Cement and Concrete Research 126 (2019) 105918

Contents lists available at ScienceDirect

Cement and Concrete Research

journal homepage: www.elsevier.com/locate/cemconres

The influence of the filler effect on the sulfate requirement of blended T

cements
Franco Zunino , Karen Scrivener
⁎

Laboratory of Construction Materials, EPFL STI IMX LMC, École Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland

ARTICLE INFO ABSTRACT

Keywords: In some blended cements, the optimum amount of sulfate addition differs from that observed in OPC. This study
A Hydration aims to understand the mechanism behind the impact of two SCMs, namely metakaolin and limestone, on the
A Kinetics sulfate balance of blended cements. No significant impact of the aluminum content of these SCMs was observed.
B Calcium-Silicate-Hydrate (C-S-H) Instead, it is observed that the filler effect of the SCM, which accelerates the reaction of alite, is the main factor
D Sulfate
impacting the sulfate balance. As the rate of precipitation of C-S-H is increased, more sulfate is adsorbed by the
D Metakaolin
C-S-H and consequently, the depletion of gypsum is reached earlier in time during the hydration process. A
relationship between heat release at the onset of the aluminate peak and the gypsum content of the system was
established.

1. Introduction reported in blended mixtures with fly ash [6,7].

The main purpose of adding gypsum to OPC and blended cements is
Calcium sulfate is added to clinker, during grinding, to control the to control the reaction of C3A and thus to prevent flash setting [8]. In
reaction of the aluminate phase. This addition is usually referred to as the presence of gypsum, the hydration reaction of C3A shifts from the
gypsum (Calcium sulfate dihydrate, CaSO4·2H2O) although anhydrite precipitation of calcium aluminate hydrates and ettringite is formed
and hemihydrates (plaster) may also be present. The optimum sulfate [9]. As more gypsum is incorporated in the system, more retardation of
addition is that which gives the greatest strength Ordinary Portland the C3A reaction can be achieved [10]. Usually, the adjustment of
cement (OPC) (at 1 or 28 days). Nowadays, blended cements are more gypsum content is made based on the measurement of compressive
common than OPC. These contain supplementary cementitious mate- strength at 24 h in mortar specimens or by isothermal calorimetry [11].
rials (SCMs) such as fine limestone, granulated ground blast furnace However, the level of gypsum required to prevent flash setting is
slag, fly ash, calcined clay, natural pozzolans and silica fume, in addi- usually less than the content leading to the highest 24-h strength [8].
tion to clinker and gypsum. Climate change and the push for a more The mechanism by which gypsum slows down the reaction of C3A
sustainable concrete industry have accelerated the adoption of SCMs as has been extensively studied, but is still a matter of debate. Originally it
a means to reduce the clinker factor and consequently, the carbon was suggested that a diffusion barrier is formed of ettringite or CaeAl
footprint of cement manufacture [1]. hydrates that retards the dissolution of C3A [12,13], but recent work
Limestone calcined clay cements (LC3) are one of the promising shows that there is no diffusion barrier and it is more tenable that the
alternatives for high performance sustainable cements [2]. LC3 in- adsorption of sulfate (or CaeS complexes) on C3A active sites slows
corporates significant amounts of metakaolin, a reactive aluminosili- down the C3A dissolution [14–16].
cate phase [3]. When made in the laboratory, by mixing OPC with Quennoz et al. observed significant differences in the time of
calcined clay and limestone, it has been observed that LC3 requires gypsum depletion between pure C3A and C3S/C3A systems, indicating
additional gypsum on top of that contained in the OPC to obtain the that C-S-H influences the availability of sulfate in these systems. Their
highest compressive strength at 24 h [4,5]. If the system is not adjusted results showed that the S/Ca ratio in C-S-H, measured by SEM-EDS, is
for sulfate, the aluminate peak may occur before the alite peak, which considerably higher (between 0.04 and 0.1) before the aluminate peak
consequently becomes smaller and broader. The incorporation of ad- in C3S/C3A mixtures. After the aluminate peak, the concentration of
ditional gypsum shifts the aluminate peak to later times and the alite sulfate in C-S-H decreases to about one fourth of the value observed
peak recovers. The need for sulfate content adjustment has also been initially [17]. Similar observations were reported by Berodier and

⁎
Corresponding author.
E-mail address: franco.zunino@epfl.ch (F. Zunino).

