in both phases) and
let the average mesoscopic strain tensor of the solid phase Es be given as
= 12 (Cs I);
(2.1)
where Cs = FTs Fs with Fs = grad xs denoting the deformation gradient of the solid
Es
phase (with grad denoting the differentiation with respect to a material particle on
the mesoscale).
Assume that on the mesoscale the solid and fluid phases are incompressible,
nonheat conducting, and that the adsorbed water is nonviscous. By postulating
constitutive dependence of the free energies in the form A = A (T; Es ) ( = l; s)
and using the Coleman and Noll method of exploiting the entropy inequality [21],
Murad and Cushman [62] obtained the following constitutive equations for the
stress tensors t
l tl = l p`I;
sts = spsI + tes + tls;
(2.2)
(2.3)
where the tensors tes and tls are defined by

@As FT ; tl = F @Al FT
tes = s s Fs
@ Es s s l l s @ Es s
(2.4)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.7
76
MARCIO
and p is the thermodynamic pressure of the -phase. For incompressible media,

the ColemanNoll method is applied to a modified entropy inequality obtained by
adding to the original entropy inequality the continuity equations premultiplied by
Lagrangian multipliers as constraints (see [53, 61, 62] for details). The thermodynamic pressures p turn out to be identified with the Lagrange multipliers. In
the compressible case it is postulated A = A (T; ; Es ) and application of
the ColemanNoll method yields the same constitutive equations (2.2) and (2.3)
with the exception that the thermodynamic pressures have their classical definition
p = ( )2 (@A =@ ).
In the constitutive theory of Murad and Cushman [62] it was postulated that
Al depends on Es and Es was included in the list of independent variables (see
also Bennethum and Cushman [12]). This implies the mesoscale thermodynamics
of the adsorbed water differ from that of a bulk phase fluid and also, unlike
non-swelling granular media, leads to the appearance of the new tensor tls . The
inclusion of Es allows for strain induced flow of the adsorbed water at the
mesoscale. This dependence of Al on the mesoscale solid strain is more general
than that of Achanta etal. [2] and Murad et al. [61], who postulated Al (T; l ). Their
constitutive theory was based on the experimental observations of Low [56] relating
the behavior of the adsorbed water to the platelet separation h. The assumption
that Al depends on l describes accurately swelling particles with interlayer spaces
. In this range the adsorbed fluid can withstand the hydrostatic
between 25 and 100 A
swelling pressure but not shear stress. The additional dependence of Al on shear
deformations is motivated by the fact that for small platelet separation h (less
), the adsorbed fluid molecules become more
than 10 molecular diameters or 25 A
ordered and arrange themselves in layers parallel to the surface, the film becomes
structured, inhomogeneous, anisotropic, its effective viscosity rises dramatically
and is able to sustain a shear stress even at equilibrium (see, e.g., Israelachvili
et al. [42], Schoen et al. [65], Cushman [22]).
The physical interpretation of (2.3) can be obtained by comparing it with the
analogous results of Hassanizadeh and Gray [35] for nonswelling granular media.
As we shall illustrate next, the difference between these two types of media is
the stress tls for swelling media which arises due to the additional constitutive
assumption Al = Al (Es ). We shall illustrate that tls governs physico-chemical
stresses within the clay particles and may also be identified with the swelling
pressure.
2.2. PHYSICAL INTERPRETATION OF THE NEW TENSOR tls FOR SWELLING MEDIA
Equation (2.3) is crucial in the present formulation since it contains important
information on the constitutive behavior of the solid phase stress tensor for the
swelling particles. To exploit its physical significance let us introduce the total
particle stress tensor t and the particle thermodynamic pressure p as
t = s ts + l tl ;
p = l pl + sps:
(2.5)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.8
77
By adding (2.2) and (2.3) and using (2.5) we obtain

t + pI = tes + tls :
(2.6)
The above result gives important insight into the stress in swelling particles. To
elucidate this consider a fixed solid strain Es and define the bulk phase B to be fluid
unaffected by the solid phase (e.g. nonswelling granular media). By definition, the
free energy of a bulk fluid AB does not change with the proximity of the solid and
therefore is independent of Es . Therefore tls = 0 and noting tes only depends on the
fixed solid strain Es , (2.6) reduces to
tB
+ pB I = tes ;
(2.7)
where now the subscript B is used to denote the corresponding property for a
non-swelling granular medium. Equation (2.7) has been derived by Hassanizadeh
and Gray [35] within the context of hybrid mixture theory applied to nonswelling
granular media. In classical soil mechanics the above result resembles in form
Terzaghis effective stress principle at the mesoscale for nonswelling media with
pB and tes normally referred to as pore pressure (or bulk phase pressure) and
effective stress tensor. In classical stress analysis of nonswelling media the pore
pressure pB has a similar definition to p in (2.5) except that it is assumed equal to
both thermodynamic fluid and solid pressures, i.e. (see, e.g., [16, 35])
pB = sps + l pl = pl = ps;
(for a nonswelling medium)
(2.8)
The effective stress tensor tes measures stresses induced by mineral to mineral
contact and primarily controls the deformation of nonswelling systems such as
sands, silts, and low and medium plastic clays such as kaolinite or illite. The
modified effective stress principle (2.6) for swelling media has the additional term,
tls . Unlike coarse-grained soils, whose stress mechanisms are primarily controlled
by the contact stresses tes , swelling clays such as montmorillonite contain the
additional stress component tls which governs the deformation of the swelling
particles. Clearly this additional intra-particle stress results from the presence of
adsorbed water within the particles. It is of physico-chemical nature and can be
viewed as a stress structural component arising from surface hydration. Whence, as
in Murad and Cushman [62], we henceforth denote tls the hydration stress tensor.
An important consequence of (2.6) is the partition of the total particle stress tensor
into its platelet (tes ) and adsorbed water (tls ) components. Consequently we can
overcome some limitations in the works of Lambe [48] and Hueckel [40] where it
is assumed that only one stress exists in the platelets and adsorbed water, which
is measured as the difference between the total macroscopic stress and bulk phase
pressure.
We next relate the hydration stress tensor tls to the swelling pressure which
has for many years been used to study swelling clays (see Low [56, 57]). Figure 2
tipm1248.tex; 29/07/1997; 6:53; v.7; p.9
78
MARCIO
Figure 2. Swelling experiment (from Low [57]).
depicts a classical reverse osmosis swelling pressure experiment performed by Low

wherein bulk water is separated from a well ordered parallel clay platelet-adsorbed
water mixture by a semipermeable membrane. Due to the hydrophilic interaction
between the adsorbed water and the clay minerals the clay tends to swell as water
penetrates the region between its superimposed layers and forces them apart. In
this experiment an overburden pressure P is applied normally to the clay-water
mixture and the average interlayer separation, h, of the platelets is measured. The
excess in the overburden pressure relative to the bulk pressure pB is the swelling
pressure , at equilibrium
P
pB :
(2.9)
Considering pB = patm , with patm denoting the atmospheric pressure, Low examined the equilibrium swelling pressure of different montmorillonites clays saturated with adsorbed water. For incompressible fluid he found that the dimensionless
swelling pressure (=patm ) satisfies the empirical relation

+ 1 = exp
e e

= B exp
;
e
(2.10)
where e = l =(1 l ) is the void fraction, e is the void fraction when = 0

(i.e. when P = pB ), is a constant that is related to the specific surface area
and the cation exchange capacity, B = exp( =e ), and the notation for the
dimensionless swelling pressure has been maintained.
In what follows we pursue a more general definition of the mesoscale swelling
pressure which retains the same physical interpretation as (2.9) under the equilibrium conditions of the swelling experiment depicted in Figure 2. For simplicity
we assume incompressibility and first note that in Lows swelling pressure experiment, the reference bulk phase pressure pB is defined in the domain occupied by the
bulk water. Unfortunately the generalization of Lows definition (2.9) to the case
where particles undergo nonequilibrium processes requires a pointwise definition
for (x; t). We thus pursue a local definition for relative to a reference virtual
tipm1248.tex; 29/07/1997; 6:53; v.7; p.10
79
bulk fluid which shall be locally constructed. To this end begin by defining the
chemical potential density (Gibbs free energy) of the the adsorbed and bulk fluids
G A + 1p; = l; B:
(2.11)
where, for simplicity the notation = l = B has been used when the fluids are
assumed incompressible. We then invoke a classical Gibbsian result which states

that at equilibrium, the chemical potentials of a single constituent coexisting in
two phases are equal (see, e.g., Callen [20]). Using the maximum entropy principle
Callen [20] illustrated this result in a classical example of osmotic water-pressure
difference across a semipermeable membrane and showed that the chemical potential of the water is constant in this experiment. The same idea can easily be applied
to Lows swelling experiment to show that at equilibrium the chemical potential of
the adsorbed water is equal to that of the bulk water, i.e. Gl = GB . We make use
of this result to characterize the virtual local reference bulk fluid (denoted by the
subscript B ) where (x; t) 0: This reference bulk water is constructed at instantaneous equilibrium with the adsorbed water such that their chemical potentials are
equal, and the swelling pressure (x; t) locally represents a pressure excess due to
the interaction of the water with the clay. In other words, would be zero if the
properties of the water were unaffected by the interaction with the solid phase, as
in the case of a bulk fluid. If we denote the free energy of the reference bulk fluid
by AB , the postulate Gl = GB together with (2.11) gives
Al + 1pl = AB + 1 pB :
(2.12)
The above result provides a partial relation between the thermodynamical properties
of the adsorbed water and reference bulk fluid. To complete the characterization
of this local reference state recall that, in the absence of thermal effects, Al only
depends on l in Lows experiment [58]. Denote l = e =(1 e ) as the volume
fraction defined in Lows relation (2.10) for which = 0 with l = l . At l
adsorbed water behaves as a bulk fluid and, hence, Al (l ) = AB . This, combined
with (2.12) yields
pB = pl + (Al AB ) = pl
l
l
@Al ds:
@s
(2.13)
Since Al is a function of l , the above result furnishes a definition for the reference
bulk phase pressure pB in terms of fpl ; l ; l g. Together with (2.7) this provides
the local characterization of pB and tB for a fixed solid deformation. We now
redefine the swelling pressure locally relative to pB . Begin by noting that definition
(2.9) is restricted to an equilibrium well ordered parallel platelet arrangement, in
which there is no mineral to mineral contact effective stresses. One may generalize
the swelling pressure concept to particles of curved shape, to incorporate particle
shear stress and nonequilibrium viscous effects. To do so we introduce a vectorial
tipm1248.tex; 29/07/1997; 6:53; v.7; p.11
80
MARCIO
definition for , namely the swelling stress vector which we shall henceforth denote
in boldface. For a mesoscopic surface of unit normal n, for tB given as in (2.7),
and pB defined in (2.13), define locally as
(x; n; t)
(t tB )n:
(2.14)
This mesoscopic definition is consistent with the microscale vectorial definition

