Cement and Concrete Research 74 (2015) 3543

Contents lists available at ScienceDirect

Cement and Concrete Research

journal homepage: http://ees.elsevier.com/CEMCON/default.asp

Acid attack on hydrated cement Effect of mineral acids on the

degradation process
T. Gutberlet, H. Hilbig, R.E. Beddoe
Centre for Building Materials (cbm), Technische Universitt Mnchen, Baumbachstrae 7, 81245 Munich, Germany

a r t i c l e i n f o a b s t r a c t

Article history: During acid attack on concrete structural components, a degraded layer develops whose properties as a protective
Received 12 January 2015 barrier are decisive for durability. 29Si NMR spectroscopy and 27Al NMR spectroscopy were used with XRD to inves-
Accepted 16 March 2015 tigate the degraded layer on hardened cement paste exposed to HCl and H2SO4. The layer comprises an amorphous
Available online 22 April 2015
silica gel with framework silicates, geminate and single silanol groups in which Si is substituted by Al. Amorphous
Al(OH)3 and Fe(OH)3 are present. The gel forms by polycondensation and cross-linking of C-A-S-H chains at AlO4
Keywords:
Acid attack
bridging tetrahedra. In the transition zone between the degraded layer and the undamaged material, portlandite
Degraded layer dissolves and Ca is removed from the C-A-S-H phases maintaining their polymer structure at rst. With HCl,
29
Si NMR monosulphate in the transition zone is converted into Friedel's salt and ettringite. With H2SO4, gypsum precipitates
27
Al NMR near the degradation front reducing the thickness of the transition zone and the rate of degradation.
2015 Elsevier Ltd. All rights reserved.

1. Introduction
Concrete structures in industry and agriculture are frequently
exposed to aggressive aqueous media. As well as shortening service
life and increasing expenditure for renewal or maintenance, this is
ecologically disadvantageous with regard to the conservation of resources
and the reduction of CO2 emissions. Acid attack is extremely diverse. For
example, containers for liquid manure, silage, biomass for methane
production, deicing agent run-off at airports can be exposed to a variety
of organic acids. Cooling towers, sewers, structures for geothermal energy
production, concrete pavement and industrial plant can be attacked by
inorganic acids, see state-of-the art report by Alexander et al. [1] and [2,3].
During acid attack, dissolution of the hardened binder paste matrix
and, if soluble, the aggregate results in the formation of a visible degraded
layer, low in mechanical strength and high in porosity, on the concrete
surface. Acid attack is often characterized by a light-brown discolouration
at the degradation front which is probably due to the precipitation of
ferric hydroxide [4]. The rate of degradation is governed primarily by
the diffusion of acid species through the degraded layer to the
undamaged concrete and is therefore accelerated markedly by the
loss of degraded surface material caused by the mechanical action
of, for example, cleaning and mixing processes or agitation by fast
liquid ow [5]. Thus the residual mechanical strength of the degraded
layer and its bonding to the undamaged material are major factors
affecting the acid resistance of concrete [6]. Consequently, testing
Corresponding author. Tel.: +49 89 289 27117; fax: +49 89 289 27069.
E-mail address: robin.beddoe@tum.de (R.E. Beddoe).
methods for acid resistance [7,8] often include systematic brushing of
the concrete surface to simulate abrasion.
Depending on the pH of the acid and its type, zones of different
mineralogical composition form during degradation [4,9,10]. Since the
OH concentration of the pore solution in the undamaged material is
many orders of magnitude higher than the proton concentration in
the reaction zone of the acid [11], the neutralization of the protons
occurs at the degraded layer/undamaged concrete interfacial region.
As degradation proceeds, rst portlandite dissolves transferring calcium
ions to the pore solution. A solution-lled space is produced thus
increasing porosity. At the same time porosity can be reduced by the
precipitation of the corresponding calcium salt of the acid depending
on its solubility, e.g. gypsum (CaSO42H2O) in the case of sulphuric
acid. However, precipitation does not always improve acid resistance
because expansive precipitate with a large molar volume can enhance
erosion, e.g. calcium citrate as observed by Larreur-Cayol et al. [12].
Once portlandite is depleted, the pH of the pore solution falls below
about 12.4 and, along with the dissolution of AFm and AFt, decalcication
of the calcium silicate hydrate phases (C-S-H) proceeds causing a substan-
tial decrease in mechanical strength and a further increase in porosity [1].
As a result, the diffusivity of the acid species is typically an order of
magnitude higher in the degraded layer compared with the undamaged
material [13,14]. The exchange of calcium in C-S-H with acid protons
separates water molecules linking the silicate chains [1517] until
ultimately an amorphous silica gel forms. Depending on the pH of the
acid, the degraded material also contains residual aluminium (pH N 3 to
4) and iron (pH N 1 to 2), probably in the form of hydroxides and,
depending on its solubility, the calcium salt of the acid [4].

http://dx.doi.org/10.1016/j.cemconres.2015.03.011
0008-8846/ 2015 Elsevier Ltd. All rights reserved.
36 T. Gutberlet et al. / Cement and Concrete Research 74 (2015) 3543

