Cement and Concrete Research 59 (2014) 118–138

Contents lists available at ScienceDirect

Cement and Concrete Research

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

Ion-specific effects influencing the dissolution of tricalcium silicate

L. Nicoleau a,⁎, E. Schreiner b,1, A. Nonat c
a
BASF Research Construction Materials & Systems, BASF Construction Chemicals GmbH, 83308 Trostberg, Germany
b
BASF Materials & Systems, BASF SE, 67056 Ludwigshafen, Germany
c
Institut Carnot de Bourgogne, UMR6303 CNRS, 9 avenue Alain Savary, BP 47870, 21078 Dijon Cedex, France

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

Article history: It has been recently demonstrated that the dissolution kinetics of tricalcium silicate (C3S) is driven by the devi-
Received 22 October 2013 ation from its solubility equilibrium. In this article, special attention is paid to ions relevant in cement chemistry
Accepted 26 February 2014 likely to interact with C3S. In order to determine whether specific effects occur at the interface C3S–water,
Available online 25 March 2014
particular efforts have been made to model ion activities using Pitzer's model. It has been found that monovalent
cations and monovalent anions interact very little with the surface of C3S. On the other side, divalent anions like
Keywords:
Dissolution
sulfate slow down the dissolution more strongly by modifying the surface charging of C3S. Third, aluminate ions
Tricalcium silicate covalently bind to surface silicate monomers and inhibit the dissolution in mildly alkaline conditions. The forma-
Ion specific tion and the breaking of these bonds depend on pH and on [Ca2+]. Thermodynamic calculations performed using
Pitzer equations DFT combined with the COSMO-RS solvation method support the experimental findings.
Alumino-silicate formation © 2014 Elsevier Ltd. All rights reserved.

1. Introduction ion–ion interactions in solution or due to some interactions at the inter-

The dissolution of tricalcium and dicalcium silicates has been recent-
ly rationalized with the help of a thermodynamic framework which
provided a plausible explanation for the relative low solubility of these
phases during their hydration in typical cement conditions [1]. The
authors focused on the role of the deviation from the solubility equilib-
rium, i.e., the activities of calcium, hydroxide and silicate ions, on the
dissolution rate of alite (tricalcium silicate or C3S containing some
impurities). In the mentioned article, the ion activities have been calcu-
lated within the Debye–Hückel theory which is reliable for moderate
ionic strength. A typical cement pore solution is however much more
complex and other ions can influence the dissolution of these anhy-
drous phases. Namely, the solution is rich in alkali and sulfate ions and
contains a non-negligible amount of aluminum ions, known for their
strong interactions with siliceous surfaces. Also, a large proportion of
monovalent anions like nitrate or chloride ions are introduced if some
chemical admixtures have been used in order to improve the setting
or the hardening properties of the cementitious materials. All these
ions constitute a set of factors likely to influence the dissolution rate of
tricalcium silicate (C3S) in cement conditions. If they are established,
specific interactions can come from two origins: they might be due to
⁎ Corresponding author. Tel.: +49 8621862734; fax: +49 862166502734.
E-mail addresses: eduard.schreiner@basf.com (E. Schreiner),
andre.nonat@u-bourgogne.fr (A. Nonat).
1
Tel.: +49 6216043021.
face C3S–water. The first cause will require a better estimation of the
activity coefficients than with the Debye–Hückel framework in order
to calculate the deviation from solubility equilibrium. The second
cause will lead to stabilization or destabilization of the dissolving sur-
face by ions bound to this one according to two types of bonding:
ionic or covalent.
The species in this work are covered in four different sections since
categorized in four classes: the alkali ions (Na+, K+, Cs+), the monova-
lent anions (Cl−, NO− 3 , etc.), the (divalent) sulfate anions and finally the
aluminate anions. Experimentally, we proceeded as previously [1], i.e.,
the dissolution experiments were performed in diluted-enough condi-
tions undersaturated with respect to most of the hydrates likely to
precipitate if higher concentrations would have been used. In these con-
ditions, a pure dissolution regime is observed. The activities of the ions
in solution are estimated according to the Debye–Hückel's and Pitzer's
models. The last one describes the ion–ion interactions in solution and
the comparison between both models allows an appreciation of the
importance of the ion-specificity due to these interactions. Ion-specific
effects have not been in focus of studies concerned with the dissolution
of cement phases although they are important to a wealth of phenome-
na in biology, chemistry and physics [2–5]. These selective interactions
are often ranked into the so-called Hofmeister series which qualitatively
describes ion-specificity observed in many experimental fields.
Due to their ability to covalently bind to silicate species, one should
distinguish aluminum ions from other ions. Aluminum is dissolved from
anhydrous aluminous solid phases but the aluminum concentration
remains low in the pore solution because of the low solubility of alumi-
nate hydrates. Its interactions with the cementitious silicate phases

http://dx.doi.org/10.1016/j.cemconres.2014.02.006
0008-8846/© 2014 Elsevier Ltd. All rights reserved.
 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138 119

Fig. 1. Grain size distributions of both tricalcium silicates used in this study.

have been long the subject of various studies [6–9]. More generally, it is
well recognized that aluminum ions have a perturbing effect on the sil-
icate reactivity [10–14]. Nevertheless, the authors are not aware of any
publication on the influence of aluminate ions. In spite of their identifi-
cation in cement, there is often a lack of thermodynamical data
concerning the hydrated alumino-silicate species as well as a thorough
description of their structure. In undersaturated conditions [11], it was
demonstrated that aluminate ions strongly interact with quartz surfaces
and slow down quartz dissolution. A model is proposed which is based
on a mechanism involving the co-adsorption of aluminum ions. The
identification of hexa-coordinated aluminum species and maybe also
penta-coordinated species in similar systems [15,16] indicates that
chemical reactions possibly driven by surface effects can occur. The
conclusions of Bickmore et al. [11] were drawn from experiments
concerning quartz (a tectosilicate) and therefore only usable with ut-
most care for inosilicates as C3S. Swaddle [17] reviewed the formation
of silicate complexes of aluminum. High concentration of these species
may be found in alkaline media with a significant lifetime. Nevertheless,
due to their large variety, their instability and the lability of aluminum, it
is difficult to quantity them. Facing a lack of experimental and thermo-
dynamic data regarding the reactivity of silicate monomers with alumi-
num ions, we also conducted a modeling work using density functional
theory (DFT) in conjunction with the COSMO-RS solvation model to
theoretically describe the formation of detected alumino-silicate species
and to argue for their thermodynamical stability in conditions encoun-
tered in this study.
2. Technical section
2.1. Materials
Two types of tricalcium silicates were used in this study. The first
type, with which most of the experiments were carried out, is the
same monoclinic polymorph as used in the previous study [1] (denoted
C3S-m). Since this polymorph contains a small quantity of aluminum
which could be cumbersome for quantitative NMR and XPS measure-
ments of the aluminum amount which has been bound, a pure tricalcium
silicate devoid of aluminum was also used. This sample will be noted C3S–

Fig. 2. Evolution of the concentration of Si and Ca during the dissolution of 0.1 g/L C3S-m in 5.5 mM of Ca(OH)2 solution at 20 °C. The bold arrow represents the straight-line used for the
determination of the dissolution rate.
120 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138
Al free in the following. In Fig. 1, both particle size distributions are shown
(obtained by light scattering in ethanol with a MasterSizer 2000 from
Malvern Instruments). The mean particle diameters D averaged over
the surface are 4.67 and 4.33 μm for C3S-m and C3S–Al free, respectively.
Assuming a density ρ = 3.21 g/cm3 and using the relation S = 6 / (ρ · D),
we calculate a specific surface area S of 0.40 and 0.43 m2/g for C3S-m and
C3S–Al free respectively. The different salts used in this study are provided
by Aldrich and are of at least 98% purity grade. The water used for the dis-
solution experiments and the preparation of the different solutions is
twice distilled, resulting in a conductivity of σ = 0.05 μS/cm.
2.2. Experimental techniques
2.2.1. Ion concentration analysis by ICP spectrometry
In analogy to previously published procedure [1], i.e. under conditions
avoiding the C–S–H precipitation, all experiments were performed with
the same in-line experimental set-up as described previously in detail,
namely at 20 °C with 0.075 g of solid dissolved in 750 mL, hence a solid
fraction of 0.1 g/L. The continuous monitoring of ion concentrations is
started just before introducing the alite powder in the reactor. The solu-
tion takes approximately 50 s to reach the nebulizer and it enables to
get 3–4 measurements per minute. Therefore, the 2 or 3 first measure-
ments relate to the composition of the solution before the addition of
solid. It makes possible checking the composition of the solution prior
to C3S addition.
Si, Ca and Al concentrations are always recorded and the concentra-
tions of other ions only when present. The suspensions sometimes
contain some calcium hydroxide which hinders a precise determination
of the little quantity of calcium dissolved from tricalcium silicate
over time. The calculated Si concentration is therefore the most accu-
rate output data, hence the dissolution rate is determined from the
curve [Si] = f(t). If feasible, matrix matching is used for the calibra-
tion of Si and Al. Matrix matching is the preparation of standards
with the same solution used for the dissolution experiments. Matrix
matching is important for the minimization of the side-effects from
perturbing elements on the analyzed elements as well as when the
salt concentration is very high: the viscosity of the solution can lead to
significant deviation of the solution flow rate and to different droplet
sizes in the nebulized aerosol which in turn could result in intensity
variations.
Fig. 3. Evolution of the dissolution rate of alite in 11 mM of Ca(OH)2 solution versus the concentration of alkali salts. (*) The dissolution rate is theoretically calculated with the ion con-
centrations corresponding to the NaCl series using the fit obtained in [1] within the Debye–Hückel theory. At high concentrations of NaCl, the experimental dissolution rates do not match
the calculated ones with DH.
As mentioned above, the dissolution rate is obtained from the slope
of the curve [Si] = f(t). The rate is approximated by the tangent to the
curve represented by the 3–4 first points after the 1st minute, i.e., the
points corresponding to the very early dissolution. An illustrative exam-
ple (Fig. 2) shows the dissolution of C3S-m in 5.5 mM Ca(OH)2 solution.
For the calculation of the interfacial dissolution rate, the specific surface
area is needed. During this short period of time the quantity of dissolved
material is small and we may neglect the variation of the surface area as
a first approximation. The surface area is then considered constant and
equal to the initial one calculated from the particle size distribution. By
convention, the dissolution rates are defined as negative values but due
to numerous logarithm plots in this article the positive scale was used.
2.2.2. 27Al solid-state Nuclear Magnetic Resonance (NMR) spectroscopy
27
Al NMR spectroscopy has been used in this study in order (1) to
know if aluminum is bound to the tricalcium silicate and (2) to have
semi-quantitative information about the coordination state of aluminum.
The NMR experiments were performed on a Bruker Avance Spectrometer
at 78.2 MHz 27Al resonance frequency (300 MHz (H0 = 7.1 T)). The spec-
tra were acquired with a pulse width of 0.58 ms for field strength of
71 kHz. The spinning speed was of about 10 kHz, the relaxation time of
0.5 s, the temperature 298°K and typically 17,408 scans were done.
Chemical shifts are reported with respect to AlCl3 solution as an external
standard. In this article, we use the following nomenclature AlIV, AlV and
AlVI to respectively indicate the 4-fold, 5-fold and 6-fold coordinated alu-
minum species. We aimed to avoid any misunderstanding of the NMR
spectra (also true for the XPS spectra) due to the presence of aluminum
in the anhydrous phase, dedicated dissolution experiments were run for
the NMR and XPS analyses. Those ones were performed on the C3S–Al
free sample originally almost devoid of aluminum. The set-up for the dis-
solution experiments is the same. After 10 min of dissolution, the solution
is filtered through a 0.8 micron filter and then washed with 100 mL eth-
anol (99.9% purity grade). The powder is collected and dried at 80 °C
overnight. Limited by the size of the dissolution reactor (750 mL), we re-
peated the full experiment until collecting about 1.5 g of dried powder.
2.2.3. X-ray photoelectron spectroscopy (XPS)
The same samples as for NMR are used for XPS. The XPS analyses
were carried out with a Phi Versa Probe 5000 spectrometer using mono-
chromatic Al Kα radiation (49 W) under ultrahigh vacuum conditions.
XPS can detect all elements except hydrogen and helium, probes the
 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138 121

