Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

Cement and Concrete Research 124 (2019) 105823

Contents lists available at ScienceDirect

Cement and Concrete Research


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

Advances in understanding cement hydration mechanisms T


a,⁎ a b a
Karen Scrivener , Alexandre Ouzia , Patrick Juilland , Aslam Kunhi Mohamed
a
EPFL STI IMX LMC, MXG 230, Station 12, CH-1015 Lausanne, Switzerland
b
SIKA Technology, Zürich, Switzerland

A R T I C LE I N FO A B S T R A C T

Keywords: Progress in understanding hydration mechanisms of alite and Portland cement is reviewed. Up to the end of the
Cement induction period, dissolution rates determined by the undersaturation of the solution dominate the reaction, but,
Hydration better understanding is needed about the alite solution interface. The main heat evolution peak hydration is
Induction period dominated by the growth of outer C-S-H with a spiky or “needle” like morphology. Growth is rapid over several
Main hydration peak
hours (acceleration period) and then slows (deceleration period). At later ages the consumption of water and
Later ages
lack of water filled pores above about 10 nm, along with the consumption of anhydrous material are major
factors leading to the continual reduction in the rate of reaction. There is no evidence that diffusion becomes the
rate controlling mechanism even at this stage. The microstructure of cement differs significantly from that of
alite, largely due to the influence of alumina on C-S-H growth and distribution.

1. Introduction arguments. We will try to highlight the pros and cons for each hy-
pothesis.
This paper reviews our understanding of the kinetics and mechan-
isms of cement hydration. It follows closely from papers written for the 2.1.1. Protective membrane theory
two previous Chemistry of Cement Conferences [1,2]. The main focus is This long-standing hypothesis is the oldest and raises the funda-
on the hydration of the tricalcium silicate phase, which dominates the mental question: “What is the state of the solid-water interface between
hydration process, especially at early ages. Nevertheless, there are im- the C3S surface and its surrounding pore solution?” The main argument
portant differences between alite and Portland cement, particularly in in support of the protective membrane theory is the indirect thermo-
terms of microstructural development, which are discussed. dynamic consideration: to reconcile the apparent low reactivity ob-
served experimentally during the induction period with the high C3S
2. Hydration of C3S/alite solubility calculated from its enthalpy of formation [3–10]. Based on
this discrepancy, several authors have proposed that a metastable C-S-H
The mechanisms governing the hydration of calcium silicate are hydrate forms which covers the reacting surface preventing extremely
usually discussed in terms of the heat evolution curve, Fig. 1. Although fast dissolution that is otherwise expected.
this curve has been divided into up to 7 regions, it is more straight- The seminal paper of Gartner and Jennings [11] proposed that so-
forward to discuss 3 periods: I - up to the end of the induction period; II lution data in the literature supported the existence of two types of C-S-
- the main hydration peak; III - hydration after the main peak. H – one metastable curve corresponding to solutions present during the
early hydration of C3S and one corresponding to stable C-S-H, present in
2.1. Period I: up to the end of the induction period preparation of C-S-H from laboratory reagents or during later hydration
of C3S. Although, re-examination of the data in this paper [11] does not
Since the last review article 4 years ago [2], no new mechanisms seem to show such a clear distinction in these two categories. In a re-
have been proposed to explain the origin of the induction period. The cent study [12] the composition of pore solution of hydrating C3S with
new literature focuses on the two main hypotheses: the protective water to cement ratios varying from 0:4 to 0.75 was carefully measured.
membrane theory and the dissolution one. Despite new experimental Up to 5% of hydration, all calculated activities fell on the same solu-
data, the situation hasn't changed much with both theories having their bility curve attributed to a metastable C-S-H with a C/S of either 1.23 or
partisans choosing essentially to highlight their most apposite 1.44 depending if the existence of a calcium-silicate complex is


Corresponding author.
E-mail address: karen.scrivener@epfl.ch (K. Scrivener).

https://doi.org/10.1016/j.cemconres.2019.105823
Received 21 June 2019; Received in revised form 18 July 2019; Accepted 18 July 2019
0008-8846/ © 2019 Elsevier Ltd. All rights reserved.
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

I II III

Fig. 1. Typical heat evolution curve of alite, broken up into 3 periods as discussed here.

considered or not. decrease non-linearly as the undersaturation of the system decreases,


Several studies provide evidence that some hydrates do form rapidly i.e. as the system is moving towards equilibrium. These experiments are
upon contact with water [13–15] and these early hydrates may control all performed in conditions diluted enough to prevent the precipitation
the pore solution at this stage (see also Section 2.1.4). What is ques- of hydrates. Therefore, the rates obtained are not necessarily quanti-
tionable is the formation of a hydrate film with enough coverage to give tatively representative of dissolution rates taking place in paste where
a diffusion barrier which so dramatically reduces the rate of C3S reac- hydrates precipitate at roughly the same time. Such rates are very
tion. To be effective, the hydrate layer would also need to be tightly difficult to measure in real hydration conditions as it is difficult to
bound to the reactive grain. Such a membrane has never been observed isolate or determine experimentally the “free” surface of C3S which is
experimentally despite progress in experimental tools and methods. dissolving. In real hydration conditions dissolution and precipitation of
Indeed, it seems very unlikely that such a continuous film of a few hydrates take place simultaneously and at the same rate (since the so-
nanometres could form uniformly through a dissolution precipitation lution cannot store a significant amount of dissolved material). The
process, generally agreed to be the case for cementitious phases. The presence of hydrates on the reacting surface might well interfere with
formation of a layer continuous over the whole surface and tightly the dissolution process.
bound to the substrate could probably only result from a topological Another difference observed, between pure dissolution experiments
reaction with an incongruent dissolution as observed in the case of and real hydrating conditions, is the surface topography. The mor-
certain poorly reactive glassy materials [16–22]. It has however been phology of the pits observed in pure dissolution experiments is very
shown that dissolution of C3S is congruent irrespective of the initial shallow exhibiting a lens shape whereas pits observed on C3S grains
saturation state of the solution [23]. The required epitaxy with the C3S during the course of hydration are usually very deep and irregular in
surface also seems unlikely when the different crystallographic or- shape [23] (Fig. 2). This implies that dissolution rates obtained from
ientations exposed on grain surfaces and the radically different crystal pure dissolution experiments may not be transposed directly to rates of
structures of C3S and C-S-H are considered. dissolution in hydrating conditions and some effort is still required to
Another issue regarding this approach is that it fails at explaining explain these differences. Indeed, the surface of C3S hydrating in paste
the annealing experiments on C3S (thermal treatment of the anhydrous conditions is exposed to a solution which evolves over time. This means
powder [24]) which exhibit much longer induction periods without a that it may pass through different regimes of dissolution leading to a
change in the driving force, i.e. difference in chemical potential be- “history” dependent topography.
tween the surrounding pore solution and the reacting surface. In any case dissolution experiments qualitatively show that rates of
Moreover, if the rate of dissolution has to be controlled by a dissolution do indeed decrease significantly as the system evolves to-
membrane of metastable product, what happens after the end of the wards equilibrium. The interested readers are invited to consult [35] for
induction period? At the end of the induction period it is considered more information on this topic.
that stable C-S-H forms, which should cause the metastable layer to
disappear. However, the reaction rate predicted from the enthalpy of 2.1.3. How to distinguish between these theories
formation is still much higher than those observed [25]. In a set of simulation experiments, Bullard and Flatt tried to find
The apparent discrepancy between the low reactivity observed ways to distinguish between the passivation layer theory (i.e. protective
during the induction period, and the high C3S solubility calculated membrane theory) and the site deactivation hypothesis (i.e. dissolution
through the enthalpy of formation, can be the better explained by the theory) [36]. Their results indicated that both theories can lead to si-
concept of superficial hydration of the interface introduced by Barret mulations that reproduce typical experimental features, such as the
[26]. This proposes that the exposed silicate species of the reacting C3S heat evolution and the evolution of the pore solution composition.
are hydroxylated (protonation of the silicate groups) and partially de- However, when they looked at the impact of inhibition of Portlandite
protonated as the pH increases over the course of hydration. Such a precipitation on the hydration kinetics (all other things being equal),
vision has been supported by Pustovgar et al. [27] using 29Si{1H} their simulations indicated very different kinetics according to which of
CPMAS NMR methods, which indicate Q0 silicates in close proximity to the two theories was used. A strong retardation in kinetics was found in
water. Their results corroborate the ones from Rodgers et al. [28], with the case of the dissolution theory whereas no retardation was observed
a major difference in the interpretation, since the hydroxylated Q0 in the case of the protective membrane theory. This can be understood,
species found during the induction period, and previously attributed to in the case of the dissolution theory, by the fact if CH cannot pre-
the protective membrane, are present prior to contact with bulk water cipitate, there is no sink for calcium and hydroxide ions and the solu-
and show no evolution during the induction period. tion composition remains at a low undersaturation with respect to C3S
with a low rate of dissolution. On the other hand, the protective
2.1.2. Dissolution controlled by undersaturation membrane relies on the conversion of a metastable C-S-H to a more
Several articles on C3S dissolution have been published over the stable one and in that case, the inhibition of Portlandite precipitation
past few years [23,24,29–35] and summarized in [35]. It has been would have no impact on that conversion thus there would be no re-
shown experimentally that the rates of dissolution of pure C3S and alite tardation. Bullard and Flatt suggested that identifying an additive,

2
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

Fig. 2. TEM images of OPC hydrated for (a) 12 h and (b) 24 h. From [64.