In terms of the pH distribution over depth, the degradation After 28 d, the samples were removed from the acid, dried at 40 C,
process may be described by two zones which gradually move into weighed and prepared for the different investigations. Parallel reference
the undamaged material as time progresses, see for example [18,19]. samples, not exposed to acid, were prepared in the same manner and
stored in a closed container lled with inert gas until analysis.
a) A porous degraded layer of low mechanical strength containing
silica gel in which the pH of the pore solution is near that of the 2.2. Measurements
external acid. The degraded layer is often light in colour.
b) A transition zone in which the pH of the pore solution increases All the samples were nely ground in isopropyl alcohol. For the NMR
investigations, zirconia rotors (4 and 7 mm for 29Al, and 27Si NMR,
continuously until levelling off at the value for the undamaged
respectively) were lled with the dry powder. Spectra were recorded
core material. Portlandite dissolution and C-S-H decalcication
with a Bruker Avance 300 spectrometer (magnetic eld strength
occur in the transition zone. The transition zone includes the 7.0455 T) in MAS mode (magic angle spinning) using the single pulse
mechanically sound region in which portlandite dissolves and the technique and the settings listed in Table 1. The 29Si NMR spectra
C-S-H phases are essentially intact, i.e. zone 2 as dened by Bertron were deconvoluted using the Bruker WINNMR software. The interpreta-
et al. [18]. tion of the spectra was carried out according to [23,24] using the Qn
nomenclature. To avoid ambiguities, the Qn nomenclature will be written
with small letters (qn) in the interpretation of the 27Al NMR spectra.
As already pointed out in earlier publications [20,21], the resistance
The samples used for NMR were also used for qualitative XRD
of concrete to acid attack is determined primarily by the properties of
investigations and chemical analysis with ICP-OES. The diffraction
the degraded layer rather than the undamaged concrete. Thus measures
data were collected using an XRD 3003 TT instrument of GE Sensing &
in concrete design which increase the mechanical resistance of the
Inspection Technologies with conguration and Cu K radiation
degraded layer to abrasion and lower its porosity should prolong service
( = 1.54 ). Intensities were recorded over scattering angles between
life. As observed by Oueslati and Duchesne [22], the replacement of
5 and 70 at steps of 0.02, with each step measured for 6 s. A number
Portland cement by supplementary cementitious materials is benecial
of additional samples were mixed with 10 wt.% ZnO as an internal
for acid resistance because this lowers the calcium content of the
standard for quantitative analysis using the Rietveld renement. For
hardened binder paste reducing the porosity of the degraded material.
the chemical analysis of the solid material, samples were fused at
Knowledge of the changes occurring in the hardened binder matrix
1000 C for 15 min with three times their mass of lithium metaborate.
during acid attack is therefore a key to the production of concrete
The loss on ignition was documented and the annealed melt pellets
with improved acid resistance.
digested in 1 M nitric acid. The solutions were then diluted up to
This paper focuses on the changes in chemical and phase composition
100 mL with ultrapure water. The corresponding storage solutions
occurring during acid attack on hardened ordinary Portland cement paste
were additionally acidied with nitric acid to avoid matrix effects in
which lead to the formation of the degraded layer. Attack by mineral acids
the analysis and diluted to 1 l with water. Quantitative chemical analysis
forming highly or scarcely soluble calcium salts is represented by
of the solutions was performed using inductively coupled plasma optical
experiments with hydrochloric and sulphuric acid, respectively. Samples
emission spectroscopy (ICP-OES) in a HORIBA Jobin Yvon Ultima 2
were stored at controlled pH values of 2, 3 and 4 because the solubility
ICP-OES.
of aluminium in C-A-S-H and Al(OH)3 and therefore the composition of
the degraded layer are expected to depend on pH. Results are presented
3. Results and discussion
for investigations with 29Si and 27Al MAS NMR as well as chemical
analysis (ICP-OES) and X-ray diffraction.
3.1. Degradation in the autotitrator

2. Experimental The appearance of the sample disks after 28 d in hydrochloric acid

depended on pH (Fig. 1). At pH 4 they were visually similar to the
2.1. Materials reference samples, whereas at pH 3 brownish areas were evident on
the surface. The samples stored at pH 2 were much lighter in colour
A cement paste was prepared at a water to cement ratio (w/c) of 0.4 and possessed a low mechanical strength. Breaking the samples
using ordinary Portland cement CEM I 42.5 R, (20.62 wt.% SiO2, revealed a layered structure over the thickness of the disks.
61.93 wt.% CaO, 5.23 wt.% Al2O3, 3.49 wt.% Fe2O3, LOI 1.51 wt.%) and After 28 d in sulphuric acid at pH 4 (Fig. 2), the surface of the disks also
deionized water. Immediately after mixing, the paste was poured into exhibited brownish areas of discolouration. At pH 3, a well-dened
cylindrical polyethylene vials 30 mm in diameter which were vibrated white layer (veried as gypsum by XRD) had formed in the surface
and sealed with a cap. After 90 days' storage at 20 C, the hardened region. At pH 2 the gypsum layer was thicker and the brownish region
cement paste cylinders were cut into disks 3.00 mm in thickness using no longer visible on the surface, but was adjacent to the non-degraded
a high-precision saw. core.
The degradation process occurring with hydrochloric and sulphuric The titration data in Fig. 3, left, show that the rate of acid consumption
acids was studied at constant pH 2, 3 and 4 using an automatic titrator is, on the whole, much slower for H2SO4 compared with HCl. After 21 d at
with an automatic sample changer. It was therefore not necessary to pH 2 in HCl, acid addition was no longer necessary, indicating that the
renew the acid during storage in order to maintain acid strength. This maximum degree of degradation had been reached. In contrast, at pH 3
is also in better agreement with the exposure conditions which often and 4 and even at pH 2 for H2SO4 the degradation process had not
prevail in eld practice. For each sample position, three disks were terminated after 28 d.
weighed and placed in a special sample holder in a capped beaker
with 150 mL acid which was stirred with a magnetic spin bar. The pH
Table 1
of the acid solution at each sample position was measured and Settings used for NMR experiments.
readjusted at intervals of 9 min by titrating with 1 M HCl or 0.5 M
Type Resonance Rotation Repetition No. of Shifts
H2SO4. The volume of acid added was recorded continuously. Between
frequency [MHz] speed [kHz] time [s] scans relative to
the measurements, the pH electrode was cleaned automatically by
27
rinsing in distilled water. Later the titration interval was extended to Al 78.21 14 0.5 2000 [Al(H2O)6]3+
29
Si 59.63 5 5.0 15,000 Si(CH3)4
90 min.
 T. Gutberlet et al. / Cement and Concrete Research 74 (2015) 3543 37

pH 4 pH 3 pH 2

Fig. 1. Top: degraded samples after storage in HCl for 28 d at pH 4, 3 and 2 from left to right. Bottom: micrographs of the corresponding cross-sections.