for the disjoining pressure proposed by Kralchevsky and Ivanov [47] and Ivanov
and Kralchevsky [43] for curved thin films undergoing nonequilibrium processes
where viscous effects are important. In addition, (2.14) reduces to Lows swelling
pressure definition (2.9) in the swelling experiment at equilibrium. To show this
recall that in the swelling pressure experiment of Figure 2, t = P I (P denotes
the overburden pressure) and the effective component vanishes for the arrangement
of parallel platelets. Using this in (2.7) we get tB = pB I and (2.14) reduces to
the classical swelling pressure relation
= n with = P pB . Note that in
general (t tB ) may have off-diagonal components and consequently may also
have a tangential component to the mesoscale surface.
An open question is the role the excess in thermodynamic pressure of the
adsorbed fluid, relative to the local reference bulk phase pressure (pl pB ), plays
during particle consolidation. Unlike Lows swelling pressure , which has Derjaguins disjoining pressure as a microscopic counterpart, the excess pl pB does
not exhibit any microscale analogy. The reason is the different thermodynamic
representation adopted at the mesoscale (e.g. Al = Al (T; l ; Es ) for compressible
media) than that of thin films. This latter microscopic formulation usually adopts a
different Legendre transformation in which the Helmholtz free energy is replaced
by the Gibbs energy as a thermodynamic potential (see Derjaguin et al. [25]), or
even in the free energy representation the microscopic film density is eliminated
from the list of independent variables (see e.g. Li [51, 52]).
Further, note that since pl affects the total particle thermodynamic pressure, p,
through (2.5), we are led to also quantify an excess in the total particle pressure p
relative to the bulk phase pB . Henceforth, we shall refer to this difference as the
excess in pore pressure, (B ), i.e.
B p pB :
(2.15)
In analogy to the swelling pressure, the above definition reflects locally the excess
in pore pressure due to the interaction between the adsorbed water and the clay
minerals. In other words, B would be zero if the properties of the water were
unaffected by the interaction with the solid phase.
Next we proceed to derive a relation between the hydration stress tensor tls and
the swelling stress vector and the excess in pore pressure B . To this end define
0(x; t; n) B (x; t)n (x; t; n):
(2.16)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.12
81
Applying (2.6) to a mesoscale surface of normal n, and using definitions (2.7),

(2.14) and (2.15) we find
tls n
(t tes + pI)n
(t tes + (pB + p pB )I)n
(t tB + (p pB )I)n
B n = 0 :
=
=
=
=
(2.17)
Hence, projection of tls onto the normal to a mesoscale surface may be interpreted
as the difference between the excess in pore pressure and swelling stress vector.
The above result leads to an alternative way of writing the modified Terzaghis
effective principle (2.6) in terms of the reference bulk phase pressure pB , rather
than the particle thermodynamical pressure p. Define the tensor
t
= tls
B I
(2.18)
and combine with (2.17) to get

t n =
:
(2.19)
We may think of t as a swelling stress tensor, since in analogy to the classical

Cauchy argument, the projection of t onto a mesoscale surface of unit normal n
gives the swelling stress vector .
Further combining (2.6) with (2.15) and (2.18) we get
= pI + tes + tls
= (p B )I + tes + tls
= pB I + tes + t :
B I
(2.20)
This result is an alternative form of t which expresses the mesoscopic modified

effective stress principle (2.6) with p replaced by pB . Physico-chemical forces
in (2.20) are measured by the swelling stress tensor t . This alternative way of
expressing the modified Terzaghis principle resembles in form some heuristic
modified effective stress principles for clays discussed in, e.g., Sridharan and Rao
[70] or Lambe [48]. Historically, physico-chemical forces have heuristically been
modeled at the macroscale through the addition of a term to Terzaghis principle
which measures the effect of net repulsive (RI) and attractive (AI) forces between
particles. This stress is commonly denoted by (R A)I. Denoting the intra-particle
mesoscopic counterpart of these stresses as (r a), Equation (2.20) is a first rational
attempt at a rigorous derivation of the modified Terzaghis principle. From (2.20)
we have r a = t which shows that the net attractive-repulsive intra particle
forces arising from surface hydration are governed by the swelling stress tensor
t . Hence, we can reproduce the basic mechanical models for stress partitioning
between solid and fluid phases discussed for example in Sridharan and Rao [70] and
tipm1248.tex; 29/07/1997; 6:53; v.7; p.13
82
MARCIO
Lambe [48]. The effective stress concept for soils with physico-chemical stresses
and the role of hydration forces in carrying the total load has been controversial
and consequently more than one definition for effective stress has been proposed.
For example, Lambe [48] defined effective stresses as the difference between total
stress and pore pressure. Using Lambes definition in (2.20) the effective stress is
the sum of the mineral contact and swelling stress components. On the other hand
Sridharan and Rao [70] argued against Lambes definition and defined effective
stress as the mineral-mineral contact stress as it controls the resistance against
failure. Hence, if we define the mesoscopic effective stress tensors tL and tSR in
the sense of Lambe and Sridharan and Rao respectively by
tSR
= tes ;
tL
= tes + t :
Then the modified effective stress principle at the mesoscale is

t=
pB I + tL = pB I + tSR + t = pB I + tSR + r
a:
(2.21)
As we shall see further in Section 6, this result has an macroscopic analogy.

Consequently some controversial aspects in stress analysis in cohesive soils are
clarified within the current approach. Note that the deviatoric part of t does
not necessarily vanish. This suggests that in general r a is a full rank tensor.
The presence of off-diagonal components in t may be important at low moisture
contents where according to Schoen et al. [65], Israelachvili [42] and Cushman [22]
the adsorbed fluid and the solid surface may support shear forces at equilibrium,
and consequently the swelling stress vector may also have a tangential component
at the mesoscopic surface. As we shall show next, if we exclude the range of
moisture where the adsorbed fluid can support shear stresses, then r a reduces to
a multiple of the identity, i.e. r a = (r a)I and the swelling stress vector acts
normal to the surface, i.e. = n.
2.3. SWELLING AND HYDRATION STRESSES AT MODERATE MOISTURE CONTENT

A simplified scalar concept of hydration and swelling stresses can be obtained
by considering a moderate moisture content greater than that occupied by 10 fluid
monolayers. In this range the adsorbed water can not withstand shear at equilibrium
and therefore a well ordered particle composed of parallel platelets as depicted in
Lows experiment can only compress or expand and thus support no shear forces.
As we will show, in the range of moderate moisture content, the only term in the
right hand side of (2.6) and (2.20) with nonzero off-diagonal components is the
effective stress tensor tes . The hydration and swelling stress tensors tls and t reduce
to multiples of the identity. The assumption of moderate moisture content can be
easily imposed by postulating that Al does not depend on the deviatoric part of the
solid strain Es . On the other hand, unlike the bulk liquid, dependence of Al upon
the volumetric strain is still retained (see Achanta et al. [2] or Murad et al. [61]).
tipm1248.tex; 29/07/1997; 6:53; v.7; p.14
83
Thus, let Js = det Fs be the Jacobian of the incompressible solid motion whose
material derivative satisfies [35, 62]
Ds(Jss) = 0:
Dt
Denote the volume fraction of the reference configuration by s

integration
= s (Xs). After
Jss = s:
(2.22)
Hence Js governs the volumetric mesoscopic deformation of the solid phase. In the
range of moderate moisture content, as in [2, 61], we assume that the free energy
of the adsorbed fluid depends on volumetric strains by postulating Al = Al (Js ) or
Al = Al (l ) since they are coupled by (2.22). Using (2.1) in (2.4) together with
the identity (@Js2 =@ Cs )Cs = Js2 I (Eringen [28]) and (2.22) we get
tls
@Al FT = 2 @Al @Js2 C