which could selectively inhibit the precipitation of portlandite while observed in hydrating conditions. Taking this anisotropy of dissolution
not affecting C3S nor C-S-H, could decisively differentiate between the into account is also be an important step in modelling approaches since
two theories. so far only isotropic dissolution is considered.
In a recent study Juilland and Gallucci [37] found that sucrose
mainly acts as an inhibitor of portlandite nucleation and growth 2.1.4. C-S-H: new insights
without having much impact on other parallel reactions involving si- Recently, synthetic C-S-H precipitation studies [42–44] have ob-
licates for pH up to 12.7, i.e. C-S-H nucleation and growth as well as served a nanoglobular morphology (Fig. 3). It may be speculated that
pure C3S dissolution. These results remain valid for pure C3S hydration this could be the metastable product conjectured to form in the first few
in paste as the measured pH did not exceed 12.8, keeping the fraction of minutes of C3S hydration reaction. Kumar et al. [42,45] synthesized
deprotonated sucrose molecules low enough to limit their calcium nanoglobular C-S-H at solution conditions with pH < 11. X-ray dif-
mediated adsorption on silicate surfaces [38]. As predicted by the dis- fraction showed that these nanoglobules are amorphous [45]. The au-
solution theory, a strong retardation was observed in this system and thors also observed that it has a Ca/Si ratio < 1.25 from 29Si DNP NMR
these experiments provide a strong argument in favour of the dissolu- studies and does not follow the ‘dreierketten’ silicate chain structure
tion theory hypothesis. that is generally found for C-S-H (Fig. 4) [42].
The work of Marchon et al. [39] on retardation induced by PCE also Schönlein et al. [43] observed similar results by mixing calcium
provides indirect evidence for the dissolution theory. In their work pure nitrate and sodium silicate in a syringe that were filtered and collected
dissolution experiments showed a strong impact of the PCE on the after a few minutes. The authors illustrated the metastable nature of the
measured rate of dissolution for undersaturation conditions enabling nanoglobular C-S-H which transformed into nanofoils when it is filtered
the calcium mediated adsorption of these molecules (2 orders of mag- after about 15–30 min of reaction. Here, the pH is not controlled during
nitude lower in the most pronounced case). This reduction in dissolu-
tion rate allowed the retardation behaviour seen in annealing experi-
ments to be exactly reproduced with a defined dosage of a PCE [40].
Pourchez et al. also proposed that PCE would impact the dissolution
process of C3S and Nicoleau came to a similar conclusion for latexes
[41].
In light of these findings, the idea that reaction is limited by a
protective membrane as a diffusion barrier should be set aside. The
hypothesis of the dissolution theory as a mechanism responsible for the
induction period as proposed originally [24] remains valid. Never-
theless, more work is needed to characterise the solution composition
and the early hydrate formation, particularly a better understanding of
the reacting alite surface is needed, i.e. hydroxylated state, etch pit
formation, step retreat, kink site formation and crystallographic defects.
Indeed, TEM images during the main hydration peak (Fig. 2), reveal
a strong anisotropy in the front of dissolution on reacting alite grains.
These features clearly highlight the importance of crystallographic or-
ientation. This has been quantitatively assessed by Robin et al. [34]
showing that strong differences in rate of dissolution are found for a
same degree of undersaturation. Furthermore, the evolution of the rate
of dissolution as a function of the undersaturation did not follow a
single trend. This means that the relative contribution of each plane of
different crystallographic orientation to the overall rate of dissolution
can vary as a function of the undersaturation. An important next step
would be to establish whether the fastest dissolving crystallographic
orientation analysed in pure dissolution experiments are the same as Fig. 3. Globular C-S-H structure from [42].

3
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

Fig. 5. Alite grain with C-S-H “needles” growing on surface around the time of
the main heat evolution peak. Within the red circles, the surface of the alite.
From [58]. (For interpretation of the references to colour in this figure legend,
Fig. 4. Typical foil like morphology of C-S-H grown in solution [42]. the reader is referred to the web version of this article.)

their synthesis and not reported, so cannot be compared with the ex- 2.2.1. The diffusion layer theory
periments of Kumar et al. [42] where the morphology is a function of The diffusion layer theory postulates that as C-S-H precipitates on
the pH which in-turn determines the relative abundance of the different the surface of the cement grains, it forms a thickening layer that ulti-
species in solution. mately becomes thick enough to act as a diffusion barrier. Many models
Krautwurst et al. [44] also observed such morphology which they before 2000 rely on this hypothesis, in particular HYMOSTRUC [55].
called spheroids, which crystallized at later ages (> 600 min, in their Even if many textbooks refer to this hypothesis as plausible, the evi-
solution conditions) to form ordered layers within these globules. dence built up against it over the last decade is overwhelming.
However their final product is a crystalline C-S-H and such a high de- The arguments against the diffusion theory were already detailed in
gree of crystallinity is generally not found in the real system. the previous ICCC articles [1,2]:
It could be supposed that during the initial minutes of dissolution,
nanoglobules are formed as a metastable phase as the pH builds up. 1. Existence of gaps (as in Fig. 2)
When the pH is above 12, then the nanofoils of C-S-H form as illustrated
by Kumar et al. [45]. Nonetheless, is seems very unlikely that such a If the outer C-S-H shell acted as a diffusion barrier, then the con-
globular product (if present on the surface) would significantly inhibit centration of species and supersaturation within the gap would be
dissolution. higher, so that precipitation should happen inside it rather than outside
Apart from this anomoulous globular structure, C-S-H synthesized the shell at the end of the “needles”. Yet precipitation carries on at the
from solution is always reported to have a foil like structure as shown in ends of the needles, not in the gap, up to the end of the first day, which
Fig. 4. This foil morphology reflects the underlying layer structure of is well after the peak.
the C-S-H which experiments and atomistic modelling show is a heavily
defective tobermorite structure [42,46,47]. C-S-H growing on grains of 2. No change in activation energy
alite, cement or filler may occasionally have a fairly foil-like appear-
ance, but more usually the foils seem to merge together to give the If there was a transition of mechanism at the peak time, then the
typical spiky or “needle” morphology seen in Fig. 5. Despite the ap- activation energy of the first mechanism, during the acceleration, and
pearance of quasi one-dimensional growth of these needles, they are the second mechanism, during the deceleration, would likely be dif-
still based on the layer structure of defective tobermorite. ferent. However, the activation energy, after 4 h, has been measured to
be constant not only throughout the whole peak but up to a few days
[56], indicating no change of mechanism during the first few days.
2.2. Period II: the main hydration peak

3. Unrealistic and changing values to fit the data of Costoya


In discussing the mechanisms governing the main peak it is im-
portant to stress that just being able to fit this curve does not validate
Ten years ago, Bishnoi modelled the diffusion hypothesis through
the model used to generate the curve. This point is underlined by the
his then newly developed platform μic (pronounced “mike”) [57] and
fact that Costoya's [48] PSD experiment has been fitted by at least five
showed that the C-S-H diffusion coefficient had to be varied over one
completely contradictory models [49–53]. As a consequence, hydration
order of magnitude to fit tricalcium silicate powders having different
models must be, in addition, assessed according to their physical,
fineness [52]. Such a variation of the diffusion parameters appears ad
chemical and microstructural soundness.
hoc and required only to save the hypothesis given that no variation of
In the past, the two most popular hypotheses used to explain the
any C-S-H properties (which are forming from solution) has ever been
heat evolution peak were the diffusion layer and impingement hy-
measured to depend on the powder fineness.
potheses. As discussed below, neither of these hypotheses can explain
To these we add a fourth, which is that a continuous layer is not
the experimental evidence. Two more recent hypotheses, confined
formed at the time of the peak as illustrated in Fig. 5. This figure
growth and dissolution limitation hypotheses, are then introduced and
shows a representative alite grain close to the main hydration peak, at
critically discussed. Finally, we describe the C-S-H growth hypothesis
this stage quite significant parts of the grain surface (circled) are visible
proposed by Bazzoni [54] and already presented in 2014 [2], which has
between the needle like growths. With this level of porosity, the
now received quantitative support thanks to the needle model [50].

4
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

Fig. 6. Schematic illustration of Avrami type nucleation and growth. [52]. The rate of growth is proportional to the surface area, so initially increases. When the solid
phases start to impinge the free surface for growth decreases and so is the rate.