https://doi.org/10.1016/j.cemconres.2019.105918
Received 7 August 2019; Received in revised form 30 September 2019; Accepted 4 October 2019
Available online 03 November 2019
0008-8846/ © 2019 Elsevier Ltd. All rights reserved.
F. Zunino and K. Scrivener Cement and Concrete Research 126 (2019) 105918

Scrivener in OPC blended with quartz [18], who also observed a change
in C-S-H morphology due to sulfate adsorption. Both results are also in
good agreement with the measurements by STEM-EDS reported by
Gallucci et al. [19]. These observations show that sulfate not only in-
teracts with C3A during the acceleration period but also with C-S-H.
Despite the primary effect of gypsum addition being to modify the
reaction of C3A, no clear relationship between the content of aluminates
phases and the gypsum requirement for proper sulfation of a system has
been established. A comprehensive approach to sulfate balance that
accounts for the adsorption of sulfate by C-S-H during hydration is re-
quired to better understand the hydration kinetics of blended cements.
Furthermore, the contribution of the aluminate phases from SCMs
themselves, such as metakaolin, to the sulfate amount required to en-
sure a properly sulfated system needs to be better understood. This
paper looks at the effect of a variety of SCMs on the sulfate balance.

2. Materials and methods

2.1. Materials

A commercial ordinary Portland cement (OPC) conforming to EN

197-1 as CEM I 42.5R was used in this study. The chemical (by XRF)
and phase (by XRD/Rietveld) composition is shown in Table 1. The
calcined clay was a high purity metakaolin (MK, 95% purity from Fig. 1. Particle size distributions of materials used in this study.
Burgess) to avoid side effects from secondary materials. Two limestone
powders (calcite) with different particle sizes were also used in this Table 2
study: the coarser material is noted as cLS and the finer one as fLS. In Distribution values, span (width), specific surface area and specific gravity of
both cases, the content of calcium carbonate was above 98% as mea- raw materials.
sured by thermogravimetric analysis (TGA). The particle size distribu-
OPC MK cLS fLS
tion (PSD) of these materials, measured by laser diffraction, is shown in
Fig. 1. The optical model parameters and the dispersant were selected DV90 (μm) 41.42 20.17 7.93 6.72
following the recommendations given in [20]. Further details can be DV50 (μm) 14.22 5.13 3.93 1.77
found in Table 2. DV10 (μm) 1.67 0.54 1.80 0.24
Span (−) 2.80 3.83 1.56 3.66
The specific surface area (SSA) of each raw material was measured
SSA (m2/g) 1.41 13.56 3.60 8.96
by nitrogen adsorption, using the BET method. In all cases, samples of Sp. Gravity (g/cm3) 3.09 2.20 2.72 2.72
around 1.5 g were degassed for 2 h at 200 °C under a N2 flux before the
measurement. The specific gravity was measured using a liquid pycn-
ometer with isopropanol as solvent. incorporate metakaolin. The reference LC3 system contained 55% OPC,
15% limestone (cLS) and 30% metakaolin (LC3-50) on a mass basis plus
2.2. Mixture design different amounts of added gypsum (see Table 3). Different grades of
clay (metakaolin content) were simulated by combining the 95% me-
The influence of the additional SSA introduced by the incorporation takaolin material with limestone, as detailed in Table 3. Following this
of SCMs (filler effect) on the sulfate balance of the system was com- approach, LC3-50 systems with clay grades ranging from 95% to 20%
pared to the effect of the additional aluminates sourced from the ad- were prepared. From this dataset, the systems with 95%, 63% and 40%
dition of metakaolin. For this purpose, OPC blends with different were selected for further study with different gypsum contents. Based
limestone replacements were compared to LC3 systems that additionally on the computed SSA for each LC3 system, an equivalent blend of OPC
and fine limestone (fLS) was prepared, with a proportion intended to
Table 1 replicate the same binder surface area of each LC3 system. As these
Chemical (from XRF) and phase composition (from XRD) of OPC, metakaolin systems do not contain any additional aluminates, the effect of the
and LS. additional aluminates sourced from metakaolin can be decoupled from
OPC Metakaolin LS the contribution of the filler effect alone. The details of the mixtures are
shown in Table 3. The water-to-binder ratio (w/b) was fixed at 0.4 by
SiO2 19.511 52.00 0.11 mass, for all systems.
Al2O3 4.42 43.80 0.00
Fe2O3 3.12 0.33 0.04
CaO 63.85 0.03 54.96
Na2O 0.19 0.14 0.06 2.3. Experimental methods
K2O 0.83 0.29 0.00
MnO 0.05 0.01 0.00 The heat evolution was measured in a TAM Air isothermal calori-
TiO2 0.31 1.53 0.00
meter at 20 °C for up to 7 days. Paste samples were mixed with a high
MgO 2.10 0.01 0.15
P2O5 0.33 0.16 0.00 shear mixer at 1600 rpm for 2 min, 10 g of paste was placed in a glass
SO3 3.25 0.10 0.03 ampoule, sealed and introduced in the calorimeter. The onset of the
LOI 1.54 1.47 42.5 aluminate peak was identified using the heat flow derivative, as shown
C3S 66.5 – – in Fig. 2 left.
C2S 4.0 – –
To separate the heat flow from the aluminate reaction, the heat
C3A 4.9 – –
C4AF 9.6 – – evolution from the same systems were measured, but with an extra 5%
of gypsum to retard the aluminate peak and give the full shape of the