Fig. 4. Evolution of the dissolution rate of alite in 11 mM Ca(OH)2 solution containing various concentrations of alkali chloride salts in function of the ion activity product relative to the C3S
dissolution. (*) The theoretical curve is calculated with the experimental fit obtained in [1] only considering the ionic strength effect on ln Π. The empty symbols correspond to ion activity
products calculated with the Debye–Hückel theory and the full symbols correspond to ion activity products calculated with the Pitzer's model.

surface of the sample to a depth of ~ 10 nm, and has detection limits 2.3. Thermodynamical calculations at high salinity
ranging from 0.1 to 0.5 at.% depending on the element. The instrument
work function was calibrated to give a binding energy of 84.00 eV for The driving force for the C3S dissolution, and consequently for the rate,
the Au 4f7/2 line of metallic gold and the spectrometer dispersion was is the deviation from the solubility equilibrium (ΔGC3S − ΔG°C3S). The hy-
adjusted to give a binding energy of 932.62 eV for the Cu 2p3/2 line of droxylated surface of C3S, denoted as (Ca3SiO5)SH, dissolved according to
metallic copper. The built-in Phi charge neutralizer system was used the following equation:
on all specimens. To minimize the effects of differential charging, all
samples were mounted insulated against ground. Survey scan analyses 2þ −
ðCa3 SiO5 ÞSH ¼ 3Ca þ H4 SiO4 þ 6OH 1
were carried out with an analysis spot of 200 μm diameter, a pass energy
of 117 eV and an energy step size of 0.5 eV. High resolution analyses
were carried out on the same analysis area with a pass energy of 0
23.5 eV and an energy step size of 0.1 eV. Spectra have been charge ΔGC3 S −ΔGC3 S ¼ RT ln ΠC3 S 2
corrected to the main line of the carbon 1s spectrum set to 284.8 eV as
a typical value quoted for the energy of adventitious carbon. C1s- where R is the gas constant, T the temperature, ΔG° the free energy at
spectra were analyzed using CasaXPS software version 2.3.16 using Shir- solubility equilibrium and ΠC3S the ion activity product defined as
ley background subtraction using a sum function of 90%-Gaussian–10%- follows:
Lorentzian as provided from the software. Each measure was averaged
 
on three spots and the measured deviations between the 3 spots 2þ 3 − 6
ΠC3 S ¼ Ca ðH4 SiO4 ÞðOH Þ : 3
represent the reported standard deviations.

Fig. 5. Evolution of the dissolution rate of tricalcium silicate in solution containing 11 mM of calcium hydroxide and different quantities of 1:1 sodium salt. The empty symbols correspond
to ion activity products calculated with the DH theory and the full symbols correspond to ion activity products calculated with the Pitzer's model. The symbols ○, ◊, □, △, + and ⨂
correspond to NO− − − − − −
3 , Cl , I , AcO , SCN and ClO3 ions respectively.
122 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138

Fig. 6. Dissolution rates of C3S-m in different concentrations of calcium hydroxide and sodium sulfate.

The estimation of the deviation from equilibrium depends on the
way to determine the ion activities in solution. In the previous article
[1], we used the extended Debye–Hückel (DH) equation to calculate
the activities of Ca2 +, OH− and H4SiO4. However, the extended DH
equation accounts only for the ionic strength but completely ne-
glects the ion–ion interactions in solution at the origin, for example,
of significant ion pairing in concentrated electrolyte solutions. By
contrast, Pitzer developed a set of equations which enables the cal-
culus of the activity coefficients up to 3–4 M. It is not the purpose
of the present article to elaborate on the set of equations in detail
but the interested reader may find all developments in the following
reference [18]. Due to a lack of experimental data, the necessary
Pitzer's coefficients are not known for all the species which hampers
a broader use of Pitzer's model. Fortunately, for the system consid-
ered here most of the coefficients are already tabulated. The Pitzer's
equations are implemented in the software Phreeqc [19] and the rel-
evant Pitzer's coefficients are given in Table 4. All experimental re-
sults, the dissolution rates as well as the ion concentration and the
corresponding ion activities calculated according to both models
are presented in Table 5.
2.4. Quantum chemical modeling
The considered reactions take place at the interface between
the mineral and water, posing a challenging problem for quantum-
chemical treatment. Short of performing explicit free energy calcula-
tions using computationally demanding methods like metadynamics,
one has to resort to some cluster models. Since the primary interest is
in specific bonds which are formed and broken, minimal models for
the bond in question were chosen as a proxy for the real reaction, e.g.
a dimer formed by a silicate ion and an aluminate ion to model Si\O\Al
bonds. Nonspecific solvent effects are accounted for with the use of the
COSMO-RS solvation model.
In particular, all reaction energies were calculated using a recently
developed workflow [20] which is based on density functional theory
(DFT) combined with the COSMO-RS model [21,22] in order to
account for solvation effects. The workflow shows an accuracy of about
2 kcal/mol at relatively acceptable calculation cost. In a nutshell, one cal-
culates the energetics of the chemical reaction in the gas phase and adds a
solvation treatment of the individual chemical species involved in the re-
action. All structures were optimized using a SV(P) [23] basis set and the

Fig. 7. Variation of the dissolution rate of C3S-m according to the logarithm of the activity product (calculated with DH equation) when the tricalcium silicate is dissolved in solutions
containing different concentrations of calcium hydroxide and sodium sulfate. The bigger dots correspond to experiments containing no sulfates and the smaller the dots, the higher the
concentration of sulfate ions. The arrow joining the dots of the 5.5 mM Ca(OH)2 curve indicates the direction for higher concentrations of sodium sulfate. We clearly see that this
arrow departs from the dashed line (reference curve without any specific effects).
 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138
BP86 [24–26] functional in a COSMO environment mimicking an ideal
conductor [27]. The solvation was treated using the COSMO-RS method,
for which single point calculations were performed at the BP86/def-
TZVP level for the previously obtained structures. For the description of
the actual chemical reaction in the gas phase, energies were calculated
using the more accurate M06-2X functional [28] employing a def2-TZVP
basis set [29]. All the BP86 calculations were performed using the pro-
gram package TURBOMOLE [30], while for M06-2X calculations NWChem
was used [31]. The chemical potentials for all the solvated species were
obtained using the COSMOtherm software [32].
3. Experimental results
3.1. Influence of alkali ions
C3S-m samples were dissolved in 11 mmol/L calcium hydroxide
solution containing different concentrations of NaCl, KCl and CsCl, (10,
100, 500 and 1000 mmol/L). The evolution of the dissolution rate
according to the concentration of alkali chloride is plotted in Fig. 3.
The addition of a soluble salt leads to an acceleration of the dissolution.
Most of this effect can be attributed to the increase of the ionic strength
and consequently to the decrease of the activity coefficients of Ca2+, OH−
and H4SiO4 or in other words an increase of the deviation from equilibri-
um. However, the acceleration at constant salt concentration varies with
the salt: it is always higher for NaCl than KCl and CsCl. The measured
rates can be compared to the theoretical rates calculated with the dissolu-
tion rate law obtained without any salt [1] in which the deviation of equi-
librium is calculated according to the extended DH equation (dotted line
in Fig. 3). First of all, at identical ionic strength, i.e., for the same devia-
tion from equilibrium calculated according to DH, the dissolution is
not accelerated or can even decrease. Second, the dissolution rates in
the presence of potassium and cesium chlorides depart much more
from the theoretical curve indicating some specificity. After using the
Pitzer's equations, it can be seen in Fig. 4 that NaCl and KCl curves rough-
ly match the theoretical curve. It means that most of the ion specificity is
due to ionic interactions in solution for these two alkali ions. Concerning
cesium ions, it still remains a significant departure indicating an interac-
tion affecting the surface of alite.
3.2. Influence of monovalent anions
We varied the type of monovalent counter-anion of sodium salts
solutions (200 mmol/L) added to 11 mM calcium hydroxide. The results
for these experiments are presented in Fig. 5. The interfacial rate of dis-
solution is plotted in function of the ion activity product calculated
according to both DH and Pitzer's equations. Variations obtained with
different concentrations of NaCl are also plotted in order to see the evo-
lution of the dissolution rate on a larger range of ln Π. If the extended
DH equation is used to calculate Log Π, the dissolution rate significantly
varies with respect to the type of anion, alighting ion specific effects.
Within the experimental incertitude, we can identify three groups
according to the departure from the calculated reference curve:
(I) NaCl, Na–acetate, NaNO3; (II) NaClO3; and (III) NaI and NaSCN.
Just looking at this ranking, it seems that the direct Hofmeister
series for anions is respected [5]. Using the Pitzer's equations to
calculate the activities for Cl−, NO− −
3 , I , AcO
−
and ClO− 3
3 , differences
are strongly reduced since all points tend to merge on a master curve.
The specificity is mostly due to ionic interactions in solution rather
than at the interface C3S–water.
3
The calculations are performed only with these anions since the Pitzer's coefficients of
others are not available.
123
3.3. Influence of sulfate ions
During the hydration of alite in typical ordinary Portland cement the
concentration of sulfate is high and sulfates ions have been shown to
specifically interact with C–S–H [33,34]. We therefore studied sulfate
ions as well: Na2SO4 was added in different concentrations (50–100
and 200 mM for the final solution concentration) to solutions of increas-
ing calcium hydroxide concentrations (0, 3, 5.5, 8, 11 and 20 mM for the
final solution concentration). Experiments were also performed in a mix
of calcium and sodium hydroxide (5.5 and 100 mM respectively). The
experimental dissolution rates are plotted in function of the sulfate con-
centration in Fig. 6. Except in the case of pure water and sodium hydrox-
ide, the evolution of the rate in function of the sulfate concentration
shows a maximum. The sulfate concentration at the maximum rate in-
creases with the calcium hydroxide concentration suggesting specific
effects due to interactions between both.
It is well known that the addition of sulfate ions to calcium solution
yields CaSO04 species in solution and it thus leads to an important de-
crease of the activity of Ca2+ for a given calcium concentration. Soluble
sulfate salts therefore increases the deviation from the solubility equi-
librium of C3S and the dissolution should be accordingly accelerated.
Fig. 7 represents the plot of the dissolution rate as function of the devi-
ation from equilibrium. Here Log Π can be reasonably calculated with
DH because ion–ion interactions are strongly dominated by the formation
of CaSO04 well described by a solution equilibrium in Phreeqc's database.
This equilibrium is sufficient to well account for the decrease of the
Ca2+ activity in the presence of sulfate ions. Indeed, the calculation
with Pitzer's model gives very similar results (see Fig. 20). As expected,
the deviation from equilibrium is increasingly higher with the concentra-
tion of sulfates but the dissolution rate does not obey to the same mono-
tonic evolution. A much stronger specific effect than the one observed
with monovalent anions is revealed and it cannot be attributed to ionic
interactions in solution since most of them are taken into account in the
calculus of the activities. Therefore, a strong specificity at the interface
C3S/water exists with sulfate ions.
3.4. Influence of aluminate ions
The low solubility of calcium aluminate hydrates or even gibbsite
imposes limitations with respect to the maximum concentration of Al
which can be used without reaching the supersaturation of these
hydrates. A particular attention had to be paid in this respect: variations
of the aluminum concentration have been carefully examined.
3.4.1. Influence of Al concentration according to pH
The pH of solutions with different quantities of aluminum chloride
(from 0.2 to 2 mM) has been varied with NaOH. As an example, the
dissolution curves in the case of 10 mM of NaOH are presented in
Fig. 8. Looking at graphs A and B, it can be clearly seen that the addition
of aluminum slows down the dissolution as soon as alite is starting to
dissolve. In the solution containing 2 mM of AlCl3, the dissolution is
even almost totally inhibited: a quasi-concentration plateau of silicate
is reached 2.5 min after the beginning of the dissolution experiment.
The concentrations of Si and Ca follow the same evolution and the
aluminum concentration (chart C of Fig. 8) is always constant over
time, i.e., the decrease and even the inhibition of dissolution cannot be
attributed to the precipitation of an alumino-silicate in solution.
The same kind of experiments has been performed with different
concentrations of sodium hydroxide ranging from 10 to 100 mmol/L.
The evolution of the dissolution rate (estimated from the [Si] = f(t)
curves) as function of the Al concentration are reported in Fig. 9. The fig-
ure reveals that the influence of aluminum on dissolution is strongly de-
pendent of the initial pH. The higher the pH, the lower the inhibiting
effect of aluminum. For instance with 1 mM of AlCl3, the dissolution of
alite is almost unchanged in 100 mM NaOH whereas it is divided by
about a factor 2 in 10 mM NaOH. In this section, we did not represent
124 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138