transport of species through the layer could not be slowed down en- et al. [53] extended this to the full particle size distribution. However,
ough. both predict 2 μm grains to release more heat at 24 h than any other
Once inner product starts to form, which occurs from about one day, grains, but such small grains have already vanished by the peak time.
thick rims of product eventually form around the reacting grains as in The model of Masoero et al. [51] also extended the model of Biernaki
Fig. 17. It is an open question as to whether these layers act as diffusion and Xie, with a different treatment of the boundary for C-S-H growth:
barriers, this will be discussed later (Section 2.3.4). when tested on Costoya's PSD experiment this predicts C-S-H thick-
nesses of 1 μm, which is more than twice the upper bound measured by
2.2.2. Nucleation and growth with impingement Costoya: (100–400 nm).
This theory proposes that the progressive impingement of C-S-H
hydrates from neighbouring grains lead to the transition from accel- 2.2.4. Another view: dissolution theory and the coalescence of etch pits
eration to deceleration. The rate of reaction is assumed proportional to As discussed earlier, dissolution controlled by the undersaturation
the surface of the growing hydrates. As hydrates impinge against each of the solution, is the dominant mechanism during up to the end of the
other the surface in contact with water decreases and so does the rate. induction period. Nicoleau and Nonat [33] proposed that this me-
This idea originated in the context of metal melts and was modelled by chanism continues to determine the rate of reaction throughout the
Avrami [59]: Fig. 6 [52] illustrates this theory. main heat evolution peak. To summarise their views, the increasing rate
Modifications of this basic idea have been made to account for the of reaction during the acceleration period is due to the opening of etch
fact that the hydrates nucleate on the surface of the cement grains, pits which causes an increase in the alite contact surface area. Because
rather than homogeneously in space: so-called Boundary Nucleation the flux of dissolution is proportional to the surface area of the dissol-
and Growth (BNG) [60]. BNG models were further refined to consider a ving substance (here alite), etch pit opening results in higher exposed
C-S-H densification mechanism [52] or anisotropic growth [49]. alite surface and thus heat flow. At some point, the surface decreases,
Many studies have tried to fit Avrami or BNG type models to kinetic for example by etch pits coalescence, which leads to the transition to
data, but different parameters emerge for each data set and there is a the deceleration period.
lack of consideration as to whether the fitting parameters have any The main argument that dissolution is controlling the reaction rate,
physical meaning. For example, the fits to the Costoya data [48] by is that the undersaturation of the C3S continually decreases throughout
Scherer et al. [49] require the growth rate of C-S-H to vary by a factor of the main heat evolution peak, but this is not a definitive argument for a
2.5 with the particle size of the cement and give thicknesses for the C-S- rate controlling step. If dissolution was rate controlling, the rate of
H layer at the peak > 2 times higher than the thicknesses observed reaction would continually decrease (no deceleration period) and the
experimentally. observed decrease can only be explained by a decrease in the surface
However, the most conclusive evidence against the impingement area. Etch pits have been observed during the whole hydration peak
theory is the very weak effect of water to cement ratio on the main [25,62], but there is no disappearance of the etch pits in the decel-
hydration peak. The main hydration peak barely evolves from water to eration period. This is clear in Fig. 2, where the roughness of the re-
cement ratio 0.4 to 0.8 as pointed out by Kirby and Biernacki in [61]. If acting grain at 12 h (near the peak) is qualitatively the same as at 24 h
impingement was to explain the transition from acceleration to decel- (at the end of the peak). Given the very different rates between the peak
eration then the interparticle distance, which is directly linked to the and 24 h the roughness of the grains would need to decrease dramati-
water-to-cement ratio, should influence it; yet this not the case. More cally (to around 10–15% of the roughness at the peak), if the change in
dramatic are the experiments of Garrault et al. [62] in dilute suspen- rate comes only from the amount of alite surface dissolving. Even at
sion, with water-to-cement ratio of 50, where the interparticle distance 28 days, Fig. 17, the grains show a high roughness, with a crystal-
is then in the order of tens of microns, well beyond any C-S-H thickness lographic orientation. As discussed in Section 2.3.4 the impact of water
measured at any time, and where a main hydration peak is still ob- to cement ratio on the later kinetics also contradicts the argument that
served at the same time and same height. dissolution is the sole rate controlling step.

2.2.3. The confined-growth hypothesis and related models 2.2.5. C-S-H growth: the needle model
Following the criticism of deceleration due to impingement, several The work of Bazzoni [54] indicated that during the acceleration
authors proposed that growth of C-S-H was restricted to happen no period, clusters of C-S-H (needles) progressively nucleate on the cement
more than one micron far from the grain surface in agreement with SEM grains and grow rapidly to a certain length. The deceleration period
observations [51,53,63]. Although these models are closer to SEM ob- corresponds to a drastic slowing down of the growth rate of the C-S-H,
servations, the imposition of the distance limit for hydrate formation is which at this point covers most of the surface of the grain (Fig. 7).
completely arbitrary. However, in a sense it anticipates the limitation A key element in the development of this new hypothesis is the
on the length of the growth of the C-S-H needles, which is discussed impact of small amounts of zinc doping in alite [65]. This experiment
later. These models also suffer from other shortcomings. Biernaki and was repeated by Li [66] and the key results are shown in Fig. 8. The rate
Xie [63] elaborated a model based on a single grain size and Honorio of heat evolution shows that zinc doping has quite different effects on

5
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

that for undoped alite or C3S very similar parameters for the C-S-H
nucleation rates, growth rates and needle length at the end of the fast
grow period could be used to simulate the kinetics of several different
preparations of alite or C3S.
The needle model quantitatively shows that the peak coincides with
the time at which most of the needles have nucleated and are in their
fast growth regime. The progressive coverage of the grain surface AND
the slowdown of the needle growth together account for the transition.
From a mechanistic point of view, the main hydration peak should
therefore be seen as a single episode, the separation into acceleration
and deceleration period is arbitrary and a heritage of Kondo and Ueda
[67] who were the first to qualitatively describe the calorimetry curve
of alite hydration in 1968. The model shows there is no qualitative and
sudden change of mechanism, only a progressive one which starts at the
onset of the acceleration and ends at about one day. This is why in
Fig. 1, we proposed to group the whole main peak in a single hydration
stage.

2.2.6. Key insights from the needle model

• Impact of dissolution of small grains


Experiments of fractionated powders [68] and other observations
[69] indicate that grains below about 5–7 μm are totally consumed by
Fig. 7. Evolution of the needle length with time for an alite paste at w/c = 0.7. the time of the peak. This led to the postulation that the disappearance
From [54].
of these small grains may be a contributing factor to the deceleration
period [70]. The needle model [50] shows that the dissolution of the
the period 1 (up to the end of the induction period) and period II (main small grains actually does not change the overall shape of the peak and
heat evolution peak). The end of the induction period is delayed, pos- does not alone cause the transition from acceleration to deceleration.
sibly due to an impact of zinc on the alite dissolution, but also by the Nevertheless, it does sharpen and shift it to the left whenever the PSD
inhibition of the nucleation of C-S-H, seen in the micrographs at 1 h. In incorporates a large fraction of grains below 10 μm. The experiments of
contrast, the main heat evolution peak is greatly enhanced, with an Bergold et al. [71] illustrate this effect: they studied very fine alite
increase in the cumulative heat at one day of around 65%. The en- powders and observed a shift of several hours of the peak time.
hancement of the main heat evolution peak corresponds to much longer
C-S-H needles in the case of the zinc doping, seen in the micrographs.
A new model was developed by Ouzia [50] to validate the hy-
• Kinetics of the acceleration period are a linear function of
surface area
pothesis that the reaction kinetics during the main heat evolution peak
are dominated by C-S-H growth. The model takes all its input para- Whenever the dissolution of the small grain is negligible, the DoH
meters from experimental measurements and has no free fitting para- and heat flow, HF, were shown to be linear, not just monotonic, with the
meters. The use of an analytical, rather than numerical approach means needle length and initial specific surface of the powder:
that calculations can be made rapidly and that both realistic grain η1
shapes (form factor and surface roughness) and realistic particle size DoH (t ) = Δr H ∗ ∗ Malite ∗ Sspe ∗ l24h ∗ f (kn , s, tc , t )
η3
distributions can be accounted for. The disadvantage is that the mi-
crostructure is not simulated so impingements cannot be explicitly η1 df (kn , s, tc , t )
calculated. HF (t ) = Δr H ∗ ∗ Malite ∗ Sspe ∗ l24h ∗
η3 dt
The model simulates the growth of C-S-H as cylinders (called nee-
dles) perpendicular to the surface of the cement grain (Fig. 9) and where ΔrH is the enthalpy of reaction, η1 and η3 are the volumetric
computes the volume of C-S-H by summing the individual needles vo- stoichiometric coefficients of the alite hydration equation, Malite the
lumes over all grains. The volume of alite, water and C-S-H are com- volumetric mass of alite, Sspe the initial specific surface of the powder,
puted from the alite hydration reaction: l24h the needle length at 24 h, kn the nucleation rate, s the needle base
surface (needles are assumed to be cylinders) and tc the characteristic
η1 Valite + η2 VH → η3 VCSH + η4 VCH
needle growth rate, and f is the bell-shape curve defined by the un-
Finally, the heat of this reaction is computed and compared with the derlying algorithm (see [50] for details).
experiments. The required input parameters are the nucleation and As a consequence, when calorimetry curves are replotted per unit of
growth rates of C-S-H with time, estimated from SEM and TEM images initial specific surface, they show an interesting trend: Fig. 10. Coarse
and the PSD, form and roughness of the anhydrous grains, from laser enough powders exactly align on the same curve whereas powders
granulometry and SEM images, which incorporate a large fraction of grains below 10 μm are sharpened
The model was tested on 20 sets of experimental data (14 in [50] and shifted left, though their acceleration periods still align. This re-
and 6 more in [58]), covering ranges of particle sizes and alite doping presentation per unit of surface also illustrates the influence of the
experiments. Experiments, which is a wider set of experiments than any dissolution of the small grains on the main hydration peak. These re-
other alite hydration model yet published. Moreover, by contrast with sults are consistent with experimental observations [687273].
all previous models, the parameters were measured and thus con-
strained within experimental confidence intervals. Across the range of • The influence of Zn on nucleation and growth
experiments, the average error on the heat flow and DoH were re-
spectively of 16.2% and 10.1%. The model was able to capture varia- The needle model shows that the difference in needle length is
tions of the peak height by a factor of 30. It was particularly interesting sufficient to quantitatively account for the impact of zinc doping on the

6
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

Fig. 8. Impact of zinc doping (3%) on hydration kinetics and microstructure of C-S-H. From the heat evolution curve the end of the induction period is delayed, while
the main hydration peak is dramatically enhanced. The large difference in the height of the main peak corresponds to a drastic change in the morphology of the C-S-H
needles, which are much longer and more clustered in the case of the zinc doped alite. Taken from [66].

reaction kinetics, Fig. 11. three phenomena together: dissolution, transport and precipitation. As
discussed earlier, one aspect, which needs to be understood better, is
how the reacting alite surface evolves and how the dissolution is af-
2.2.7. Next challenges
fected, by the deposition of hydrates on the surface and by the presence
To understand the mechanisms behind the observed nucleation and
of ions in the vicinity of the surface.
growth rates is clearly one of the next challenges. So far, most models
As a step in this direction Bellmann and Scherer [74] measured
take these parameters as input parameters but do not endeavour to
growth rates of C-S-H as a function of the solution concentration. In
explain them in terms of more fundamental mechanisms. Why is there a
Naber et al. [25] they used this date to calculate the expected rate of
burst of nucleation that lasts about 4 h, not 1 h, not 10 h? Why do
reaction from the measured pore solution evolution. The good agree-
needles initially grow fast and then slow down? Why does zinc promote
ment with the experimental rates, adds further support to the argument
the C-S-H growth?
that C-S-H formation strongly influences the hydration kinetics during
Because the driving force for precipitation is the supersaturation,
the main heat evolution peak.
and because the supersaturation is dictated by the balance between the
Another important contribution has been the modelling of C-S-H
dissolution and precipitation, future models should try to couple these

7
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

of the C-S-H with the “needle” morphology if much faster so this


dominates. As the rate of growth of the needles slows down, the rate of
formation of the inner product becomes comparable so both form at a
similar rate. However more work is needed in this area.