2
F. Zunino and K. Scrivener Cement and Concrete Research 126 (2019) 105918

Table 3
Mixture proportions of OPC, LC3-50 and OPC + fLS systems.
Mixture OPC (%) MK (%) LS (%) fLS (%) Added gypsum (%) Added SO3 (%) Total SO3 (%)b w/c w/b SSA (m2/g)

OPC 100 0.0 0.0 0.0 0.0 0.0 2.45 0.40 0.40 1.41
LC3-50 MK95 55-G 30.0 15.0 0.0 G (0 to 7) 0.0 to 3.26 1.34 to 4.43 0.73a 0.40 5.36
LC3-50 MK71 55-G 22.5 22.5 0.0 G (1.2) 0.56 1.87 0.73a 0.40 4.62
LC3-50 MK63 55-G 20.0 25.0 0.0 G (0, 0.5, 1, 1.5) 0.0 to 0.70 1.34 to 2.01 0.73a 0.40 4.37
LC3-50 MK47 55-G 15.0 30.0 0.0 G (0.85) 0.40 1.72 0.73a 0.40 3.88
LC3-50 MK40 55-G 12.6 32.4 0.0 G (0, 0.4, 0.8, 1) 0.0 to 0.47 1.34 to 1.79 0.73a 0.40 3.64
LC3-50 MK20 55-G 6.3 38.7 0.0 G (0.35) 0.16 1.50 0.73a 0.40 3.02
OPC-fLS SSA MK95 47.7 0.0 0.0 52.3 0.0 0.0 1.17 0.84 0.40 5.36
OPC-fLS SSA MK63 60.8 0.0 0.0 39.2 0.0 0.0 1.49 0.66 0.40 4.37
OPC-fLS SSA MK40 70.5 0.0 0.0 29.5 0.0 0.0 1.72 0.57 0.40 3.64

a
Value of w/c computed considering an OPC content of 55%. The value will vary as the concentration of added gypsum is increased.
b
Computed considering the SO3 content of OPC measured as calcium sulfate minerals by QXRD and the OPC fractions given in this table.

alite peak, which was then subtracted from the overall curves, as shown
in Fig. 2 right for an LC3 system. The area associated with the aluminate
peak was computed as the difference in total heat of the original mix-
ture and that of the over sulfated system, between the maximum of the
alite peak and 24 h, as this was the time when the ettringite content was
measured by XRD. Furthermore, in-situ XRD measurements showed
that in all cases, the maximum amount of ettringite formed was reached
before this time.
X-ray diffraction (XRD) measurements were carried out on freshly
cut slices of hardened paste at 24 h of hydration to determine the
amounts of ettringite and AFm formed after the aluminate peak.
Rietveld refinement was conducted using the HighScore Plus v4.8
software. The slices were analyzed in a Bragg-Brentano configuration in
a PANalytical X'pert pro diffractometer working at 45 kV and 40 mA
with a copper source. A 1/2° soller slit was used, and scans were ac-
quired between 7 and 70° 2θ in 14 min, equivalent to a step size of
0.0167° 2θ. The external standard method was used to compute the K-
factor of the device and account for the amorphous phases present [21].
A scan of a rutile standard was acquired in the same conditions after the
Fig. 3. Heat flow curve of OPC highlighting the main hydration events during
experiments for this purpose. the first 24 h.
For in-situ XRD experiments, paste specimens were prepared in the
same manner as the isothermal calorimetry samples. In this case, a 1°
3. Results and discussion
soller slit was used. The 14 min scans were collected every 30 min up to
48 h of hydration. The paste samples were mounted on a disc-shaped
3.1. Early hydration of Portland cement
sample holder, and covered with a 12.7 μm thick Kapton film that was
secured on top with a ring. The temperature was controlled at the base
The heat flow curve for the OPC used in this study is shown in Fig. 3.
of the sample holder with a Peltier stage and kept constant at 20 °C.
The main kinetic events related to the alite and aluminate reactions are
highlighted. Previous studies have shown that C3S dissolves quickly