Fig. 8. Evolution of concentrations of silicon (A), calcium (B) and aluminum (C) during the dissolution of 0.1 g/L of tricalcium silicate in 10 mM NaOH solution containing different quan-
tities of AlCl3. In chart (A), arrows represent the initial slope of dissolution curves used for the determination of the initial dissolution rate.

the evolution of the dissolution rates in function of the deviation consumes 4 mol of hydroxide. This acidic effect of Al should even de-
from equilibrium as we did above since the specific effect of alumi- part more the solution from the solubility equilibrium of tricalcium
num is evident. In this pH range, the most stable aluminum species silicate and therefore should have an accelerating effect, not a decel-
in solution is Al(OH)− 4 , i.e., each mole of aluminum chloride erating effect as observed.
 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138
Fig. 9. Evolution of the initial dissolution rates of tricalcium silicate (0.1 g/L) according to the concentration of AlCl3 added to different NaOH solutions.
In addition, the strong dependency of the dissolution inhibition
according to pH is well illustrated by the following experiment. The disso-
lution is started in 100 mM NaOH solution to which 2 mM AlCl3 has been
added, such conditions do not affect a lot the dissolution rate (see Fig. 9).
After 5 min, 8.44 mL of a concentrated HNO3 solution is added in order to
reduce pH to an equivalent of 10 mM NaOH (Fig. 10A & B). The dissolution
rate briefly increases due to the pH decrease and then the dissolution is
rapidly inhibited in a similar way as the dissolution experiment started
in the 10 mM NaOH + 2 mM AlCl3 solution. No significant change is
observed regarding the Al concentration after the addition of HNO3
(Fig. 10B), indicating that the inhibition is not related to the precipitation
of an Al-containing species in solution. Moreover, it is worthwhile to point
out that this inhibition of the dissolution cannot be attributed to any
steady state between the C3S dissolution and the precipitation of any Si-
containing species since a precipitation should lead to a drop of the silicon
concentration instead of the observed plateau. Among all inhibitory spe-
cific effects studied in this paper, aluminate ions surely show the largest
one. In low alkaline conditions, 500 μmol of Al divides the dissolution
rate roughly by a factor 2.
3.4.2. Reversibility of the inhibitory effect
The inhibition is assumed to originate from the interaction between the
surface of tricalcium silicate and the adsorbed/bound aluminum species.
We further studied the stability of such bonds. For this purpose, the dissolu-
tion in 10 mM NaOH + 2 mM AlCl3 is chosen as reference because of the
strong inhibition of the dissolution in this case. Since the interactions be-
tween the surface and Al species are pH-dependent, we varied the pH dur-
ing this experiment, once by increasing it and once by decreasing it.
3.4.2.1. pH increase. A small quantity of a 2 M NaOH solution is added
10 min after the start of the experiment (when the dissolution is strong-
ly inhibited). This quantity is chosen in order to obtain a solution equiv-
alent to 100 mM of NaOH. The result is presented in Fig. 11: the
dissolution rate after the addition is not as high as if the dissolution
would have begun in 100 mM NaOH + 2 mM AlCl3. The corresponding
dissolution rates are shown in Table 1 (alkaline conditions) wherein the
rates are extracted from the curves of Fig. 11. The times, at which the
dissolution rates are estimated, are chosen in such a way to have very
similar deviation from equilibrium (ln Π = −65.4). A similar degree
of dissolution is also reached for the three experiments. Even if the dis-
solution is partly released after the subsequent addition of NaOH, the
dissolution rate remains roughly one order of magnitude lower than
125
in the comparative experiment (in 100 mM of NaOH + 2 mM AlCl3
added at t = 0). This indicates that the strong inhibitory effect of Al
ions is not totally reversible and that the formed Al bonds cannot be
easily broken by an increase of pH.
3.4.2.2. pH decrease. The second test for the reversibility has been per-
formed by the addition of nitric acid in order to decrease the pH down to
4.4. Fig. 12 represents the evolutions of the concentrations of (A) silicon
and (B) aluminum. Conversely to the previous case, the addition of HNO3
briefly unblocks the dissolution: a rapid release of silicon is observed
(Fig. 12A). However, a plateau of [Si] is then rapidly established. The transi-
tion between the rapid silicon release and the second plateau, namely the
time elapsed between the 11th and 14th minutes, is accompanied by a
drop of aluminum (Fig. 12B). This behavior seems to indicate that the inhib-
itory effect of Al is neutralized by the addition of acid and followed by the
precipitation of an alumino-silicate leading to a constant concentration of Si.
The dissolution rates, calculated at a similar degree of dissolution and
consequently after different dissolution times, are reported in Table 1
(acidic conditions). The corresponding ion activity products are also cal-
culated and varied mainly due to pH. The ln Π of the solution 10 mM +
2 mM AlCl3 after the addition of nitric acid lies between the ln Πs of the
dissolution experiment started in pure water and started in 10 mM
HNO3. Accordingly, the same tendency is observed for the dissolution
rates, i.e., the addition of nitric acid has completely annihilated the inhib-
itory effect of Al. This indicates that the Al bonds formed in the 10 mM
NaOH solution can be broken in acidic conditions. One can hypothesize
that the occurrence of a second concentration plateau of Si at 15 min is
due to the re-formation of bonds since pH increased to 5.5.
In order to identify and quantify the aluminum bound to the surface
of C3S, X.P.S. and NMR experiments were performed. Those analyses
were realized on aluminum-free tricalcium silicate samples (C3S–Al
free) as it is after washing with ethanol (1), and also, after partial
dissolution in 10 mM NaOH + 2 mM AlCl3 solution (2) or in 100 mM
NaOH + 2 mM AlCl3 solution (3). For the dissolved samples (2 & 3)
the operation is repeated many times in order to collect enough solid.
In order to be sure of the good reproducibility of the experiment, the
ion concentrations have been monitored until the filtration of the solu-
tion (done at 10 min) and always been compared to reference curves
where dissolutions were pursued until the 35th minute (Figs. 18 and
19). The dissolution curves of C3S-m and C3S–Al free are very similar,
which enables the comparison of NMR (and XPS) results obtained
with C3S–Al free to all other results obtained with C3S-m.
126 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138

Fig. 10. Evolution of concentrations of silicon (A) and aluminum (B) during the dissolution of 0.1 g/L of tricalcium silicate started in a solution containing 100 mM NaOH and 2 mM AlCl3.
After 5 min, the solution is slightly acidified by the addition of concentrated HNO3. The pH is then equivalent to the dissolution experiment started in 10 mM NaOH + 2 mM AlCl3.