2.3. Period III: later ages

As discussed in the previous ICCC papers [1,2], there are still very
few studies which have investigated the kinetics and mechanisms
Fig. 9. Schematic illustration of the needle model algorithm. At time step 1 a governing the later hydration period, after the main hydration peak.
first generation of needle nucleates (in red). At step 2, a second generation The space filling hypothesis was raised as a leading mechanism at later
(blue) of needles nucleates while the first one grow fast. Gradually the surface is
ages [2]. The fact that hydration is ultimately limited by the space was
covered and the needles grow slower and slower. The peak time corresponds to
made clear in the seminal work of Powers [77,78] and is a fundamental
the time where most of the generations have nucleated and needles are still in
their fast growth regime. (For interpretation of the references to colour in this assumption in the famous Power's model, where anhydrous cement
figure legend, the reader is referred to the web version of this article.) remains unreacted below a w/c of 0.38. If this was not the case then
cement pastes with a water to cement ratio below about 0.38 would
eventually expand significantly when stored in water, which is known
precipitation from solution concentrations by Andalabi et al. [75]. The
not to occur.
past two decades of research on mineral nucleation and growth has
However, the work of Berodier indicated that “lack of space” starts
revealed a more diverse picture of these phenomena as demonstrated in
to impact the hydration kinetics after only a few days [79]. Whereas the
the recent textbook of Van Driessche, 2017 [76]. Common minerals
influence of w/c on the hydration for the first day is negligible, its
such as calcium carbonate, calcium sulfate, iron oxides, silica and
impact on the later ages is dramatic: 0.6 w/c PC pastes exhibit 15%
alumina have been shown to follow the non-classical nucleation theory.
more DoH at 28 days than 0.4 w/c ones. The impact of water to cement
New concepts have been brought forward like nucleation pathway,
ratio on later kinetics is clearly illustrated for alite hydration in Fig. 12.
clusters, liquid/amorphous precursors, crystalline intermediates, nu-
There are two (non-exclusive) ways in which the lack of space can
clei/crystallites aggregation etc., to describe the complex nucleation
limit hydration, considering that hydrates precipitate from solution and
and growth pathways of these minerals. The old “classical pathway”
so can only grow in water filled spaces. Either the volume of water
thus appears to be just one among other pathways, and not necessarily
available for precipitation is depleted or the remaining water filled
the most energetically favorable.
pores are smaller a critical pore size for precipitation.
The paper of Andalabi et al. [75] suggests C-S-H precipitation em-
braces two overlapping phenomena: primary nucleation and growth,
secondary nucleation and growth. The primary nucleation and growth 2.3.1. Space filling limitation by lack of water
generate crystallites of the order of 3 to 6 nm, the secondary nucleation The space available for precipitation is space occupied by water, so
generate crystallites in the vicinity of already formed ones. The model the original water to cement ratio controls the space available for
parameters - interfacial tension, cohesion energy, growth rate constant, precipitation. Recently, Ouzia [58] showed quantitatively that the vo-
kinetic order of growth and aspect ratio – are fitted to the evolution of lume opened by the gap, which forms between the hydrate shell and the
calcium concentrations with time measured during synthetic C-S-H reacting grain has a very important impact on the water filling of spaces
precipitation. The values obtained show good agreement with those in sealed samples. Because the water flows from the external space (the
obtained experimentally by other researchers. In particular, the C-S-H space outside of the C-S-H shells) to the gaps, the water in the external
particles are found to about 3–6 nm thick and about 100 nm wide space may be exhausted and precipitation stopped there.
sheets, which matches very well TEM observations of synthetic C-S-H Fig. 13 illustrates this effect on a 10 μm grain at w/c = 0.32 (the
[42,45]. This method provides a promising approach to look at the volume of water is then exactly identical to the volume of alite). When
impact of different parameters – such a solution composition and par- the gap is 1 μm, the grain has lost 50% of its mass so that half of the
ticularly other ions (sulfate, alumina, zinc, etc.). initial water has flowed in. To this must then be subtracted the amount
In this section, it has been argued that the growth of outer C-S-H as of water that reacted with alite to form the hydrates. This computation,
needles is the dominant influence on the hydration during the main shows there is no more water in the external space, which implies that
heat evolution peak. Towards the end of this period the formation of the outer hydrates cannot grow anymore.
“inner” product starts, in the regions originally occupied by the anhy- This leads to a rule of thumb for space filling:
drous grains. This inner product has an isotropic morphology, quite DoHexternal − space − filled = 1.6 w/c
different from that of the outer product, which appears granular in TEM
images. The start of the formation of inner product could be related to Whenever the DoH is equal or superior to 1.6 w/c, there is no more
the change in growth rate of the outer C-S-H. Initially the rate of growth external water available and precipitation then stops in the external

Fig. 10. Simulated calorimetry curve computed for


different log-normal PSD whose mode varied from
0.7 μm to 91.6 μm. On the right, the curves have
been normalized by surface area and show the im-
pact of the small grain dissolution: it is negligible for
PSD coarser than 11.5 μm (all curves exactly align)
but for the finer PSD, it gradually decreases and
shifts the peak left. The fact that all curves align
during the acceleration period show that alite sur-
faces chemically react in the same manner in-
dependently of the PSD: the surfaces of small grains
is not more, or less, reactive than the surface of big
grains, from [50].

8
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

Fig. 11. Comparison of experimental results of Bazzoni et al. [65] for coarse (left) and fine (right) alite with and without zinc, with simulations from needle model
[50]. Left plot: coarse powder, right one: fine powder.

space. aluminate of different curvatures using the equation above compared to


This equation was corroborated at 20 °C and 40 °C as Fig. 12 shows: the measured pore solution [83]. This indicates that the pore solution
a clear inflection point is seen on the 0.32 w/c systems cross can only sustain growth in pores above about 10 nm in size.
1.6 ∗ 0.32 = 52% of DoH. Of course, the calculation in the example above oversimplifies the
real situation. The equation above only applies when all pores are water
filled. In reality almost all cementitious materials are only partially
2.3.2. Space filling limitation by a limiting critical pore size
saturated, because voids are formed after setting due to chemical
In addition, more evidence has emerged indicating the importance
shrinkage. The existence of menisci causes under-pressure in the pore
of the lack of water filled pores greater than a limiting critical pore size
solution, which will affect the equilibrium. Furthermore, the energy of
[8081]. The role of pore size on crystal growth can be understood from
crystals is usually anisotropic. Nevertheless, the close agreement be-
the schematic shown in Fig. 14. This shows a crystal which has grown
tween the lack of water filled pores above a certain size observed ex-
to fill the central (large) pore and can only continue to grow into to
perimentally and the calculation is remarkable. More work is needed to
smaller pores of radius rc. To do this the crystal must increase its cur-
better understand the mechanisms acting here and the balance between
vature, which requires a higher concentration of ions in a solution in
growth, which appears to stop at the certain point, without producing
equilibrium with it.
expansion and the closely related phenomenon where a continued state
If the system is completely saturated the Ionic activity product (K)
of supersaturation in the solution leads to the exertion of crystallisation
of the solution at equilibrium with a crystal of radius r is given by the
pressure and macroscopic expansion, such as in the case of sulfate at-
Freundlich equation [82]:
tack (internal and external) [85-88] and periclase hydration.
RT K ⎞ 2y
∙ln ⎜⎛ ⎟ =
Vm K sp rc 2.3.3. The predominant role of outer C-S-H in the later ages
⎝ ⎠
Returning to the kinetics of alite hydration, the needle model as
where: R is the gas constant, T is temperature, Vm is the molar volume presented in [50] was later expanded to the later age kinetics in [58]. In
of the crystal, γ the interfacial energy of the crystal, rc the radius of the this extension the filling of space and impingements between hydrates
crystal-liquid interface, K the ion activity product of the solution and was dealt with by a statistical treatment. To support this development
Ksp the equilibrium ion activity product of a flat crystal surface. of the model, the growth of outer C-S-H needles after the main peak was
This phenomenon of the limitation of water filled pores for crystal quantified on SEM images of polished surfaces. Although the rate of
growth is most obvious in the case of the formation of carboaluminate growth slows down dramatically, it was seen that they grow as much
phases (hemi (Hc) and mono carboaluminate (Mc)) in ternary blends of from 1 to 28 days as during the first day. As a consequence, the volume
Portland cement, calcined clay and limestone (LC3), Fig. 15 (left) [80]. of outer C-S-H produced in the later ages is not negligible. The com-
As the calcined kaolinite content of the calcined clay component in- putations show, that on average there is in total twice as much outer C-
creases the amount of Hc and Mc pass through a maximum and then S-H as inner C-S-H. This may sound surprising when one considers the
decreases. The time at which the Hc and Mc stop forming, corresponds appearance of SEM polished sections (e.g. Fig. 17) but the following
to the time where the pore entry size (by MIP) is refined to below about argument may help: by 24 h, typical pastes have reached 50% of DoH.
10 nm. The Fig. 15 (right) shows a calculation for the ion activity This hydration results in almost all outer C-S-H. There is a further 30%
product of the solution in equilibrium with a crystals of monocarbo of DoH up to 28 days, the product of which are about equally