A plot of mass loss versus the total amount of acid protons added for
both types of acid is presented in Fig. 3, right. The solid line in the gure
corresponds to the calculated mass loss according to the neutralization
reaction of the acid protons with hydroxide ions in Ca(OH)2. It is in good
agreement with the experimental data. According to XRD analysis, the
reference samples contain 17 wt.% portlandite which would require
4.6 mmol/g H+ for complete dissolution (dotted lines in Fig. 3, right).
The loss of mass of a degraded sample beyond about 17 wt.% is therefore
mainly due to the decalcication of C-S-H phases. Thus HCl and
H2SO4 attack results primarily in the dissolution of portlandite and
portlandite bound in the C-S-H phases.
In order to increase the degree of degradation at pH 3 and 4, the
storage period of an additional batch of samples in HCl was extended to
91 d. After this period, the disks exposed to pH 4 were similar in
appearance to those stored for 28 d at pH 3. In both cases, the same
amount of acid was titrated (7.3 mL/g). The disks exposed to pH 3
now possessed a well-dened degraded layer after the addition of a
total of 20.1 mL/g. It was now possible to detach the degraded layer
from the inner material for separate analysis of both regions, the latter
containing material from the transition zone.
3.2. Analysis of the reference sample
The chemical composition of the hardened Portland cement paste
reference is listed in Table 2 and the corresponding XRD pattern plotted
in Fig. 4. XRD revealed the presence of residual clinker phases (not
labelled), ettringite, AFm, portlandite and calcite, the last occurring
because it was not possible to avoid carbonation completely.
The 29Si NMR spectrum for the reference material, Fig. 5, left,
contains signals from C3SQ0 and C2SQ0 sites between chemical shifts
of 69 and 73 ppm due to unreacted clinker, C-S-H phases with
incorporated aluminium, i.e. C-A-S-H phases, at 79 ppm (CASHQ1),
82 ppm (CASHQ2(1Al)) and 85 ppm (CASHQ2). The percentages
of silicate tetrahedra in these sites obtained from the deconvolution of
the spectrum are presented in Table 3. Using the signal intensities and
the well-known equations in [25] for aluminium in the bridging
tetrahedra of C-S-H dreierketten, the mean chain length of C-A-S-H
in the reference material is 3.6 and the Al/Si ratio 0.06. For comparison,
the overall Al/Si ratio based on the chemical analysis of the cement is
0.31. Based on the 29Si NMR spectrum, the degree of hydration of the
silicate phases was estimated at 84%.

pH 4 pH 3 pH 2

Fig. 2. Top: degraded samples after storage in H2SO4 for 28 d at pH 4, 3 and 2 from left to right. Bottom: micrographs of the corresponding cross-sections.
38 T. Gutberlet et al. / Cement and Concrete Research 74 (2015) 3543

Fig. 3. Left: protons added by 1 M HCl or 0.5 M H2SO4 during 28 d storage to maintain the given pH. The values are with respect to initial dry sample weight. Right: comparison with mass
loss after 28 d. The mass of precipitated gypsum (5, 18 and 48 wt.% gypsum for pH 2, 3 and 4, respectively, estimated by XRD) is subtracted in the case of H2SO4. a) Data point for prolonged
storage to 91 d in HCl at pH 3.

The reference 27Al NMR spectrum in Fig. 5, right, consists of signals
from ettringite and AFm [26] and a broad peak near 70 ppm which is
attributed to q2(2Si) sites in agreement with the presence of Q2(1Al)
sites for C-A-S-H bridging tetrahedra in the 29Si NMR spectrum. The
chemical shifts attributed by Mueller et al. [27] to the structural units
q1(1Si), q2(2Si) and q3(3Si) in aliuminosilicate solutions are 74, 69
and 64 ppm, respectively. A chemical shift of 55 ppm for q4(4Si) was
published by Davidovits [28] for geopolymeric systems.
3.3. Degradation in hydrochloric acid
The chemical compositions of the degraded disks with respect to
their dry weight before exposure to HCl were calculated from the
chemical analyses of the dry samples after acid exposure using the
appropriate dry weights before and after storage, Table 2. Mass loss is
mainly due to the release of calcium into the storage solution and,
only at the lowest pH, due to the signicant release of aluminium and
iron. In this case, the residual solid consists essentially of silicates with
small amounts of iron and aluminium. The alkali content of the degraded
material was, in all cases, almost zero.
The phase changes occurring during 28 d degradation are illustrated
qualitatively by the X-ray patterns in Fig. 4. Although a well-dened
surface degraded layer was not apparent (Fig. 1), compositional changes
have nonetheless occurred throughout the thickness of the samples
stored at pH 3 and 4. Portlandite is no longer present indicating that
the pH of the pore solution is below about 12.4. Spraying the surface
of the fractured disks with a phenolphthalein solution showed that
the pH was still above 9, i.e. the main part of these samples consisted
of transition zone material. Due to chloride penetration, most of the
monosulphate converted into Friedel's salt and ettringite as summa-
rized by the following equation.
3C3 ACaSO4  12H2 O 2 CaCl2 16H2 O C3 ACaSO4 3
 32H2 O 2 C3 ACaCl2  10H2 O: 1
The additional calcium needed for the reaction is supplied by the
dissolution of portlandite and, potentially, from the decalcication of
C-A-S-H. The formation of Friedel's salt in the transition zone during
the storage of hardened Portland cement paste in HCl was also observed
by Israel et al. [10].
The absence of sharp diffraction lines and the broad peak for the
samples stored at pH 2 are characteristic of the amorphous silicate
phases (silica gel) known to form at high degrees of degradation.
The XRD patterns measured with the inner material of the separated
sample stored for 91 d in HCl at pH 3 revealed a high degree of degrada-
tion with poorly crystalline phases which could not be clearly identied
owing to low scattering intensity and a markedly amorphous signal. The
degraded layer material was completely amorphous as expected.
3.3.1. Degradation products after 28 d at pH 3 and 4
Fig. 5, left, shows the effect of pH on the 29Si NMR spectra recorded for
the samples. At pH 3 and 4, signicant microstructural differences com-
pared with the reference are apparent, although, as already mentioned,

Table 2
Chemical composition of the samples after 28 d or 91 d in HCl at different pH with respect
to dry weight before storage.