l 2
@ Cs s
@Js @ Cs s
@A
@A @
= 2l Js2 2l I = 2l Js2 l s2 I
@Js
@s @Js
l s @Al I = @Al I
=
l s
Js @s
@l
= 2l Fs
(2.23)
Together with (2.18) this shows that in the range of moderate moisture contents tls
and t reduce to multiples of the identity. Hence, we are led to introduce the scalar
components of tls and t , namely the hydration pressure p and swelling pressure
, as
l
p = @A
@ ;
l
=
1
3 tr
t :
(2.24)
Using the above definition for p in (2.23), the hydration stress tensor is given by
tls
= p l s I = p l (1 l )I:
(2.25)
Since t is a multiple of the identity, using definition (2.24) in (2.19) we have
= n with
t
= I = (p l s
B )I
(2.26)
where (2.18) and (2.25) have also been used in the last equality. In addition, using
(2.25) in (2.3) and (2.6), respectively, along with definition (2.15) yields
ts
= ( ps + p l )I + (s ) 1 tes
(2.27)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.15
84
MARCIO
tes
= ( p + p l s )I = ( pB + p l s
B )I
= (pB + )I;
(2.28)
where (2.26) has also been used in the last equality in (2.28). The above result is
our modified mesoscopic Terzaghis effective principle in the range of moderate
moisture content. When comparing with (2.21) we now have r a = (r a)I =
I which shows that intra-particle net attractive-repulsive forces are governed by
the swelling pressure. Moreover, in applying (2.28) to Lows swelling experiment
and recalling that t = P I and tes = 0 because the platelets are ordered, we
reproduce the classical swelling pressure definition = P pB .
2.4. EQUILIBRIUM RESULTS
Our aim in this subsection is to obtain a relation between hydration forces and
swelling pressure at equilibrium. Then we can make use of Lows experimental
relation (2.10) to derive the dependence of the hydration pressure on the volume
fraction, i.e. the relation p = p (l ). Following Truesdell and Toupin [73], it is
postulated that at equilibrium entropy is a maximum and entropy generation is
a minimum. Application of these conditions to the entropy inequality yields at
equilibrium (see Murad and Cushman [62] for details)
pl = ps = p:
(2.29)
The above result states that at equilibrium, the thermodynamic pressures of the
solid and adsorbed fluid phases are equal. Recall that this reproduces (2.8) which
is a result that has been extensively used in the theory of granular nonswelling
media even at nonequilibrium (see, e.g., [17, 35]). As we shall see in the next
subsection the equality between pl and ps may not necessarily hold in swelling
systems away from equilibrium. Moreover, using (2.29) in (2.15) and combining
with (2.13) yields
B = pl
pB =
l
l
p(s) ds:
(2.30)
If we combine the above result with (2.26) we get
= B
l sp =
l
l
p(s) ds pl (1 l ):
(2.31)
Since p has the thermodynamical definition (2.24), the above result provides
an alternative thermodynamic definition for the swelling pressure at equilibrium.
Using Lows relationship (2.10) in (2.31), one can determine the relation
p = p(l ) and therefore Lows experimental result for may provide alternative
tipm1248.tex; 29/07/1997; 6:53; v.7; p.16
85
ways of measuring hydration forces in clays. This procedure is somewhat different

from that of Achanta et al. [2] and Murad et al. [61] where Lows result was
reproduced by neglecting the stresses in the solid phase. Differentiating (2.31) with
respect to l we get
l ddp + 2p + (1 1 ) dd = 0:

l
l
l
Whence
d(2l p )
dl
l
(1
d
l ) dl
= 0;
which after integration and using p (l ) = (l ) = 0 yields

l
2l p (l ) =
(1
l
d(s)
s) ds ds:
Using Lows result for in the right-hand side we can derive a relation for p (l ).
Consider for simplicity = B = 1 in (2.10) and denote by Ei (x) the exponentialintegral function defined as
Ei (X )
X exp( )
d
= ln X +
Xn :
n=1 nn!
We then have
l p =
2
l
l
s2(1 s) exp
l
s ds

l

e exp e d(1=e) = Ei e
l

1 l
= Ei Ei
l ;
where Ei = Ei(1=e ) = Ei ((1 l )=l )). The above relation is a first attempt
=
l
to develop constitutive relations for hydration forces in swelling clay particles as

they appear measured by the relation p = p (l ).
2.5. NEAR-EQUILIBRIUM RESULTS
We begin by presenting the near equilibrium results of Murad and Cushman [62] in
the range of moderate moisture content. These results were derived by linearizing
tipm1248.tex; 29/07/1997; 6:53; v.7; p.17
86
MARCIO
the entropy inequality about equilibrium. In particular, when linearizing about

fvl;s; Dsl =Dtg, where vl;s vl vs and Ds=Dt @=@t + vs denote the
velocity of the adsorbed water relative to the solid phase and material derivative
following the solid phase respectively, the following results were obtained
l vl;s =
Kl (
rpl + prl) ;
(2.32)
s l
pl ps = DDt
;
(2.33)
where Kl and are material coefficients and for simplicity gravity has been
neglected. Equation (2.32) is the mesoscopic Darcys law for the adsorbed water
with Kl = Kl (l ) denoting the permeability tensor of the clay particles. In addition
to a pressure gradient, the above form of Darcys law contains a gradient of a
generalized interaction potential which accounts for swelling. The appearance
of this additional term is consistent with the fact that volume fraction gradients
provide a potential for adsorbed water flow in a swelling medium. From (2.32) we
can overcome the limitations of the works of Ma and Hueckel [59], and Hueckel
[40] where the adsorbed water is often termed immobile water and considered
part of the solid phase. Further, note that using definitions (2.11) and (2.24) in
(2.32) we have by the chain rule
l vl;s =
Kl (
rpl + rAl) =
rGl:
Kl
(2.34)
The above reproduces the well known result that the gradient of the chemical
potential provides the generalized force for flow of matter, i.e. matter tends to
flow from regions of high chemical potential to regions of low chemical potential.
Alternatively, recall from the classical thermodynamics of Stokesian fluids (Eringen
[28]) that in the absence of thermal effects, AB is constant for an incompressible
bulk fluid. We can then make use of (2.12) and rewrite Darcys law in its classical
form in terms of the gradient of reference bulk fluid as follows
l vl;s =
Kl (
rpB + rAB ) =
Kl
rpB :
(2.35)
Equations (2.32), (2.34) and (2.35) consist of alternative forms of writing Darcys
law for the adsorbed water flow. As we shall illustrate in the next sections the
adoption of a particular form is somewhat related to the choice of primary variables
in the set of governing equations for the particles.
We now turn to the physical interpretation of (2.33).Using (2.33) in (2.5) we
also have
s l
p = pl s DDt
:
(2.36)
Using (2.33) and (2.36) in (2.27) and (2.28) respectively yields

ts
s l
pl + pl + DDt
I + (s )
tes
tipm1248.tex; 29/07/1997; 6:53; v.7; p.18
87
s l
I:
pl + pl s + s DDt
(2.37)
tes
Equation (2.36) tells us that near equilibrium, the thermodynamic pressure of the
adsorbed fluid and solid phases are not necessarily equal. Thus, the commonly
assumed equality between pl and ps for granular nonswelling media (2.8) may
not necessarily hold, especially for swelling systems. The coefficient may be
thought of as a retardation factor which among other effects, accounts for the reordering of the adsorbed water molecules as they are disturbed, i.e. an entropic
effect (see Bennethum et al. [14]). If this is the only source of retardation, then it
follows that for a granular medium, 0, since there is very little ordering of the
bulk liquid phase in such a medium. The evaluation of requires experimental
study. In a different fashion, some information on this coefficient can be obtained
by averaging the constitutive relations for the nonequilibrium disjoining pressure
of microscopic thin liquid films (see [47, 43]). To this end use (2.36) in (2.15) along
with (2.13), and obtain the following near equilibrium relation for B
B = p pB = pl
=
l
l
s l
pB s DDt
s l
:
p(s) ds s DDt
(2.38)
When combined with (2.26) this yields for the swelling pressure
=
l
l
s l
p(s) ds pl s s DDt
:
(2.39)
Hence, we may think of as composed of two parts. A static (equilibrium) component (eq ) measured by the first two terms in the right-hand side and a viscous
(non-equilibrium) component (neq ) measured by the last term. The motivation for
this decomposition is based on a similar microscopic result proposed Kralchevsky
and Ivanov [47] and Ivanov and Kralchevsky [43] for the viscous disjoining pressure of thin films away from equilibrium. After neglecting convective effects and
using conservation of mass @l =@t + l div vl = 0 we have
eq =
l
l
p (s) ds p l s
@
neq = s l = s l div vl :
@t
Kralchevsky and Ivanov [47] have advocated that the microscopic counterpart
of the purely viscous nonequilibrium component neq accounts for the excess in
the viscosities of the thin film relative to the bulk phase. Thus, one can extend this
argument to the mesoscale and possibly identify the coefficient s l with the
tipm1248.tex; 29/07/1997; 6:53; v.7; p.19
88
MARCIO
difference between the averaged mesoscopic volumetric viscosity of adsorbed and

bulk water. This averaged excess in viscosity, which we shall denote by l;B , was
also measured by Low [54] who experimentally obtained the following analogous
relation to (2.10)

l;B = B exp
= exp (1 l ) ;
B
e
l
where is another characteristic constant that depends on the nature of the montmorillonite. Thus l;B =(l s ) and the above provides a first attempt to
measure the non-equilibrium coefficient of the viscous disjoining pressure neq
in the average sense. Of course, this claim is subject to experimental validation.
3. Linearized Governing Equations for Clay Particles and Bulk Water
The infinitesimal theory for the clay particles is obtained following the standard
linearization procedure: Assume that particles are initially homogeneous, isotropic
and at equilibrium. Expand A ( = l; s) in a Taylor series about equilibrium and
retain quadratic terms in A and linear terms in the set of governing equations. In
particular, if we assume that As is an isotropic function of Es , depending only on
its invariants to fulfill the usual objectivity requirements (Eringen [28]), then the
linearization procedure is exactly analogous to that of the classical linear isotropic
elasticity theory [28]. Let us consider that the clay particles are initially at an
equilibrium state given by Es = 0, l = l and s = s (s = 1 l ) and let Al =
Al (l ) denote the free energy of the adsorbed fluid at the reference configuration.
For simplicity assume initially a well ordered parallel platelet arrangement within
each particle such that the reference configuration is free of effective stresses.
Let p = p (l ), K l I = Kl (l ) and = s (l ) and let the strain tensor be
identified with its linearized form
Es
= rs us;
(3.1)
where s us = 1=2( us + uTs ), with us denoting the displacement of the solid

phase. Let fs ; s g denote the pair of Lame coefficients of the platelet matrix, and
let denote a material coefficient of the adsorbed water. Postulate the quadratic
expansions
ssAs = 2s (tr Es)2 + s tr E2s;