Fig. 12. Hydration kinetics for alite at 20 °C (left)


and 40 °C (right) at different water to cement ratios
and for different alite finenesses. The continuous and
dashed lines are superimposed up to a given DoH
beyond which they abruptly separate because of the
space filling. The horizontal dashed lines indicate the
rule of thumb prediction for external space filling:
80% line corresponds to the w/c = 0.5, the 51.2% to
the w/c = 0.32 and the 41.6% to the w/c = 0.26.
The black one in particular deals with 0.32 w/c ratio
exactly crosses the point where the 0.32 curves di-
verge from the 0.5 ones. From [58].

9
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

Fig. 13. First image from the left: equatorial of a


10 μm spherical alite grain at w/c = 0.32. At w/
c = 0.32, the volume of water equals the volume of
alite. On the second image, a 1 μm gap opens. On the
third one, half of the initial water has flowed in: a
1 μm gap on a 10 μm grains means that the grain has
lost about half its volume, this equals half of the in-
itial volume of water at this w/c. Fourth image: the
water consumed by the hydration is then subtracted,
the remaining water is not sufficient to fill even the
inner space; this would mean that hydrates can no
longer grow in the outer space. From [58].

systems should have thicker C-S-H shells and therefore react more
slowly; they should release less heat and thus have smaller slopes. If the
inner C-S-H rim acted as a diffusion barrier there would be significant
concentration gradients across this layer, which would be reflected in
compositional gradients of the C-S-H. EDS maps for the C-S-H rim,
Fig. 16, do not indicate any compositional gradient across the rim.
Similarly, dissolution cannot be alone rate controlling. Because a
higher DoH implies a lower alite surface, high w/c systems should re-
lease less heat and exhibit a lesser slope than low w/c ones. This does
not mean that diffusion or dissolution have no influence on the later
ages kinetics, only that they are not alone rate-controlling.
On the other hand, the disappearance of the small grains becomes
more and more important in the later ages. By definition 100% DoH
means that all grains have reacted, so that the closer the system is to
complete hydration, the more significant the impact of the dis-
appearance of small grains.
This discussion indicates that, at least some of the stages of hydra-
tion, require a combination of mechanisms to account for the experi-
mentally observed kinetics. Space filling, precipitation and dissolution
surely influence the later ages kinetics; the role played by transport and
the pore solution concentration remain open questions. Models that
Fig. 14. Schematic illustration of crystal growing into a small pore (adapted
from [84]). Note there is a thin liquid film between the growing crystal and the attempt to fully understand the later age should couple these different
pore wall. phenomena. However, it will by complicated to determine how me-
chanism combine, since even with single mechanisms it is possible to fit
hydration kinetics by adjusting parameters. As discussed further at the
distributed between outer and inner C-S-H this amounts to roughly
end or tis paper, a systematic approach and a broad data are critical to
twice as much outer C-S-H than inner C-S-H. The inner C-S-H is a
avoid falling in the “fitting trap”.
function of the outer C-S-H produced: the outer C-S-H opens the gap,
which then controls the space in which the inner C-S-H can form.
3. From alite to cement

2.3.4. Perspectives: the later ages kinetics are not controlled by a single 3.1. The microstructures are very different
overwhelming mechanism but a combination of them
The dataset built by Ouzia (Fig. 12) leads to the rejection of the Despite the focus of most work, and this review, on the mechanisms
diffusion layer theory and dissolution hypotheses as alone sufficient governing the hydration of alite, alite is actually NOT an ideal model
to account for the kinetics at later ages. On Fig. 12, the continuous lines for the hydration of cement, particularly where microstructure is con-
(high w/c curves) always have a higher DoH and a steeper slope. This is cerned. This is illustrated by Fig. 17, which compares the mature mi-
particularly visible for the 20 °C experiments. If the inner C-S-H acted as crostructure of pastes of alite and cement paste, both at a water to
a diffusion barrier controlling the reaction rate, then higher DoH cement ratios of 0.4. The main differences are the in the distribution of

Fig. 15. The amount of mono (Mc) and Hemi (Hc) carbo aluminate phases formed in different LC3 blends from Avet et al. [80] (left). On the right image, calculation
of the minimum pore radius in which Mc would be expected to grow according to the measured pore solution 83 and Freundlich equation [83] is shown.

10
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

Fig. 16. EDS mapping of Ca/Si ratio across C-S-H rim.

Fig. 17. Microstructure and mercury intrusion porosimetry curves comparing alite to Portland Cement. MIP curves from [89]

11
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

Fig. 18. Impact of sodium aluminate on the hydration of alite from [92].

the calcium hydroxide masses and in filling of the originally water filled block dissolution sites.
space. Large open pores remain in the case of alite as confirmed in the This well-known inhibition of the C3A reaction by gypsum is im-
MIP curves from Mota Gasso [89]. portant to have correct behaviour of a cement. It allows the main heat
The reasons for these pronounced differences were previously noted evolution peak of alite to occur before the aluminate phase. Otherwise
by Kjellsen and Justnes [69] and investigated by Gallucci and Scrivener the reaction of the alite is severely inhibited (undersulfation). When the
in 2007 [90]. They showed that neither the addition of gypsum nor gypsum is exhausted, the reaction of C3A (and the ferrite phase) re-
alkalis to C3S significantly modified the microstructure, even though starts. However, sulfate ions are absorbed on the C-S-H during the first
these both modify the kinetics of hydration as studied in more detail by part of the reaction. So, after the depletion of gypsum, there is first a
Moto Gasso et al. [91]. The commercial cement microstructure could be new formation of ettringite, corresponding to the usually seen shoulder
very well reproduced by a polyphase laboratory clinker of C3S and C3A peak and the AFm phases (mono sulfate or mono or hemi carbo alu-
with the addition of gypsum (essential to control the C3A reaction). This minate) form later, usually after several days. This was detailed in the
confirms that the difference cannot be attributed to some aspect of previous ICCC paper [2], from the studies of Quennoz [97] and was
commercial clinker production (such as quenching or grinding) and confirmed more recently by Jansen et al. [98]
indicates that the presence of alumina has an important impact on the It is known that the additions of supplementary cementitious ma-
nucleation and growth of C-S-H. terials, such as slag, fly ash or calcined clay require adjustment of the
Recent work of Pustovgar et al. [92] investigated the impact of level of sulfate addition in order to avoid undersulfation. It is usually
alumina in more detail by the addition of small amounts of NaAlO2. As considered that this adjustment depends on the amount of alumina in
illustrated in Fig. 18, taken from this paper, NaAlO2 significantly en- these additions. However recent work by Zunino [99] shows that the
hances the main hydration peak. In the original paper the authors fineness of the addition is also very important and that even an addition
propose that the effect of alumina is related to the existence of a frac- of limestone changes the amount of sulfate needed. This is due to the
tion of very fine reactive particles (< 0.2 μm). However, it is interesting “filler” effect, whereby the addition of any fine material increases the
to note, that according to the needle model discussed above, this en- rate of reaction of the alite. This increased reaction of alite means more
hancement of the main hydration peak should correspond to the growth sulfate is absorbed by the C-S-H during the main heat peak and there-
of C-S-H outer product to a longer length in the fast growth phase. fore the sulfate depletion peak occurs earlier.
Micrographs published in [93] do show that the C-S-H needles are
longer in the case of additions of aluminium, but more work would be 3.2.1. Influence of SCMS on hydration kinetics
needed to distinguish between cause and effect. Berodier and Scrivener [101] explained that the general trend of
The growth of longer C-S-H needles has two important effects: First acceleration of the hydration of the clinker in the presence of SCMs is
the infilling of the space between the grain occurs faster and is more due to the shearing conditions generated in these systems. However,
efficient. Second the formation of more outer product during the peak limestone provides better conditions for the nucleation and growth of
will increase the size of the gap between the hydrate shell and the re- C-S-H compared to other systems. From zeta-potential measurements,
acting grain, compared to the case of alite as has been frequently noted Ouyang et al. [102 observed that the calcium ions have higher affinity
experimentally [68,69]. for adsorption on limestone filler, having a positive zeta-potential at
low concentrations of calcium compared to micronized sand particles
3.2. New insights on the inhibition of C3A and factors affecting necessary (quartz-like system). A calcite surface would favour chemical interac-
sulfate addition tions with Ca2+ ions thereby reducing its mobility possibly through
strong chemical bonds and increasing the probability of formation of
Recently some nice studies by Geng, Myer, Monteiro and others stable C-S-H nuclei. In the case of micronized sand, which according to
[94,95] have brought new insights on the inhibition of the C3A reaction the authors have similar surface charge to C-S-H due to its similar iso
in the presence of gypsum. By a combination of synchrotron-based electric point (IEP), Ca2+ ions are not chemically adsorbed and here,
methods, they were able to show that: electrostatic and Van der Waals' forces are involved which reduces the
mobility of Ca2+ compared to the bulk, but still more mobile when
• Ettringite is the only (stable) hydration product in induction period compared to chemically adsorbed species.
• Ettringite preferably nucleates on C A surfaces
3 Pourchet et al. [103] measured the variation of shear yield stress
• Ettringite grows to limited diameters and lengths and zeta-potential on calcite particles as a function of concentration of
• There is no evidence that ettringite is a diffusion barrier. calcium ions in the solution. The authors observed a decrease in the
shear yield stress with an increasing zeta-potential measurement which
This work indirectly supports the hypothesis of Minard et al. [96] is a result of the stronger repulsive interaction forces exerted by the
that the reaction of C3A is inhibited by the specific adsorption of the adsorbed calcium inducing a better dispersion of the particles and thus
sulfate and/or calcium ions on the surface of the grains of C3A which a lower yield stress. The authors observed that this is in accordance

12
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

4.3. Period III

• The filling of space becomes a key factor in this period.


