Cement and Concrete Research 42 (2012) 1317–1326

Contents lists available at SciVerse ScienceDirect

Cement and Concrete Research

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

Accelerated carbonation testing of alkali-activated binders significantly

underestimates service life: The role of pore solution chemistry
Susan A. Bernal a, John L. Provis a,⁎, David G. Brice b, Adam Kilcullen a, b,
Peter Duxson b, Jannie S.J. van Deventer b
a
Department of Chemical and Biomolecular Engineering, The University of Melbourne, Victoria 3010, Australia
b
Zeobond Pty Ltd., P.O. Box 210, Somerton, Victoria 3062, Australia

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

Article history: The carbonation resistance of alkali-activated binders is often tested via accelerated test protocols designed
Received 23 December 2011 for Portland cements, without questioning whether the tests replicate the mechanisms observed in service.
Accepted 23 July 2012 Thus, validation of accelerated methods is required to enable realistic prediction of material performance.
Changes in pore solution equilibria cause the formation of sodium bicarbonates during accelerated carbon-
Keywords:
ation, compared with hydrous sodium carbonates in natural carbonation. This shifts the carbonation mecha-
Alkali-activated cement (D)
Durability (C)
nism to favour more rapid reaction progress, to give a higher apparent degree of acceleration (compared to
Carbonation (C) natural conditions) than in Portland cements. The pore solution pH under accelerated carbonation is signif-
Pore solution (B) icantly lower than at natural CO2 concentrations, leading to a falsely short predicted service life (time to
Thermodynamic calculations (B) expected corrosion of embedded steel), as natural CO2 concentrations appear not to reduce the pH below
10. Thus, accelerated carbonation testing is unduly aggressive towards alkali-activated binders, and test
results must be cautiously interpreted.
© 2012 Elsevier Ltd. All rights reserved.

1. Introduction
Carbonation of cement and concrete materials in service has been
identified as a critical problem in many parts of the world, particular-
ly where the environment is warm and relatively humid. In order to
predict, in a short period of time, how a material will perform in ser-
vice when subjected to CO2 exposure, accelerated carbonation testing
of binders, mortars and concretes is widely applied to the study of
materials based on both Portland cement [1–5] and alkali-activated
[6–12] binders. Carbonation of alkali-activated binders has been iden-
tified from accelerated testing to be a potentially significant problem
in some important applications, with laboratory tests often showing
relatively high carbonation rates [6–11]. However, alkali-activated con-
cretes removed from service after extended periods have not in general
shown problems related to carbonation [8,13,14]. This indicates that the
outcome of the accelerated test is not as accurate a predictor of in-
service performance of alkali-activated materials as it is for Portland
cement-based systems. Establishing and validating durability testing
methods for alkali-activated concretes remains the major obstacle to
their commercial adoption in demanding structural applications, and
ultimately their acceptance in national and international regulatory
standards for structural concrete [15,16]. Therefore, the availability of
⁎ Corresponding author. Tel.: +61 3 8344 8755; fax: +61 3 8344 4153.
E-mail address: jprovis@unimelb.edu.au (J.L. Provis).
a reliable method for predicting the service life of alkali-activated con-
crete under carbonating conditions is an important issue which needs
to be addressed.
In all widely applied tests, acceleration of carbonation is achieved
by increasing the concentration of CO2 in the atmosphere surround-
ing the specimens under a controlled relative humidity (RH). Howev-
er, the testing methodologies have mostly not been validated in detail
by comparison to natural carbonation of comparable samples, and
there is still ongoing discussion regarding the most applicable testing
methodologies and conditions [4], with no testing methodology or
conditions being universally preferred. The work of Ho and Lewis
[2] provided a detailed study of Portland cement carbonation in ser-
vice, as well as direct comparisons with the results of accelerated test-
ing, which has given some validation for the accelerated carbonation
testing conditions used in that study (4% CO2, 50% RH, 23 °C) as ap-
plied to the particular materials investigated. However, there are
three key areas in which information is lacking with regard to the un-
derstanding of the carbonation of alkali-activated binders under ser-
vice conditions:
– CO2 concentrations ranging from 1% to 100% CO2, relative humidities
from 50% to 90% and temperatures from 5 to 37 °C have been reported
as test conditions for accelerated carbonation of Portland cement
concrete [4]; it remains to be determined in which environments
alkali-activated materials show carbonation mechanisms which accu-
rately replicate natural carbonation mechanisms. Castellote et al. [17]

0008-8846/$ – see front matter © 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.cemconres.2012.07.002
1318 S.A. Bernal et al. / Cement and Concrete Research 42 (2012) 1317–1326
showed that CO2 concentrations exceeding 3% lead to changes in the
mechanism of Portland cement carbonation when compared to natu-
ral carbonation, but no similar studies have been published for
alkali-activated binders.
– The relationship between accelerated carbonation rates and in-service
carbonation rates in alkali-activated concretes, which is essential in
using the results of accelerated carbonation testing to calculate cover
depths required for a given service life, is unknown. Ho and Lewis
[2] showed that, for a range of Portland cement concrete mixes,
1 week at 4% CO2, 50% RH and 23 °C gave a carbonation depth equiv-
alent to 1 year indoors in ambient air in a laboratory environment, but
also noted that exposure outdoors under uncontrolled conditions
gave as much as a factor of 3 difference (reduction) in natural carbon-
ation rate. Carbonation rates are well known to be sensitive to wet/dry
cycling, humidity, temperature and other exposure parameters
[4,5,18], which means that a detailed mechanistic understanding of
the carbonation phenomenon itself is necessary for such predictions
to be made with any confidence.
– During natural carbonation the binder structure is evolving over the
extended time periods (years) in contrast with accelerated tests
which are generally conducted on relatively young concretes for
short time periods (weeks). There is thus a need to incorporate an
understanding of the influence of gel ageing into the analysis of nat-
ural carbonation, but the structural changes taking place during
long-term ageing of alkali-activated binders remain incompletely
understood on a mechanistic level.
The carbonation of Portland cement concrete proceeds predomi-
nantly by the reaction of atmospheric CO2 (via its dissolution in the
water present in the pores) with the calcium in Ca(OH)2 and/or
Ca-rich C―S―H phases in the binder, to form calcium carbonate,
mainly as the calcite polymorph [17]. Alkali-activated binders are
synthesised with lower calcium content (including systems with
no calcium), and thus show a binder structure based on gels
which are enriched in alkalis and aluminium and lower in Ca/Si
ratio than those which are observed in hydrated Portland cements
[15,19,20]. This is expected to lead to differences in the carbonation
mechanisms depending on the chemistry of the binder source ma-
terials, which will influence both the rate of natural carbonation
and the degree of acceleration of carbonation induced by a given
CO2 concentration.
The fact that the pore solution chemistry in alkali-activated binders
is dominated by alkali hydroxides [20–23] will lead to differences in the
CO2 uptake rates, as these solutions are well known to absorb CO2 from
gas streams or the air [24], to form alkali carbonates and/or bicarbon-
ates. The low to zero Ca content and high degree of crosslinking of the
binder gel in alkali-activated binders is also likely to be relevant, as
the degradation of gels by decalcification will be influenced by these
parameters. The maturity of the alkali activated concrete is anticipated
to play a significant role in determining the accelerated carbonation
rates, considering the modification of both gel densification and the
structure of the pore network. These issues will be addressed in detail
in future studies, while the focus of this paper is specifically on the
role of the pore solution chemistry in alkali-activated slags, containing
significant levels of Ca, but much lower than in Portland cement, in
determining accelerated carbonation rates and mechanisms. As the
carbonation of Portland cement is to a large extent controlled by chem-
ical reactions involving the solid hydration products, the influence of
the pore solution on the carbonation rate has rarely been analysed.
Most studies have taken the reverse approach, and instead focused on
the influence of carbonation on pore solution chemistry [18,25]. In
this paper, we address the differences in carbonation rates induced by
accelerated carbonation of alkali-activated binders when compared
with Portland cement binders, and show that this is related to differ-
ences in alkali carbonate phase chemistry. The thermodynamic basis
for these differences is discussed, and comments are provided regarding
the validity of accelerated carbonation testing procedures in analysing
the durability of alkali-activated materials.
2. Experimental program
2.1. Concrete samples
A Colombian granulated blast furnace slag (GBFS 1) from the factory
Acerías Paz del Río was used as the primary raw material for the produc-
tion of concrete specimens. The basicity coefficient (Kb = CaO + MgO/
SiO2 + Al2O3) and the quality coefficient (Kq = CaO + MgO + Al2O3/
SiO2 + TiO2) based on the chemical composition (Table 1) were 1.01
and 1.92, respectively. Its specific gravity was 2900 kg/m3 and Blaine
fineness was 399 m 2/kg. The particle size range, determined through
laser granulometry, was 0.1–74 μm, with a d50 of 15 μm. The alkaline
activating solution was formulated by blending a commercial sodium
silicate solution with 31.7 wt.% SiO2, 12.32 wt.% Na2O and 56.9 wt.%
H2O, and 50 wt.% NaOH solution, to reach a silicate modulus (Ms =
SiO2/Na2O) of 1.3, and a concentration of activation of 4 wt.% Na2O
with respect to the mass of slag.
The concretes were formulated with slag contents of 340 kg/m3,
412 kg/m3 and 512 kg/m3 and a water/(slag+anhydrous activator)
ratio of 0.48. Crushed gravel and river sand were used as coarse and
fine aggregates in the manufacture of the concretes. The coarse aggregate
was of 19 mm maximum size, with a specific gravity of 2790 kg/m3 and
absorption of 1.23%. The specific gravity, absorption, and fineness mod-
ulus of the fine aggregate were 2450 kg/m3, 3.75% and 2.57, respectively.
Cylindrical specimens were produced, 76.2 mm in diameter and 152 mm
in length, and the ends covered with an acrylic resin to give solely radial
penetration of CO2. Samples were cured for 28 days at 25 °C and RH
>90% prior to the start of testing. Accelerated carbonation exposure
was carried out in a controlled environmental chamber at 25 °C, with
7% CO2 and 65% RH. Natural carbonation exposure was carried out in a
laboratory in Cali, Colombia, with average daily conditions approximately
25±5 °C and 75% RH.
As a reference material, Portland cement concretes were also pro-
duced, using a commercial Colombian Portland cement type I, which
includes 25% limestone. Detailed information regarding the mix de-
signs, along with analysis of the engineering and durability properties
of the concrete mixes assessed in this study, is reported elsewhere
[26].
2.2. Paste samples
To enable detailed analysis of binder mineralogy, paste samples were
synthesised from a different supply of granulated blast furnace slag
(GBFS 2), whose basicity coefficient Kb and quality coefficient Kq based
on the chemical composition (Table 1) were 1.00 and 1.79 respectively.
Its specific gravity was 2800 kg/m3 and Blaine fineness 450 m2/kg. In
order to regulate strength development, ordinary Portland cement
(OPC) type GP was incorporated in the binders as a minor admixture
(4% of total binder). All raw materials were supplied by Zeobond Pty
Ltd., Australia.
Table 1
Compositions of the GBFS supplies used, from X-ray fluorescence analysis. L.O.I. is loss
on ignition at 1000 °C.
Component (mass % as oxide) GBFS 1 GBFS 2
SiO2 31.08 33.80
Al2O3 13.98 13.68
Fe2O3 3.09 0.40
CaO 43.92 42.56
MgO 1.79 5.34
SO3 0.66 0.83
Others 3.49 1.45
L.O.I. 2.08 1.81
 S.A. Bernal et al. / Cement and Concrete Research 42 (2012) 1317–1326 1319