Fig. 2. Identification of the aluminate peak onset using the derivative of the heat flow (left) and decoupling procedure applied to compute the aluminate peak area,
shown as an example in the case of LC3-50 MK95 (right).

3
F. Zunino and K. Scrivener Cement and Concrete Research 126 (2019) 105918

Fig. 4. Heat flow of LC3-50 MK95 systems with different amounts of gypsum Fig. 5. Heat flow of LC3-50 blends incorporating fine calcined clay with 50%
addition. metakaolin and the MK material (95% metakaolin).

during the acceleration period of the alite reaction. In parallel, during aluminate from the metakaolin does not explain the observed changes
the alite peak, ettringite is precipitated continuously and consequently in the aluminate peak.
the gypsum content decreases. In addition, sulfate becomes adsorbed on The system which has the higher SSA clay (50% metakaolin con-
C-S-H [17]. When the solid gypsum is depleted from the system, there is tent) shows more acceleration and enhancement of the alite peak, re-
an increase in slope of the heat evolution. This leads to the second lated to an increased filler effect of the clay fraction. Simultaneously,
(aluminate) peak, which is linked to an accelerated formation of et- this system exhibits more acceleration of the aluminate reaction. This
tringite during this period (second ettringite formation as described in suggests that the filler effect in LC3 due to the incorporation of fine clay
[17,22]), associated with a higher rate of C3A dissolution. leading to the acceleration of the C3S reaction, might be linked to the
acceleration of the aluminate peak (i.e., gypsum depletion).
3.2. Early-age hydration in LC3 systems Furthermore, this suggests that the degree of reaction of C3S (and
consequently, of C-S-H precipitation) and the onset time of the alumi-
Fig. 4 shows the calorimetry results for LC3-50 MK95 systems nate peak are related.
compared to OPC, normalized per gram of clinker. A significant ac-
celeration (time shift) and enhancement (height) of the aluminate re- 3.3. Early-age hydration in OPC + limestone systems
action is observed in all LC3 systems as compared to the reference OPC.
For the system without additional gypsum added (LC3-50 0% Add. To decouple the possible contribution of the additional aluminates
Gypsum), it is observed that the enhanced aluminate peak is almost phases from the filler effect, additional experiments were carried out in
super imposed on the alite peak, leading to a situation close to under- systems of OPC with fine limestone. In these cases, there are no addi-
sulfation. tional aluminate phases in the system. As shown in Table 3, the SSA of
The addition of 1 to 2% of additional gypsum leads to a situation these systems was matched to each of the reference LC3–50 systems by
where the aluminate peak is delayed and the alite and aluminate peaks varying the OPC and fLS proportions. The isothermal calorimetry re-
are separated similar to the pattern observed in OPC. As the aluminate sults are shown in Fig. 6. As observed, the same acceleration and en-
peak shifts to the right, it becomes lower and broader. hancement of the aluminate peak seen in LC3 is observed in these
In addition, a strong acceleration and enhancement of the alite systems. Moreover, the degree of acceleration of the aluminate reaction
(first) peak is also observed in all LC3 systems, which is linked to the is related to the observed enhancement of the alite peak, in agreement
increased shear rate between particles and the surface provided by the with the results on LC3-50 blends shown in Fig. 5.
addition of limestone and metakaolin [18]. The filler contribution of The effect on the aluminate peak has been observed in
limestone and metakaolin in LC3 appears to be independent of the OPC + limestone and OPC + quartz systems [24] and those with silica
gypsum addition. Other LC3 systems studied exhibit the same behavior fume [25]. Thus, the acceleration of the aluminate peak appears to be
as the LC3-50 MK95 systems shown in Fig. 4 (Appendix B). linked to the amount of surface area supplied to the system, rather than
The incorporation of metakaolin provides another source of alu- the aluminum content of the added material.
minum to the system in addition to C3A. In previous studies, it has been
suggested that the intensity of the aluminate peak in LC3 is linked to the 3.4. Mechanism relating the rate of C3S reaction and the occurrence of the
amount of metakaolin in the system [23]. However, this correlation is aluminate peak
not systematic as seen in Fig. 5. The calorimetry curves for LC3 systems
with pure (95%) metakaolin and a calcined clay with only 50% meta- The results indicate that the effects of SCMs on the aluminate peak
kaolin content are compared. By deconvoluting the calorimetry curves, are mostly explained by the specific surface area of the material, as seen
it was seen that the total heat associated with the aluminate peak differs in the results for OPC + fLS, rather than its chemical composition
by less than 5% between the two systems, and is actually slightly higher (aluminum content). As the filler effect influences hydration by en-
in the system with lower metakaolin content. This further indicates that hancing/accelerating the hydration of C3S and consequently, the rate of