A = A + p ( ) + (
l
l
l )2 :
Then the linearized forms of (2.4) and (2.24) are

tes
= s tr Es I + 2sEs ;
p = p + (l l ):
(3.2)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.20
89
Equation (3.2) is nothing but the mesoscopic version of the classical Biots linear
0
elastic constitutive equation for the effective stresses. Denote by fB ; ; g the
values of fB ; ; 0 g at the equilibrium reference state obtained by setting l = l
in (2.30) and (2.31) together with definition (2.16). We then have
B =
l
l
0 = B
p (s) ds;
= B
pl s;
= p l s :
(3.3)
Introduce the functions f (l ) and g (l ) as

g(l ) = l s ddp +p(s
l l =l

l );
f (l ) = g(l ) + p:
The linearized forms of (2.38), (2.39) and (2.16) are
B = B
s l
p(l l ) DDt
;
s l
pl s (g + p)(l l ) DDt
D
= f (l l ) s l ;
Dt
0 = p l s = p l s + g(l l ) :
= B
(3.4)
(3.5)
We are now ready for our mesoscopic linearized governing equation in the clay
particle domain. By neglecting all inertial and convective effects, the linearized
mass balances for the solid and fluid phases reduce to
@l + div v = 0;
@t l l
@s + div @ us = 0:
s
@t
@t
After adding them up and using the constraint s + l =
rewritten in terms of the percolation velocity ql l vl;s as
div ql + div
@ u s = 0;
@t
1, the above can be
@l + div q = 0:
@t s l
From the constitutive equations (2.32), (2.37), (3.2) and (3.5) together with the
balance laws, for Es as in (3.1), our system of linearized equations governing the
swelling clay particles written in terms of the unknowns fus ; ql ; tes ; l ; 0 ; pl ; tg is
tipm1248.tex; 29/07/1997; 6:53; v.7; p.21
90
MARCIO
Mass Conservation of the Adsorbed Water
@l + div q = 0:
@t s l
Total Mass Conservation
div ql + div
@ u s = 0:
@t
Total Momentum Balance

div t = 0:
Total Particle Stress Constitutive Equation
t=

l
plI + tes + 0 + @
@t I:
Linearized Effective Stress Constitutive Relation

tes
= s div us I + 2srs us:
Linearized Hydration Stress Constitutive Relation
0 = p l s + g (l
l ):
Modified Darcys Law for the Adsorbed Water

ql
= Kl (rpl + p rl ):
4. Mesoscopic Problem for Clay Particles and Bulk Water

Let
l and
f denote the clay particle and bulk water domains respectively, and
let be the interface between them. For given l and s and a set of coefficients
fK l ; ; s ; s; p; gg at the initial equilibrium state, the above system of linearized
equations governs the swelling of the particles in
l . In addition, following earlier
work, [61, 62], the slow Newtonian movement of the bulk phase is governed by
the classical Stokes problem
div tf
tf
= 0 in
f ;
= pf I + 2f rs vf in
f ;
div vf
= 0 in
f :
tipm1248.tex; 29/07/1997; 6:53; v.7; p.22
91
In the above system the corresponding bulk water properties in the bulk phase
domain
f are now denoted by the subscript f to distinguish from the subscript B
used for the reference bulk fluid in the particle domain
l . The above system is supplemented by boundary conditions on the particle-bulk water interface. Continuity
of mass and normal component of the stress tensor on leads to
@ us n
@t
ql n = v f
on
tn = tf n on
where n is the unit normal exterior to

l. Moreover, following [61, 62], an additional
boundary condition is proposed based on the continuity of the variable whose
gradient is the driving force for fluid flow. According to Darcys law (2.34) the
variable is the chemical potential. Hence, as in [61, 62] we postulate
Gl = Gf
on
(4.6)
where following definition (2.11), Gf

leading to (2.13) we can rewrite (4.6) as
pl pf =
l
l
p (s) ds
Al + 1pf . By the same arguments
on
(4.7)
or alternatively using (2.30)
pB = pf
on
(4.8)
The boundary condition (4.8) expresses continuity in the reference bulk fluid pressure. Continuity in the fluid pressure on the boundary between a non-swelling
granular medium and a free fluid has been previously used (see, e.g., Levy and
Sanchez-Palencia [49], Douglas and Arbogast [27]). In contrast to the continuity
between pB and pf , we may note from (4.7) that a discontinuity between pl and
pf arises due to the hydration pressure p. In the case of a granular medium where
p = 0, from (2.13) we have pl = pB and then the classical continuity in fluid
pressure is recovered.
5. Alternative Mesoscopic Formulation
An alternative way to formulate our mesoscopic initial/boundary value problem
can be obtained by replacing fpl ; 0 g by fpB ; g in the set of primary unknowns.
In linearizing (2.38) and using (3.3) these quantities are related via
pB pl =
l
l
p(s) ds = B + p (l l ):
(5.1)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.23
92
MARCIO
Using (2.28), (2.35) and (3.4) the alternative mesoscopic formulation in terms of
the primary unknowns fus ; ql ; tes ; l ; ; pB ; tg and ftf ; vf ; pf g is given by
div tf
tf
= 0 in
f ;
= pf I + 2f rs vf in
f ;
div vf
= 0 in
f ;
div t = 0 in
l ;
pB I + tes I in
l ;
tes = s div us I + 2s rs us in
l ;
@
= f (l l ) l in
l ;
@t
@ us = 0 in
;
div ql + div
l
@t
@l + div q = 0 in
;
l
l
@t s
ql = K l rpB in
l ;
tn = tf n on ;
ql n = vf;s n on ;
pB = pf on ;
l = l in
l; t = 0;
div us = 0 in
l ; t = 0:
After obtaining pB and l within this formulation, pl can be evaluated in a postt=
processing approach using (5.1).
6. Macroscale Behavior: Two Scale Asymptotic Expansions

In this section we use the homogenization procedure to upscale the mesoscopic
results derived in the previous section to the macroscale. Our swelling clay at
the macroscale is idealized as a bounded domain
" with a periodic structure.
Following the general framework of the homogenization procedure, described,
for example, in Bensoussan et al. [15] and Sanchez-Palencia [64], we introduce
tipm1248.tex; 29/07/1997; 6:53; v.7; p.24
93
Figure 3. Two elements of equivalent clay soils.
Figure 4. The reference cell
Q and its distribution over the homogenized soil.
mesoscopic and macroscopic lengths, denoted by l and L, which characterize

the mesoscopic size of the period and the field respectively. Their ratio " l=L.
Consider
" as the union of disjoint parallelepiped cells, Q", congruent to a standard
Q consisting of the union of several clay particles Ql completely surrounded by
a connected bulk water domain Qf . Let the systems of bulk phase water and clay
particles in
" be denoted by
"f and
"l , respectively. The "-model on
" consists
of the mesoscopic governing equations of Section 5 on each subdomain
"f and
"l. Our starting point, " = 1, corresponds to our mesoscopic model. For " < 1
a swelling clay soil is posited wherein the centers of the bulk phase channels are
located "-times the reference distance apart, though congruent to the reference
cell (Figure 3). The homogenized model for the macroscopic clay soil is obtained
by letting " ! 0 while the lattice extends to infinity. As we shall show next
the limit model (" ! 0) consists of a distributed model with microstructure in
which the macroscopic swelling clay soil is viewed as two coexisting systems: one
representing the clay particles and the other representing the bulk water. The picture
corresponding to the limiting model is depicted in Figure 4, where a mesoscopic
cell Q is assigned to each point x of the macroscopic bulk phase domain.
The approach developed next is similar to that proposed by Arbogast and coworkers [35, 27] for flow in naturally fractured reservoirs. For simplicity we
consider the limiting case of the clay geometry wherein the clay particle system
is disconnected. Following the terminology of fissured media this geometry is
termed totally fissured medium (TFM). Since particles are completely isolated
tipm1248.tex; 29/07/1997; 6:53; v.7; p.25
94
MARCIO
from each other by the bulk phase fluid there is no direct mass and momentum
transfer from particle to particle. Instead the adsorbed water first flows into the
bulk phase where it passes into another particle or remains in the bulk phase. As
a consequence, most of the flow passes through the bulk phase, while the storage
of the fluid takes place in the system of clay particles. This picture corresponds
to an idealized clay soil wherein clay particles are highly ordered so that there is
no solid-solid contact. In reality, the clay is not well ordered and there is some
particle-particle contact. A porous medium which exhibits interaction between
particles through their interfaces is termed partially fissured medium (PFM). In
PFM particles are connected to neighboring particles, so that a percentage of the
water flow passes through particle interconnections and therefore particles are not
only coupled indirectly through the bulk phase system. The modeling of PFM
requires an additional coexisting system that governs vicinal water flow from
particle to particle. The homogenization tools for upscaling such media requires
more complexity (see Douglas et al. [26] and Showalter [67]) and will be saved for
a latter occasion.
A crucial point in the analysis of dual porosity models is the proper scaling of
the coefficients by appropriate powers of ". The idea is to conserve flow in some
sense and consequently avoid degeneration of the governing equations as " ! 0.
Following Arbogast and co-workers [5, 27], this is done by considering the scaling
"
law K l = K l "2 . This scaling has the effect of making the particles progressively
less permeable as " ! 0 and consequently preserves the secondary particle bulk
phase flux. In addition, recalling the standard homogenization procedure of the
Stokes problem, the bulk water viscosity coefficient f is also rescaled by "2 (see
Auriault [7], Sanchez-Palencia [64]).
The upscaling is achieved by considering every property to be of the form f (x; y)
(where x and y denote the macroscopic and mesoscopic coordinates, respectively,
with y = " 1 x) and then postulating two scale asymptotic expansions for the
set u" consisting of primary unknowns. We then expand our set of unknowns
fus; ql ; tes; l ; ; pB ; tg and ftf ; vf ; pf g in terms of the perturbation parameter "
u"
= u0 + "u1 + "2 u2 +
with the coefficients ui ,