The alkali activator used was a solid anhydrous sodium metasilicate

with 49.1 wt.% SiO2 and 50.9 wt.% Na2O. This was dissolved in the mix
water and allowed to cool to room temperature prior to preparation
of the specimens. A water/binder ratio of 0.40 was used to prepare
50 mm paste cube specimens, which were cured in sealed bags at
23 °C. After 14 days of curing, the samples were removed from the sealed
bags and stored exposed to air on a laboratory bench top (temperature
23±5 °C and relative humidity 55±30%) for 3 years. Due to the low po-
rosity of the paste cubes, the centres of the cubes remained uncarbonated
throughout this time, and these are used as reference samples.
An additional set of samples was fabricated with the same mix design,
cured for 7 days, crushed using a mortar and pestle, and then exposed to
CO2 for 7 days under controlled environmental conditions using an accel-
erated carbonation chamber with a precision of ±0.2% CO2 concentration.
The exposure conditions used were CO2 concentrations of 1.0%, 3.0% and
5.0%, a temperature of 23±2 °C, and a relative humidity of 65%±5%. It is
also noted that it is unknown whether the crushing of the sample prior to
accelerated carbonation exposure induces any changes in the carbonation
mechanism, but this step is required to achieve rapid and uniform carbon-
ation of the paste.
Carbonated and uncarbonated pastes were analysed by X-ray dif-
fraction (XRD) using a Bruker D8 Advance instrument with Ni-filtered
CuKα radiation. The tests were conducted with a step size of 0.020°
and a count time of 4 seconds per step.

3. Results and discussion

3.1. Relationship between natural and accelerated carbonation results in

alkali-activated concretes

To calibrate an accelerated test, it is important to relate the rate of

degradation induced under accelerated exposure conditions to the
rate of degradation experienced in service. Fig. 1 shows a comparison
of the carbonation depths observed using the phenolphthalein meth-
od for alkali-activated slag (AAS) and Portland cement (OPC) con-
cretes exposed to natural (indoor) carbonation conditions and to
accelerated (7% CO2) conditions.
Fig. 1. Relationship between natural (in a laboratory in Cali, Colombia; average daily
conditions approx. 25 ± 5 °C and 75% RH) and accelerated (7% CO2, 65% RH, exposure
times of (a) 10 or (b) 16 days) carbonation rates in alkali-activated slag (AAS) and
Portland cement (OPC) concrete mixes, with a w/b = 0.48 and a range of paste volumes.

Fig. 1 demonstrates that there are important differences in perfor-

mance between AAS and OPC concretes under natural and accelerated
carbonation conditions. If an accelerated carbonation test is to accurately
replicate what is happening under natural carbonation conditions, it is
expected that, independent of the nature of the material assessed, a linear
relationship should be identified when correlating natural and accelerat-
ed carbonation results. It is immediately evident that the degree of accel-
eration (i.e. the aggressiveness of the accelerated test) observed in the
AAS systems when increasing the CO2 concentration above a natural con-
centration is significantly higher than when OPC concretes are subjected
to the same elevated CO2 conditions. This has major implications for the
comparative analysis of the two types of binder system by accelerated
carbonation testing; if an alkali-activated concrete and a Portland cement
concrete show the same carbonation depth in an accelerated test, the
durability of the alkali-activated material under natural carbonation
conditions (and thus the service life) would be considerably greater.
This also implies that it is very likely that an alkali-activated material,
although exhibiting a higher susceptibility to accelerated carbonation
than a Portland cement-based material exposed to the same elevated
CO2 conditions, can perform equally well in service, which correlates
well with observations in the literature [8].
Based on this observation, it seems important to assess in detail the
differences between the mechanisms of carbonation of Portland cement
and alkali-activated binders. As noted above, hydrated Portland cement
carbonates to form CaCO3, mainly as calcite, leaving a residual silica-rich
gel [5,17]. Some CO2 is also proposed to be adsorbed onto the C―S―H
gel [27], and the formation of amorphous or poorly-crystalline CaCO3
[28,29] is possible. However, the nature of the phases formed in the
CaO―CO2―H2O system is not particularly strongly influenced by CO2
partial pressure in the range of relevance here (although the absolute
solubility is well known to vary as a function of pCO2), and there is little
probability of formation of hydrous calcium carbonate phases at room
temperature. Conversely, the Na2O―CO2―H2O phase diagram, which
is of relevance to alkali-activated binder pore solution chemistry, is
much more complex in terms of potential hydrous phases and phase
boundaries. This necessitates a detailed analysis of the carbonation
products observed in alkali-activated binders under both natural and
accelerated carbonation conditions.
3.2. Alkali-activated binder carbonation products as a function of CO2
concentration
X-ray diffraction data collected from alkali-activated slag paste
samples at 14 days of age, and from the uncarbonated region of the
specimen after 3 years of exposure to air, are presented in Fig. 2.
The main crystalline phases present in these samples correspond to
the remnant unreacted slag and include quartz (SiO2; Powder Diffrac-
tion File (PDF) card #01-070-2536), åkermanite (Ca2MgSi2O7, PDF
#00-035-0592) and gehlenite (Ca2Al2SiO7, PDF #00-035-0755). Peaks
corresponding to three different types of calcium silicate hydrates are ob-
served: dicalcium silicate hydrate (Ca2SiO4∙3H2O, PDF #00-029-0374),
tobermorite (Ca5(Si6O16)(OH2), PDF #01-089-6458) and hillebrandite
(Ca2(SiO3)(OH2), PDF #00-042-0538), similar to the calcium silicate hy-
drate phases identified in hydrated Portland cement [30]. These are all
poorly crystalline, and are marked as a single C―S―H contribution in
Fig. 2. Hydrotalcite (Mg6Al2(CO3)(OH)16∙4H2O, PDF #00-041-1428) is
1320 S.A. Bernal et al. / Cement and Concrete Research 42 (2012) 1317–1326