3.5. NMR spectra
27
Al NMR spectra of samples (1), (2) and (3) as well as a reference
sample of Gibbsite Al(OH)3 (4) were collected, the four spectra are re-
ported in Fig. 13. The spectrum of Al(OH)3 is typical and in accordance
with literature data [35]. The two resonances at ~ 6 and ~ − 15 ppm
are characteristic of two octahedral aluminum sites present in Gibbsite
coordinated by 6 hydroxyl groups each but differing by their quadrupo-
lar coupling constant. The spectrum related to the pure C3S sample indi-
cates only a very small quantity of aluminum by the presence of a
resonance peak at ~ 80 ppm. This can be possibly attributed to some
metal impurities deposited on C3S grains during milling. Spectra related
to both C3S samples dissolved in the presence of AlCl3 indicate the pres-
ence of a significant amount of aluminum bound to the solid. The reso-
nance lines are broad mainly due to quadrupolar interactions. The
sample dissolved in 10 mM NaOH + 2 mM AlCl3 shows the highest in-
tensities. Even if a real quantitative comparison between 27Al spectra is
hazardous due to the quadrupolar character of Al, we note at least one
order of magnitude between the line intensities. This clearly means
that more aluminum is bound to the C3S sample dissolved in 10 mM
NaOH solution than to the one dissolved in 100 mM NaOH. The amount
of Al revealed by NMR seems to be connected to the inhibition of the
dissolution since this latter was readily reduced in 10 mM NaOH
solution and only slightly affected in 100 mM.
Three centerbands can be distinguished in NMR experiment C3S (2)
‘10 mM NaOH + 2 mM AlCl3’ at roughly 55, 30 and 2 ppm corresponding
to tetra-, penta- and hexa-coordinations respectively. Even if this spec-
trum significantly differs from Gibbsite in the range of chemical shifts
characteristic of six fold-coordinated aluminum species, we cannot totally
exclude the precipitation of a small amount of Al(OH)3 in sample (2). The
employed conditions exclude the precipitation of C–S–H as well as C–A–
S–H which is basically confirmed also by NMR [6,9]. The most abundant
coordination of Al in alumino-silicate species is tetrahedral but 5 and 6 co-
ordinations are also recurrent in anhydrous phases [36,37] as well as in
hydrated phases [38]. When we refer to (calcium) aluminum silicate
chemistry, the simultaneous presence of the 4, 5 and 6 coordination is fre-
quent and it is verified for various materials: in natural crystals as
metakaolin [37], vesuvianite [39], and andalusite [36], in CaO–Al2O3–
SiO2 tecto-aluminosilicate and per-aluminous glasses [40], in
tobermorite-like C–A–S–H [6], in hydrated white Portland cement [41]
 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138
Fig. 11. Evolution of the degree of dissolution of alite over time in different alkaline solutions. The black vertical arrow symbolizes the addition time of NaOH which increases the concen-
tration of the 10 mM NaOH solution to 100 mM.
and in so-called geopolymer systems [42]. To the best of our knowledge, it
was never reported for the reaction of C3S with an aluminate solution
when the precipitation of C–A–S–H is excluded.
The sample (3) shows less aluminum and the coordination has also
changed compared to sample (2). Indeed, the resonance line of AlVI
(at 8.4 ppm) is relatively higher than the resonance line of AlIV
(at ~60 ppm) and the 5-coordinate 27Al species has vanished. For sample
(3), the AlIV line at 60 ppm is shifted to higher chemical shift compared to
sample (2) (55 ppm) which can be interpreted as a lower degree of con-
nectivity to silicate units [9]. The AlVI line position at 8.4 ppm may be at-
tributed to Al species expected to be associated with strongly adsorbed
water [43] or to small amount of Gibbsite precipitated locally on the C3S
surface only [15]. However, the absence of a line or shoulder at
−15 ppm is questioning the precipitation of Gibbsite. It is therefore diffi-
cult to surely assign this position to an aluminum species in our case.
3.6. Spectrometry of X-ray photoelectrons (XPS)
The element concentrations (O, Ca, Si, Al and Na) are reported in
Fig. 14. Significant amount of carbon was also detected, probably due to
the rest of ethanol adsorbed onto the surface and to the superficial
carbonatation of samples. Some traces of chloride and fluorine were also
collected. All of these elements were subtracted from the superficial com-
position and the other elements have been then renormalized to 100%.
The non-dissolved C3S sample contains a very small amount of alu-
minum (1.9%) corroborating the NMR result. The ratio Ca/Si is close to
3, which is the expected ratio of the original composition. In sample
(2) about 9% of aluminum is found within the first atomic layers. The
quantity of silicon is approximately the same as in sample (1) but the
calcium decreased, i.e., the Ca/Si ratio is close to 2. Only 1.7% of Na
is found which is probably due to the charge compensation of the
deprotonated silicates. The deficit of calcium puts forward that alumi-
num has possibly replaced some calcium ions in the coordination
sphere of silicate monomers originally present in alite. It would mean
that some Si\O\Al bonds have been formed. Sample (3) contains
about the same quantity of Al as in sample (2) but the Ca to Si ratio is
again close to 3. The aluminum concentrations found by XPS in samples
(2) and (3) drastically differ from what could be expected from the
semi-quantitative analysis by NMR. This leads to the conclusion that
aluminum in sample (3) is only present at the outermost surface.
Since the Ca/Si ratio is almost kept identical to the composition of alite,
127
it suggests that aluminum ions are adsorbed on the C3S surface instead
of being inserted in the first atomic layers of the alite structure. Adsorbed
aluminum species could explain the resonance line found at ~8 ppm.
4. Discussion
The ion-specific effects studied in this paper are likely to occur
during the cement hydration since all ionic species investigated in this
article are typical components of ordinary cements. NMR results in
10 mM of NaOH strongly suggest the formation of covalent bonds be-
tween Si and Al. Thus, the interactions are possibly of two types, ionic
or covalent; the results are discussed according to both.
4.1. Ionic interactions
4.1.1. The monovalent cations and anions
Most of the specific effects on the dissolution rates in Figs. 3 and 5 are
due to ion–ion interactions in solution since the specificity disappears
when the rates are plotted in function of the ion activity product calcu-
lated with Pitzer's equations. The weak deviation from the master curve
observed in Fig. 4 in the case of cesium could be either due to the fact
that the ion–ion interactions are not perfectly described by the Pitzer's
coefficients, or to a real interaction with the surface. Indeed, in the
case of C–S–H, Vialis-Terrisse et al. found a shift of the isoelectric point
towards higher pCa in the presence of cesium [44]. They demonstrated
that at equivalent concentrations of Ca(OH)2, the addition of CsCl
decreases the overcompensation of the negative surface charge, i.e.,
the ζ-potential decreases to lower positive values. Since the overcom-
pensation is due to the condensation of the divalent calcium cations
onto the negatively charge C–S–H surface because of strong electrostatic
interactions [45]. This result indicates that Cs+ ions competitively
adsorb on C–S–H and, most probably, on C3S too. The consequence of
this adsorption is discussed in the next section.
4.1.2. Divalent anions, the case of sulfate ions
Like cesium ions, it is well established that sulfates ions are specifi-
cally adsorbed on C–S–H [33,46,47]. Potassium sulfate also shifts the
isoelectric point of C–S–H towards higher pCa, indicating a specific
adsorption as well [34]. In calcium hydroxide solutions, sulfate ions
are adsorbed on C–S–H because of the prior adsorption of calcium.
The anion adsorption is therefore dependent of the buildup of the
128 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138

Fig. 12. Evolution of the silicon concentration (A) and the aluminum concentration (B) over time during the dissolution of tricalcium silicate. The black vertical arrow symbolizes the time
of addition of nitric acid used for the pH decrease.

overcompensating calcium layer, itself influenced by the surface charge
density (i.e. the pH) and the calcium concentration in solution. Divalent
co-anions even build a second ionic layer at 5–10 Å from the surface
[48,49].In the same manner as the competition Cs+/Ca2+, the adsorp-
tion of divalent anions also has consequences on the surface charge. It
is demonstrated that for a given pH, the deprotonation of the surface
silanol groups (at the origin of the negative charge) is much easier in
the presence of divalent than monovalent cations because of the charge
screening [50]. It follows that at same pH and same pCa, the surface
charge is more negative in the presence of cesium. The adsorption of sul-
fate has an analogous consequence: by forming neutral CaSO04 species the
charge screening ensured by calcium cations is strongly reduced and the
surface charge becomes more negative [50]. The data obtained on C–S–H
surfaces are extendable to C3S surfaces since we found quite similar elec-
trokinetic properties. Indeed the plots of Fig. 15 for C3S relate the same
evolutions as those published by Nachbaur [34].
The effect of the surface charge on the dissolution rate of natural
silicate minerals is well documented. The minimum rate corresponds
to the point of zero charge and the higher the surface charge, the
faster the dissolution rate [51,52]. Unfortunately, it is not possible
to determine experimentally this point for di- and tri-calcium sili-
cates because of their high dissolution rate in low-alkaline or acidic
range. However, the increase of the solubility of C 2S with pH, i.e.
with increasing surface deprotonation was established [1]. It indi-
cates that C3S and C2S dissolution behaves in a similar way as other
alumino-silicates.

Table 1
Dissolution rates calculated from the curves represented in Fig. 11 and in Fig. 12 after similar degree of dissolution in alkaline conditions and in acidic conditions.

Experiment Time [min] ln Π Degree of dissolution [%] Rate [pmol/m2/s]

Alkaline conditions 100 mM NaOH 1.7 −65.4 10% 17.0

100 mM NaOH + 2 mM AlCl3 2.6 −65.4 11% 11.0
[10 mM NaOH − N100 mM NaOH] + 2 mM AlCl3 18.7 −65.4 7% 1.1
Acidic conditions 10 mM HNO3 0.9 −199.7 26% 157
Pure water 1.3 −82.4 23% 69
[10 mM NaOH − N10 mM HNO3] + 2 mM AlCl3 12.1 −137.4 26% 81
 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138 129

Fig. 13. 27Al NMR spectra obtained with different solutions of C3S and gibbsite. The asterisks, black circles and lozenges (*, ●, ◊) indicate the centerbands for tetra-, penta- and hexa-
coordinated Al species respectively. The other bands, a priori unassigned, are indicated with a dagger (†). The corresponding chemical shifts are annotated. On right-hand side, the values
in parenthesis are the vertical expansion factors employed for the corresponding spectra. The spectra are also arbitrarily shifted in the y-direction for the sake of clarity.

In experiments revealing the strong effect of sulfate as well as the
weak effect of cesium, we can extract the variation of the dissolution
rates ΔRi specifically induced by sulfate or cesium. For the same ionic
activity product Πi, the departures ΔRi between the experimental
dissolution rates RSO4,Cs measured in the presence of sulfate or cesium
ion and the rates R0 related to the law without specific interactions
and symbolized by a dashed line (Figs. 4 & 7) have been calculated as
follows:
0
ΔRi ¼ R ð ln Πi Þ−RSO4 ;Cs ð lnΠi Þ:
ΔRi are plotted versus the activity of calcium in solution (Fig. 16). All
points tend to merge on a master curve, showing a sigmoid shape sim-
ilar to a titration curve and confirming the key role of free calcium ions.
Fig. 16 reveals that for a given ion activity product, the lower the Ca2+
activity, the higher the diminution of the rate. A lower activity of Ca2+
also means a reduced screening which in turn leads to a more protonat-
ed surface comparatively to experiments without sulfate or cesium. In
the presence of sulfate (or cesium) because of the higher degree of pro-
tonation of the surface, the solubility of the C3S surface partially
hydroxylated becomes lower. At higher pH values, the degree of proton-
ation is less sensitive to the calcium activity or in other words the addi-
tion of sulfates will influence the dissolution rate less. It is confirmed by
the absence of a maximum value with the sulfate concentration in the
curve 100 mM NaOH + 5.5 mM Ca(OH)2 (Fig. 6).
4.2. Covalent bonds. The case of aluminate ions
The decrease of the dissolution rate of C3S caused by the addition of
4 aluminum salt is quite significant. The NMR and XPS results suggest cova-
lent bonds between the surface silicates and aluminate ions. Considering
the large decrease of the dissolution rate (especially at mildly alkaline
pH), it would mean that the formed Si\O\Al bonds limit the dissolution,
similarly to what has been observed for other alumino-silicate materials
[10–12,53–55]. Reactions involving such bonds and their cleavage have
been well studied at acidic and neutral pH but less so in alkaline condi-
tions. Yet, Bickmore et al. [11] showed that the dissolution rate of quartz
is more affected by aluminum in low alkaline conditions than in high al-
kaline conditions, similar to the findings in this study. Whereas the role
of aluminum bonding has been demonstrated in the dissolution of

Fig. 14. Element distribution of O, Ca, Si, Al and Na species determined by XPS on the three C3S different samples.
130 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138

Fig. 15. ζ-potential evolution of C3S in various concentrations of sodium sulfate plotted in function of the activity of Ca2+.

relatively low-soluble aluminosilicate materials (most of tectosilicates), it
has never been highlighted in dissolution processes of such soluble mate-
rials like C3S (an inosilicate). DFT calculations offer additional rationale for
the interpretations arising from experiments. Before the discussion of
possible interfacial reactions at the surface of C3S, the main observations
reported in the experimental section are recalled:
(1) The slow-down of C3S dissolution starts at pH = 5.5 which is
about the pH where Al(OH)− 4 is present in solution. The dissolu-
tion is still more inhibited in mildly alkaline solutions. We will
therefore consider that Al(OH)− 4 is the main reactive species for
the pH range encountered here.
(2) The dissolution rate is less inhibited by aluminum when pH
increases. The formation of Si\O\Al bonds should not be favored
by an increase of pH.
(3) NMR spectra have revealed the presence of tetra- and hexa-
coordinated aluminum species at high pH and also the presence
of the penta-coordination at lower basic pH (in 10 mM NaOH).
The presence and therefore the stabilization of the AlV species
can be an indicator of the kinetic path leading to the inhibition
since the dissolution is more inhibited in 10 mM NaOH conditions.
(4) Once Si\O\Al bonds are formed, a subsequent increase of pH
does not fully reverse the inhibitory effect of aluminum.
DFT modeling can shed light on reactions occurring at the interface
between the C3S surface and water in the presence of aluminate ions.
Yet, for the sake of pragmatism, the study on Si\O\Al bond formation
was divided into two sections. The first part is dedicated to the reaction
of a protonated or partly deprotonated silicate cluster with Al(OH)−
4 and
to the stability of the obtained species. The second part is dedicated to
the formation of AlV and AlVI.
4.2.1. Formation and stability of Si\O\Al clusters
The energetics of relevant reactions is reported in Table 2. Chemical
reactions and the corresponding calculated energies are given for the
hydrolysis. It can be seen that the condensation between a
fully protonated silicate and Al(OH) − 4 is possible. It corre-
sponds to the reaction − R 2 a and/or − R 2 b , depending on how
the water molecule is eliminated (ΔG = [− 8; − 5] kJ/mol). The
higher the deprotonation of silicate species, the less favorable the con-
densation. − R3 and − R4 represent the condensation between an

Fig. 16. Evolution of the difference ΔRi between the dissolution rates with and without sulfate/cesium salts for a same deviation from equilibrium in function of the activity of Ca2+ in
solution. Positive ΔRi corresponds to a decrease of the dissolution rate.
 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138 131

Table 2
Enthalpies of reaction involving silicate and aluminate species calculated by DFT. The attack on Si or Al side is indicated below the main reaction.