4
F. Zunino and K. Scrivener Cement and Concrete Research 126 (2019) 105918

Fig. 6. Heat flow of OPC + fLS systems with matched SSA to the different LC3- Fig. 7. Heat released at the aluminate peak (A.P.) onset versus gypsum content
50 blends studied. of the system. P-value of regression 5.89·10−14.

C-S-H precipitation [18], a relationship between C-S-H precipitation

rate and sulfate depletion can be inferred.
The adsorption of sulfate by C-S-H is proportional to the sulfate
concentration in solution [26]; up to the depletion of solid gypsum in
the system this is high. Quennoz [17] and Berodier [27] showed that a
significant amount of sulfate can be detected by microanalysis in the C-
S-H. They reported S/Ca ratios ranging from 0.04 to 0.1 prior to the
aluminate peak. While neither of these studies was able to measure the
sulfate content in C-S-H before 10 h due to the low quantity of hydra-
tion products available, both clearly show that the S/Ca ratio of C-S-H
remains high until the onset of the aluminate peak.
The acceleration of the silicate reaction leads to a faster precipita-
tion of C-S-H, more C-S-H formation implies more sulfate adsorbed, and
therefore, the earlier the gypsum depletion point will be reached. The
aluminate peak is triggered by the depletion of sulfate from the solu-
tion, which drops quickly (by about 20 times) after the solid gypsum is
consumed. Before gypsum depletion, the sulfate concertation is de-
termined by the gypsum solubility, while after no gypsum or other
sulfate sources are available, the sulfate concentration is limited by its
equilibrium with ettringite [28].
The drop in the sulfate concentration in solution causes an accel-
eration of the C3A dissolution while simultaneously sulfate ions are
desorbed from the C-S-H, becoming available to react during a second Fig. 8. In-situ XRD results showing the content of the main phases participating
formation of ettringite (aluminate peak). in the reaction during the first 48 h of hydration for LC3-50 MK-95. Solid line
The total heat released during the acceleration period, before indicates the associated heat flow curve, and dashed lines indicate the alumi-
nate peak onset and provide an eye-guide for the change in ettringite pre-
gypsum depletion, can be used as an estimate of the amount of C-S-H
cipitation rate.
precipitated. Fig. 7 shows the relationship between heat release at the
time of the aluminate peak onset and the SO3 content for all the systems
studied; this shows a clear linear relationship as expected. 3.5. Hydration products of the aluminate reaction in LC3
Another aspect to notice is that the regression shown in Fig. 7
crosses the SO3 axis at a positive value. This indicates that there is some The in-situ XRD measurements, Fig. 8 for the LC3-50 MK-95 system
SO3 consumption, in addition to the sorption on the C-S-H. This can be (see Appendix A for OPC and fLS systems), indicate that the main phase
attributed to the sulfate associated with the ettringite precipitation formed during the aluminate peak is ettringite, in agreement with
before the aluminate peak, as observed in the in-situ XRD experiments. previous studies on OPC [18,29]. It can be observed that ettringite is
The total heat shown in Fig. 7 also includes this heat associated with detected from the beginning of the reaction. It has been previously
ettringite precipitation before the aluminate peak. This amount is si- observed that the amount of metakaolin does not influence the et-
milar among the different types of systems studied, as seen in the in-situ tringite content in LC3 systems, being rather determined by the added
results (Fig. 8 and Appendix A). sulfate content [23]. After the maximum amount of ettringite is
reached, hemicarboaluminate (Hc) is formed.
The deconvolution of the calorimetry curves was done in all systems