Fig. 2. X-ray diffractograms of (a) alkali-activated slag after 14 days of curing, and (b) the same sample after 3 years.

also identified as a reaction product, which is coherent with other studies
[31–34] of the activation of slags containing åkermanite and significant
contents of magnesium.
An aluminosilicate with a zeolitic structure (gismondine,
CaAl2Si2O8∙ 4H2O, PDF #00-020-0452) is observed in specimens
after 14 days of curing, although the intensities of the peaks
assigned to this phase are reduced after 3 years. At this time,
a (Ca,Na)-aluminosilicate product with a different zeolitic struc-
ture (thomsonite, NaCa2Al5Si5O20∙ 6H2O, PDF #00-046-1448) is
identified. Gismondine has been observed by high resolution X-ray dif-
fraction as a reaction product in low magnesium silicate-activated slag
binders [9], and gismondine and thomsonite are known to be formed
under very similar conditions via natural or laboratory hydrothermal pro-
cesses [35–37]. The two zeolites observed here have similar Si/Al ratios,
and it appears that the gismondine has been partially converted to the
more alkali-rich thomsonite during an extended period of exposure to
the alkali-rich pore solution environment.
A small amount of calcite is present in the 14-day and 3-year
uncarbonated specimens, associated with a slight degree of carbonation
of the activating solution and/or fresh paste before or during mixing of
the samples. In specimens with 3 years of curing, a small content of
vaterite is also identified. It is therefore likely that this is a metastable
product formed in these systems indicating the very first stages of
binder carbonation, consistent with recent data for Portland cements
carbonated under supercritical conditions, where vaterite was in
some cases observed as a rim ahead of the measured “carbonation
zone” [38].
To provide a point of comparison between the natural and acceler-
ated carbonation conditions, Fig. 3 shows the XRD data obtained for
naturally carbonated alkali-activated slag, and specimens subjected
to accelerated carbonation at CO2 concentrations of 1%, 3% and 5%,
for 7 days. The naturally carbonated specimen displays a range of
carbonate phases: CaCO3 in three crystalline polymorphs (calcite, vaterite
and aragonite), most probably derived from the carbonation of the
C―S―H type products, the hydrous sodium carbonate phase natron
(Na2CO3∙10H2O, PDF #00-015-0800) attributed to the carbonation of
the remnant alkali supplied by the activator and present in the pore solu-
tion, and several carbonate double salts (pirssonite (Na2Ca(CO3)2∙2H2O),

Fig. 3. X-ray diffractograms of carbonated alkali-activated slag pastes: (a) carbonated under natural conditions for 3 years, and where accelerated carbonation was induced by ex-
posing the specimens to (b) 1% CO2, (c) 3% CO2 and (d) 5% CO2.
 S.A. Bernal et al. / Cement and Concrete Research 42 (2012) 1317–1326 1321