Ref. pH pH pH pH 3/91 d pH 3/91 d

4/28 d 3/28 d 2/28 d inner degraded
material layer

Composition
[mmol/100 g]
Ca 850 515 397 4.1 265 17
Si 243 250 227 256 215 207
Al 76 76 73 2.5 68 29
Fe 34 35 34 17 31 25
Mg 31 32 27 0.6 49 6.2
S 31 31 29 0.4 27 2.7
Mass loss[wt.%] 19.9 29.1 76.7 71.6
Added acid 570 730 1937 2010
Fig. 4. XRD patterns for samples before and after storage in HCl for 28 d. Peaks labelled with
[mmol H+/100 g]
E = ettringite, M = monosulphate, F = Friedel's salt, P = portlandite, C = calcite.
 T. Gutberlet et al. / Cement and Concrete Research 74 (2015) 3543 39

Fig. 5. Left: 29Si NMR spectra of the samples after 28 d in HCl with approximate ranges for Qn sites. Right corresponding 27Al NMR spectra with chemical shifts qn(nSi).

a distinct degraded layer was not visible. A reduction in the amount of
residual clinker is indicated by a decrease in Q0 signal intensity. Since
the Q2, not the Q1 signal, is now predominant, the length of the silicate
chains increased from 3.6 (reference) to 5.2 (pH 4) or 5.6 tetrahedra
(pH 3) during degradation, i.e. polymerisation or, more precisely,
polycondensation of C-A-S-H chains occurred. In a study on the
structure of C-S-H in hydrated tricalcium silicate by Chen et al. [16],
29
Si NMR measurements showed that the decalcication of hydrated
C3S paste results in an increase in the mean chain length of C-S-H.
Decalcication explains the polycondensation process of C-A-S-H
chains observed in the present work. In Table 3, the Al/Si ratio increases
slightly from 0.06 (reference) to 0.08 for pH 3 and 4 while the Al/Si
determined by chemical analysis remains virtually the same as the
reference value. Thus, at this stage of the degradation process, aluminium
remains in the bridging tetrahedra of C-A-S-H, where it contributes to
cross-linking between C-A-S-H chains.
The signal intensities at 13 and 9 ppm in the 27Al NMR spectra for
degradation at pH 3 and 4, Fig. 5, right, reveal an increase in ettringite
at the expense of AFm in agreement with the changes in the XRD
patterns. The chemical shift of the peak originally near 70 ppm for
aluminate bridging tetrahedra in C-A-S-H moves towards about
60 ppm and is now in the ranges of q3(3Si) and q4(4Si) sites [27,28].
Hence the AlO4 tetrahedra are connected to more SiO 4 tetrahedra
indicating a certain amount of cross-linking of the C-A-S-H chains
producing Q3(1Al) and, to a lesser extent, Q4(1Al) sites not yet prominent
in the 29Si NMR spectra.
Thus during degradation at pH 3 or 4, the acid ions penetrated the
whole sample causing changes in both the crystalline and C-A-S-H
phases transforming the hardened cement paste into the transition
zone material described above.
3.3.2. Degradation products after 91 d at pH 3 and 4
Fig. 6 compares the NMR results obtained for the pH 3 disks after
28 d (mainly transition zone material) with the spectra for the samples
separated into degraded layer and inner material after 91 d. Although
not visually damaged, the 29Si NMR spectrum for inner material
changed signicantly as expected for material in the transition
zone. The Q1 peak is much smaller and the chemical shift of Q2 has
moved by about 2 ppm. Based on the deconvolution, the mean
chain length calculated from the Q1, Q2 and Q2(1Al) intensities has
increased to 21.7. The broad shoulder extending to about 115 ppm
on the right of the Q2 peak is most likely produced by cross-linking
between the C-A-S-H chains. The 29Si NMR spectrum of the sample
stored for 91 d at pH 4 (not shown here) is similar to the pH 3 spectrum,
but with less intensity in the shoulder region. This sample also consists
of transition zone material. In this case, the calculated mean chain
length of the C-A-S-H phases is 11.2.

Table 3
Percentages of aluminosilicate 29Si NMR sites in hardened Portland cement paste (w/c = 0.4) after storage in HCl.

Site Position Reference pH 4 28 d pH 3 28 d pH 2 28 d pH 3 91 da

[ppm] [%] [%] [%] [%] [%]