Reactions ΔG [kJ/mol] ΔH [kJ/mol]

R1 −2.7 −22.0

Attack of a water molecule on Si

R2a 5.5 −28.4

Attack of a water molecule on Si

R2b 8.3 −55.5

Attack of a water molecule on Al

R3a −3.5 11.5

Attack of a water molecule on Si

R3b −18.2 −28.2

Attack of a water molecule on Al

R4a −18.8 32.9

Attack of a water molecule on Si

R4b −20.3 −4.3

Attack of a water molecule on Al

R5 38.4 1.0

R6 −3.0 14.2

aluminate and either 1 fold or 2 fold deprotonated silicate monomers
respectively. Both reactions are energetically unfavorable. Our theoret-
ical observations rationalize the experimental observation of alumino-
silicate complexes in 0.015 mol/L NaOH and not in 0.1 mol/L NaOH
done by Yokohama et al. [56]. It also supports our observation that alu-
minum ion is inhibiting the dissolution less when pH increases since the
degree of deprotonation of silicate groups increases with pH. This can be
assigned to the dominance of the electrostatic repulsion between an
aluminate ion and the deprotonated silicates. For comparison, the con-
densation between two fully protonated silicates (−R1) is rather difficult
(ΔG = 2.7 kJ/mol), indicating that only aluminate ions easily condensate
with silicate. We can also expect that, in always neutral or mildly alkaline
conditions, the newly formed Si\O\Al species can further react
with another protonated silicate giving Si\O\Al\O\Si trimers.
The free enthalpy of such a reaction has to be close to the free en-
thalpy calculated for − R2.
Once Si\O\Al or Si\O\Al\O\Si species are formed, an increase
of pH shows that the inhibition of dissolution is only partially reversed.
In alkaline conditions, two ways are possible for the bond breaking: ei-
ther through the cleavage by a water molecule or by a hydroxide ion. In
the case of the hydrolysis by water, reactions R2, R3 and R4 are relevant.
Similarly to the previous conclusion, the cleavage is favorable only if

Table 3
Enthalpies of reaction involving aluminosilicate dimers calculated by DFT.

Reactions ΔG [kJ/mol] ΔH [kJ/mol]

R7 71.6 −16.4

R8 Instable product

R9 = −R5 −41 26.1

R10 Instable product

R11 57 137.6

R12 37.8 132.5

R13 −9.2 463

132 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138
Fig. 17. Relative evolution of the dissolution rate of C3S in function of the concentration of AlCl3. All the rates are normalized with respect to the experiments without AlCl3 (R0).
silicates are deprotonated (R3 and R4). In alkaline conditions, silicates
are indeed partially deprotonated; it means that Si\O\Al bonds can
be broken by water if pH increases.
The reaction leading to a 1 fold deprotonated silicate and an alumi-
nate is energetically favorable (− R5 + R6) ΔG = − 38 − 3 =
− 41 kJ/mol. Compared to reaction − R3b, one can conclude that the
cleavage by a hydroxide ion is more favorable than that by a water mol-
ecule. Moreover, the addition of one hydroxide to aluminum leads to a
stable 5-fold coordinated aluminum species, revealed by NMR. By con-
trast, no stable AlV species was obtained through the addition of
water. The ΔG of formation of Si\O\AlV is −38 kJ/mol which is signif-
icantly higher than the energy gain obtained by the cleavage by water
(− 18.2 kJ/mol). It indicates that the formation of Si\O\AlV should
be significantly favored compared to the cleavage of Si\O\AlIV by
water. Yet, the newly formed Si\O\AlV species should further react
(R6, ΔG = − 3 kJ) and decompose into one protonated silicate and
one aluminate. Considering this result, Si\O\Al bonds should not be
stable in alkaline conditions, which is basically true since it is well-
known that alumino-silicate materials dissolve faster when pH
increases.
In spite of the apparent contradiction with the experimental obser-
vation of the persistence of these bonds, two reasons can lead to a strong
diminution of the probability that R5 and R6 occur in experiments. First
of all, at high pH, the surface of C3S will be strongly deprotonated lead-
ing to a high surface charge density. This high negative charge will
hinder the approach of hydroxide ion and therefore the formation of
doubly charged AlV species. Assuming that OH− ion overcomes the
potential barrier leading to the formation of two fold negatively charged
Al species, the close presence of divalent Ca2+ cations can compensate
the additional charge created by the addition of OH− and therefore
limit the cleavage reaction. Similarly, North et al. [57] demonstrated
that lifetimes of aluminosilicate solution species, i.e., the stability before
precipitation, depend on the nature of the alkali metal ion. The stabiliza-
tion by calcium ions will be discussed in the following text part.
4.2.2. Stability and formation of highly coordinated aluminum AlV and
AlVI species
Revealed by NMR, the presence of AlV and AlVI species on C3S surface
points out that aluminum ions cannot be only adsorbed or co-adsorbed
but can also form covalent bonds. The aluminum coordination is multi-
ple, either in solids (AlVI) [41] or in aqueous solution species [54]. Larger
aqueous poly-aluminum hydroxide species like Keggin structures can
even be formed [58]. Basically, there are two ways to increase the coor-
dination of AlIV, either by the addition of water molecules and/or by the
addition of hydroxide ions. We investigate both. In above section, we al-
ready hypothesized that the coordination of a divalent calcium ion
could stabilize the charge. The addition of Ca is therefore also investigat-
ed. For the sake of simplicity, we only considered the fully protonated
Si\O\AlIV species as reactant. Results are reported in Table 3.
The addition of one and two water molecules to Si\O\AlIV is
respectively rather costly (R7, ΔG = 72 kJ/mol) and impossible (R8, no
stable species can be formed). Therefore, it seems impossible that 5
and 6-fold coordination of aluminum is due to the addition of water.
The picture is different if hydroxide ions are considered as reactions R9
and R10 demonstrate. Indeed, as concluded above the addition of one
hydroxide ion leads to a stable Si\O\AlV cluster (R9 = −R5) but the
formation of Si\O\AlVI species is still impossible (R10) by this way.
Whereas AlV species is stable, no indication of the stability of AlVI is
found without additional stabilization.
Although the first coordination of calcium is known to be rather ver-
satile, here an octahedral shell was chosen [59,60]. In addition, as the
dominant calcium species is Ca2+ at such pH, we considered only the
coordination by 6 water molecules and not by one or two hydroxide
plus 5 or 4 water molecules respectively. The calculations have therefore
focused on the possible stabilization of Si\O\AlV and Si\O\AlVI species
by the addition of one hexa-coordinated calcium cation in the coordina-
tion sphere of aluminum. R12 and R13 correspond respectively to the
formation of one Si\O\AlV\Ca and one Si\O\AlVI\Ca species. Sur-
prisingly, the presence of Ca2+ in the vicinity cannot stabilize the
Si\O\AlV cluster (ΔG(R12) largely positive) whereas the Si\O\AlVI
cluster becomes stable (ΔG = −9 kJ/mol). The free enthalpy of reaction
R13 is more negative than the free enthalpy of reaction R6, meaning that
the addition of OH− ions to the Si\O\AlIV cluster should lead at the
end to a Si\O\AlVI species instead of being cleaved into an aluminate
and a silicate, provided that the solution contains calcium. Reaction R13
brings strong support to the persistence of Si\O\Al bonds in alkaline
conditions as well as a possible explanation for the existence of AlVI. By
contrast, AlV species are not further stabilized by calcium which in turn
confirms their presence only over a sharp pH range (see NMR patterns
Fig. 13) since AlVI clusters revealed to become even more stable in alkaline
conditions in the presence of calcium. The bound Ca ion partly coordinat-
ed with water molecules can explain the presence of the 27Al NMR signal
at 8 ppm corresponding to a 6-fold coordinated aluminum associated
with strongly bound water molecules.
 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138
Because of the key role of calcium ions during the stabilization of
Si\O\Al bonds and thus in the inhibition of the dissolution, it is inter-
esting to compare the dissolution rate of C3S in the presence of NaOH or
Ca(OH)2 and AlCl3 at same pH (Fig. 17). The retarding effect of alumi-
num on dissolution is less pronounced upon the addition of 100 mM
NaCl to 10 mM NaOH solution. The higher ionic strength favors the
deprotonation and therefore not the addition of Al(OH)− 4 at the surface.
The same should be true for experiments containing 5.5 mM of Ca(OH)2
because of the known higher deprotonation degree in the presence of
divalent ions. Nevertheless, a stronger inhibition of the dissolution is ob-
served. It comforts the hypothesis that calcium ions stabilize directly the
AlVI species and indirectly the AlV species since the R6 is not the most
favored reaction anymore.
5. Conclusion
Most ions studied here show selective interactions and influence the
interfacial dissolution rate of C3S. These differences observed in kinetics
have different origins. We summarized below our results according to
the four classes as defined in the Introduction:
(1) Monovalent anions are the species showing the weakest pertur-
bations and it is mainly due to ion–ion interactions in solution
which change a little the activity of calcium and therefore the
(true) deviation from equilibrium.
(2) To a greater extent than monovalent anions, the monovalent cat-
ions also present specific ion–ion interactions in solution but can
also out-compete with calcium ions at the surface of C3S which
becomes more protonated upon the addition of alkalis since Ca
ions are replaced. Compared to sodium and potassium, cesium
ions have a stronger affinity for the C3S surface and tend to re-
duce the dissolution rate.
(3) The higher valence of sulfate ions augments their capacity to
form ion pairs with the adsorbed calcium ion layer. This phe-
nomenon significantly contributes to increase the protonation
of the surface in mildly alkaline conditions. These effects origi-
nate from electrostatic interactions and are connected to the sur-
face charge of C3S: the lower the charge, the lower the solubility.
(4) Contrary to ions discussed above which are physically adsorbed,
Appendices
Fig. 18. Evolution of the Ca, Si and Al concentrations during the dissolution of C3S–Al free in 10 mM NaOH + 2 mM AlCl3.
133
aluminate ions undergo covalent bonding. This is especially
remarkable at moderate alkaline conditions. A strong inhibition
of the dissolution is due to the condensation of an alumino-
silicate species at the outermost surface of C3S. Despite that
these Si\O\Al bonds are preferentially formed in low alkaline
conditions; they are stabilized at higher pH by calcium ions in
the coordination sphere of aluminum ions.
As far the cement hydration is concerned, potassium and sodium
ions are by large more wide-spread in typical cement compositions
than other alkali ions. Variations of the dissolution rate – for the
same deviation from equilibrium – remain very small. The effect on
the rate is mainly due to the variation of the ionic strength and can
be estimated by the Debye–Hückel approximation, provided that
the salt concentration remains below 0.5 mol/L. Above, which is un-
typical in ordinary cement, a more sophisticated model (like Pitzer's)
is necessary. Regarding the anions, nitrate chloride and acetate ions
are by far the most common monovalent anions in a typical cement
paste especially if hardening accelerators are used. Taking into ac-
count the small differences observed between the three anions, we
can neglect the specific effects. In contrast, sulfate ions show a strong
specific effect originating from an interaction with the surface. It re-
mains difficult to predict the variation of the dissolution rate. Indeed,
no simple analytical model predicts the degree of protonation of the
surface as function of the solution composition (except some simula-
tions [61]) and the exact relation between the degree of protonation
and the solubility is not established for C3S yet. Aluminate ions, even
at concentrations below 1 mmol/L, exhibit an inhibitory effect which
can be strong enough to affect the dissolution during the very first
seconds/minutes of the hydration of C 3S. Indeed, during this short
period of time, the conditions for making Si\O\Al bonds are met
due to the rapid dissolution of aluminate anhydrous phases present
in cement clinker. This point could be a factor influencing the dura-
tion of the so-called dormant period; a phenomenon not fully ex-
plained and still debated [8,62,63].
Acknowledgment
The authors are grateful to S. Hirth and K. Seidel (BASF SE) for the
XPS and NMR measurements.
134 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138

Fig. 19. Evolution of the Ca, Si and Al concentrations during the dissolution of C3S–Al free in 100 mM NaOH + 2 mM AlCl3.