PDF #00-022-0476), huntite (Mg3Ca(CO3)4; PDF #00-014-0409) and water from the naturally-carbonated samples, while the accelerated
dolomite (CaMg(CO3)2; PDF #01-083-1530), where the Mg-containing carbonation tests were conducted under a slightly higher (but con-
phases are consistent with the carbonation of hydrotalcite or other trolled) relative humidity, is likely to contribute to the carbonate
Mg-containing phases. Detailed discussion of the influence of natural and phase differences between the samples exposed to different conditions.
accelerated carbonation on the silicate and aluminosilicate binder phases The molar volume of pirssonite (102.58 cm3/mol) is lower than that of
will be presented in a subsequent publication [39]; the focus of this paper gaylussite (148.95 cm3/mol) [41], which may partially counteract the
is specifically on the carbonate phases. differences between natron and the less-hydrous bicarbonates as
The XRD results (Fig. 3) also reveal that the crystalline carbonation discussed above, but this difference is much less pronounced and so
products formed in specimens exposed to high concentrations of CO2 would only ever provide a partial compensation for the loss of natron
differ both in phase and in concentration from those which are as a pore-filling product.
obtained under natural carbonation conditions. The strong calcite
peak at 29.42° 2θ is much less evident under accelerated carbonation 3.3. Effect of carbonation on the pore solution of alkali-activated binders
conditions, and substantial increases in the intensities of peaks asso-
ciated with vaterite and aragonite are observed with increasing CO2 To understand the pore solution carbonation chemistry of alkali-
concentrations. This is most likely to be due to kinetic effects, as the activated binders, it is potentially instructive, as a first approximation, to
metastable CaCO3 polymorphs vaterite and aragonite will be the represent the solution as falling within the system Na+―OH−―CO32−
first to form, and will then convert to calcite as the reaction time ex- ―HCO3−―CO2―H2O. This neglects the contributions of Ca, Mg, Si, S and
tends, according to the Ostwald Step Rule [40]. This is consistent with Al, which are also identified in the pore solutions of alkali-activated
the presence of calcite as the primary CaCO3 phase in the carbonated binders, and discussion related to some of these elements will be
region (and vaterite in the “uncarbonated” region) of the sample presented later in this paper. There have been a number of published
exposed to natural carbonation conditions for 3 years, as discussed studies of phase equilibria in the sodium carbonate/bicarbonate system
above. In another study [9], calcite was the only CaCO3 polymorph in a wide variety of contexts, and probably the most relevant studies are
identified after accelerated carbonation of alkali-activated slag paste those which have been aimed at understanding phase equilibria in alka-
over 1000 h (a much longer timeframe than was used here), which line brine lakes and salt deposits in the Americas and Africa. Fig. 4 is
is consistent with this mechanism. adapted from one such study [46], and displays data for the system Na2-
The other important differences observed in Fig. 3 are the changes CO3―NaHCO3―H2O; i.e., without the addition of extra NaOH to the sys-
in the sodium-containing carbonate phases between the naturally tem. This diagram shows that the invariant point between natron, trona
carbonated samples and those subjected to accelerated carbonation.
No bicarbonates were observed in the samples exposed to natural
carbonation, while nahcolite (NaHCO3, PDF #00-015-0700) was nota-
ble at elevated CO2 concentrations. This change, coupled with the
presence of some thermonatrite rather than natron in the samples,
will have important consequences related to two important aspects
of concrete durability. The first of these consequences is that the dif-
ference in solid phases indicates a marked change in the dissolved
carbonate/bicarbonate equilibrium. This is important in determining
the pH of the carbonated pore solution, and therefore the potential
development of corrosion in the metallic reinforcement of structural
concrete. This issue will be explored in detail in the following sections
of this paper, with the aid of thermodynamic modelling tools.
The second important consequence of the change in phase prevalence
is that natron binds a very large quantity of water (5 water molecules per
Na+ cation), while thermonatrite has only 10% of this water content, and
nahcolite is anhydrous. This means that natron has a very large molar
volume (196.56 cm3/mol), while thermonatrite (54.96 cm3/mol) and
nahcolite (38.66 cm3/mol) [41] will fill much less space, and thus provide
a greatly reduced degree of pore blockage as they grow within the pore
space of an alkali-activated binder. Pore blockage is important in reducing
diffusion through all cementitious binders, and particularly in the case of
alkali-activated materials, where the binder phases are less space-filling
than in the case of Portland cement hydrates, and where reductions in
permeability (achieved either through reduced pore volume or increased
tortuosity) are essential in extending the time to corrosion of embedded
steel reinforcing elements [15,42].
It is also of interest that pirssonite (Na2Ca(CO3)2∙2H2O) is the Na―Ca
carbonate phase observed in the naturally-carbonated specimens,
where the more-hydrated gaylussite (Na2Ca(CO3)2∙5H2O) is instead ob-
served in the specimens subjected to accelerated carbonation. This is
likely to be related to the differences in water availability between the
sample sets; the relative prevalences of gaylussite and pirssonite are
known to be determined by aH2O and temperature [43–45]. As the sam-
ples were subjected to similar temperature regimes (23± 5 °C), falling
in the region in which gaylussite and pirssonite may both be formed in
Fig. 4. Phase diagram as a function of temperature and gas phase CO2 concentration for
a humid atmosphere, the former as a stable product and the latter meta- the Na2CO3―NaHCO3―CO2―H2O system in an air atmosphere at ambient pressure;
stable [43], it is unlikely that temperature is the determining parameter. after [46]. The dashed line represents a temperature of 23 °C, as used in the experi-
The fact that natron formation and atmospheric drying removed free mental work reported here.
1322 S.A. Bernal et al. / Cement and Concrete Research 42 (2012) 1317–1326
and nahcolite (in equilibrium with an aqueous solution) is present at
around 20 °C and 0.15% CO2. More recent experiments have indicated
that it is possible that this point may in fact occur at slightly lower T
and pCO2 conditions [47], but the phase diagram presented as Fig. 4 is
the most complete available in the open literature at present.
For a temperature of 23 °C, corresponding to the experimental con-
ditions used here (marked as a dotted line in Fig. 4), natron would be
expected in a natural environment (~0.03–0.04%), but increasing this
by as little as a factor of two would push the equilibrium into the region
in which trona is favoured. It is noted that trona has previously been
observed as a dominant phase in accelerated carbonation testing of
alkali-activated slag binders (at 3% CO2, 20 °C and 65% RH) [9], although
only for some of the mix designs discussed in that paper, and the prox-
imity of the test conditions to the phase boundaries as shown in Fig. 4
suggests that the influence of some of the other system components
may be significant. Unfortunately, the availability of thermodynamic in-
formation regarding the influence of alkaline earth metals, aluminates,
silicates, sulfates and sulfides on the phase equilibria in alkali carbon-
ate–bicarbonate systems is limited. The addition of 10 wt.% metakaolin
(as an additional alumina source) into the binder systems in [9] was
seen to entirely suppress trona formation as a consequence of the influ-
ence of dissolved alumina, potentially via the binding of sodium in
aluminosilicate gels and thus reduced alkali availability for carbonate
formation.
Fig. 4 also shows that a further increase in CO2 concentration, moving
beyond 0.2%, would then give nahcolite formation instead of natron or
trona. This indicates not only a change in solid phase formation, but also
a fundamental shift in solution chemistry, from a regime in which carbon-
ate is the main dissolved form of CO2 present to a bicarbonate-dominated
regime. This will be important in the understanding of pore solution pH,
which is critical to the protection of steel reinforcing elements in an
alkali-activated binder, and will be addressed in more detail later in this
paper.
To provide further insight into the aqueous solutions and solid alkali
carbonate phases formed in the pore networks of alkali-activated
binders (which are enriched in NaOH compared to those depicted in
Fig. 4) under natural and accelerated carbonation conditions, a series
of thermodynamic simulations have been undertaken. These calcula-
tions are conducted using the HSC Chemistry 4.1 software package
[48], with a Gibbs Energy minimisation solver routine. Each simulation
represents a system with 1 kg of water at 25 °C, containing specified
quantities of dissolved solids, in equilibrium with at least 2000 m 3 of
air at atmospheric pressure (representing an open system) with concen-
trations of CO2 corresponding to natural or accelerated carbonation con-
ditions. The temperature of 25 °C was selected because of the much
greater availability and reliability of Pitzer activity parameters at this
temperature when compared with any other; the results are very similar
when calculations are carried out at 23 °C. Although the software does
not explicitly enforce the Gibbs phase rule, this has been implemented
manually by forbidding the formation of solid phases exceeding the
number indicated by this rule (i.e. a maximum of two solid alkali car-
bonate phases is allowed, as none of the CO2 concentrations simulated
correspond to the natron–trona–nahcolite invariant point in Fig. 4).
Full details of the thermodynamic model, including all parameters
used, are given in Appendix A. The results of the simulations at different
pCO2 levels are shown in Fig. 5.
Fig. 5 shows that at 23 °C, natron is predicted to be dominant at low
pCO2, and nahcolite at high pCO2, consistent with the experimental ob-
servations in alkali-activated slag systems as discussed in Section 3.2
above. The point at which this change in the preferred carbonate
phase takes place is predicted to be around 0.1–0.15% CO2, consistent
with the data in Fig. 4. There is not a distinct region of trona prevalence
in these simulations, which may be due to the higher NaOH concentra-
tion in these simulations compared to the systems depicted in Fig. 4, or
may be related to differences in the thermodynamic database and pa-
rameters used in the computations. To confirm that the database and
Fig. 5. Phases (and relative concentrations) calculated to be formed when 1 mol/kg
NaOH solution is exposed to pseudo-open system gas environments with CO2 present
at different concentrations in dry air, at 25 °C.
calculation procedure give at least a qualitatively accurate representa-
tion of the phase diagram, some calculations were conducted at higher
temperatures, showing that (as expected, based on Fig. 4) trona and
thermonatrite become important at 50 °C. Simulations of 10 mol/kg
NaOH solutions were also carried out, and gave results (in terms of
relative phase prevalences) which were essentially identical to those
shown for 1 mol/kg NaOH solutions in Fig. 5.
The inclusion of additional components in the simulation to more
closely replicate the actual pore solution of an alkali-activated binder
leads to some minor changes in the solid phases formed, but the sodium
carbonate/bicarbonate phases are relatively unchanged by this inclusion.
Here, the additional components in the pore solution are represented by
the inclusion of 10 mmol/kg CaO and 10 mmol/kg MgO in the initial
definition of the simulated systems. The results of the calculations,
expressed as the quantities of solid carbonates deposited as a function
of the quantity of NaOH added to the system at natural (0.04%) and ele-
vated (4%) CO2 concentrations, are given in Fig. 6. As was observed in
Figs. 4 and 5, natron is the primary sodium carbonate phase formed at
ambient pCO2, while nahcolite is dominant at 4% CO2. The trends in rela-
tive concentrations of dissolved CO32−and HCO3− ions are markedly differ-
ent between the two CO2 partial pressures, with carbonate ions much
more prominent at ambient CO2 concentrations than at elevated concen-
tration, although the precipitation of calcium and magnesium carbonates
(rather than bicarbonates) appears also to play a role in regulating this
ratio in the aqueous phase. This point will be revisited below, with regard
to the discussion of the pH of the solutions.
At low NaOH content and either ambient or elevated CO2 concentra-
tions, Mg is predicted to precipitate as the mixed Mg―Ca carbonate
phase huntite (CaMg3(CO3)4), whereas magnesite (MgCO3) is favoured
at higher NaOH concentrations. In the XRD data presented in Fig. 3,
huntite was observed as a carbonation product, and the calculations
appear consistent with this observation. The majority of the calcium
present is predicted to form calcite under all conditions, in addition to
the huntite as discussed. The metastable aragonite and vaterite phases
of CaCO3, which were identified by XRD here, are not predicted because
they are not thermodynamically stable. The Ca concentrations simulat-
ed here are actually notably higher than those which have been reported
in pore solutions extracted from alkali-activated binders (which tend to
be closer to 1 mM, or even lower) [20–23], indicating that the extensive
formation of calcium carbonate phases identified in Section 3.2 is likely
also to be related to decalcification of the gel during carbonation. The
 S.A. Bernal et al. / Cement and Concrete Research 42 (2012) 1317–1326
Fig. 6. Calculated concentrations of key species present in, or precipitated from, simu-
lated alkali-activated slag pore solutions (10 mmol/kg Ca, 10 mmol/kg Mg) as a func-
tion of NaOH concentration, under (a) natural (0.04% CO2) and (b) accelerated
carbonation (4% CO2) conditions. Dashed lines are dissolved species, solid grey lines
are alkaline earth carbonates, and solid black lines are alkali metal carbonates.
formation of gaylussite or pirssonite was not observed in these calcula-
tions because these phases are not included in the database of the
simulation software used here [48]; this is noted as a limitation of the
current calculations. Königsberger et al. [49] commented specifically
on the limited extent and quality of rigorous thermodynamic data avail-
able in the CaO―Na2O―CO2―H2O system, and noted the need for more
detailed experimental data (such as those reported in [47]) to remedy
some of the uncertainties in these relationships.
Additional simulations were conducted in the presence of Si at var-
ious concentrations up to 1000 mmol/kg, and also with the presence of
a small quantity of Al, but the presence of these elements did not show a
strong influence on the nature of the carbonate phases formed, or the
pore solution chemistry (carbonate/bicarbonate ratio and pH). For this
reason, it is postulated that the absence of silicate and aluminate species
from the calculations presented here does not prevent the development
of a reasonable explanation for the observed trends in the experimental
data, and that the use of this approximation to the real pore solution
chemistry of these materials is appropriate. The value of the calculations
conducted in the presence of Si and Al was restricted to some extent by
1323
the lack of a rigorous description of C―(A)―S―H gel phases within the
HSC Chemistry 4.1 simulation software environment, meaning that an
alternative method of simulating the carbonation process taking place
at the pore solution–gel interface must be developed for alkali-activated
binder systems; work in this area is ongoing at present.
3.4. Implications for durability of reinforced alkali-activated concretes
Probably the most economically important consequence of the degra-
dation of concrete by carbonation is the corrosion of embedded steel
reinforcing components, leading to failure of the concrete due to cracking
either as the corroding steel expands or as the loss of the structural prop-
erties of the steel transfers excessive load to the concrete. The loss of
mechanical properties of the binder which can often accompany
carbonation is also important, but in general it is assumed that the
initiation of steel corrosion is responsible, in a practical sense, for the
designated end of the service life of the material or structure. Carbonation
of Portland cement reduces the pH to a level where steel is depassivated,
and is thus highly problematic with regard to steel corrosion. The stan-
dard indicator used to determine carbonation depths in concretes is phe-
nolphthalein [50], which shows a colour change from pink/purple to
colourless corresponding to the pH at which carbonate is converted to bi-
carbonate in solution [51]. This pH also corresponds approximately to the
pH at which steel is depassivated in pore solution environments, thus
supporting its use as a measure of carbonation depth.
However, the XRD and simulation results presented above show that
under natural carbonation conditions, the phases precipitated from the
pore solution of alkali-activated binders are dominated by alkali carbon-
ates rather than bicarbonates, and so the issue of exactly how far the pH is
likely to drop during natural carbonation becomes important. The calcu-
lated alkalinity and carbonate/bicarbonate ratios in the simulated pore
solutions (which precipitate solid phases as shown in Fig. 6) are given
in Fig. 7. The carbonate/bicarbonate ratios are generally >1 for the natu-
ral carbonation conditions, and b0.1 for the accelerated carbonation con-
ditions. This indicates that the pore solution chemistry is in good
agreement with the nature of the solid phases which precipitate, and
thus that the nature of these phases (as readily identifiable by XRD)
may be used as an experimental tool to provide insight into the chemistry
of the pore solution, which is very difficult to obtain directly.
It should be noted that the alkalinity curves in Fig. 7 are plotted
according to the definition of pH as − log10(aH+), using the correct
activity coefficients as calculated by the Pitzer equations. The activity
coefficient of H + in the concentration range of interest here is always
less than 1, meaning that the estimation of pH directly from hydrogen
ion concentrations will provide an underestimation of its true value.
Fig. 7. Plot of pH (solid lines) and carbonate/bicarbonate ratio (dashed lines) as a function of
NaOH addition and CO2 concentration, for simulated alkali-activated binder pore solutions
(10 mmol/kg Ca, 10 mmol/kg Mg). Black lines are at 0.04% CO2, and grey lines at 4% CO2.
1324 S.A. Bernal et al. / Cement and Concrete Research 42 (2012) 1317–1326