C2SQ0 71 5.6 3.3 3.3

C3SQ0 69 to 73 10.8 1.7 3.4
CASHQ1 79 49.8 39.1 34.6 1.0
CASHQ2(1Al) 82 10.3 14.7 14.0 0.23
CASHQ2 85 23.6 41.5 41.8
Q2(1Al) 86 0.8 3.8
Q2 92 5.8 6.9
Q3(1Al) 95 1.0 10.1
Q3 101 34.2 27.9
Q4(1Al) 104 1.9 16.0
Q4 110 - 56.3 34.2
NMR analysis:
Al/Si 0.06 0.08 0.08 0.009c 0.079c
Mean chain length 3.6 5.2 5.6
NBO [mol O / mol Si + Al] 0.48 0.56b
Chemical analysis:
Al/Si 0.31 0.31 0.32 0.010 0.141
Ca/Si 3.49 2.06 1.75 0.016 0.084
Counterions [charge/Si + Al atom] 0.24d 0.52d
a
Degraded layer separated from surface of samples.
b
Calculated without C-A-S-H sites.
c
Calculated for silica gel (Q2(1Al) to Q4).
d
Including Fe3+ in the network counterions.
40 T. Gutberlet et al. / Cement and Concrete Research 74 (2015) 3543
Fig. 6. Left: 29Si NMR spectra of the samples (separated into inner material and degraded layer) after 91 d in HCl at pH 3 and (not separated) after 28 d. Right: corresponding 27Al NMR spectra.
According to the 27Al NMR spectra, the inner material of the pH 3
samples contains remnants of AFm and ettringite (Fig. 6, right). The
peak for aluminosilicates in the 27Al NMR spectrum has shifted from
about 60 ppm (28 d) to 55 ppm (91 d) indicating a predominance of
q4(4Si) sites.
Similar to the 29Si NMR results for the samples stored at pH 2 (see
detailed discussion below), the silicates in the degraded layer detached
from the pH 3 samples are mainly at Q3 and Q4 sites (Fig. 6, left).
However, in the present case, the Q3 signal is more intense than the
Q4 signal. This and the broad shoulder to the left of the Q3 peak shows
that the degradation process at pH 3 is less advanced than at pH 2.
The 27Al NMR spectrum of the degraded layer detached from the
pH 3 samples exhibits two distinct peaks. The chemical shift and
shape of the peak at 5 ppm indicate the presence of poorly crystalline
aluminium hydroxides, see [29]. The other peak is due to aluminium
mainly at q4(4Si) sites. Based on the intensities of these peaks, about
75% of the residual Al is incorporated in the silica gel network.
Since the Al/Si ratio of the degraded layer at pH 3 calculated from the
chemical analysis is relatively high (0.141, Table 3) signicant contribu-
tions from Qn(1Al) to the 29Si NMR signals are likely. Because about 75%
of the available aluminium is incorporated in the structure of the alumi-
nosilicate gel phases, their Al/Si ratio is therefore 0.11.
3.3.3. Degradation products after 28 d at pH 2
The 29Si NMR spectrum for the disks stored at pH 2 (Fig. 5, left)
reveals, like the XRD pattern (Fig. 4), extensive changes in microstructure.
The signals from the original C-A-S-H chains are no longer present and the
main signals now occur at chemical shifts of 100 and 110 ppm
corresponding to silicate tetrahedra in Q3 and Q4 environments. The
shape of the 29Si NMR spectrum is very similar to a spectrum published
by Lutz et al. [30] who synthesised aluminium-free silica gels at pH 2.
Differences in the signal intensities are linked to the aluminium content
of the present samples. Chen et al. [16] also observed similar signals
from silica gel produced by the decalcication of hydrated C3S paste to a
Ca/Si ratio of 0.17. The chemical analysis of the present material yields a
Fig. 7. 29Si NMR spectra including the deconvolution results of the samples stored in HCl. Left: at pH 2 for 28 d. Right: at pH 3 for 91 d (degraded layer separated from surface). See Table 3
for the assignment of the deconvolution peaks.
much lower Ca/Si ratio of 0.016. Thus after 28 d at pH 2, the samples
completely degraded to an amorphous silica gel.
In the 27Al NMR spectrum for pH 2 (Fig. 5, right), the lines for
ettringite and AFm are absent and aluminium in q 4(4Si) sites is
predominant. The presence of aluminium in octahedral coordination
would result in a sharp peak at 0 ppm which is not present in Fig. 5,
or in the case of poorly crystalline aluminium hydroxides, a broad
peak near 5 ppm [29,31]. Since such a peak is not clearly identiable
in the spectrum, a maximum content of 5% Al(OH)3 is assumed.
Compared with pH 3 and 4, storage for 28 d at pH 2 results in
extensive disintegration of the C-A-S-H polymer system to produce
a disordered silicate structure similar to silica gel, but incorporating
a small amount of residual aluminium (Al/Si = 0.01) substituting for
Si in the network structure and/or as poorly crystalline aluminium
hydroxide. The formation of the gel structure is controlled by Al
(as described above) and Ca ions whose removal links hydroxide
groups (forming H2O) increasing the interconnectivity of the silicate
system. The gel is probably combined with loosely bound ferric
hydrates producing the characteristic brownish colour.
The results for the extended storage period show that the general
degradation process of hardened Portland cement paste in HCl at pH 3
or 4 is similar to the faster reaction at pH 2. After 91 d at pH 3, a well-
dened degraded layer formed, but with more aluminium incorporated
in the silicagel structure and in Al(OH)3 than at pH 2.
3.3.4. Microstructural characterisation of the degradation products
Owing to the considerable overlap of NMR signals from Qn and
n+1
Q (1Al), see for example [23] and the broad peaks of the degradation
products, it is difcult to distinguish between contributions from Q2 and
Q3(1Al) or Q3 and Q4(1Al) sites in the deconvolution without additional
27
information. According to the Al NMR spectrum for the degradation
material produced at pH 2, at least 95% of the total aluminium is in the
silicate phases (in q4(4Si) coordination) which, based on the chemical
analysis, results in an Al/Si ratio of 0.009. This means that the 29Si NMR
signals from Qn(1Al) sites account for only approximately 3.7% of the
total signal intensity from the silicate gel sites. Signals from Qn(2Al)
 T. Gutberlet et al. / Cement and Concrete Research 74 (2015) 3543 41