Table 4
Pitzer's coefficients for the ion–ion interactions encountered in this work. β(0), β(1), β(2) and Cϕ are the cation–anion interaction parameters for the Pitzer aqueous model. θ is the cation–
cation and anion–anion interaction parameter. Ψ is the anion–anion–cation and cation–cation–anion interaction parameter. The parameters relative to interactions with involving neutral
species are not treated. The α parameter for the Pitzer aqueous model has the default value of 2 for any electrolyte containing a monovalent ion. For electrolytes containing polyvalent ions,
two parameters α1 and α2 are used. For 2–2 electrolytes the default values are α1 = 1.4 and α2 = 12. In all cases, b parameter is fixed to 1.2. (See the reference [19] for the detailed
description.) The other acido-basic equilibriums in solution are taken into account through the usual Debye–Hückel approach. All necessary equilibrium constants used in the Debye–
Hückel approach are referenced in Table 10 of [1].

Interactions β(0) β(1) β(2) C* Interactions cation–cation & θ Interactions cation–cation–anion & Ψ
cation–anion anion–anion cation–anion–anion

Na+ + Cl− 0.0765 0.2664 0.00127 Na+ + K+ −0.012 Na+ + K+ + Cl− −0.0018
K+ + Cl− 0.04835 0.2122 −0.00084 Ca2+ + Na+ 0.07 Na+ + K+ + SO2− 4 −0.01
Cs+ + Cl− 0.038 0.04 −0.0012 Ca2+ + K+ 0.032 Na+ + Ca2+ + Cl− −0.007
Ca2+ + Cl− 0.3159 1.614 −0.00034 Na+ + H+ 0.036 Na+ + Ca2+ + SO2− 4 −0.055
H+ + Cl− 0.1775 0.2945 0.0008 K+ + H+ 0.005 Na+ + H+ + Cl− −0.004
Na+ + I− 0.1195 0.32 0.0018 Ca2+ + H+ 0.092 + +
Na + H + HSO4 −
−0.0129
Ca2+ + I− 0.437925 1.80675 −0.05 Cl− + NO−
3 0.016 K+ + Ca2+ + Cl− −0.025
Na+ + SO2−4 0.01958 1.113 0.00497 Cl− + SO2−
4 0.02 Cs+ + Ca2+ + Cl− −0.035
Ca2+ + SO2−4 0.2 3.1973 −54.24 −
Cl + HSO− 4 −0.006 K+ + H+ + Cl− −0.011
H+ + SO2−
4 0.0298 0.5556 0.0438 Cl− + OH− −0.05 K+ + H+ + SO2−4 0.197
Na+ + HSO− 4 0.0454 0.398 OH− + SO2−4 −0.013 K+ + H+ + HSO− 4 −0.0265
Ca2+ + HSO− 4 0.2145 1.729 Ca + H + Cl−
2+ +
−0.015
+ − −
Na + OH 0.0864 0.253 0.0044 Cl + SO4 + Na+
2−
0.0014
−
+
K + OH 0.1298 0.32 0.0041 Cl− + SO2−
4 + Ca2+ −0.018
Cs+ + OH− 0.131 0.42 0.003 Cl− + HSO−4 + Na
+
−0.006
Ca2+ + OH− −0.1747 −0.2303 −5.72 Cl− + HSO−4 + H
+
0.013
Na+ + NO− 3 0.0068 0.1783 −0.00072 Cl− + OH− + Na+ −0.006
Ca2+ + NO− 3 0.2108 1.4092 −0.02014 Cl− + OH− + K+ −0.006
Na + ClO−
+
3 0.0249 0.2455 0.0004 Cl− + OH− + Cs+ −0.006
Ca + ClO−
2+
3 0.438 1.76 Cl− + OH− + Ca2+ −0.025
Na+ + CH3COO− 0.1426 0.3237 −0.00629 SO2−
4 + HSO2−4 + Na+ −0.0094
Ca + CH3COO−
2+
0.269 1.134 −0.031 SO2−
4 + HSO2−4 + K+ −0.0677
SO2−
4 + OH− + Na+ −0.009
SO2−
4 + OH− + K+ −0.05
Table 5
Activities and ion activity products related to the C3S dissolution calculated with the Debye–Hückel and the Pitzer's equations.

Experimental conditions Ion concentrations Ion activities, Debye–Huckel Ion activities, Pitzer

Type of Starting Ca Si Na, K or Cs Cl, AcO, NO3, SO4 (mM) (OH) (Ca2+) (H4SiO4) Ln n (OH) (Ca2*) (H4SO4) Ln n Dissolution rate
mineral conditions (mM) (mM) (mM) ClO3, I or (μmol/m2/s)
SCN (mM)

C3S-m In 11 mM Ca(OH)2 10.2 0.00327 0 0 0 1.57E−02 4.54E−03 8.42E−09 −59.7 1.61E−02 4.95E−03 7.69E−09 −59.4 1.67
C3S-m In 11 mM Ca(OH)2 + 10.3 2.27E−03 10 10 0 1.54E−02 4.30E−03 5.71E−09 −60.4 1.60E−02 4.63E−03 5.16E−09 −60.0 2.50
10 mM NaCl
C3S-m In 11 mM Ca(OH)2 + 10.4 3.43E−03 100 100 0 1.37E−02 3.30E−03 9.69E−09 −61.3 1.48E−02 3.41E−03 6.75E−09 −61.2 4.17
100 mM NaCl
C3S-m In 11 mM Ca(OH)2 + 10.2 5.94E−03 500 500 0 1.04E−02 2.45E−03 1.92E−08 −63.2 1.22E−02 2.53E−03 8.47E−09 −62.9 6.73
500 mM NaCl
C3S-m In 11 mM Ca(OH)2 + 10.2 9.40E−03 1000 1000 0 8.68E−03 2.40E−03 4.06E−08 −63.6 1.10E−02 2.61E−03 9.71E−09 −63.4 6.58
1000 mM NaCl

L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138

C3S-m In 11 mM Ca(OH)2 + 10.4 0.00182 10 10 0 1.56E−02 4.33E−03 4.55E−09 −60.5 1.62E−02 4.65E−03 4.08E−09 −60.2 2.24
10 mM KCl
C3S-m In 11 mM Ca(OH)2 + 10.4 0.00373 100 100 0 0.0137 3.30E−03 9.45E−09 −61.4 1.52E−02 3.31E−03 7.18E−09 −61.0 3.47
100 mM KCl
C3S-m In 11 mM Ca(OH)2 + 10.4 0.0054 500 500 0 0.0105 2.49E−03 1.71E−07 −63.0 1.41E−02 2.29E−03 6.78E−09 −62.6 6.41
500 mM KCl
C3S-m In 11 mM Ca(OH)2 + 10.4 0.00849 1000 1000 0 0.0088 2.44E−03 3.59E−08 −63.6 1.40E−02 2.11E−03 7.03E−09 −62.9 5.78
1000 mM KCl
C3S-m In 11 mM Ca(OH)2 + 10.4 3.01E−03 10 10 0 1.56E−02 4.33E−03 7.50E−09 −60.0 1.62E−02 4.62E−03 6.73E−09 −59.7 1.51
10 mM CsCl
C3S-m In 11 mM Ca(OH)2 + 10.4 3.29E−03 100 100 0 1.37E−02 3.30E−03 8.34E−09 −61.5 1.58E−02 3.17E−03 6.24E−09 −61.1 2.59
100 mM CsCl
C3S-m In 11 mM Ca(OH)2 + 10.4 4.77E−03 500 500 0 1.05E−02 2.49E−03 1.51E−08 −63.3 1.59E−02 2.06E−03 5.72E−09 −62.4 4.36
500 mM CsCl
C3S-m In 11 mM Ca(OH)2 + 10.4 7.28E−03 1000 1000 0 8.85E−03 2.44E−03 3.08E−08 −63.7 1.66E−02 1.83E−03 5.66E−09 −62.5 4.93
1000 mM CsCl
C3S-m In 11 mM Ca(OH)2 10.2 0.00327 0 0 0 1.57E−02 4.54E−03 8.42E−09 −59.7 1.61E−02 4.95E−03 7.69E−09 −59.4 1.67
C3S-m In 11 mM Ca(OH)2 + 9.5 0.00474 100 0 50 1.28E−02 1.68E−03 1.27E−08 −63.5 1.33E−02 1.66E−03 1.05E−08 −63.5 7.15
50 mM Na2SO4
C3S-m In 11 mM Ca(OH)2 + 9.7 0.0114 200 0 100 1.19E−02 1.13E−03 3.18E−08 −64.2 1.27E−02 1.18E−03 2.24E−08 −64.0 9.88
100 mM Na2SO4
C3S-m In 11 mM Ca(OH)2 + 11.4 7.78E−03 400 0 200 1.24E−02 1.23E−03 2.05E−08 −64.1 1.39E−02 9.62E−04 1.06E−08 −64.9 3.17
200 mM Na2SO4
C3S-m In 50 mM Na2SO4 0.2124 0.0708 100 0 50 2.55E−04 3.87E−05 1.00E−05 −91.6 2.67E−04 3.60E−05 8.98E−06 −91.7 117.90
C3S-m In 100 mM Na2SO4 0.351 0.117 200 0 100 3.77E−04 4.88E−05 1.14E−05 −88.5 4.05E−04 4.18E−05 9.18E−06 −88.7 129.52
C3S-m In 200 mM Na2SO4 0.2451 0.0817 400 0 200 2.34E−04 2.70E−05 1.21E−05 −93.0 2.62E−04 2.05E−05 8.37E−06 −93.6 154.65
C3S-m In water 0.165 0.055 0 0 0 2.76E−04 1.48E−04 8.95E−06 −87.2 2.76E−04 1.49E−04 9.00E−06 −87.2 69.84
C3S-m In 5.5 mM Ca(OH)2 + 5.53 0.00795 100 0 50 7.58E−03 9.91E−04 3.93E−08 −67.1 7.89E−03 9.59E−04 3.30E−08 −67.1 9.61
50 mM Na2SO4
C3S-m In 5.5 mM Ca(OH)2 + 4.99 0.00334 400 0 200 5.52E−03 5.45E−04 2.17E−08 −71.4 6.20E−03 4.19E−04 1.30E−08 −72.0 2.75
200 mM Na2SO4
C3S-m In 5.5 mM Ca(OH)2 + 4.96 0.00537 200 0 100 6.19E−03 6.81E−04 3.17E−08 −69.7 6.64E−03 5.98E−04 2.35E−08 −69.9 5.71
100 mM Na2SO4
C3S-m In 5.5 mM Ca(OH)2 5.71 0.00789 0 0 0 9.36E−03 3.06E−03 3.68E−08 −62.5 9.55E−03 3.24E−03 3.56E−08 −62.3 6.84
C3S-m In 5.5 mM Ca(OH)2 + 5.13 0.0007 100 0 0 7.77E−02 1.03E−03 1.77E−10 −58.4 8.26E−02 1.14E−03 1.26E−10 −58.1 0.78
100 mM NaOH