From Fig. 7, it is evident that the effect of carbonation on pore solution related to the natural carbonation of concrete specimens. JLP would also
pH in alkali-activated systems is highly dependent on the partial pressure like to thank E. Jagniecki (Binghampton State University) for useful dis-
of CO2; the pH does not drop below 10 under natural CO2 levels in the cussions regarding carbonate phase equilibria.
[Na+] range studied, while the pH induced by the elevated CO2 levels of
traditional accelerated-carbonation tests is almost two units lower. The Appendix A. Details of thermodynamic calculations
pH increases with NaOH concentration up to ~0.5 mol/kg, which is also
the point where Fig. 6 shows that the composition of the solution remains Table A1
approximately constant as solid alkali (bi)carbonates precipitate. Thermodynamic parameters for the phases and species simulated.
It is also important to note, from Fig. 7, that even if the NaOH concen-
Gas ΔHf0 (kJ/mol) S0 (J/mol∙K)
tration is reduced to 0.001 mol/kg, the calculated pH still remains above
Carbon dioxide CO2(g) −393.5 213.77
10. The loss of steel protection via alkali leaching from alkali-activated
Solid
binder systems thus appears to be a lesser concern with respect to car- Calcite CaCO3 −1206.9 92.90
bonation; alkali leaching may reduce the NaOH concentration, and thus Huntite CaMg3(CO3)4 −4529.6 299.53
the pH, but even when the majority of the alkalis have been removed, Magnesite MgCO3 −1095.8 65.70
the pH is not seen to drop into the region where steel depassivation Artinite Mg2(OH)2CO3∙ −2920.6 232.92
2H2O
would be expected. This is not to say that steel corrosion in alkali-
Natrite Na2CO3 −1130.8 138.78
activated binders will not ever be related to alkali leaching or carbon- Thermonatrite Na2CO3∙H2O −1432.0 168.10
ation—there may certainly be circumstances in which these phenomena Sodium carbonate Na2CO3∙7H2O −3192.7 422.20
are linked—but from these results, it is unlikely that carbonation in atmo- heptahydrate
Natron Na2CO3∙10H2O −4082.0 564.70
spheric conditions, even in combination with a significant degree of alkali
Trona Na3H(CO3)2∙ −2682.1 303.13
leaching, will reduce the pH to such an extent as to directly lead to steel 2H2O
corrosion. Conversely, under accelerated carbonation conditions and par- Nahcolite NaHCO3 −950.8 101.70
ticularly when the alkalis are leached from the material, the pore solution Aqueous
pH is certainly reduced to a level which could induce steel corrosion, and Carbonate ion CO32−(aq) −675.2 −50.00
2+
Calcium ion Ca (aq) −543.1 −56.48
it is therefore important to note that testing for steel corrosion in binders
Hydrogen ion H+(aq) 0 0
exposed to elevated CO2 environments will not give results which are Bicarbonate ion HCO3−(aq) −689.9 98.45
representative of in-service performance under atmospheric conditions. Magnesium ion Mg2+(aq) −466.0 −138.07
This is the key outcome of the work presented in this paper, which repre- Sodium ion Na+(aq) −240.3 58.41
Hydroxyl ion OH−(aq) −230.0 −10.71
sents the early stages of an extended work programme dedicated to anal-
Carbon dioxide CO2(aq) −413.84 117.61
ysis of the carbonation of alkali-activated binders under a variety of
formulation and testing conditions, aimed at providing a detailed scientif-
ic understanding of this process and how it may be tested and predicted. Table A1 presents the basic thermodynamic data used in the HSC
Chemistry 4.1 software [48] to describe the species modelled in this inves-
4. Conclusions tigation. This software package uses a Gibbs energy minimisation routine
to calculate stable phase assemblages, with the standard enthalpy of for-
Accelerated carbonation testing of alkali-activated binders has been mation and standard entropy of each component, and solution-phase
shown to provide results which are unlikely to represent the behaviour activity coefficients, as input parameters. The default database supplied
of the products under actual service conditions, due to changes in alkali with the HSC Chemistry 4.1 software was used for all thermodynamic
carbonate phase equilibria when the CO2 concentration is increased. parameters except the activity coefficients of dissolved species. All calcu-
The reaction products formed during natural and accelerated carbon- lations presented here were conducted at a temperature of 25 °C. The gas
ation testing of alkali activated slag binders are notably different, and phase also contained nitrogen and oxygen in a 79:21 ratio to simulate nat-
thermodynamic calculations support the experimental observations ural dry air, but these components were treated as inert.
related to differences in carbonate/bicarbonate equilibria and in the
nature of the carbonation products. In particular, bicarbonates favoured In the thermodynamic calculations, dolomite formation is suppressed,
over carbonates at high CO2 partial pressures. This gives a significantly as this phase is known to be kinetically very slow to precipitate [52].
greater decrease in pH upon carbonation than would be observed in Calcite is predicted to be the stable form of calcium carbonate under all
natural CO2 environments, rendering the pore solution environment conditions simulated in this paper. The relative stabilities of magnesite
in an accelerated test far more damaging to embedded steel reinforcing (MgCO3) and nesquehonite (MgCO3∙3H2O) are calculated to be very sim-
than is the case in actual service conditions. In fact, natural CO2 concen- ilar, but only the former has been observed in the XRD analyses of
trations appear unlikely to reduce the pore solution pH below 10, even alkali-activated binders conducted in our laboratory, and so this is the
when combined with alkali leaching effects, while the bicarbonate- phase considered in the calculation. Solid alkali or alkaline earth hydrox-
dominated pore solutions present at high pCO2 have a much lower ides (NaOH, brucite or portlandite) were not calculated to be stable under
pH, giving a falsely early predicted likelihood of corrosion initiation. the conditions simulated. Some of the carbonate minerals listed, such as
Accelerated carbonation testing of alkali-activated binders at elevated artinite and thermonatrite, were also not predicted to be stable under
CO2 partial pressures is a far more aggressive test when compared any of the conditions discussed in this study, but do appear in the phase
with the exposure of Portland cement to the same conditions, and assemblages under other specific sets of conditions, and so are retained
may provide outcomes which show apparent degradation much more in the computations for completeness. Following the discussion pres-
rapid and extensive than would actually be experienced in service. ented by Reardon [53] for Portland cement pore solutions, the formation
of aqueous complex ions or neutral complexes (e.g. NaCO3−(aq), CaOH+
Acknowledgements (aq), CaCO30(aq) and others) is not considered explicitly, as the effect of
the formation of these species on the solution-phase activity coefficients
This work has been funded by the Australian Research Council, is encapsulated in the fitting of the activity coefficient model parameters
through a Linkage Project co-sponsored by Zeobond Pty Ltd, and also for the dissociated ions.
including partial funding through the Particulate Fluids Processing The solution-phase activity coefficients utilised in the model were
Centre. SAB thanks E.D. Rodríguez and Professor R. Mejía de Gutierrez calculated using the Pitzer model [54]. This is a widely used semi-
(Universidad del Valle) for collaboration and support in the test program empirical correlation model for osmotic and activity coefficients,
 S.A. Bernal et al. / Cement and Concrete Research 42 (2012) 1317–1326 1325