sites are not considered because the likelihood of a silicate tetrahedron

having two neighbouring aluminate tetrahedra is about 104. Keeping
the amount of Qn(1Al) sites (3.7%) and the positions of the peaks (see
Table 3) xed during the deconvolution of the 29Si NMR spectrum yields
the percentages of the different silicate units shown in Fig. 7, left, and
Table 3. Apart from the presence of aluminium, the percentages of the
silicate units and their chemical shifts are in good agreement with
those published by Lutz et al. [30] for silica gel prepared at pH 2, i.e. 5,
37 and 50% SiO2 in Q2, Q3 and Q4 sites, respectively.
In view of the high proportion of Q3 sites in the silica gel and its low
calcium content, Ca/Si = 0.016 (Table 3) these are attributed to
primarily single silanol groups to Si(OSi)3OH, see for example [31].
The Q2 groups are probably geminate silanol groups Si(OSi)2(OH)2
where a small degree of proton substitution with calcium seems likely.
This will be considered more closely in the following.
The microstructure of the silica gel formed (which determines its
properties) can also be characterized by the number of non-bridging
oxygen atoms (NBO), a value commonly used in glass chemistry [17]
as a measure of glass durability. It describes the degree of networking
identifying the number of oxygen atoms in a silicate (aluminate) Fig. 8. XRD patterns for samples stored in H2SO4 for 28 d compared to the reference
tetrahedron not bridging to other tetrahedra. It can be used to calculate before storage. Peaks labelled with E = ettringite, M = monosulphate, G = gypsum,
P = portlandite, C = calcite.
the charge balance between the silicate network and its counterions.
In the case of the degraded material produced at pH 2, NBO was
determined to be n = 0.48 per silicon and aluminium atom for q4(4Si)
coordination and using the listed percentages of 29Si NMR sites. Based
on the chemical analysis of the degraded material and assuming 95% 3.4. Degradation in sulphuric acid
of the total aluminium in the silicate structure, the total charge zC of
the counterions per silicon and aluminium atom is 0.24 / (Si + Al). The chemical compositions of the degraded samples with respect to
Since each aluminium tetrahedron adds one negative charge to the net- their dry weight before exposure to H2SO4 are presented in Table 4.
work, the additional negative charge per silicon and aluminium atom is Whereas in the case of HCl the chemical composition reects the
given by molar fraction of aluminium XAl. Finally, the excess charge is dissolving effect of the acid attack, the data in the table are affected by
the precipitation of gypsum.
zE n zC X Al 2 The X-ray patterns in Fig. 8 show the presence of portlandite,
ettringite and monosulphate in all the degraded samples, even in
yielding zE = 0.25 / (Si + Al) which is compensated by protons. Since those stored at pH 2 where no distinct broad peak due to amorphous sil-
the framework silica groups (Q4: Si(OSi)4) are saturated, the Q2 and Q3 ica phases is apparent. Gypsum precipitation is more pronounced at
tetraheda contain on average roughly 0.61 / (Si + Al) hydroxide groups. lower pH values.
In this calculation, iron was counted as an integrated network modier Compared with the reference sample, the 29Si NMR spectra of the
in the silica gel. Alternately, if the iron is considered to be precipitated in samples after sulphuric acid storage (Fig. 9, left) exhibit changes in the
the matrix as Fe(OH)3 and therefore not taken into account in the Q2, Q2(1Al) and Q1 signals equivalent to just a slight increase in mean
calculation, the total charge of the counterions would be 0.04 / chain length of the C-A-S-H phases (from 3.6 to at most 4.2) while the
(Si + Al) and thus the fraction of hydroxide groups in the silica gel Al/Si ratio remains virtually unchanged irrespective of the storage
1.1 / (Si + Al). pH. The intensities of all the signals from the CS and C-A-S-H phases
In the case of the degraded layer material produced by extended decrease progressively with pH to the benet of the Q3 and Q4 signal
storage at pH 3, deconvolution of the spectrum (Fig. 7, right) yields intensities, i.e. with the degree of degradation (Table 5). The position
the data in Table 3 from which the NBO of the silica gel was calculated and shape of the Q3 and Q4 signals are very similar to those observed
to be n = 0.56 / (Si + Al). Using Eq. (2) and including iron in the coun- with HCl pH2.
terions, the excess charge is zE = 0.14 / (Si + Al) and therefore 0.30 The 27Al NMR spectra (Fig. 9, right) behave in a similar manner. The
hydroxide groups are on average bonded to the Q2 and Q3 tetrahedra. ratio of ettringite to AFm changes from the initial reference value to a
If iron is considered to exist as Fe(OH)3, the value is 1.02. higher value on exposure to the acid which remains about the same
for all the pH investigated. At the same time, the peak at 70 ppm of
the reference sample increases in intensity and moves with decreasing
Table 4
pH towards the lower chemical shift expected for aluminium in
Chemical composition of the samples after 28 d storage in H2SO4 at different pH with
respect to dry weight before storage. framework silica environments, i.e. q4(4Si).
Because portlandite was detected by XRD in the samples, regions
Reference pH 4 pH 3 pH 2
are present where the composition of the C-A-S-H phases has not
Composition mmol/100 g mmol/100 g mmol/100 g mmol/100 g signicantly changed. This means that the 29Si NMR peaks of the C-A-
Ca 850 654 533 563 S-H phases represent a mixture of unchanged C-A-S-H phases (as in
Si 243 223 207 188 the reference) and partially degraded C-A-S-H phases (in the transition
Al 76 71 50 36 zone) with a lower Q1/Q2 ratio. Using the 29Si NMR spectra, the
Fe 34 31 26 23
thickness of the transition zone was estimated to be about 0.2 and
Mg 31 28 19 14
S 31 71 118 263 0.8 mm after 28 d at pH 2 and 4, respectively.
Mass loss wt.% wt.% wt.% The results indicate that the changes in the C-A-S-H phases leading
10.5 19.2 4.9 up to silica gel formation during sulphuric acid attack are basically the
Added acid mmol H+/100 g mmol H+/100 g mmol H+/100 g same as for hydrochloric acid, the rate of degradation being generally
370 670 1100
slower with H2SO4. For a given thickness of the degraded layer, the
42 T. Gutberlet et al. / Cement and Concrete Research 74 (2015) 3543

Fig. 9. Left: 29Si NMR spectra of the samples after 28 d in H2SO4. Right: corresponding 27Al NMR spectra.