(continued on next page)

135
 136
Table 5 (continued)
Experimental conditions Ion concentrations Ion activities, Debye–Huckel Ion activities, Pitzer

Type of Starting Ca Si Na, K or Cs Cl, AcO, NO3, SO4 (mM) (OH) (Ca2+) (H4SiO4) Ln n (OH) (Ca2*) (H4SO4) Ln n Dissolution rate
mineral conditions (mM) (mM) (mM) ClO3, I or (μmol/m2/s)
SCN (mM)

C3S-m In 5.5 mM Ca(OH)2 + 5.57 0.00232 200 0 50 7.03E−02 7.45E−04 6.02E−10 −58.8 7.66E−02 7.09E−04 3.46E−10 −59.0 1.70
100 mM NaOH +
50 mM Na2SO4
C3S-m In 5.5 mM Ca(OH)2 + 5.59 0.00529 300 0 100 6.54E−02 6.29E−04 1.47E−09 −58.8 7.30E−02 5.41E−04 6.84E−10 −59.4 2.46
100 mM NaOH +
100 mM Na2SO4
C3S-m In 5.5 mM Ca(OH)2 + 5.38 0.00671 500 0 200 5.89E−02 5.14E−04 2.24E−09 −59.6 6.84E−02 3.82E−04 6.94E−10 −60.8 3.07
100 mM NaOH +
200 mM Na2SO4
C3S-m In 3 mM Ca(OH)2 3.24 1.53E−02 0 0 0 5.58E−03 2.01E−03 1.29E−07 −65.6 5.66E−03 2.08E−03 1.26E−07 −65.5 12.30