5
F. Zunino and K. Scrivener Cement and Concrete Research 126 (2019) 105918

Fig. 9. XRD patterns at 24 h of hydration collected from fresh slices. Ett = ettringite; Hc = hemicarboaluminate; Mc = monocarboaluminate; F = ferrite.

in order to compare the computed heat release associated with the

aluminate peak with the amount of ettringite (and AFm phases) formed
at 24 h measured on fresh slices by XRD (Fig. 9). In the case of blended
cements, the precipitation of an AFm phase (hemicarboaluminate) was
observed before 24 h. As it was not possible to isolate the heat release
associated to the formation of ettringite and AFm during this period,
they were considered together.
The heat evolved in the aluminate peak is directly related to the
amounts of aluminate phases formed, Fig. 10. Here, the ettringite pre-
cipitated during the alite peak is also included, but it is not accounted
for in the heat released during the aluminate peak. As the in-situ XRD
results show, this amount is similar among the different system studied,
and thus the heat release during the aluminate peak can still be related
to the overall ettringite precipitated.
Although the peak is enhanced even in the systems with only
limestone, where all the alumina must come from the clinker compo-
nent, the heats for the systems with metakaolin are generally higher
indicating some contribution from the MK reaction to this peak. By
comparison of Fig. 4 and Fig. 6, it can be seen that more acceleration of
the aluminate peak can be achieved with the incorporation of limestone
alone as compared to LC3 for the same amount of surface area added. Fig. 10. Ettringite (AFt) and AFm (Hc + Mc) content measured at 24 h versus
This could be explained by limestone being a better surface for C-S-H computed heat associated to the aluminate peak (A.P.). P-value of regression
nucleation [18], in agreement with the enhancements observed in the 8.67·10−8.
alite peaks in both systems. On the other hand, more ettringite is
formed in the LC3 system, even considering that the second formation
sustained by the aluminates from metakaolin well beyond the max-
of ettringite occurs later. In addition, the replacement level of OPC by
imum availability of C3A.
fLS are lower compared to LC3-50, thus, more C3A is available. As
Fig. 11 shows schematically the mechanisms behind the sulfate
aluminates are required to form ettringite, these can be partially
balance in OPC and blended cements, considering the contribution of
sourced from metakaolin. Further experiments on LC3 with high
the filler effect and the adsorption of sulfate into C-S-H.
gypsum contents showed that the precipitation of ettringite can be

6
F. Zunino and K. Scrivener Cement and Concrete Research 126 (2019) 105918

Fig. 11. Diagram of proposed mechanism involving sulfate adsorption on C-S-H during the acceleration period. Expected phase assemblage (top) and heat flow
profile (bottom).

4. Conclusions to deplete sulfate linearly increases with the initial sulfate content of
the system.
The results presented here relate the filler effects of SCMs to the
calcium sulfate addition required for proper sulfation. The adsorption 4. After gypsum is depleted, sulfate is desorbed from C-S-H following
of sulfate by C-S-H is the important factor linking these two things. the decrease in the concentration in the pore solution, and becomes
Based on the results presented, the following conclusions can be available to precipitate additional ettringite. The heat release during
drawn: the aluminate peak can be solely explained by the precipitation of
ettringite and AFm.
1. Supplementary cementitious materials can have a significant impact
on the sulfate balance of blended cements due to their filler effect. Declaration of Competing Interest
The impact is related to additional surface area, rather than the
chemical composition of the material. In particular, the aluminate The authors declare that they have no known competing financial
content of calcined clay does not influence the sulfate requirement interests or personal relationships that could have appeared to influ-
for proper sulfation. ence the work reported in this paper.