• More work is needed to understand how the progressive filling of
space impacts the rate of reaction.
• There is no strong evidence that the diffusion of ions through pro-
duct layers is rate limiting.
• Neither is the rate of dissolution rate limiting in this period.
• It would be very interesting to know more about spatial variations in
the composition of the pore solution (diffusion gradients), but this
does not seem experimentally possible at present.

4.4. From alite to cement

• The main difference between the microstructure from the hydration


of cement compared to C3S is the more even and efficient filling of
the originally water filled space. This leads to much finer pores in
Fig. 19. Heat flow measured for Portland cement (PC) compared with blended cement pastes.
systems with 50% replacement of limestone (L), quartz (Qz) and fly ash (FA).
All the curves are normalized to the mass of PC in the system with an effective
• The presence of alumina seems to be implicated in these differences.
One explanation could be that its presence leads (directly or in-
water to solid ratio of 0.75. Adapted from [100]. directly) to the growth of longer needles during the main heat
evolution peak, which would also explain the formation of much
with the model of Scales [104] which predicts a linear decrease of shear larger gaps between the hydrate shells and the reacting grains.
yield stress with the square of zeta-potential. However, beyond a cer- • Nevertheless, there is considerably more work to be done, which
tain limit of calcium concentration, zeta-potential plateaus and the considers the effect of all the other ionic species, including, for ex-
yield stress slightly increases due to the screening effect induced by the ample sulfate.
resulting high ionic strength of the solution. As Pourchet et al. [103] • The depletion of sulfate after the main heat peak of alite is impacted
and Ouyeng et al. [102] worked on model systems, blended cements by the amount of C-S-H formed, as sulfate ions are absorbed on C-S-
bring even more complexities than these simple cases as observed by H during the first hours of hydration and later released leading to
Schöler et al. [105] (Fig. 19). Here, for the systems with limestone, the the formation of ettringite in the shoulder aluminate peak, before
dominating mechanism can still be attributed to the high affinity of formation of AFm phases.
Ca2+ providing an ‘easy’ pathway for C-S-H nucleation.
4.5. The importance of models
4. Summary and perspectives
The complexity of the phenomena occurring during cement hydra-
4.1. Period I tion requires models to be built to validate or refute the hypotheses
advanced to explain hydration. Hypotheses that are not expressed in
• There is more and more evidence against the existence of any in- quantitative form in a model have only a very weak force of conviction.
However, so far, most publications dealing with hydration models
hibiting layer forming by dissolution and precipitation of a type of
C-S-H on the surface. only show their model to satisfactorily fit a few calorimetry curves with
• Nevertheless, the first C-S-H precipitates may be metastable and a “relatively” small variations of the input parameters. The mechanisms
may completely be wrong, like the Avrami or diffusion-based models
influence the pore solution concentration.
• This leaves the theory of dissolution controlled by undersaturation for instance, and still the model fits the calorimetry curves.
Alternatively, existing models based on completely different mechan-
as the most plausible explanation for the induction period.
• Work is still needed to determine the state, and effective interfacial isms are still able to fit the exact same set of calorimetry experiments.
Therefore, the ability for a model to fit calorimetry curves is necessary
energy of alite upon contact with water.
• The formation of etch pits also needs to be studied in more detail, but not sufficient to validate the underlying hypothesis.
Therefore, in our opinion, the methodology to move forward needs
particularly the very deep and regular erosion seen on certain alite
surfaces, as illustrated in Fig. 2. to be sharpened and new criteria put forward such as:

4.2. Period II i) The ability to fit a large enough dataset. As a rule of thumb


models should at least be able to fit twice the number experiments
• Neither diffusion, nor impingement, nor the dissolution of small than input parameters. It is also important to study several different
variables in a dataset. For example, it was not enough to vary just
grains, explain or control the transition from the acceleration to the
deceleration period. the particle size of alite as in the experiments of Costoya [48], only
• The growth rate of outer C-S-H dominates this period. by varying also the water to cement ratio and the doping of alite
• More work is need to understand the factors affecting the growth of was it possible to refute different hypotheses and develop the C-S-H
growth (needle) mode, which could explain all the experimental
C-S-H, why does it slow down and why can some elements like zinc
and aluminium have such a great impact on the length of the data.
“needles” at the slow down. ii) All the key experimental parameters that are known to influence
• Dissolution and growth are coupled through the pore solution and the reaction and commonly vary in real life applications should be
considered. At least the PSD for alite and amount and PSD each
this coupling should be better incorporated into quantitative
models. phase assemblage for cement. Models dealing with the later ages
• As in period I the role of etch pits and surface roughness needs to be should additionally take the w/c into account as it has a strong
effect on the reaction [106].
better understood in this period.
iii) The ability to explain simultaneously results from several