L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138

C3S-m In 3 mM Ca(OH)2 + 2.97 1.52E−02 100 0 50 4.11E−03 5.37E−04 1.44E−07 −71.3 4.29E−03 5.10E−04 1.25E−07 −71.3 22.28
50 mM Na2SO4
C3S-m In 3 mM Ca(OH)2 + 3.07 1.81E−02 200 0 100 3.84E−03 4.24E−04 1.77E−07 −72.2 4.13E−03 3.69E−04 1.35E−07 −72.5 18.72
100 mM Na2SO4
C3S-m In 3 mM Ca(OH)2 + 3.27 1.59E−02 400 0 200 3.62E−03 3.58E−04 1.64E−07 −73.2 4.07E−03 2.74E−04 1.01E−07 −73.7 17.70
200 mM Na2SO4
C3S-m In 100 mM NaOH 0.0459 1.53E−02 100 0 0 7.28E−02 9.63E−06 4.37E−09 −69.6 7.77E−02 1.05E−05 3.24E−09 −69.3 13.89
C3S-m In 100 mM NaOH + 0.0591 1.97E−02 200 0 50 6.45E−02 8.05E−06 5.86E−09 −70.6 7.07E−02 7.48E−06 3.50E−09 −70.8 17.22
50 mM Na2SO4
C3S-m In 100 mM NaOH + 0.0624 2.08E−02 300 0 100 5.97E−02 7.12E−06 6.67E−09 −71.3 6.70E−02 6.00E−06 3.21E−09 −71.8 18.46
100 mM Na2SO4
C3S-m In 100 mM NaOH + 0.2484 8.28E−02 500 0 200 5.39E−02 2.40E−05 3.15E−08 −66.7 6.29E−02 1.76E−05 1.02E−08 −67.8 22.28
200 mM Na2SO4
C3S-m In 8 mM Ca(OH)2 8.25 1.37E−03 0 0 0 1.30E−02 3.94E−03 4.39E−09 −61.9 1.33E−02 4.25E−03 4.11E−09 −61.6 3.34
C3S-m In 8 mM Ca(OH)2 + 7.3 7.91E−03 100 0 50 9.93E−03 1.30E−03 2.87E−08 −65.0 1.03E−02 1.27E−03 2.39E−08 −65.0 8.02
50 mM Na2SO4
C3S-m In 8 mM Ca(OH)2 + 7.68 9.40E−03 200 0 100 9.50E−03 1.05E−03 3.41E−08 −65.7 1.02E−02 9.32E−04 2.46E−08 −66.0 6.52
100 mM Na2SO4
C3S-m In 8 mM Ca(OH)2 + 7.87 0.00301 400 0 200 8.65E−03 8.54E−04 1.20E−08 −67.9 9.69E−03 6.62E−04 6.68E−09 −68.6 3.07
200 mM Na2SO4
C3S-m In 20 mM Ca(OH)2 19.9 8.82E−04 0 0 0 2.78E−02 6.84E−03 1.06E−09 −57.1 2.87E−02 7.75E−03 9.01E−10 −56.7 0.58
C3S-m In 20 mM Ca(OH)2 + 18.8 4.84E−04 100 0 50 2.43E−02 3.19E−03 5.91E−10 −60.8 2.51E−02 3.30E−03 4.46E−10 −60.8 1.18
50 mM Na2SO4
C3S-m In 20 mM Ca(OH)2 + 19.8 2.60E−03 200 0 100 2.36E−02 2.61E−03 3.11E−09 −59.9 2.50E−02 2.44E−03 1.97E−09 −60.2 1.88
100 mM Na2SO4
C3S-m In 11 mM Ca(OH)2 + 10.5 4.32E−03 200 200 0 1.26E−02 2.92E−03 1.11E−08 −62.1 0.01433 0.00186 7.37E−09 −63.1 4.06
200 mM NaSCN
C3S-m In 11 mM Ca(OH)2 + 9.89 6.08E−03 200 200 0 1.19E−02 2.77E−03 1.71E−08 −62.1 0.01344 0.00267 1.12E−08 −61.9 5.17
200 mM NaNO3
C3S-m In 11 mM Ca(OH)2 + 10.4 5.23E−03 200 200 0 1.25E−02 2.89E−03 1.39E−08 −61.9 0.01391 0.00296 9.27E−09 −61.6 5.28
200 mM NaCl
C3S-m In 11 mM Ca(OH)2 + 10.3 4.07E−03 200 200 0 1.24E−02 2.87E−03 1.09E−08 −62.2 0.01407 0.00258 7.11E−09 −62.2 5.36
200 mM acetate
C3S-m In 11 mM Ca(OH)2 + 10.7 3.19E−03 200 200 0 1.29E−02 2.96E−03 8.22E−09 −62.2 0.01459 0.00328 5.30E−09 −61.6 4.08
200 mM NaI
C3S-m In 11 mM Ca(OH)2 + 10.8 3.28E−03 200 200 0 1.30E−02 2.99E−03 8.36E−09 −62.1 0.01472 0.00333 5.39E−09 −61.5 4.69
200 mM NaClO3
 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138
Fig. 20. Variation of the dissolution rate of C3S-m according to ln Π (calculated with Pitzer's equations) when the tricalcium silicate is dissolved in solutions containing different concen-
trations of calcium hydroxide and different concentrations of sodium sulfate. The bigger dots correspond to experiments containing no sulfates and then the smaller the dots, the higher the
concentration of sulfate ions.
References
[1] L. Nicoleau, A. Nonat, D. Perrey, The di- and tricalcium silicate dissolutions, Cem.
Concr. Res. 47 (2013) 14–30.
[2] R.L. Baldwin, How Hofmeister ion interactions affect protein stability, Biophys. J. 71
(1996) 2056–2063.
[3] M.G. Cacace, E.M. Landau, J.J. Ramsden, The Hofmeister series: salt and solvent
effects on interfacial phenomena, Q. Rev. Biophys. 30 (3) (1997) 241–277.
[4] W. Kunz, P. Lo Nostro, B.W. Nimham, The present state of affairs with Hofmeister
effects, Curr. Opin. Colloid Interface Sci. 9 (2004) 1–18.
[5] Y. Zhang, P.S. Cremer, Interactions between macromolecules and ions: the
Hofmeister series, Curr. Opin. Chem. Biol. 10 (2006) 658–663.
[6] X. Pardal, F. Brunet, T. Charpentier, I. Pochard, A. Nonat, 27Al and 29Si solid-state
NMR characterization of calcium aluminosilicate hydrate, Inorg. Chem. 51 (2012)
1827–1836.
[7] G.K. Sun, J.F. Young, R.J. Kirkpatrick, The role of Al in C–S–H: NMR, XRD, and compo-
sitional results for precipitated samples, Cem. Concr. Res. 36 (2006) 18–29.
[8] F. Begarin, S. Garrault, A. Nonat, L. Nicoleau, Hydration of alite containing aluminum,
Adv. Appl. Ceram. 110 (3) (2011) 127–130.
[9] P. Faucon, T. Charpentier, A. Nonat, J.C. Petit, Triple-quantum two-dimensional 27Al
Magic Angle Nuclear magnetic resonance study of the aluminum incorporation in
calcium silicate hydrates, J. Am. Chem. Soc. 120 (1998) 12075–12082.
[10] R.K. Iler, Effect of adsorbed alumina on the solubility of amorphous silica in
water, J. Colloid Interface Sci. 43 (2) (1973) 399–408.
[11] B.R. Bickmore, K.L. Nagy, A.K. Gray, A.R. Brinkerhoff, The effect of Al(OH)− 4 on the
dissolution rate of quartz, Geochim. Cosmochim. Acta 70 (2006) 290–305.
[12] T. Chappex, K.L. Scrivener, The effect of aluminum in solution on the dissolution of
amorphous silica and its relation to cementitious systems, J. Am. Ceram. Soc. 96
(2) (2013) 592–597.
[13] L. Charlet, P.W. Schindler, L. Spadini, G. Furrer, M. Zysset, Cation adsorption on
oxides and clays: the aluminum case, Aquat. Sci. 55 (4) (1993) 291–303.
[14] P. Van Capellen, S. Dixit, J. van Beusekom, Biogenic silica dissolution in the oceans:
reconciling experimental and field-based dissolution rates, Glob. Biogeochem.
Cycles 16 (4) (2002) 23-1–23-10.
[15] J.R. Houston, J.L. Herberg, R.S. Maxwell, S.A. Carroll, Association of dissolved alumi-
num with silica: connecting molecular structure to surface reactivity using NMR,
Geochim. Cosmochim. Acta 72 (2008) 3326–3337.
[16] F.J. Doucet, C. Schneider, S.J. Bones, A. Kretchmer, I. Moss, P. Tekely, C. Exley, The for-
mation of hydroxylaluminosilicates of geochemical and biological significance,
Geochim. Cosmochim. Acta 65 (15) (2001) 2461–2467.
[17] T.W. Swaddle, Silicate complexes of aluminum(III) in aqueous systems, Coord.
Chem. Rev. 219–221 (2001) 665–686.
[18] K.S. Pitzer, Thermodynamics, 3rd ed. McGraw-Hill, New York, 1995.
[19] L.N. Plummer, D.L. Parkhust, G.W. Fleming, S.A. Dunkle, A computer program incor-
porating Pitzer's equations for calculation of geochemical reactions in brines, Water
Resour. Investig. Rep. 88 (1988) 4153.
[20] P. Deglmann, S. Schenk, Thermodynamics of chemical reactions with COSMO-RS:
the extreme case of charge separation or recombination, Comput. Chem. 33
(2012) 1304–1320.
[21] A. Klamt, Conductor-like screening model for real solvents: a new approach to the
quantitative calculation of solvation phenomena, J. Phys. Chem. 99 (1995) 224–2235.
[22] A. Klamt, F. Eckert, COSMO-RS: a novel and efficient method for the a priori predic-
tion of thermophysical data of liquids, Fluid Phase Equilib. 172 (2000) 43–72.
137
[23] K. Eichkorn, O. Treutler, H. Ohm, M. Haser, R. Ahrlrichs, Auxiliary basis sets to
approximate Coulomb potentials, Chem. Phys. Lett. 242 (1995) 652–660.
[24] A.D. Becke, Density-functional exchange-energy approximation with correct
asymptotic behavior, Phys. Rev. A 38 (6) (1988) 3098–3100.
[25] J. Perdew, Density-functional approximation for the correlation energy of the inho-
mogeneous electron gas, Phys. Rev. B 33 (12) (1986) 8822–8824.
[26] S.H. Vosko, L. Wilk, M. Nusair, Accurate spin-dependent electron liquid correlation
energies for local spin density calculation: a critical analysis, Can. J. Phys. 58 (8)
(1980) 1200–1211.
[27] A. Klamt, G. Schüürmann, COSMO: a new approach to dielectric constant screening
in solvents with explicit expressions for the screening energy and its gradient,
J. Chem. Soc., Perkin Trans. 2 (1993) 799–805.
[28] Y. Zhao, D.G. Truhlar, The Mo6 suite of density functionals for main group thermo-
chemistry, thermodynamical kinetics, noncovalent interactions, excited states, and
transition elements: two new functionals and systematic testing of four Mo6-
class functionals and 12 other functionals, Theor. Chem. Accounts 120 (2008)
215–241.
[29] F. Weigend, R. Ahlrichs, Balanced basis sets of split valence, triple zeta valence and
quadrupole zeta valence quality for H to Rn: design and assessment of accuracy,
Phys. Chem. Chem. Phys. 7 (2005) 3297–3305.
[30] R. Ahlrichs, M. Bär, M. Häser, H. Horn, C. Kölmel, Electronic structure calculations on
workstation computers: the program system turbomole, Chem. Phys. Lett. 162 (3)
(1989) 165–169.
[31] E.J. Bylaska, W.A. de Jong, N. Govind, K. Kowalski, T.P. Straatsma, M. Valiev, D. Wang,
E. Apra, T.L. Windus, J. Hammond, P. Nichols, S. Hirata, M.T. Hackler, Y. Zhao, P.D.
Fan, R.J. Harrison, M. Dupuis, D.M.A. Smith, J. Nieplocha, V. Tipparaju, M. Krishnan,
A. Vazquez-Mayagoitia, Q. Wu, T. Van Voorhis, A.A. Auer, M. Nooijen, L.D. Crosby,
E. Brown, G. Cisneros, G.I. Fann, H. Fruchtl, J. Garza, K. Hirao, R. Kendall, J.A.
Nichols, K. Tsemekhman, K. Wolinski, J. Anchell, D. Bernholdt, P. Borowski, T.
Clark, D. Clerc, H. Dachsel, M. Deegan, K. Dyall, D. Elwood, E. Glendening, M.
Gutowski, A. Hess, J. Jaffe, B. Johnson, J. Ju, R. Kobayashi, R. Kutteh, Z. Lin, R.
Littlefield, X. Long, B. Meng, T. Nakajima, S. Niu, L. Pollack, M. Rosing, G. Sandrone,
M. Stave, H. Taylor, G. Thomas, J. van Lenthe, A. Wong, Z. Zhang, NWChem, A
Computational Chemistry Package for Parallel Computers, Version 5.1.1, Pacific
Northwest National Laboratory, Richland, Washington, 2009.
[32] COSMOlogic GmbH & Co., K.G. Leverkusen, Germany, Available at http://www.
cosmologic.de (Version C2.1, Release 01.11).
[33] L. Divet, R. Randriambololona, Delayed ettringite formation: the effect of tempera-
ture and basicity on the interaction of sulphate and C–S–H phase, Cem. Concr. Res.
28 (3) (1998) 357–363.
[34] L. Nachbaur, P.-C. Nkinamubanzi, A. Nonat, J.-C. Mutin, Electrokinetic properties
which control the coagulation of silicate cement suspensions during early age
hydration, J. Colloid Interface Sci. 202 (1998) 261–268.
[35] A. Vyalikh, K. Zesewitz, U. Scheler, Hydrogen bonds and local symmetry in the
crystal structure of gibbsite, Magn. Reson. Chem. 48 (2010) 877–881.
[36] E. Lippmaa, A. Samoson, M. Mägi, High-resolution 27Al NMR of alumino-silicates,
J. Am. Chem. Soc. 108 (1986) 1730–1735.
[37] M.E. Smith, Application of 27Al NMR techniques to structure determination in solids,
Appl. Magn. Reson. 4 (1993) 1–64.
[38] B.M. De Witte, P.J. Grobet, J.B. Uytterhoeven, Pentacoordinated aluminum in
noncalcined amorphous aluminosilicates, prepared in alkaline and acid medium,
J. Phys. Chem. 99 (1995) 6961–6965.
[39] B.L. Phillips, F.M. Allen, R.J. Kirkpatrick, High-resolution solid-state 27Al NMR
spectroscopy of Mg-rich vesuvianite, Am. Mineral. 72 (1987) 1190–1194.
138 L. Nicoleau et al. / Cement and Concrete Research 59 (2014) 118–138
[40] D.R. Neuville, L. Cormier, A.-M. Flank, V. Briois, D. Massiot, Al speciation and Ca en-
vironment in calcium aluminosilicate glasses and crystals by Al and Ca K-edge X-ray
absorption spectroscopy, Chem. Geol. 213 (2004) 153–163.
[41] M.D. Andersen, H.J. Jakobsen, J. Skibsted, A new aluminium–hydrate species in hy-
drated Portland cements characterized by 27Al and 29Si MAS NMR spectroscopy,
Cem. Concr. Res. 36 (2006) 3–17.
[42] F. Zibouche, H. Kerdjoudj, J.-B. d'Espinose de Lacaillerie, H. Van Damme,
Geopolymers from Algerian metakaolin. Influence of secondary minerals, Appl.
Clay Sci. 43 (2009) 453–458.
[43] G. Le Saout, E. Lécolier, A. Rivereau, H. Zanni, Chemical structure of cement aged at
normal and elevated temperatures and pressures Part I. Class G oilwell cement,
Cem. Concr. Res. 36 (2006) 71–78.
[44] H. Vialis-Terrisse, A. Nonat, J.-C. Mutin, Zeta-potential study of calcium silicate hy-
drates interacting with alkaline cations, J. Colloid Interface Sci. 244 (2001) 58–65.
[45] C. Labbez, A. Nonat, I. Pochard, B. Jönsson, Experimental and theoretical evidence
of overcharging of calcium silicate hydrate, J. Colloid Interface Sci. 309 (2007)
303–307.
[46] Y. Fu, P. Xie, P. Gu, J.J. Beaudoin, Effect of temperature on sulphate adsorption/
desorption by tricalcium silicate hydrates, Cem. Concr. Res. 24 (8) (1994) 1428–1432.
[47] R. Barbarulo, H. Peycelon, S. Prene, Experimental study and modelling of sulfate
sorption on calcium silicate hydrates, Ann. Chim. 28 (2003) S5–S10.
[48] C. Labbez, I. Pochard, B. Jönsson, A. Nonat, C–S–H/solution interface: experimental
and Monte Carlo studies, Cem. Concr. Res. 41 (2011) 161–168.
[49] B. Jönsson, A. Nonat, C. Labbez, B. Cabane, H. Wennerström, Controlling the cohesion
of cement paste, Langmuir 21 (2005) 9211–9221.
[50] C. Labbez, B. Jönsson, I. Pochard, A. Nonat, B. Cabane, Surface charge density and
electrokinetic potential of highly charged minerals: experiments and Monte Carlo
simulations on calcium silicate hydrate, J. Phys. Chem. B 110 (18) (2006) 9219–9230.
[51] P.V. Brady, J.V. Walther, Controls on silicate dissolution rates in neutral and basic pH
solutions at 25 °C, Geochim. Cosmochim. Acta 53 (1989) 2823–2830.
[52] S. Brantley, Chapter 5 in “Kinetics of Water–Rock Interaction”, in: Susan L. Brantley,
James D. Kubicki, Art F. White (Eds.), Springer, 2007.
[53] J.J. Denny, W.D. Robson, D.A. Irwin, The prevention of silicosis by metallic aluminum,
Can. Med. Assoc. J. 37 (1) (1937) 1–11.
[54] C.P. Morrow, S. Nangia, B.J. Garrison, Ab initio investigation of dissolution mecha-
nisms in aluminosilicate minerals, J. Phys. Chem. 113 (2009) 1343–1352.
[55] J.C. Lewin, The dissolution of silica from diatom walls, Geochim. Cosmochim. Acta 21
(1961) 182–193.
[56] T. Yokohama, S. Kinoshita, H. Wakita, T. Tarutani, 27Al NMR study on the interaction
between aluminate and silicate ions in alkaline solution, Bull. Chem. Soc. Jpn. 61 (3)
(1988) 1002–1004.
[57] M.R. North, M.A. Fleischer, T.W. Swaddle, Precipitation from alkaline aqueous
aluminosilicate solutions, Can. J. Chem. 79 (1) (2001) 75–79.
[58] W.H. Casey, Large aqueous aluminum hydroxide molecules, Chem. Rev. 106 (1)
(2005) 1–16.
[59] J.E. Enderdy, Diffraction studies of aqueous ionic solutions, in: M.-C. Bellisent-Funel,
G.W. Neilson (Eds.), The Physics and Chemistry of Aqueous Solutions, NATO ASI
SeriesReidel, 1987.
[60] F. Bruni, S. Imberti, R. Mancinelli, M.A. Ricci, Aqueous solutions of divalent chlorides:
ions hydration shell and water structure, J. Chem. Phys. 136 (6) (2012) 064520-
1–064520-7.
[61] C. Labbez, I. Pochard, A. Nonat, B. Jönsson, Colloidal behavior of Cnanohydrates in ce-
ment paste CONMOD, Proceedings of the Symposium on concrete modelling (22
juin 2010) pp. 29–32 Lausanne, Suisse.
[62] F. Bellmann, D. Damidot, B. Möser, J. Skibsted, Improved evidence for the existence
of an intermediate phase during hydration of tricalcium silicate, Cem. Concr. Res. 40
(2010) 875–884.
[63] P. Juilland, E. Gallucci, R. Flatt, K. Scrivener, Dissolution theory applied to the induc-
tion period in alite hydration, Cem. Concr. Res. 40 (2010) 831–844