-periodic in y. Inserting the above expansions into the set
of mesoscopic governing equations with the differential operator @=@x replaced by
@=@x + " 1 @=@y we obtain, after a formal matching of the powers of ", successive
cell problems. For the bulk water we have
divy t0f
= 0;
(6.1)
divx t0f
+ divy t1f = 0;
(6.2)
t0f
= p0f I;
(6.3)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.26
= p1f I + 2f rsy v0f ;
t1f
95
(6.4)
divy v0f
= 0;
(6.5)
divx v0f
+ divy v1f = 0;
(6.6)
and for the clay particles
s4yy u0s + (s + s)ry divy u0s = 0;

divy t0
= 0;
divx t0 + divy t1
(6.7)
(6.8)
= 0;
(6.9)
t0
= (p0B + 0 )I + tes0 ;
(6.10)
t1
= (p1B + 1 )I + tes1 ;
(6.11)
tes0
= s(divx u0s + divy u1s )I + 2s(rsx u0s + rsy u1s );
0
l
l ) @
@t ;
@ u0s + div @ u1s = 0;
divy q1l + divx
y
@t
@t
@0l + div q1 = 0;
s y l
@t
q0l = 0;
0 = f (0l
q1l
= K l ry p0B ;
(6.12)
(6.13)
(6.14)
(6.15)
(6.16)
(6.17)
along with the boundary conditions
@ u0s n = 0 on ;
@t

@
u1s
1
1
ql n = vf
@t n on ;
(2srsy u0s + s divy u0s I)n = 0 on ;

v0f
(t0
t0f )n = 0 on
(6.18)
(6.19)
(6.20)
(6.21)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.27
96
MARCIO
(t1
t1f )n = 0 on
p0B = p0f
(6.22)
on
(6.23)
and initial conditions
0l = l ;
in
l ; t = 0;
divx u0s + divy u1s
(6.24)
= 0; in
l ; t = 0:
(6.25)
Next we formally collect our homogenized results. Recall that within the above
alternative formulation pl was replaced by pB and therefore a post-processing is
still required for evaluation of p0l
6.1. DARCYS LAW FOR THE BULK WATER FLOW
Using (6.1) in (6:3) we have t0f = p0f (x; t)I. In addition, noting that u0s satisfies
the Neumann problem given by (6.7) and boundary condition (6.20), we have
u0s = u0s (x; t). The macroscopic Darcys law for the bulk water relative to the solid
phase follows from the well-known upscaling of the Stokes problem (6.2)(6.5)
together with boundary condition (6.18) (see, e.g., Auriault [7], Sanchez-Palencia
[64]). Introducing the mean value operator
= j
j
d
i ;
i = l; f
and defining the macroscopic volume fractions of the particles and bulk phase,
respectively, by n = j
j=j
j; = l; f , and the averaged bulk phase velocity
relative to the solid phase by qff 0 vff 0 nf @ u0s =@t, we have
qff 0
= Kf rp0f ;
which is the classical Darcys law governing the macroscopic bulk water movement.
If we assume that the swelling medium is isotropic at the macroscale then the scalar
Kf denotes the macroscopic hydraulic conductivity for the bulk phase flow.
6.2. MODIFIED TERZAGHIS EFFECTIVE STRESS
To derive the macroscopic modified Terzaghis effective stress we apply the mean
value operator to (6.9)(6.11), use the boundary condition (6.22) together with
(6.2), (6.3), the divergence theorem, and periodicity assumptions to obtain
tipm1248.tex; 29/07/1997; 6:53; v.7; p.28
= divx tees 0
divx et
j
j
j
j
= j
j
=
rxpB rx
f
e0
(1 + p1B )I)n d
(tes1
t1f n d
j
j
97
divy t1f d
f
divx t0f d
f
= nf rx p0f :
Defining the total macroscopic stress tensor and bulk phase pressure as
(
T=
tf
t
f ;
in
l ;
in
Pf =
pf
pB
f ;
in
l ;
in
the above result together with boundary conditions (6.21) and (6.23) yield
0
e
divx T
= 0; Te = et0
0
nf p0f I = (Pff 0 + e 0 )I + tees0 ;
(6.26)
where the averaged form of (6.10) has also been used. This is similar in form to
the modified effective stress principle of Sridharan and Rao [70]
Pf I + tes + (R A)I
e 0 I. This shows that, in analogy with the mesoscale results, if
with (R A)I =
T=
we assume that swelling is governed by surface hydration, then the net attractiverepulsive forces between the clay particles are governed by the macroscopic
swelling pressure. The approach presented herein provides a rational attempt to
model constitutive responses associated with physico-chemical forces in clays.
According to Sridharan [68, 69, 70], the magnitude of the two last components
of the right-hand side of (6.26) varies with the type of clay considered and the
moisture content. For example, for coarse-grained soils such as sands, silts and
low and non-expansive medium plastic clays such as kaolinite or illite, the stress
0
mechanisms are primarily controlled by the contact stresses tees . While for active
plastic smectitic clays such as bentonite and montmorillonite the stress mechanisms
e 0 I.
appear to be governed by the swelling stress component
6.3. MASS BALANCE

Finally, we derive the macroscopic mass balance by averaging (6.6) and (6.14)
using the boundary condition (6.19) along with the divergence theorem and the
tipm1248.tex; 29/07/1997; 6:53; v.7; p.29
98
MARCIO
periodicity assumption to get
g 0
divx vf
j
j
= j
j
= j
j
= j
j
divy v1f d
v1f n d
!
@ u1
ql + s n d
@t
@ u1
divy q1l + divy s
@t
l
0
nf )divx @@tus :
= (1
Whence
divx
@ u0s + div qf 0 = 0;
x f
@t
which is our macroscopic mass balance equation.

6.4. BOUNDARY AND INITIAL CONDITION
From (6.23) and using (6.3) and (6.10) in (6.21), our set of homogenized boundary
conditions is
(tes0
(0 + p0B )I)n = p0f n on ;
p0B = p0f
on
which implies (tes0 0 I)n = 0 on . This, together with the initial conditions
(6.24) and (6.25) establishes our macroscopic model.
6.5. DUAL POROSITY MODEL
The application of the formal homogenization procedure presented above leads to
the following dual porosity model: Find fu0s ; p0f g functions of (x; t) and fq0f ; p0B ; t0 ;
T0 ; u1s ; q1l ; tes0 ; 0 ; 0l g functions of (x; y; t) such that
0
e
divx T
= 0;
e
T
= et0
nf p0f I;
@ u0s = 0;
@t
0
0
qff = Kf rx pf ; x 2
f ; t > 0;
divx qff 0 + divx
tipm1248.tex; 29/07/1997; 6:53; v.7; p.30
99
and
t0 = (p0B + 0 )I + tes0 = 0;
divy t0 = 0;
tes 0 = s (divx u0s + divy u1s )I + 2s ( sx u0s + sy u1s );
0
1
0
l
l ) @
@t ; ql = K l ry pB ;
@0l + div q1 = 0; div q1 + div @ u1s = div @ u0x ;
y l
y
x
s y l
@t
@t
@t
y 2
l ; t > 0
0 = f (0l
along with the boundary and initial conditions
(tes0 0 I)n = 0 on ;
p0B = p0f on ;
0l = l ; in
l ; t = 0;
divx u0s + divy u1s = 0; in
l ; t = 0:
(6.27)
In terms of fu0s ; p0f g and fu1s ; p0B g the above system can be rewritten as

nl s4xx s + (s + s + fs)rxdivx

u0
rxpf = F(x; t)

@
divx u0s
s + s rx @t
u0
@ u0s K 4 p0 = 0; x 2
; t > 0;
f xx f
f
@t
s u1 ( + f )r divgu1 +
F(x; t) = 2s divx rg
s
s x y s
y s
divx
+rx (pfB 0
(6.28)
nlp0f ) srx div@ty us
g 1
and
s4yy u1s + (s + s + fs)ry divy u1s ry p0B +

@ divy u1s = 0;
+ s ry
@t
@
u1s
0
divy
@t K l 4yy pB = G(x; t);
0
G(x; t) = divx @@tus (x; t); y 2
l ; t > 0;
(6.29)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.31
100
MARCIO
with boundary conditions

@
u1s
(s + fs )divy s + + s div
2s r
@t I n

@
u0s
s
0
0
= 2s rx us + (s + fs )divx us + sdiv
@t I n
p0B = p0f on :

s u1 +
y s
u1
on
After solving the above system for the unknowns fu0s ; p0f g and fu1s ; p0B g, we can
evaluate p0l using an upscaled version of the post processing (5.1) in the following
manner
p0l = p0B + B p (0l l ) = p0B + B ps(divy u1s + divx u0s );

where (6.14) and (6.15) have been used in the last step.
We have arrived at a homogenized dual porosity model which shows a macroscopic bulk phase momentum equation influenced at the mesoscale through a
macroscopically distributed source/sink vectorial momentum transfer function
F(x; t). In addition, the macroscopic flow takes place in the bulk phase and influences the adsorbed water flow within the mesoscale particles through another
macroscopically distributed source/sink scalar transfer function, G(x; t).
7. Reduction to Some Classical Poroelastic Models
The transfer functions F(x; t) and G(x; t) in the dual porosity model arose as a
consequence of treating the clay particles as a two-phase system, which allowed
us to capture the effects of the adsorbed water and its influence upon macroscopic
swelling. This is in direct contrast with some classical linear models for consolidation which have been developed at the macroscale and have treated the clay particles
as one phase (see, e.g., Biot [16, 17]) and Sridharan and Rao [70] for granular and
swelling media respectively). Within these phenomenological approaches the presence of adsorbed water within the particles is overlooked and consequently particle
swelling and secondary mesoscopic adsorbed-bulk water mass and momentum
transfer are neglected. A notable consequence of our approach is that it reduces to
some of these classical models when we neglect the adsorbed water within the particles. Adsorbed water effects can be overlooked by neglecting its mesoscopic flow
and the nonequilibrium swelling pressure component neq . These assumptions can
be imposed by dropping the nonequilibrium coefficients, i.e. setting K l = = 0.
Consequently, from the mass balances and Darcys law of the mesoscopic system
of governing equations (Section 5) we get ql = div us = 0 and = l , which
means that the clay particles are now viewed as a single incompressible solid phase.
In addition, if we a assume that the reference configuration is free of hydration
forces, i.e. = 0, then particle incompressibility and (3.4) leads to = 0 and
tipm1248.tex; 29/07/1997; 6:53; v.7; p.32
101
thus there is an absence of hydration forces. Consequently, the mesoscale system