2. Sulfate ions are adsorbed in the C-S-H precipitated due to the on- Acknowledgements
going hydration of C3S. The increase in the rate of reaction of C3S
due to the filler effect accelerates the precipitation rate of C-S-H, The authors would like to acknowledge financial support by the Swiss
increasing the amount of sulfate adsorbed. The depletion of gypsum Agency of Development and Cooperation (SDC) grant 81026665. The Swiss
occurs earlier and consequently the aluminate peak shifts left. federal commission for scholarships for foreign students (FCS) is acknowl-
edged for supporting Franco Zunino's studies through scholarship
3. The amount of heat released at the onset of the aluminate peak is 2016.0719. The authors also acknowledge the advice and comments of Dr.
linearly correlated with the sulfate content of the system for a given Xuerun Li on the analysis of the in situ XRD data, and Mr. George Fairlie for
binder composition. This means that the amount of C-S-H required his help in conducting some of the experiments shown in this paper.

7
F. Zunino and K. Scrivener Cement and Concrete Research 126 (2019) 105918

Appendix A. In-situ XRD results for OPC and OPC + fLS systems studied

Fig. A.1. In-situ XRD results showing the content of the main phases participating in the reaction during the first 48 h of hydration for OPC.

Fig. A.2. In-situ XRD results showing the content of the main phases participating in the reaction during the first 48 h of hydration for OPC + fLS SSA MK95.

8
F. Zunino and K. Scrivener Cement and Concrete Research 126 (2019) 105918

Appendix B. Calorimetry curves of additional LC3 systems

Fig. B.1. Heat flow of LC3-50 MK63 systems with different amounts of gypsum addition.

Fig. B.2. Heat flow of LC3-50 MK40 systems with different amounts of gypsum addition.

References high volume fly ash mixtures using chemically inert fillers, Constr. Build. Mater.
161 (2018) 221–228, https://doi.org/10.1016/j.conbuildmat.2017.11.132.
[7] M.D. Niemuth, L. Barcelo, J. Weiss, Effect of fly ash on optimum sulfate levels
[1] B. Lothenbach, K. Scrivener, R.D. Hooton, Supplementary cementitious materials, measured using heat and strength at early ages, Adv. Civ. Eng. Mater. 1 (2012)
Cem. Concr. Res. 41 (2011) 1244–1256, https://doi.org/10.1016/j.cemconres. 20120012, , https://doi.org/10.1520/acem20120012.
2010.12.001. [8] W. Lerch, The influence of gypsum on the hydration and properties of Portland
[2] K. Scrivener, F. Martirena, S. Bishnoi, S. Maity, Calcined clay limestone cements cement pastes, Am. Soc. Test. Mater. J. 12 (1946) 1–41.
(LC3), Cem. Concr. Res. (2017) 1–8, https://doi.org/10.1016/j.cemconres.2017.08. [9] RILEM T.C., The hydration of tricalcium aluminate and tetracalcium aluminoferrite
017. in the presence of calcium sulfate, Mater. Struct. 19 (1986) 137–147, https://doi.
[3] M. Antoni, Investigation of Cement Substitution by Combined Addition of Calcined org/10.1007/BF02481758.
Clays and Limestone, École Polytechnique Fédérale de Lausanne, 2011. [10] A. Quennoz, K.L. Scrivener, Hydration of C3A–gypsum systems, Cem. Concr. Res.
[4] K.L. Scrivener, F. Avet, H. Maraghechi, F. Zunino, J. Ston, A. Favier, et al., 42 (2012) 1032–1041, https://doi.org/10.1016/j.cemconres.2012.04.005.
Impacting factors and properties of Limestone Calcined Clay Cements (LC3), Green [11] ASTM C563, Standard Guide for Approximation of Optimum SO3 in Hydraulic
Mater (2018), https://doi.org/10.1680/jgrma.18.00029. Cement, (2018), https://doi.org/10.1520/C0563-18.2.
[5] M. Antoni, J. Rossen, F. Martirena, K. Scrivener, Cement substitution by a combi- [12] W. Corstanje, H.N. Stevels, J.M. Stevels, Hydration reactions in pastes C3S
nation of metakaolin and limestone, Cem. Concr. Res. 42 (2012) 1579–1589, +C3A+CaSO4 2aq + water at 25 °C, Cem. Concr. Res. 4 (1974) 417–431.
https://doi.org/10.1016/j.cemconres.2012.09.006. [13] M. Collepardi, G. Baldini, M. Pauri, Tricalcium aluminate hydration in the presence
[6] I. De la Varga, J. Castro, D.P. Bentz, F. Zunino, J. Weiss, Evaluating the hydration of of lime, gypsum or sodium sulfate, Cem. Concr. Res. 8 (1978) 571–580.