degree of degradation is higher with H2SO4 than with HCl; q4(4Si) rath- with a predominant tetrahedral structure. This layer also contains,
er than q3(3Si) sites predominate. depending on the pH of the acid, poorly crystalline aluminium
The degradation process with sulphuric acid differs signicantly hydroxide and amorphous ferric hydroxide.
from that of hydrochloric acid owing to gypsum precipitation which b) A transition zone between the degraded layer and the undamaged
a) removes acid anions from the pore solution near the degradation material in which portlandite dissolves and calcium is removed
front and b) on account of its volume reduces the diffusion ux of from the C-A-S-H phases while maintaining their polymer structure
protons to greater depths. This reduces the thickness of the transition at rst.
zone and the rate of degradation. After 28 d at pH 2, the degraded
layer and the transition zone were estimated to be 0.5 and 0.2 mm
in thickness, respectively. At pH 4 the thicknesses were b 0.1 and During acid attack, the decalcication of the C-A-S-H phases in the
0.8 mm, respectively. For comparison, with HCl at pH 2, the degraded transition zone leads to polycondensation processes which increase
layer and the transition zone were both over 1.5 mm in thickness because the average C-A-S-H chain length from about 3.6 to values as high as
the sample was completely degraded. At pH 4 the thicknesses were b0.1 22 tetrahedra. At the same time cross-linking between AlO4 and SiO4
and N 1.5 mm, respectively. tetrahedra in adjacent C-A-S-H chains begins to form a network
structure. At this stage of the degradation process, aluminium remains
4. Conclusions incorporated in the C-A-S-H phases so the Al/Si ratio (about 0.06)
does not differ appreciably from that of the undamaged material.
Hydrated Portland cement slices (w/c = 0.4) were exposed to As the degree of degradation increases, the C-A-S-H phases transform
hydrochloric and sulphuric acids at different pH in order to into a silica gel network with low NBO values. The basic silica gel structure
investigate the changes occurring in the hardened cement paste comprises framework silicates Si(OSi) 4 and single silanol groups
matrix of concrete during attack by mineral acids with soluble and Si(OSi) 3OH as well a small amount of geminate silanol groups
scarcely soluble calcium salts. Automatic titration combined with a Si(OSi) 2(OH) 2 . The degree of substitution of silicon by residual
multi-sampling system was used to keep the pH constant at 2, 3 or aluminium depends on the pH of the acid. The amount of non-bonded
4 and enable systematic investigation of a larger number of sample/ oxygen atoms and aluminium ions in the network as well as residual
acid combinations. Changes in phase composition were determined calcium counterions require that charge is balanced on average by
using 29Si and 27Al MAS NMR spectroscopy in combination with X-ray about one hydroxide group per SiO4 or AlO4 tetrahedron.
diffraction and chemical analysis with ICP-OES. Exposure to HCl pH 2 produced a degraded layer containing an
The exposure of hardened cement paste to mineral acids results in amorphous silica gel with Si substituted by 0.9% Al and containing
changes in microstructure extending into the material well beyond 1.6% Ca (with respect to total molar Al and Si content), as well as traces
the light-brown surface layer in the degraded surface layer characteristic of Mg, S, Na and K. At pH 3 the gel contained 9.6% Al, 7.6% Ca, 2.7% Mg
of acid attack. Acid attack may be described by two zones which move and 1.2% S. A total of 25% of the residual aluminium content of the de-
into the undamaged core material as degradation proceeds. graded material was in amorphous Al(OH)3. Extended storage at pH 3
produced a degraded layer with a similar silica gel, but with a higher
a) A degraded surface layer (optically visible) of high porosity and low
Al content and NBO value indicating a lower degree of degradation.
in mechanical strength containing amorphous aluminosilica gel
The thicknesses of the degraded layer and transition zone are affected
signicantly by the acid anion. During exposure to hydrochloric acid,
Table 5 chloride ions diffuse far beyond the degraded layer increasing the
Percentages of aluminosilicate sites in hardened Portland cement paste (w/c = 0.4) after thickness of the transition zone where monosulphate is converted into
28 d in H2SO4. Friedel's salt and ettringite. Additional calcium is supplied by portlandite
Site Reference pH 4 pH 3 pH 2
dissolution and the decalcication of C-A-S-H. The changes in the micro-
structure of the decalcied C-A-S-H phases after portlandite dissolution
[%] [%] [%] [%]
which lead up to the nal silica gel are distributed over the transition
CSQ0 16.3 13.3 12.0 7.8 zone.
CASHQ1 49.8 43.4 33.0 24.4
In contrast, exposure to sulphuric acid results in the precipitation of
CASHQ2(1Al) 10.3 8.3 6.4 5.0
CASHQ2 23.6 35.3 18.7 16.2 gypsum at the front of the degraded layer which a) reduces the diffusion
Degr. product (Q3 + Q4) 0 0 29.4 46.6 ux of sulphate ions into the transition zone thus reducing its thickness
NMR analysis: and b) reduces the diffusion of acid protons to the transition zone and
Al/Si (C-A-S-H) 0.06 0.05 0.06 0.05 therefore the rate of degradation. According to the NMR results, the
Mean chain length (C-A-S-H) 3.6 4.2 3.7 3.9
microstructure of decalcied C-A-S-H phases in the transition zone
 T. Gutberlet et al. / Cement and Concrete Research 74 (2015) 3543
varies, on average, little as degradation proceeds. The main conversion
of the decalcied C-A-S-H to the aluminosilica gel appears to occur
within a narrow region located in the transition zone near the front of
the degraded layer.