of Section 5, in the clay particle domain, reduces to the following incompressible
elasticity problem in terms of fus ; pB g
s4us
rpB = 0
in
l ;
= 0 in
l ;
( pB I + tes )n = tf n on :
div us
Note that, unlike the system of mesoscopic equations of Section 5 in

l , the
problem is now governed by an elliptic equation. This assumption implies that
transient effects within the particles related to the adsorbed water are overlooked.
This means that particles reach equilibrium instantaneously as they are disturbed
by the bulk phase. The above system exhibits a somewhat different behavior than
that of Section 5. Its well posedness is ensured provided the pressure boundary
condition (4.8) is omitted. It is also well known that the upscaling of the above
system coupled with the Stokesian motion of the bulk water leads to the classical
Biot model of consolidation of linear elastic granular media (see, e.g., Auriault [6]).
Alternatively, one can also derive this upscaled result by imposing the assumptions
K l = = 0 in the cell problem (6.29). In doing so this system reduces to
s4yy u1s
ry pB = 0
l ;
divy u1s + divx u0s = 0 in
l ;
( p0B I + 2s(rsx u0s + rsy u1s ))n = p0f n on :
0
in
The homogenization of this problem is similar to that of the Stokes and incompressible elasticity problems with Neumann boundary conditions which have been
extensively discussed in the literature (see, e.g., Auriault [6, 8]). We shall omit the
details of the homogenization procedure. By linearity the solution, u1s , is given by
u1s = (y) sx u0s + (y)p0f + u(x; t);
where the vectorial coefficients

and are particular solutions corresponding
to sx u0s = I; p0f = 0 and sx u0s = ; p0f = 1, respectively. By redefinition of
coefficients, from the first equation in (7) we also have
p0B = (y) : rsxu0s + (y)p0f + p(x; t);

where and denote the redefined tensorial and scalar coefficients. Using the
above relations together with the assumption = 0, the macroscopic problem
(6.28) yields
divx C sx u0s
divx
rxp0f = 0;
@ u0s K 4 p0 = 0;
f xx f
@t
x 2
f ;
tipm1248.tex; 29/07/1997; 6:53; v.7; p.33
102
MARCIO
where C and also denote new tensorial averaged coefficients. The above system
consists essentially of Biots consolidation model of linear elastic media. Thus we
have shown that in neglecting the adsorbed water and viewing the clay particles as a
single incompressible phase, physico-chemical effects vanish and the dual porosity
model reduces to a classical model based on poroelasticity.
8. Memory Effects
Within the context of dual porosity models applied to fractured porous media it has
been shown by Arbogast [3], Peszynska [63] and Hornung and Showalter [38] that
application of the Greens function method may reduce the dual porosity equations
to a single integro-differential equation of Volterra type with a convolution kernel. In
this section we exploit this idea and illustrate that application of the same technique
to the dual porosity model of Section 6 leads to a similar long-term memory effect
0
e . This result
in the constitutive equation of the macroscopic total stress tensor T
may provide a rational basis for explaining several secondary consolidation stress
constitutive equations of creep type. Begin by decomposing the total macroscopic
stress tensor into its spherical and deviatoric parts
0 13 tr T0
and S0
T0 0I:
For convenience in what follows we shall concentrate our analysis on the macroscopic constitutive behavior of the volumetric component 0 I. For simplicity we
also restrict our analysis to the interior of the clay particles and neglect boundary
effect. These assumptions are made purely for ease of exposition and the results
which follow can be extended to the general case with shear stresses and boundary
effects. The general analysis requires further manipulations and will be pursued in
a future article. Our purpose here is only to illustrate that a straightforward application of the Greens function method reduces the linearized dual porosity model of
Section 6 to a macroscopic integro-differential equations in which the constitutive
behavior for the total stress appears governed by viscoelastic equation exhibiting
long term memory.
Let e1 = divy u1s denote the volumetric deformation within each cell. For
simplicity, assume that particles are initially incompressible, i.e. e1 (x; y; 0) = 0.
Taking the divergence of the first equation in (6.29), for each x 2
f , this system
can be rewritten in terms of the following equation for e1
1
@e1 = G(x; t) in Q ;
a4yy e1 + b4yy @e
(8.1)
l
@t @t
e1 = 0 in Ql; t = 0;
where a = K l (s + 2s + s f ) and b = s K l . This equation shows that the
volume fraction is governed by a linear KelvinVoigt viscoelastic equation, with
tipm1248.tex; 29/07/1997; 6:53; v.7; p.34
103
an additional external damping term. In a fashion similar to Arbogast [3], a more

complete understanding of the above system can be achieved in terms of the Greens
function method. Define the auxiliary Greens function g (y; t) as the solution of
the problem
@g = (t);
a4yy g + b4yy @g
@t @t
g = 0 in Ql ; t = 0;
(8.2)
where (t) denotes the Dirac measure in the t variable. The solution of problem
(8.1) can be written in terms of g . Define the Laplace transform
Lf (t) =
exp(
st)f (t) dt = fb(s):
In the Laplace space problems (8.1) and (8.2) are given by
(a + bs)4yy eb1
seb1 = Gb (x; s)
and
(a + bs)4yy gb sgb = 1:
By linearity eb1 and gb are related via
eb1 (x; y; s) = gb(y; s)Gb (x; s):

Hence, application of the inverse Laplace transform yields
e1 (x; y; t) =
g(y; t ) divx @@us (x; ) d;

0
which, after averaging, gives
ee1 (x; t) =
ge(t ) divx @@us (x; ) d

0
(8.3)
and
@ ee1 (x; t) = @ Z t ge(t ) div @ u0s (x; ) d:

x
@t
@t 0
@
(8.4)
tipm1248.tex; 29/07/1997; 6:53; v.7; p.35
104
MARCIO
Taking the trace of (6.26) using (6.12)-(6.15) we obtain
e 0 = Pff 0 + 13 tr tees0 e 0

0
2s
f
(n div
= P + +
s
+f (el
l
nl l ) + @@t
e
= (Pff + nl ) + s +
0
x us + ee
) nl +
2s
3
+ s f (nl divx u0s + ee1 ) +
@ ee1 + n div @ u0s :

@t l x @t
Hence, introducing the coefficients c = (s + 2s =3 + s f ) and d = s we get

1
0
@
u0s
0
0
f
e = (Pf + nl ) + nl c divx us + d divx @t + cee1 + d @@tee : (8.5)
+ s
Using (8.3) and (8.4) for the two last terms in the right-hand side of (8.5) we obtain
e 0 = (Pff 0 + nl ) + nl c divx u0s + d divx @@tus +

0
+c
+b
ge(t )divx @@us (x; ) d +

0
@ Z t ge(t )div @ u0s (x; ) d:

x
@t 0
@
This macroscale constitutive result resembles in form some viscoelastic models