9
F. Zunino and K. Scrivener Cement and Concrete Research 126 (2019) 105918

[14] H. Minard, S. Garrault, L. Regnaud, A. Nonat, Mechanisms and parameters con- 1017/S0885715600013026.
trolling the tricalcium aluminate reactivity in the presence of gypsum, Cem. Concr. [22] K.L. Scrivener, Development of the Microstructure During the Hydration of Portland
Res. 37 (2007) 1418–1426, https://doi.org/10.1016/j.cemconres.2007.06.001. Cement, University of London, 1984.
[15] R.J. Myers, G. Geng, J. Li, E.D. Rodríguez, J. Ha, P. Kidkhunthod, et al., Role of [23] F. Avet, K. Scrivener, Investigation of the calcined kaolinite content on the hydra-
adsorption phenomena in cubic tricalcium aluminate dissolution, Langmuir 33 tion of limestone Calcined clay cement (LC3), Cem. Concr. Res. 107 (2018)
(2016) 45–55, https://doi.org/10.1021/acs.langmuir.6b03474. 124–135, https://doi.org/10.1016/j.cemconres.2018.02.016.
[16] G. Geng, R.J. Myers, Y.-S. Yu, D.A. Shapiro, R. Winarski, P.E. Levitz, et al., [24] D.P. Bentz, C.F. Ferraris, S.Z. Jones, D. Lootens, F. Zunino, Limestone and silica
Synchrotron X-ray nanotomographic and spectromicroscopic study of the tricalcium powder replacements for cement: early-age performance, Cem. Concr. Compos. 78
aluminate hydration in the presence of gypsum, Cem. Concr. Res. 111 (2018) (2017) 43–56, https://doi.org/10.1016/j.cemconcomp.2017.01.001.
130–137, https://doi.org/10.1016/j.cemconres.2018.06.002. [25] F. Lavergne, R. Belhadi, J. Carriat, A. Ben Fraj, Effect of nano-silica particles on the
[17] A. Quennoz, K.L. Scrivener, Interactions between alite and C3A-gypsum hydrations hydration, the rheology and the strength development of a blended cement paste,
in model cements, Cem. Concr. Res. 44 (2013) 46–54, https://doi.org/10.1016/j. Cem. Concr. Compos. 95 (2019) 42–55, https://doi.org/10.1016/j.cemconcomp.
cemconres.2012.10.018. 2018.10.007.
[18] E. Berodier, K. Scrivener, Understanding the filler effect on the nucleation and [26] R. Barbarulo, H. Peycelon, S. Leclercq, Chemical equilibria between C-S-H and et-
growth of C-S-H, J. Am. Ceram. Soc. 97 (2014) 3764–3773, https://doi.org/10. tringite, at 20 and 85 °C, Cem. Concr. Res. 37 (2007) 1176–1181, https://doi.org/
1111/jace.13177. 10.1016/j.cemconres.2007.04.013.
[19] E. Gallucci, P. Mathur, K. Scrivener, Microstructural development of early age hy- [27] E. Berodier, Impact of the Supplementary Cementitious Materials on the Kinetics
dration shells around cement grains, Cem. Concr. Res. 40 (2010) 4–13, https://doi. and Microstructural Development of Cement Hydration, EPFL Thesis 6417 (2015).
org/10.1016/j.cemconres.2009.09.015. [28] A. Vollpracht, B. Lothenbach, R. Snellings, J. Haufe, The pore solution of blended
[20] K. Scrivener, R. Snellings, B. Lothenbach, A Practical Guide to Microstructural cements: a review, Mater. Struct. Constr. 49 (2016) 3341–3367, https://doi.org/10.
Analysis of Cementitious Materials, (2016), p. 540, https://doi.org/10.7693/ 1617/s11527-015-0724-1.
wl20150205. [29] A. Quennoz, Hydration of C3A With Calcium Sulfate Alone and in the Presence of
[21] B.H. O’Connor, M.D. Raven, Application of the Rietveld refinement procedure in Calcium Silicate, EPFL Thesis 5035 (2011).
assaying powdered mixtures, Powder Diffract. 3 (1988) 2–6, https://doi.org/10.

10