The results show that aluminium incorporated in the C-A-S-H
phases affects the formation of silica gel in the degraded layer. The
microstructure and integrity of the degraded layer are important for
the acid resistance of concrete because they determine the ability of
the layer to remain as a protective barrier on the concrete surface during
attack. It is conceivable that the replacement of Portland cement by
supplementary cementitious materials rich in aluminium affects the
microstructure of the degraded layer favourable.
Acknowledgements
The authors thank the German Research Foundation (DFG, project
number BE 1228/9) for nancially supporting this research work.
References
[1] M.G. Alexander, A. Bertron, N. De Belie, Performance of Cement-based Materials in
Aggressive Aqueous Environments, State-of-the-Art Report, RILEM TC 211-PAE,
Springer, 2013.
[2] V. Zivica, A. Bajza, Acidic attack of cement based materials a review. Part 1. Principle of
acidic attack, Constr. Build. Mater. 15 (2001) 331340.
[3] A. Bertron, J. Duchesne, G. Escadeillas, Degradation of cement pastes by organic
acids, Mater. Struct. 40 (2007) 341354.
[4] V. Pavlk, Corrosion of hardened cement paste by acetic and nitric acids part II:
formation and chemical composition of the corrosion products layer, Cem. Concr.
Res. 24 (1994) 14951508.
[5] R.E. Beddoe, H.W. Dorner, Modelling acid attack on concrete: Part I. The essential
mechanisms, Cem. Concr. Res. 35 (2005) 23332339.
[6] C. Shi, J.A. Stegemann, Acid corrosion resistance of different cementing materials,
Cem. Concr. Res. 30 (2000) 803808.
[7] L. Lohaus, L. Petersen, R. Griese, Beton mit hohem Surewiderstand fr den
Khlturmbau eine Zwischenbilanz, Beton 9 (2009) 370379.
[8] E. Gruyaert, P. Van den Heede, M. Maes, N. De Belie, Investigation of the inuence of
blast-furnace slag on the resistance of concrete against organic acid or sulphate attack
by means of accelerated degradation tests, Cem. Concr. Res. 42 (2012) 173185.
[9] V. Pavlk, Corrosion of hardened cement paste by acetic and nitric acids part I:
calculation of corrosion depth, Cem. Concr. Res. 24 (1994) 551562.
[10] D. Israel, D.E. Macphee, E.E. Lachowski, Acid attack on pore-reduced cements, J.
Mater. Sci. 32 (1997) 41094116.
[11] V. Pavlik, S. Uncik, The rate of corrosion of hardened cement pastes and mortars
with additive of silica fume in acids, Cem. Concr. Res. 27 (1997) 17311745.
43
[12] S. Larreur-Cayol, A. Bertron, G. Escadeillas, Degradation of cement-based materials
by various organic acids in agro-industrial waste-waters, Cem. Concr. Res. 41
(2011) 882892.
[13] D.P. Bentz, E.J. Garboczi, Modelling the leaching of calcium hydroxide from cement
paste: effects on pore space percolation and diffusivity, Mater. Struct. 25 (1992)
523533.
[14] R.E. Beddoe, H.W. Dorner, Prognose von Betonschden bei Sureangriff, in: 15.
ibausil, Weimar, Germany, 2003. 08730882.
[15] K. Haga, M. Shibata, M. Hironaga, S. Tanaka, S. Nagasaki, Silicate anion structural
change in calcium silicate hydrate gel on dissolution of hydrated cement, J. Nucl.
Sci. Technol. 39 (2002) 540547.
[16] J.J. Chen, J.J. Thomas, H.F.W. Taylor, H.M. Jennings, Solubility and structure of calcium
silicate hydrate, Cem. Concr. Res. 34 (2004) 14991519.
[17] H. Scholze, Glass: Nature, Structure, and Properties, Springer-Verlag, 1991.
[18] A. Bertron, J. Duchesne, G. Escadeillas, Accelerated tests of hardened cement pastes
alteration by organic acids: analysis of the pH effect, Cem. Concr. Res. 35 (2005)
155166.
[19] R.E. Beddoe, H. Hilbig, Modelling the evolution of damage to concrete by acid attack, in:
L. Franke, G. Deckelmann, R. Espinosa-Marzal (Eds.), Simulation of Time Dependent
Degradation of Porous Materials, Final Report on Priority Programm SPP1122 Funded
by the DFG, Cuvillier Verlag Gttingen, Hamburg, 2008, pp. 275292.
[20] R.E. Beddoe, K. Schmidt, Acid attack on concrete effect of concrete composition:
part 1, Cem. Int. 7 (2009) 8894.
[21] R.E. Beddoe, K. Schmidt, Acid attack on concrete effect of concrete composition:
part 2, Cem. Int. 7 (2009) 8693.
[22] O. Oueslati, J. Duchesne, The effect of SCMs and curing time on resistance of mortars
subjected to organic acids, Cem. Concr. Res. 42 (2012) 205214.
[23] G. Engelhardt, D. Michel, High-Resolution Solid-State NMR of Silicates and Zeolites,
John Wiley & Sons Ltd., Chichester, 1987.
[24] I. Schachinger, H. Hilbig, T. Stengel, Effect of curing temperature at an early age on
the long-term strength development of UHPC, Second International Symposium
on Ultra High Performance Concrete, Kassel, Germany, 2008, pp. 205212.
[25] I.G. Richardson, The nature of C-S-H in hardened cements, Cem. Concr. Res. 29
(1999) 11311147.
[26] M.D. Andersen, H.J. Jakobsen, J. Skibsted, A new aluminium-hydrate species in
hydrated Portland cements characterized by 27Al and 29Si MAS NMR spectroscopy,
Cem. Concr. Res. 36 (2006) 317.
[27] D. Mueller, D. Hoebbel, W. Gessner, 27Al NMR studies of aluminosilicate solutions.
Inuences of the second coordination sphere on the shielding of aluminium,
Chem. Phys. Lett. 84 (1981) 2529.
[28] J. Davidovits, Chemistry of geopolymeric systems, terminology, '99 Geopolymer
International Conference, Saint-Quentin, 1999, pp. 939.
[29] T. Isobe, T. Watanabe, J.B. d'Espinose de la Caillerie, A.P. Legrand, D. Massiot, Solid-state
1
H and 27Al NMR studies of amorphous aluminum hydroxides, J. Colloid Interface Sci.
261 (2003) 320324.
[30] W. Lutz, D. Heidemann, C. Hbert, W. Wieker, Contribution of silica gels to
superimposed 29Si MAS NMR spectra of Y zeolites dealuminated by steaming, Z.
Anorg. Allg. Chem. 627 (2001) 25592564.
[31] W.E.E. Stone, G.M.S. El Shafei, J. Sanz, S.A. Selim, Association of soluble aluminum
ionic species with a silica-gel surface: a solid-state NMR study, J. Phys. Chem. 97
(1993) 1012710132.