applicable up to high ionic strength and based on an extension of the E ′ zA zB  1 1

 Eθ
θE ¼ Ω1 ð xAB Þ− Ω 1 ðx AA Þ− Ω1 ðx BB Þ − AB ð11Þ
Debye–Hückel limiting law to include second and third virial coefficients. 8I2 2 2 I
The theoretical basis, development and formulation of the Pitzer model
for activity coefficients in aqueous electrolyte solutions have been
with:
presented in detail by Pitzer [54]; Eqs. (1)–(13) are the basic expressions
comprising the Pitzer model for the activity coefficients γM and γX of cat-
∞


ions and anions respectively. The equations given are for mixed electro- x 1 2 x −y
lyte systems, generally following the widely-used notation of Harvie Ω0 ðxÞ ¼ −1 þ ∫ y 1− exp − e dy ð12Þ
4 x0 y
and Weare [55] where possible, but including the β(2) term for ion asso-
ciation in 2–2 electrolytes following Pitzer & Mayorga [56].

 ∞



2 x 1 2 x −y x −y
lnγM ¼ zM F þ ∑ ma ð2BMa þ ZC Ma Þ þ ∑ mc 2ΦMc þ ∑ ma ψMca Ω1 ðxÞ ¼ − ∫ y 1− 1 þ e exp − e dy ð13Þ
a c a 4 x y y
ð1Þ 0
þ∑ ∑ma ma0 ψaa0 M þ jzM j∑ ∑ mc ma C ca
aba0 c a

 All Pitzer model parameters used in this work have been obtained
lnγX ¼ z2X F þ ∑ mc ð2BcX þ ZC cX Þ þ ∑ ma 2ΦXa þ ∑ mc ψXac from the literature, as outlined in Table A2. The full multicomponent
c a c
ð2Þ Pitzer model was implemented using Microsoft Excel 2007, and the
þ∑ ∑mc mc0 ψcc0 X þ jzX j∑ ∑ mc ma C ca activity coefficients for all dissolved species were then transferred nu-
cbc0 c a
merically to the HSC Chemistry software for computation of phase
M and X refer to the cationic and anionic species of interest respective- equilibria.
ly, while c and a are dummy summation indexes used to indicate sums The parameters β(0), β(1), β(2), CMX and θ are fitted for each pair of
over cations or anions. I is the ionic strength, Aϕ is the Debye–Hückel pa- ions—β and CMX parameters for interactions between oppositely-
rameter (0.392 for aqueous solutions at 25 °C), and b is an empirical pa- charged ions from single-electrolyte data, and θ and ψijk for interactions
rameter set to 1.2 kg1/2.mol−1/2 following the discussion of Pitzer [54]. between like-charged species and three-species combinations, from
F is a collection of terms that depend only on ionic strength, as mixed-electrolyte data. β(2) is set to zero for interactions involving uni-
given by Eq. (3): valent species, with α1 = 2.0. For interactions of type 2–2 (i.e., between
alkali metal cations and CO32−here), α1 =1.4 and α2 = 12 are used [54].
" pffiffi #
2 
I pffiffi 0
Reliable θ and ψijk parameter values are scarce in the literature, and
ϕ
F ¼ −A pffiffi þ ln 1 þ b I þ ∑ ∑ mc ma Bca these parameters are set to zero unless shown otherwise in Table A2.
1þb I b c a
0 0
þ ∑ ∑mc mc0 Φcc0 þ ∑ ∑ma ma0 Φaa0 ð3Þ
0 0
cbc aba
Table A2
The B and Φ terms in Eqs. (1)–(3) represent measurable combina- Pitzer ion interaction parameters used in the model, and literature sources. Where pa-
rameters are not listed, zero values are used.
tions of second virial coefficients for interactions between pairs of
unlike- and like-charged species respectively. The B terms are defined Ions Single-electrolyte parameters Reference
by Eqs. (4)–(7): β(0) β(1) β(2) CMX
 pffiffi  pffiffi Na+―OH− 0.09226 0.2424 – 0.003343 [57]
ð0 Þ ð1Þ ð2Þ
BMX ¼ βMX þ βMX g α 1 I þ βMX g α 2 I ð4Þ Na+―HCO3− − 0.5535 – 0.008285 [57]
0.05876
 pffiffi  pffiffi Na+―CO32− 0.04625 1.264 – 0.002943 [57]
0 0
0
ð1 Þ
g α1 I ð2Þ
g α2 I Ca2+―OH− −0.1421 −0.3727 – − [58]a
BMX ¼ βMX þ βMX ð5Þ 0.00092
I I
Ca2+―HCO3− 0.39975 2.9775 – – [54]
Ca2+―CO32− 0.16 2.38 −69 – [59]
where:
Mg2+―OH− – – – – b