which have been proposed to model creep during secondary consolidation (see,
e.g., [19, 74]). The viscoelastic constitutive equation exhibits fading type memory
where the stress tensor depends upon the complete past history of the strain. Within
the current approach, the memory appears somewhat related to the time dependent
non-equilibrium effects within the particles, due to the adsorbed water flow. As
illustrated in Section 7, if particle deviations from equilibrium are omitted, the
memory vanishes and consequently the secondary stage is neglected and consolidation appears primarily governed by the drainage of the bulk fluid. To the authors
knowledge the proposed approach is the first rigorous derivation of fading memory
in the macroscopic constitutive theory for secondary consolidation.
9. Conclusions
A three-scale theory for swelling clay soils was presented. Clay minerals, adsorbed
water and bulk water were treated as distinct phases in order to capture swelling.
tipm1248.tex; 29/07/1997; 6:53; v.7; p.36
105
At the mesoscale the swelling particles were modeled within the framework of
hybrid mixture theory. Selection of a proper set of constitutive variables led to the
appearance of an additional hydration stress component accounting for physicochemical effects. This component is defined in terms of the change of the free
energy of the adsorbed fluid with respect to the mesoscopic strain. The relation
between this term and the swelling pressure of Low or averaged disjoining pressure
of Derjaguin was discussed. It was shown that Lows swelling experiment provides
an alternative way of measuring physico-chemical forces (R A) when hydration
forces are the dominant component of the swelling pressure.
Particle deformation at the mesoscale is governed by a viscoelastic equation of
KelvinVoigt type. The viscous term appeared related to the viscous swelling pressure component neq . According to Kralchevsky and Ivanov [47] the microscopic
counterpart of this viscous component is a result of the excess of viscosity of the
adsorbed fluid relative to the bulk phase. One can possibly extend this claim to the
mesoscale and invoke Lows experimental result [54] for the excess in the averaged
viscosity of the adsorbed water in order to obtain more precise information about
the constitutive behavior of neq . This hypothesis remains to be experimentally
validated.
Upscaling from the meso to the macroscale was pursued within the homogenization framework and led to a macroscopic model of dual porosity type in which
the macroscopic deformation and bulk phase flow are influenced at the mesoscale
through macroscopically distributed source/sink momentum transfer functions.
Reduction of the dual porosity model to a single porosity model via a Greens
function shows that the macroscopic total stress tensor has long term memory. This
memory arises due to particle nonequilibrium effects, such as the adsorbed water
flow.
Some notable consequences of the theory developed herein are that it provides
a rational basis for some heuristic modified effective stress principles for smectitic
clays and some creep models for secondary consolidation. In addition, classical
poroelastic models such as the one proposed by Biot [16, 17] can easily be reproduced upon neglecting the adsorbed water effects, i.e. by treating the clay particles
as a single solid phase.
Extensions of the theory to more complex particle constitutive behavior such as
plasticity can be obtained by postulating more general sets of independent variables
in the hybrid mixture theory. This topic will be a subject of future work.
Acknowledgements
This work was supported by the USARO/Environmental Sciences and Engineering
under contract DAAL 03-90-G-0074, the Army Engineering Waterways Experimental Station under contract DACA39-95-K-0056, and the NSF under contract
9510066-BES MM was also supported by CNPq/Brazil through proc. 300810/91-
tipm1248.tex; 29/07/1997; 6:53; v.7; p.37
106
MARCIO
1. The authors wish to thank Lynn S. Bennethum for her comments on earlier
versions of this manuscript.
References
1. Achanta, S. and Cushman, J. H.: 1994, Non-equilibrium swelling and capillary pressure relations
for colloidal systems, J. Collord Interface Sci. 168, 266268.
2. Achanta, S. and Cushman, J. H. and Okos, M.R.: 1994, On multicomponent, multiphase thermomechanics with interfaces, Int. J. Engrg. Sci. 32(11): 17171738.
3. Arbogast, T.: 1992, A simplified dual-porosity model for two-phase flow, in: T. F. Russel, R. E.
Ewing, C. A. Brebbia, W. G. Gray, and G. F. Pindar, (eds), Computational Methods in Water
Resources, Computational Mechanics Publication, Southampton, U.K., pp. 419426.
4. Arbogast, T.: 1993, Gravitational forces in dual-porosity systems, Transport in Porous Media,
13, 179220.
5. Arbogast, T. and Douglas Jr., J. and Hornung, U.: 1991, Modeling of naturally fractured reservoirs
by formal homogenization techniques, in: R. Dautray (ed.), Frontiers in Pure and Applied
Mathematics, Elsevier, Amsterdam, pp. 119
6. Auriault, J. L.: 1990, Behavior of porous saturated deformable media, in F. Darve (ed.), Geomaterials: Constitutive Equations and Modelling, Elsevier, New York, pp. 311328.
7. Auriault, J. L.: 1991, Heterogeneous media: Is an equivalent macroscopic description possible?
Int. J. Eng. Sci. 29, 785795.
8. Auriault, J. L. and Boutin, C.: 1992, Deformable porous media with double porosity. Quasi-Statics
I: Coupling effects, Transport in Porous Media 7, 6382.
9. arbour, S. L. and Fredlund, D. G.: 1989, Mechanisms of osmotic flow and volume changes in
clay soils, Can. Geotech J. 26, 551562.
10. Barden, L.: 1965, Consolidation of clay with nonlinear viscosity, Geotechnique 15, 345361.
11. Bedford, A. and Drumheller, D. S.: 1983, Theories of immiscible and structured mixtures, Int. J.
Eng. Sci. 21(8), 863960.
12. Bennethum, L. S. and Cushman, J. H.: 1996, Multiscale hybrid mixture theory for swelling
systems: Part II: Constitutive theory, Int. J. Eng. Sci. 34(2), 147169.
13. Bennethum, L. S. and Cushman, J. H.: 1996, Multiscale hybrid mixture theory for swelling
systems: Part I: Balance laws, Int. J. Eng. Sci. 34(2), 125145.
14. Bennethum, L. S., Murad, M. M. and Cushman, J. H.: 1996, Macroscale thermodynamics and the
chemical potential for swelling porous media, Center for Computational Mathematics, University
of Colorado at Denver.
15. Bensoussan, A., Lions, J. L. and Papanicolaou G.: 1978, A Symptotic Analysis for Periodic
Structures, North-Holland, Amsterdam.
16. Biot, M.: 1941, General theory of three-dimensional consolidation, J. Appl. Phys. 12, 155164.
17. Biot, M.: 1955, Theory of elasticity and consolidation for a porous anisotropic solid, J. Appl.
Phys. 26, 182185.
18. Bowen, R. M.: 1976, Theory of mixtures, in: A. C. Eringen (ed.), Continuum Physics, 3. Academic
Press, New York.
19. Budkowska, B. and Fu, Q.: 1988, Some aspects of numerical analysis of creep in layered granular
medium, Comput. Geotech. 5, 285306.
20. Allen, H.: 1980, Thermodynamics, Wiley, New York.
21. Coleman, B. D. and Noll, W.: 1963, The thermodynamics of elastic materials with heat conduction
and viscosity, Arch. Rational Mech. Anal. 13, 167178.
22. Cushman, J. H.: 1990, Molecular-scale lubrication, Nature 347, 227228.
23. Derjaguin, B. V. and Churaev, N.V.: 1978, On the question of determining the concept of
disjoining pressure and its role in the equilibrium and flow of thin films, J. Colloid Interface Sci.
66(3), 389398.
24. Derjaguin, B. V. and Churaev, N.V.: 1989, The current state of the theory of long-range surface
forces. Colloids and Surfaces 41, 223237.
25. Derjaguin, B. V., Churaev, N. V. and Muller, V. M.: 1987, Surface Forces, Plenum Press, New
York.
tipm1248.tex; 29/07/1997; 6:53; v.7; p.38
107
26. Douglas, J., Peszynska, M. and Showalter, R. E.: Single phase flow in partially fissured media,
Transport in Porous Media, in press.
27. Douglas, Jr., J. and Arbogast, T.: 1990, Dual porosity models for flow in naturally fractured
reservoirs, in: J. H. Cushman (ed.), Dynamics of Fluid in Hierarchical Porous Media, Academic
Press, New York, pp. 177222.
28. Eringen, A. C.: 1967, Mechanics of Continua. Wiley, New York.
29. Gee, M. L., Mcguiggan, P. M. and Israelachvili, J.: 18951906, 1990, Liquid to solidlike transitions of molecularly thin films under shear, J. Chem. Phys. 93(3).
30. Gibson, R. E. and Lo, K. Y.: 1961, A theory of consolidation for soils exhibiting secondary
compression. Norweg. Geotech. Inst. Publ. 41, 116.
31. Graham, J., Oswell, J. M. and Gray, M. N.: 1992, The effective stress concept in a saturated
sand-clay buffer. Canad. Geotech. J. 29, 10331043.
32. Grim, R. E.: 1968, Clay Mineralogy, McGraw-Hill, New York.
33. Hassanizadeh, S. M. and Gray, W. G.: 1979, General conservation equations for multiphase
systems: 1. Averaging procedure, Adv. Water Resour. 2, 131144.
systems: 2. Mass, momenta, energy, and entropy equations, Adv. Water Resour. 2, 191208.
systems: 3. Constitutive theory for porous media, Adv. Water Resour. 3, 2540.
36. Hassanizadeh, S. M. and Gray, W. G.: 1993, Thermodynamic basis of capillary pressure in porous
media, Water Resour. Res. 29(10), 33893405.
37. Hassanizadeh, S. M. and Gray, W. G.: 1993, Toward an improved description of the physics of
two-phase flow, Adv. Water Resour. 16, 5367.
38. Hornung, U. and Showalter, R. E.: 1990, Diffusion models for fractured medis, J. Math. Anal.
Appl. 147, 6980.
39. Hueckel, T.: 1992, On effective stress concepts and deformation in clays subjected to environmental loads, Canad. Geotech. J. 29, 11201125.
40. Hueckel, T.: 1992, Water mineral interaction in hygromechanics of clays exposed to environmental loads: a mixture theory approach, Canad. Geotech. J. 29, 10711086.
41. Israelachvili, J.: 1991, Intermolecular and Surface Forces, Academic Press, New York.
42. Israelachvili, J., Mcguiggan, P. M. and Homola, A. M.: 1988, Dynamic properties of molecularly
thin liquid films, Science 240, 189191.
43. Ivanov, I. B. and Kralchevsky, P. A.: 1988, Mechanics and thermodynamics of curved thin films,
in: I. B. Ivanov (ed.), Surfactant Science Series 29, Dekker, New York, pp. 49129.
44. Feda, J.: 1992, Creep of Soils and Related Phenomena, Developments in Geotechnical Engineering 68, Elsevier, New York.
45. Karabomi, S., Urai B. Smith, J., Heidug, W. and Oort, E.: 1996, The swelling of clays: Molecular
simulations of the hydration of montmorillonite, Science 271, 11021104.
46. Keedwell, M. J.: 1984, Rheology and Soil Mechanics, Elsevier, New York.
47. Kralchevsky, P. A. and Ivanov, I. B.: 1990, Micromechanical description of curved interfaces:
II Film surface tensions, disjoining pressure and interfacial stress balances, J. Colloid Interface
Sci. 137(1), 235252.
48. Lambe, T. W.: 1960, A mechanistic picture of shear strength in clay, in Proc. ASCE Research
Conference on Shear Strength of Cohesive Soils, Boulder, Colorado, pp. 503532.
49. Levy, T. and Palencia, S.: 1975, On boundary conditions for flow in porous media, Int. J. Engrg.
Sci. 13, 923940.
50. Lewis, R. W. and Tran, D. V.: 1989, Numerical simulation of secondary consolidation of soil:
Finite element formulation, Int. J. Anal. Methods Geomech. 13, 118.
51. Li, D.: 1993, Thermodynamics of thin liquid films, in Thermodynamics and the Design, Analysis,
and Improvement of Energy Systems, Amer. Soc. Mech. Eng., Advanced Energy Systems Division
(Publication) AES, New York.
52. Li, D. and Neumann, A. W.: 1991, Thermodynamics of contact angle phenomena in the presence
of a thin film, Advances in Colloid and Interface Sci. 36, 125151.
53. Liu, I. S.: 1972, Method of Lagrange multipliers for exploitation of the entropy principle, Arch.
Rational. Mech. Anal. 46, 131148.
tipm1248.tex; 29/07/1997; 6:53; v.7; p.39
108
MARCIO
54. Low, P. F.: 1976, Viscosity of interlayer water in montmorillonites, Soil Sci. Soc. Am. J. 40,
500505.
55. 55 Low, P. F.: 1976, Viscosity of interlayer water in montmorillonites, Soil Sci. Soc Am. J. 40,
500505.
56. Low, P. F.: 1980, The swelling of clay: II. Montmorillonite-water systems, Soils Sci. Soc. Am. J.
44(4), 667676.
57. Low, P. F.: 1987, Structural component of the swelling pressure of clays, Langmuir 3, 1825.
58. Low, P. F.: 1994, The clay/water interface and its role in the environment, in: Progress in Colloid
and Polymer Science 95, 98107.
59. Ma, C. and Hueckel, T.: 1992, Effects of inter-phase mass transfer in heated clays: A mixture
theory, Int. J. Eng. Sci. 30(11), 15671582.
60. Morgensten, N. M. and Balasubramonian, B. I.: 1980, Effects of pore fluid on the swelling of
clay-shale, in Proceedings of the 4th International Conference on Expansive Soils, Denver Colo,
pp. 190205.
61. Murad, M. A., Bennethum, L. S. and Cushman, J. H.: 1995, A Multiscale theory of swelling
porous media: I Application to one-dimensional consolidation, Transport in Porous Media 19,
93122.
62. Murad, M. A. and Cushman, J. H.: 1996, Multiscale flow and deformation in hydrophilic swelling
porous media, Int. J. Eng. Sci. 34(3), 313336.
63. Peszenska, M.: 1992, Mathematical analysis and numerical approach to flow through fissured
media, PhD thesis, Univ. Augsburg.
64. Sanchez-Palencia, E.: 1980, Non-homogeneous media and vibration theory, in J. Ehler (ed.),
Lecture Notes in Physics 127. Springer-Verlag, New York.
65. Schoen, M., Diestler, D. J. and Cushman, J. H.: 1987, Fluids in micropores. I. Structure of a
simple classical fluid in a slit- pore, J. Chem. Phys. 87(9), 54645476.
66. Showalter, R. E.: 1991, Diffusion models with microstructure, Transport in Porous Media 6,
567580.
67. Showalter, R. E.: 1992, Distributed Microstructure Models of Porous Media, Int. Series Numer.
Math. Birkhauser, Verlag, Basel, 1993, pp. 155164; J. Douglas, Jr. and U. Hornung (eds), Proc.
Oberwolfach Conference, 1992.
68. Sridharan, A.: 1991, Engineering behaviour of fine grained soils: A fundamental approach, Indian
Geotech. J. 21(1), 1136.
69. Sridharan, A.: 1991, Role of clay minerals in controlling the engineering properties of soils, Clay
Res. 10(2), 3947.
70. Sridharan, A. and Rao, G. V.: 1973, Mechanisms controlling volume change of saturated clays
and the role of the effective stress concept, Geotechnique 23(3), 359382.
71. Sridharan, A. and Rao, G. V.: 1982, Mechanisms controlling the secondary compression of clays,
Geotechnique 32(3), 249260.
72. Terzaghi, K.: 1942, Theoretical Soil Mechanics, Wiley, New York, 1942.
73. Truesdell, C. and Toupin, R.: 1960, The classical field theories, in Flugge (ed.), Handbuchder
Physik, volume III. Springer-Verlag, New York.
74. Wang, Y. C., Murti, V. and Valliappan, S.: 1992, A transient finite element analysis of linear
viscoelastic material model, Int. J. Num. Anal. Methods Geomech. 16, 265294.
75. Wilcox, R. D.: 1990, Surface area approach key to borehole stability, Oil and Gas J. Feb. 26,
6680.
76. Wray, W.: 1995, So Your Home is Built on Expansive Soils, Amer. Soc. Civil Eng., New York.
77. Zeevart, L.: 1986, Consolidation in the intergranular viscosity of highly compressible soils, in
N. Yong and F. C. Townsend (ed.), Consolidation of Soils: Testing and Evaluation. Amer. Soc.
for Testing and Materials ASTM 892, Philadelphia.
tipm1248.tex; 29/07/1997; 6:53; v.7; p.40