13
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

techniques such as SEM, ICP, NMR, XRD or MIP instead of only interfacial effects, means that the concentration of species at interfaces
calorimetry. For example, nucleation and growth rates found by varies of a scale of < 1 nm.
fitting the calorimetry curve that are within the confidence inter- In conclusion models are very important, but, when it comes to
vals measured by SEM reinforce the model and/or the underlying hydration and microstructure, we need to be much more rigorous about
hypothesis credibility. In the opposite case, as exemplified in why and how we are using them.
Section 2.2, the model and/or underlying hypothesis should be
rejected. Acknowledgements
iv) An objective fitting process. Fitting requires an error function to
be defined followed by a standard fitting method such as the gra- Scrivener, Ouzia and Kunhi Mohamed are grateful for the support of
dient method, the simplex method or genetic algorithms. In all EPFL for their work. We also thank numerous friends and colleagues for
cases, the uniqueness of the solution should be addressed. insightful discussions over the years, particularly Paul Bowen.
v) Cross-comparison by testing the model on a common dataset.
So far Costoya [48] PSD experiment has served such a purpose for References
alite hydration but only deals with five curves. Thermkhajornkit
et al. [107] has been used for testing cement hydration [53]. [1] K.L. Scrivener, A. Nonat, Hydration of cementitious materials, present and future,
However, the field is still lacking a big enough dataset dedicated to Cem. Concr. Res. 41 (2011) 651–665, https://doi.org/10.1016/j.cemconres.2011.
03.026.
properly test either alite or Portland cement hydration. [2] K.L. Scrivener, P. Juilland, P.J.M. Monteiro, Advances in understanding hydration
of Portland cement, Cem. Concr. Res. 78 (2015) 38–56, https://doi.org/10.1016/j.
Finally, a word on microstructural models, which seek to re- cemconres.2015.05.025.
[3] H.N. Stein, J.M. Stevels, Influence of silica on the hydration of 3 CaO,
present explicitly the entire microstructure. Microstructural models SiO < sub > 2, J. Appl. Chem. 14 (1964) 338–346, https://doi.org/10.1002/jctb.
such as CEMHYD3D, Hymostruc, μic [55, 57,108, have been very 5010140805.
popular since the late 1980s. the idea that we can simulate the entire [4] D.L. Kantro, S. Brunauer, C.H. Weise, Development of surface in the hydration of
calcium silicates. II. Extension of investigations to earlier and later stages of hy-
microstructure in a computer is very seductive. However, we feel it is dration, J. Phys. Chem. 66 (1962) 1804–1809, https://doi.org/10.1021/
now time to take stock and clearly identify the limitations of this ap- j100816a007.
proach. First it is clear that even with the dramatic improvement in [5] E.M. Gartner, J.M. Gaidis, J.P. Skaln (Ed.), Hydration Mechanisms, I, American
Ceramic Society, 1989.
computer power over the past decades, we are very far from the power
[6] P.W. Brown, E. Franz, G. Frohnsdorff, H.F.W. Taylor, Analyses of the aqueous
and speed which would be required to have a realistic representation of phase during early C3S hydration, Cem. Concr. Res. 14 (1984) 257–262, https://
a cementitious microstructure on the 100 μm scale. Voxel approaches doi.org/10.1016/0008-8846(84)90112-1.
quickly run into problems of the size range of particles which can be [7] J.G.M. de Jong, H.N. Stein, J.M. Stevels, Hydration of tricalcium silicate, J. Appl.
Chem. 17 (1967) 246–250, https://doi.org/10.1002/jctb.5010170902.
represented. Vector approaches necessitate the assumption of simplistic [8] R. Kondo, M. Daimon, Early hydration of tricalcium silicate: a solid reaction with
grain shapes, such as spheres. However, the major limitations of these induction and acceleration periods, J. Am. Ceram. Soc. 52 (1969) 503–508,
models is the inability to represent the pore structure in any meaningful https://doi.org/10.1111/j.1151-2916.1969.tb09203.x.
[9] H.M. Jennings, P.L. Pratt, An experimental argument for the existence of a pro-
way. tective membrane surrounding Portland cement during the induction period, Cem.
To understand this limitation, we need to return to the purpose of Concr. Res. 9 (1979) 501–506, https://doi.org/10.1016/0008-8846(79)90048-6.
microstructural models. Other than being just a nice visualisation, the [10] F. Bellmann, D. Damidot, B. Möser, J. Skibsted, Improved evidence for the ex-
istence of an intermediate phase during hydration of tricalcium silicate, Cem.
key purpose should be to predict the properties of a cementitious ma- Concr. Res. 40 (2010) 875–884, https://doi.org/10.1016/J.CEMCONRES.2010.
terial from its microstructure. Broadly speaking the properties of ce- 02.007.
mentitious materials which are of interest are mechanical properties [11] E.M. Gartner, H.M. Jennings, Thermodynamics of calcium silicate hydrates and
their solutions, J. Am. Ceram. Soc. 70 (1987) 743–749, https://doi.org/10.1111/j.
and transport properties which determine durability. 1151-2916.1987.tb04874.x.
For mechanical properties from the very first days of Concrete [12] T. Sowoidnich, F. Bellmann, D. Damidot, H. Ludwig, New insights into tricalcium
Technology it was shown that after a few days mechanical properties silicate hydration in paste, J. Am. Ceram. Soc. 102 (2018), https://doi.org/10.
1111/jace.16133 jace.16133.
depend overwhelmingly just on the amounts of the different phases e.g.
[13] D. Damidot, A. Nonat, C 3 S hydration in diluted and stirred suspensions: (I) study
Powers [77]. This either means that the way these phases are dis- of the two kinetic steps, Adv. Cem. Res. 6 (1994) 27–35, https://doi.org/10.1680/
tributed in space does not matter or that the process of hydration means adcr.1994.6.21.27.
that the way they are distributed is always statistically the same. We [14] S. Garrault-Gauffinet, A. Nonat, Experimental investigation of calcium silicate
hydrate (C-S-H) nucleation, J. Cryst. Growth 200 (1999) 565–574, https://doi.
feel that it is this later explanation which is most likely to be correct, org/10.1016/S0022-0248(99)00051-2.
but in any case, it means you do not need a numerical microstructure to [15] S. Garrault, A. Nonat, Hydrated layer formation on tricalcium and dicalcium si-
better predict mechanical properties. licate surfaces: experimental study and numerical simulations, Langmuir 17
(2001) 8131–8138.
Regarding transport properties, there is now overwhelming evi- [16] E.H. Oelkers, S.R. Gislason, The mechanism, rates and consequences of basaltic
dence that these are controlled by small pores below about 10 nm in glass dissolution: I. An experimental study of the dissolution rates of basaltic glass
size. 1H NMR shows that there are no water filled pores above this size as a function of aqueous Al, Si and oxalic acid concentration at 25 °C and pH = 3
and 11, Geochim. Cosmochim. Acta 65 (2001) 3671–3681, https://doi.org/10.
after a few days [109]. MIP confirms that the connected porosity has 1016/S0016-7037(01)00664-0.
pore entries in this range after a few days and the limitations of space of [17] C. Cailleteau, F. Angeli, F. Devreux, S. Gin, J. Jestin, P. Jollivet, et al., Insight into
hydrate grow indicate that water filled pores above this size should fill silicate-glass corrosion mechanisms, Nat. Mater. 7 (2008) 978–983, https://doi.
org/10.1038/nmat2301.
with hydrates in systems with reasonably low w/c ratios (< ~0.5). If a
[18] R.W. (Robert W. Revie, H.H. Uhlig), Uhlig's corrosion handbook, n.d.
microstructural model is to consider a reasonable particle size dis- [19] A. Barkatt, P.B. Macedo, B.C. Gibson, C.J. Montrose, Modelling of waste form
tribution of the anhydrous cement grains: on the order of 10s of μm performance and system release, MRS Proc. 44 (1984) 3, , https://doi.org/10.
1557/PROC-44-3.
then the minimal system dimension of a microstructural model should
[20] F. Delage, D. Ghaleb, J.L. Dussossoy, O. Chevallier, E. Vernaz, A mechanistic
be at least 100s of μm. This is at least four orders of magnitude larger model for understanding nuclear waste glass dissolution, J. Nucl. Mater. 190
than the size or pore which need to be considered for transport prop- (1992) 191–197, https://doi.org/10.1016/0022-3115(92)90086-Z.
erties. In three dimensions, this means a digital representation would [21] S.-B. Xing, A.C. Buechele, I.L. Pegg, Effect of surface layers on the dissolution of
nuclear waste glasses, MRS Proc. 333 (1993) 541, , https://doi.org/10.1557/
need to have at least 1012 voxels which is well beyond the capacity of PROC-333-541.
today's computers (by comparison the atomic simulations on the most [22] S. Gin, I. Ribet, M. Couillard, Role and properties of the gel formed during nuclear
powerful computers can only handle around 109 particles 110). But in glass alteration: importance of gel formation conditions, J. Nucl. Mater. 298
(2001) 1–10, https://doi.org/10.1016/S0022-3115(01)00573-6.
fact, the situation is much worse than that as, to be meaningful, we [23] L. Nicoleau, A. Nonat, D. Perrey, The di- and tricalcium silicate dissolutions, Cem.
need to consider the pore solution composition and, in each voxel, and