Mg2+―HCO3− 0.033 0.84975 – – [54]

2 −y 
g ðyÞ ¼ 1−ð1 þ yÞe ð6Þ Mg2+―CO32− 0.18 2.7 −46 – [59]
y2
Mixing parameters
" ! #
0 −2 y2 −y θij Reference ψijk Reference
g ðyÞ ¼ 2 1− 1 þ y þ e ð7Þ
y 2 Na+―H+ 0.0368 [60] Na+―OH− −0.017 [54]
―CO32−
Na+―Ca2+ 0.07 [54] Na+―HCO3− 0.002 [54]
The Φ terms describe interactions between like-charged species as
―CO32−
per Eqs. (8)–(13): Na+―Mg2+ 0.07 [54]
Ca2+ ―H+ 0.092 [54]
E
Φij ¼ θij þ θij ðIÞ ð8Þ Ca2 + ―Mg2+ 0.007 [54]
CO3 2― OH− 0.1 [54]
0 0 CO32− −0.04 [54]
E
Φij ¼ θij ðI Þ ð9Þ ―HCO3 −
a
Reference [58] also used a β(2) parameter to describe Ca2+―OH−interactions in
The unsymmetrical mixing terms EθΑΒ are defined for species A very dilute solutions, but stated that more than 99% of the variance in their regression
and B, with chargespof analysis was fitted by the β(0) parameter alone, and did not specify the value of α2 used
ffiffi the same sign but different magnitude, in for this 2:1 electrolyte. Thus, only the β(0), β(1) and CMX parameters are used here.
terms of xij ¼ 6zi zj Aϕ I: b
Most studies in the literature use zero parameters for this interaction; some instead
represent an MgOH+ ion pair interacting with anions, but parameters for the interaction
E zA zB  1 1