A Multiscale Theory of Swelling Porous Media II. Dual Porosity Models For Consolidation of Clays

Enviado por

Dados do documento

Título original

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

A Multiscale Theory of Swelling Porous Media II. Dual Porosity Models For Consolidation of Clays

Enviado por

Direitos autorais:

Formatos disponíveis

Transport in Porous Media 28: 69108, 1997.

c 1997 Kluwer Academic Publishers. Printed in the Netherlands.

A Multiscale Theory of Swelling Porous Media:

S.A. INTERPRINT: (K.B. 6) PIPS Nr.: 135627 MATHKAP

tipm1248.tex; 29/07/1997; 6:53; v.7; p.2

A MULTISCALE THEORY OF SWELLING POROUS MEDIA

tipm1248.tex; 29/07/1997; 6:53; v.7; p.3

Both the heuristic modifications of Terzaghis effective stress principle and

tipm1248.tex; 29/07/1997; 6:53; v.7; p.4

A MULTISCALE THEORY OF SWELLING POROUS MEDIA

tipm1248.tex; 29/07/1997; 6:53; v.7; p.5

Derjaguins disjoining pressure as a microscale counterpart, this new excess in

tipm1248.tex; 29/07/1997; 6:53; v.7; p.6

A MULTISCALE THEORY OF SWELLING POROUS MEDIA

Figure 1. Three-scale model for clay (from Murad et al. [61]).

2.1. NONEQUILIBRIUM RESULTS

where the tensors tes and tls are defined by

tipm1248.tex; 29/07/1997; 6:53; v.7; p.7

and p is the thermodynamic pressure of the -phase. For incompressible media,

tipm1248.tex; 29/07/1997; 6:53; v.7; p.8

A MULTISCALE THEORY OF SWELLING POROUS MEDIA

By adding (2.2) and (2.3) and using (2.5) we obtain

(for a nonswelling medium)

tipm1248.tex; 29/07/1997; 6:53; v.7; p.9

Figure 2. Swelling experiment (from Low [57]).

depicts a classical reverse osmosis swelling pressure experiment performed by Low

where e = l =(1 l ) is the void fraction, e is the void fraction when  = 0

tipm1248.tex; 29/07/1997; 6:53; v.7; p.10

A MULTISCALE THEORY OF SWELLING POROUS MEDIA

assumed incompressible. We then invoke a classical Gibbsian result which states

tipm1248.tex; 29/07/1997; 6:53; v.7; p.11

This mesoscopic definition is consistent with the microscale vectorial definition

0(x; t; n)  B (x; t)n (x; t; n):

tipm1248.tex; 29/07/1997; 6:53; v.7; p.12

A MULTISCALE THEORY OF SWELLING POROUS MEDIA

Applying (2.6) to a mesoscale surface of normal n, and using definitions (2.7),

and combine with (2.17) to get

We may think of t as a swelling stress tensor, since in analogy to the classical

This result is an alternative form of t which expresses the mesoscopic modified

tipm1248.tex; 29/07/1997; 6:53; v.7; p.13

Then the modified effective stress principle at the mesoscale is

As we shall see further in Section 6, this result has an macroscopic analogy.

2.3. SWELLING AND HYDRATION STRESSES AT MODERATE MOISTURE CONTENT

tipm1248.tex; 29/07/1997; 6:53; v.7; p.14

A MULTISCALE THEORY OF SWELLING POROUS MEDIA

Denote the volume fraction of the reference configuration by s

@Al FT = 2 @Al @Js2 C

Since t is a multiple of the identity, using definition (2.24) in (2.19) we have

tipm1248.tex; 29/07/1997; 6:53; v.7; p.15

If we combine the above result with (2.26) we get

tipm1248.tex; 29/07/1997; 6:53; v.7; p.16

A MULTISCALE THEORY OF SWELLING POROUS MEDIA

ways of measuring hydration forces in clays. This procedure is somewhat different

l ddp + 2p + (1 1  ) dd = 0:

which after integration and using p (l ) = (l ) = 0 yields

to develop constitutive relations for hydration forces in swelling clay particles as

tipm1248.tex; 29/07/1997; 6:53; v.7; p.17

the entropy inequality about equilibrium. In particular, when linearizing about

Using (2.33) and (2.36) in (2.27) and (2.28) respectively yields

tipm1248.tex; 29/07/1997; 6:53; v.7; p.18

A MULTISCALE THEORY OF SWELLING POROUS MEDIA

tipm1248.tex; 29/07/1997; 6:53; v.7; p.19

difference between the averaged mesoscopic volumetric viscosity of adsorbed and

where e = l =(1 l ) is the void fraction, e is the void fraction when = 0

0(x; t; n) B (x; t)n (x; t; n):

We may think of t as a swelling stress tensor, since in analogy to the classical

Denote the volume fraction of the reference configuration by s

@Al FT = 2 @Al @Js2 C

Since t is a multiple of the identity, using definition (2.24) in (2.19) we have

l ddp + 2p + (1 1 ) dd = 0:

which after integration and using p (l ) = (l ) = 0 yields

ssAs = 2s (tr Es)2 + s tr E2s;

Introduce the functions f (l ) and g (l ) as

f (l ) = g(l ) + p:

= s div us I + 2srs us:

where following definition (2.11), Gf

Al + 1pf . By the same arguments

= p1f I + 2f rsy v0f ;

s4yy u0s + (s + s)ry divy u0s = 0;

= s(divx u0s + divy u1s )I + 2s(rsx u0s + rsy u1s );

(1 + p1B )I)n d

nf p0f I = (Pff 0 + e 0 )I + tees0 ;

(0 + p0B )I)n = p0f n on ;

nl s4xx s + (s + s + fs)rxdivx

nlp0f ) srx div@ty us

s4yy u1s + (s + s + fs)ry divy u1s ry p0B +

p0l = p0B + B p (0l l ) = p0B + B ps(divy u1s + divx u0s );