of this complex with carbonates and bicarbonates are unavailable, while the interactions
θE ¼ Ω0 ðxAB Þ− Ω0 ðxAA Þ− Ω0 ðxBB Þ ð10Þ
4I 2 2 of Mg2+ with these anions are better-described, and so this approach is followed here.
1326 S.A. Bernal et al. / Cement and Concrete Research 42 (2012) 1317–1326
References
[1] D.W.S. Ho, R.K. Lewis, Carbonation of Concrete Incorporating Fly Ash or a Chemical
Admixture, in: ACI SP79, Fly Ash, Silica Fume, Slag & Other Mineral By-Products in
Concrete I, American Concrete Institute, Detroit, MI, 1983, pp. 333–346.
[2] D.W.S. Ho, R.K. Lewis, Carbonation of concrete and its prediction, Cem. Concr. Res.
17 (3) (1987) 489–504.
[3] M. Castellote, C. Andrade, Modelling the carbonation of cementitious matrixes by
means of the unreacted-core model, UR-CORE, Cem. Concr. Res. 38 (12) (2008)
1374–1384.
[4] F.G. da Silva, P. Helene, P. Castro-Borges, J.B.L. Liborio, Sources of variations when
comparing concrete carbonation results, J. Mater. Civ. Eng. 21 (7) (2009) 333–342.
[5] V.G. Papadakis, C.G. Vayenas, M.N. Fardis, A reaction engineering approach to the
problem of concrete carbonation, AICHE J. 35 (10) (1989) 1639–1650.
[6] T. Bakharev, J.G. Sanjayan, Y.B. Cheng, Resistance of alkali-activated slag concrete
to carbonation, Cem. Concr. Res. 31 (9) (2001) 1277–1283.
[7] M. Criado, A. Palomo, A. Fernández-Jiménez, Alkali activation of fly ashes. Part 1:
effect of curing conditions on the carbonation of the reaction products, Fuel 84
(16) (2005) 2048–2054.
[8] J. Deja, Carbonation aspects of alkali activated slag mortars and concretes, Silic.
Industr. 67 (1) (2002) 37–42.
[9] S.A. Bernal, R. Mejía de Gutierrez, V. Rose, J.L. Provis, Effect of silicate modulus and
metakaolin incorporation on the carbonation of alkali silicate-activated slags,
Cem. Concr. Res. 40 (6) (2010) 898–907.
[10] F. Puertas, M. Palacios, T. Vázquez, Carbonation process of alkali-activated slag
mortars, J. Mater. Sci. 41 (2006) 3071–3082.
[11] S.A. Bernal, R. Mejía de Gutierrez, A.L. Pedraza, J.L. Provis, E.D. Rodríguez, S.
Delvasto, Effect of binder content on the performance of alkali-activated slag con-
cretes, Cem. Concr. Res. 41 (1) (2011) 1–8.
[12] S.A. Bernal, J.L. Provis, R. Mejía de Gutiérrez, J.S.J. van Deventer (submitted for
publication). Accelerated carbonation testing of alkali-activated slag/metakaolin
blended concretes: effect of exposure conditions, Cem. Concr. Res.
[13] J. Małolepszy, J. Deja, W. Brylicki, Industrial application of slag alkaline concretes, in:
P.V. Krivenko (Ed.), Proceedings of the First International Conference on Alkaline
Cements and Concretes, VIPOL Stock Company, Kiev, Ukraine, 1994, pp. 989–1001.
[14] C. Shi, P.V. Krivenko, D.M. Roy, Alkali-Activated Cements and Concretes, Taylor &
Francis, Abingdon, UK, 2006.
[15] J.S.J. van Deventer, J.L. Provis, P. Duxson, Technical and commercial progress in
the adoption of geopolymer cement, Miner. Eng. 29 (2012) 89–104.
[16] J.S.J. van Deventer, J.L. Provis, P. Duxson, D.G. Brice, Chemical research and climate
change as drivers in the commercial adoption of alkali activated materials, Waste
Biomass Valoriz. 1 (1) (2010) 145–155.
[17] M. Castellote, L. Fernandez, C. Andrade, C. Alonso, Chemical changes and phase
analysis of OPC pastes carbonated at different CO2 concentrations, Mater. Struct.
42 (4) (2009) 515–525.
[18] B. Bary, A. Sellier, Coupled moisture—carbon dioxide–calcium transfer model for
carbonation of concrete, Cem. Concr. Res. 34 (10) (2004) 1859–1872.
[19] J.L. Provis, J.S.J. van Deventer (Eds.), Geopolymers: Structure, Processing, Proper-
ties and Industrial Applications, Woodhead, Cambridge, UK, 2009, (448 pp.).
[20] A. Gruskovnjak, B. Lothenbach, L. Holzer, R. Figi, F. Winnefeld, Hydration of
alkali-activated slag: comparison with ordinary Portland cement, Adv. Cem. Res.
18 (3) (2006) 119–128.
[21] R.R. Lloyd, J.L. Provis, J.S.J. van Deventer, Pore solution composition and alkali dif-
fusion in inorganic polymer cement, Cem. Concr. Res. 40 (9) (2010) 1386–1392.
[22] S. Song, H.M. Jennings, Pore solution chemistry of alkali-activated ground granu-
lated blast-furnace slag, Cem. Concr. Res. 29 (1999) 159–170.
[23] F. Puertas, A. Fernández-Jiménez, M.T. Blanco-Varela, Pore solution in
alkali-activated slag cement pastes. Relation to the composition and structure
of calcium silicate hydrate, Cem. Concr. Res. 34 (2004) 139–148.
[24] J.W. Payne, B.F. Dodge, Rate of absorption of carbon dioxide in water and in alka-
line media, Ind. Eng. Chem. 24 (6) (1932) 630–637.
[25] B. Huet, V. L'Hostis, L. Tricheux, H. Idrissi, Influence of alkali, silicate, and sulfate
content of carbonated concrete pore solution on mild steel corrosion behavior,
Mater. Corros. 61 (2) (2010) 111–124.
[26] E. Rodríguez, S. Bernal, R. Mejía de Gutierrez, F. Puertas, Alternative concrete
based on alkali-activated slag, Mater. Constr. 58 (291) (2008) 53–67.
[27] L. Haselbach, Potential for carbon dioxide absorption in concrete, J. Environ. Eng.
135 (6) (2009) 465–472.
[28] L. Black, C. Breen, J. Yarwood, K. Garbev, P. Stemmerman, B. Gasharova, Structural
features of C-S-H(I) and its carbonation in air — a Raman spectroscopic study.
Part II: Carbonated phases, J. Am. Ceram. Soc. 90 (3) (2007) 908–917.
[29] E.T. Stepkowska, J.L. Pérez-Rodríguez, M.J. Sayagués, J.M. Martínez-Blanes, Calcite,
vaterite and aragonite forming on cement hydration from liquid and gaseous
phase, J. Thermal. Anal. Calorim. 73 (1) (2003) 247–269.
[30] Y. Dai, J.E. Post, Crystal structure of hillebrandite: a natural analogue of calcium sili-
cate hydrate (CSH) phase in Portland cement, Am. Mineral. 80 (1995) 841–844.
[31] M. Ben Haha, B. Lothenbach, G. Le Saout, F. Winnefeld, Influence of slag chemistry
on the hydration of alkali-activated blast-furnace slag — Part I: effect of MgO,
Cem. Concr. Res. 41 (9) (2011) 955–963.
[32] A. Fernández-Jiménez, F. Puertas, Structure of calcium silicate hydrates formed in
alkaline-activated slag: influence of the type of alkaline activator, J. Am. Ceram.
Soc. 86 (8) (2003) 1389–1394.
[33] S.D. Wang, K.L. Scrivener, Hydration products of alkali-activated slag cement,
Cem. Concr. Res. 25 (3) (1995) 561–571.
[34] I.G. Richardson, A.R. Brough, G.W. Groves, C.M. Dobson, The characterization of
hardened alkali-activated blast-furnace slag pastes and the nature of the calcium
silicate hydrate (C―S―H) paste, Cem. Concr. Res. 24 (5) (1994) 813–829.
[35] Ş. Marincea, E. Bilal, J. Verkaeren, M.-L. Pascal, M. Fonteilles, Superposed parageneses
in the spurrite-, tilleyite-and gehlenite-bearing skarns from Cornet hill, Apuseni
mountains, Romania, Can. Mineral. 39 (5) (2001) 1435–1453.
[36] U. Wirsching, Experiments on the hydrothermal formation of calcium zeolites,
Clays Clay Miner. 29 (3) (1981) 171–183.
[37] T. Kol'tsova, CaAl2Si2O8-B2Al2Si2O8-H2O (B = Na, K) zeolites, Inorg. Mater. 46 (8)
(2010) 870–877.
[38] J. Corvisier, F. Brunet, A. Fabbri, S. Bernard, N. Findling, G. Rimmelé, V. Barlet-
Gouédard, O. Beyssac, B. Goffé, Raman mapping and numerical simulation of calcium
carbonates distribution in experimentally carbonated Portland-cement cores, Eur. J.
Mineral. 22 (1) (2010) 63–74.
[39] S.A. Bernal, B. Walkley, J.L. Provis, D.G. Brice, J.G. Gehman, A. Kilcullen, P. Duxson, J.S.J.
van Deventer (in preparation). Accelerated carbonation testing of alkali-activated
slag binders: Nanostructural and microstructural effects.
[40] A. Sarkar, S. Mahapatra, Synthesis of all crystalline phases of anhydrous calcium
carbonate, Cryst. Growth Des. 10 (5) (2010) 2129–2135.
[41] Institute of Experimental Mineralogy of the Russian Academy of Sciences, WWW-
MINCRYST: Crystallographic and Crystallochemical Database for Minerals and their
Structural Analogues, 2011. http://database.iem.ac.ru/mincryst/index.php.
[42] J.L. Provis, R.J. Myers, C.E. White, V. Rose, J.S.J. van Deventer, X-ray microtomography
shows pore structure and tortuosity in alkali-activated binders, Cem. Concr. Res. 42
(6) (2012) 855–864.
[43] C.R. Bury, R. Redd, The system sodium carbonate–calcium carbonate–water, J.
Chem. Soc. (1933) 1160–1162.
[44] H.P. Eugster, G.I. Smith, Mineral equilibria in the Searles Lake evaporites, California, J.
Petrol. 6 (3) (1965) 473–522.
[45] J.L. Bischoff, D.B. Herbst, R.J. Rosenbauer, Gaylussite formation at Mono lake,
California, Geochim. Cosmochim. Acta 55 (6) (1991) 1743–1747.
[46] H.P. Eugster, Sodium carbonate–bicarbonate minerals as indicators of PCO2, J. Geophys.
Res. 71 (14) (1966) 3369–3377.
[47] E. Jagniecki, T.K. Lowenstein, D. Jenkins, Sodium carbonates: temperature and
pCO2 indicators for ancient and modern alkaline saline lakes, in: 2010 Geological
Society of America Denver Annual Meeting, Denver, CO, Geological Society of
America Abstracts with Programs, vol. 42 (No. 5), 2010, p. 404.
[48] A. Roine, HSC Chemistry 4.1, Outokumpu Research Oy, Pori, Finland, 1999.
[49] E. Königsberger, L.C. Königsberger, H. Gamsjäger, Low-temperature thermodynamic
model for the system Na2CO3―MgCO3―CaCO3―H2O, Geochim. Cosmochim. Acta
63 (19–20) (1999) 3105–3119.
[50] European Committee for Standardization (CEN), Products and systems for the protec-
tion and repair of concrete structures – test methods, Determination of the carbonation
depth in a hardened concrete through the phenolphthalein method (EN 14630),
Brussels, Belgium, 2007.
[51] J. Johnston, The determination of carbonic acid, combined and free, in solution,
particularly in natural waters, J. Am. Chem. Soc. 38 (5) (1916) 947–975.
[52] D.T. Wright, D. Wacey, Sedimentary dolomite: a reality check, in: C.J.R. Braithwaite,
G. Rizzi, G. Darke (Eds.), The Geometry and Petrogenesis of Dolomite Hydrocarbon
Reservoirs, Geol. Soc. Spec. Publ., vol. 235, 2004, pp. 65–74, London.
[53] E.J. Reardon, An ion interaction model for the determination of chemical equilibria in
cement/water systems, Cem. Concr. Res. 20 (2) (1990) 175–192.
[54] K.S. Pitzer, Ion interaction approach: theory and data correlation, in: K.S. Pitzer
(Ed.), Activity Coefficients in Electrolyte Solutions, CRC Press, Boca Raton, 1991,
pp. 75–153.
[55] C.E. Harvie, J.H. Weare, The prediction of mineral solubilities in natural waters:
the Na―K―Mg―Ca―Cl―SO4―H2O system from zero to high concentration at
25°C, Geochim. Cosmochim. Acta 44 (7) (1980) 981–997.
[56] K.S. Pitzer, G. Mayorga, Thermodynamics of electrolytes. III. Activity and osmotic
coefficients for 2-2 electrolytes, J. Solut. Chem. 3 (7) (1974) 539–546.
[57] P.M. May, D. Rowland, G. Hefter, E. Königsberger, A generic and updatable Pitzer
characterization of aqueous binary electrolyte solutions at 1 bar and 25 °C, J.
Chem. Eng. Data 56 (12) (2011) 5066–5077.
[58] J. Duchesne, E.J. Reardon, Measurement and prediction of portlandite solubility in
alkali solutions, Cem. Concr. Res. 25 (5) (1995) 1043–1053.
[59] G.M. Rosenblatt, Estimation of activity coefficients in concentrated sulfite–sulfate
solutions, AIChE J. 27 (4) (1981) 619–626.
[60] H.-T. Kim, W.J. Frederick, Evaluation of Pitzer ion interaction parameters of aque-
ous electrolytes at 25°C. 2. Ternary mixing parameters, J. Chem. Eng. Data 33 (3)
(1988) 278–283.