14
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

Concr. Res. 47 (2013) 14–30, https://doi.org/10.1016/j.cemconres.2013.01.017. [50] A. Ouzia, K. Scrivener, The needle model: a new model for the main hydration
[24] P. Juilland, E. Gallucci, R. Flatt, K. Scrivener, Dissolution theory applied to the peak of alite, Cem. Concr. Res. 115 (2019) 339–360, https://doi.org/10.1016/j.
induction period in alite hydration, Cem. Concr. Res. 40 (2010) 831–844, https:// cemconres.2018.08.005.
doi.org/10.1016/j.cemconres.2010.01.012. [51] E. Masoero, J.J. Thomas, H.M. Jennings, A reaction zone hypothesis for the effects
[25] C. Naber, F. Bellmann, T. Sowoidnich, F. Goetz-Neunhoeffer, J. Neubauer, Alite of particle size and water-to-cement ratio on the early hydration kinetics of C3S, J.
dissolution and C-S-H precipitation rates during hydration, Cem. Concr. Res. 115 Am. Ceram. Soc. 97 (2013) 967–975.
(2019) 283–293, https://doi.org/10.1016/J.CEMCONRES.2018.09.001. [52] S. Bishnoi, K.L. Scrivener, Studying nucleation and growth kinetics of alite hy-
[26] P. Barret, D. Ménétrier, D. Bertrandie, Mechanism of C3S dissolution and problem dration using μic, Cem. Concr. Res. 39 (2009) 849–860, https://doi.org/10.1016/
of the congruency in the very initial period and later on, Cem. Concr. Res. 13 J.CEMCONRES.2009.07.004.
(1983) 728–738, https://doi.org/10.1016/0008-8846(83)90064-9. [53] T. Honorio, B. Bary, F. Benboudjema, S. Poyet, Modeling hydration kinetics based
[27] E. Pustovgar, R.P. Sangodkar, A.S. Andreev, M. Palacios, B.F. Chmelka, R.J. Flatt, on boundary nucleation and space-filling growth in a fixed confined zone, Cem.
et al., Understanding silicate hydration from quantitative analyses of hydrating Concr. Res. 83 (2016) 31–44.
tricalcium silicates, Nat. Commun. 7 (2016) 10952, , https://doi.org/10.1038/ [54] A. Bazzoni, Study of Early Hydration Mechanisms of Cement by Means of Electron
ncomms10952. Microscopy, EPFL, 2014.
[28] S.A. Rodger, G.W. Groves, N.J. Clayden, C.M. Dobson, Hydration of tricalcium [55] K. Van Breugel, Numerical simulation of hydration and microstructural develop-
silicate followed by 29Si NMR with cross-polarization, J. Am. Ceram. Soc. 71 ment in hardening cement-based materials (I) theory, Cem. Concr. Res. 25 (1995)
(1988) 91–96, https://doi.org/10.1111/j.1151-2916.1988.tb05823.x. 319–331.
[29] L. Nicoleau, E. Schreiner, A. Nonat, Ion-specific effects influencing the dissolution [56] J.J. Thomas, The instantaneous apparent activation energy of cement hydration
of tricalcium silicate, Cem. Concr. Res. 59 (2014) 118–138, https://doi.org/10. measured using a novel calorimetry-based method, J. Am. Ceram. Soc. 95 (2012)
1016/J.CEMCONRES.2014.02.006. 3291–3296, https://doi.org/10.1111/j.1551-2916.2012.05396.x.
[30] L. Nicoleau, M.A. Bertolim, Analytical model for the alite (C3S) dissolution to- [57] S. Bishnoi, K.L. Scrivener, mic: a new platform for modelling the hydration of
pography, J. Am. Ceram. Soc. 99 (2016) 773–786, https://doi.org/10.1111/jace. cements, Cem. Concr. Res. 39 (2009) 266–274, https://doi.org/10.1016/j.
13647. cemconres.2008.12.002.
[31] P. Juilland, E. Gallucci, Morpho-topological investigation of the mechanisms and [58] A.R.C.W.C. Ouzia, Modeling the Kinetics of the Main Peak and Later Age of Alite
kinetic regimes of alite dissolution, Cem. Concr. Res. 76 (2015) 180–191, https:// Hydration, EPFL, 2019, https://doi.org/10.5075/epfl-thesis-9499.
doi.org/10.1016/J.CEMCONRES.2015.06.001. [59] M. Avrami, Kinetics of phase change. I. General theory, J. Chem. Phys. 7 (1939)
[32] F. Bellmann, T. Sowoidnich, H.-M. Ludwig, D. Damidot, Dissolution rates during 1103–1112, https://doi.org/10.1063/1.1750380.
the early hydration of tricalcium silicate, Cem. Concr. Res. 72 (2015) 108–116, [60] J.J. Thomas, A new approach to modeling the nucleation and growth kinetics of
https://doi.org/10.1016/J.CEMCONRES.2015.02.002. tricalcium silicate hydration, J. Am. Ceram. Soc. 90 (2007) 3283–3288.
[33] L. Nicoleau, A. Nonat, A new view on the kinetics of tricalcium silicate hydration, [61] D.M. Kirby, J.J. Biernacki, The effect of water-to-cement ratio on the hydration
Cem. Concr. Res. 86 (2016) 1–11, https://doi.org/10.1016/J.CEMCONRES.2016. kinetics of tricalcium silicate cements: testing the two-step hydration hypothesis,
04.009. Cem. Concr. Res. 42 (2012) 1147–1156.
[34] V. Robin, B. Wild, D. Daval, M. Pollet-Villard, A. Nonat, L. Nicoleau, Experimental [62] S. Garrault-Gauffinet, T. Behr, A. Nonat, Formation of the C-S-H layer during early
study and numerical simulation of the dissolution anisotropy of tricalcium silicate, hydration of tricalcium silicate grains with different sizes, J. Phys. Chem. B 110
Chem. Geol. 497 (2018) 64–73, https://doi.org/10.1016/J.CHEMGEO.2018.08. (2005) 270–275.
023. [63] J.J. Biernacki, T. Xie, An advanced single particle model for C3S and alite hy-
[35] P. Juilland, L. Nicoleau, R.S. Arvidson, E. Gallucci, Advances in Dissolution dration, J. Am. Ceram. 94 (2011) 2037–2047.
Understanding and Their Implications for Cement Hydration, (2017), https://doi. [64] P.C. Mathur, Study of cementitious materials using transmission electron micro-
org/10.21809/rilemtechlett.2017.47. scopy techniques, Thesis EPFL (2007) 3759.
[36] J.W. Bullard, R.J. Flatt, New insights into the effect of calcium hydroxide pre- [65] A. Bazzoni, M. Suhua, Q. Wang, X. Shen, M. Cantoni, K.L. Scrivener, The effect of
cipitation on the kinetics of tricalcium silicate hydration, J. Am. Ceram. Soc. 93 magnesium and zinc ions on the hydration kinetics of C3S, J. Am. Ceram. Soc. 97
(2010) 1894–1903, https://doi.org/10.1111/j.1551-2916.2010.03656.x. (2014) 3684–3693, https://doi.org/10.1111/jace.13156.
[37] P. Juilland, E. Gallucci, Hindered calcium hydroxide nucleation and growth as [66] X. Li, K. Scrivener, Impact of ZnO on C3S hydration and C-S-H at early ages, Cem.
mechanism responsible for tricalcium silicate retardation in presence of sucrose, Concr. Res. (2019) Submitted.
12th ICSP Beijing Proc, ACI, 2018, pp. 143–154. [67] R. Kondo, S. Ueda, Kinetics and mechanisms of the hydration of cements, Tokyo
[38] K.I. Popov, N. Sultanova, H. Rönkkömäki, M. Hannu-Kuure, J. Jalonen, Symp, 1968.
L.H.J. Lajunen, et al., 13C NMR and electrospray ionization mass spectrometric [68] K.L. Scrivener, The Development of Microstructure During the Hydration of
study of sucrose aqueous solutions at high pH: NMR measurement of sucrose Portland Cement (Doctoral Dissertation), Imperial College, University of London,
dissociation constant, Food Chem. 96 (2006) 248–253, https://doi.org/10.1016/J. 1984.
FOODCHEM.2005.02.025. [69] K.O. Kjellsen, H. Justnes, Revisiting the microstructure of hydrated tricalcium
[39] D. Marchon, P. Juilland, E. Gallucci, L. Frunz, R.J. Flatt, Molecular and sub- silicate—a comparison to Portland cement, Cem. Concr. Compos. 26 (2004)
molecular scale effects of comb-copolymers on tri-calcium silicate reactivity: to- 947–956.
ward molecular design, J. Am. Ceram. Soc. 100 (2017) 817–841, https://doi.org/ [70] H.F.W. Taylor, Chemistry of cement hydration, 8th Int. Congr. Chem. Cem., Rio de
10.1111/jace.14695. Janerio, Brazil, 1986.
[40] S. Pourchet, C. Comparet, L. Nicoleau, A. Nonat, Influence of PC superplasticizers [71] S.T. Bergold, F. Goetz-Neunhoeffer, J. Neubauer, Mechanically activated alite:
on tricalcium silicate hydration, 12th Int. Congr. Chem. Cem. - ICCC 2007, 2007 new insights into alite hydration, Cem. Concr. Res. 76 (2015) 202–211.
Montreal, Canada. [72] N. Tenoutasse, Manifestations Thermiques de l'Hydration, Colloqium, Paris, 1969,
[41] L. Nicoleau, Interactions physico-chimiques entre le latex et les phases minérales p. 10.
constituant le ciment au cours de l'hydratation, Dijon (2004). [73] S. Mantellato, M. Palacios, R.J. Flatt, Relating early hydration, specific surface and
[42] A. Kumar, B.J. Walder, A. Kunhi Mohamed, A. Hofstetter, B. Srinivasan, flow loss of cement pastes, Mater. Struct. 52 (2019) 5, , https://doi.org/10.1617/
A.J. Rossini, et al., The atomic-level structure of cementitious calcium silicate s11527-018-1304-y.
hydrate, J. Phys. Chem. C 121 (2017) 17188–17196, https://doi.org/10.1021/acs. [74] F. Bellmann, G.W. Scherer, Analysis of C-S-H growth rates in supersaturated
jpcc.7b02439. conditions, Cem. Concr. Res. 103 (2018) 236–244, https://doi.org/10.1016/J.
[43] M. Schönlein, J. Plank, A TEM study on the very early crystallization of C-S-H in CEMCONRES.2017.05.007.
the presence of polycarboxylate superplasticizers: transformation from initial C-S- [75] M.R. Andalibi, A. Kumar, B. Srinivasan, P. Bowen, K. Scrivener, C. Ludwig, et al.,
H globules to nanofoils, Cem. Concr. Res. 106 (2018) 33–39, https://doi.org/10. On the mesoscale mechanism of synthetic calcium-silicate-hydrate precipitation: a
1016/j.cemconres.2018.01.017. population balance modeling approach, J. Mater. Chem. A 6 (2017) 363–373,
[44] N. Krautwurst, L. Nicoleau, M. Dietzsch, I. Lieberwirth, C. Labbez, A. Fernandez- https://doi.org/10.1039/C7TA08784E.
Martinez, et al., Two-step nucleation process of calcium silicate hydrate, the na- [76] A.E.S. Van Driessche, M. Kellermeier, L.G. Benning, D. Gebauer, New perspectives
nobrick of cement, Chem. Mater. 30 (2018) 2895–2904, https://doi.org/10.1021/ and growth nucleation on mineral, From Solutions Precursors to Solid Materials,
acs.chemmater.7b04245. 2017, https://doi.org/10.1007/978-3-319-45669-0_3.
[45] A. Kumar, Synthetic Calcium Silicate Hydrates, EPFL, 2017, https://doi.org/10. [77] T.C. Powers, Structure and physical properties of hardened Portland cement paste,
5075/epfl-thesis-7658. J. Am. Ceram. Soc. 41 (1958) 1–6, https://doi.org/10.1111/j.1151-2916.1958.
[46] S. Grangeon, A. Fernandez-Martinez, A. Baronnet, N. Marty, A. Poulain, E. Elkaïm, tb13494.x.
et al., Quantitative X-ray pair distribution function analysis of nanocrystalline [78] T.C. Powers, Properties of cement paste and concrete, 4th Int. Symp. Chem. Cem.,
calcium silicate hydrates: a contribution to the understanding of cement chem- Washington, 1960.
istry, J. Appl. Crystallogr. 50 (2017) 14–21, https://doi.org/10.1107/ [79] E. Berodier, K. Scrivener, Evolution of pore structure in blended systems, Cem.
S1600576716017404. Concr. Res. 73 (2015) 25–35, https://doi.org/10.1016/J.CEMCONRES.2015.02.
[47] A. Kunhi Mohamed, S.C. Parker, P. Bowen, S. Galmarini, An atomistic building 025.
block description of C-S-H-towards a realistic C-S-H model, Cem. Concr. Res. 107 [80] F. Avet, K. Scrivener, Investigation of the calcined kaolinite content on the hy-
(2018) 221–235, https://doi.org/10.1016/j.cemconres.2018.01.007. dration of limestone calcined clay cement (LC3), Cem. Concr. Res. 107 (2018)
[48] M. Costoya, Effect of Particle Size on the Hydration Kinetics and Microstructural 124–135, https://doi.org/10.1016/J.CEMCONRES.2018.02.016.
Development of Tricalcium Silicate (Doctoral Dissertation), Ecole Polytechnique [81] W. Hanpongpun, Investigation of the Use of Limestone Calcined Clay Cement
Fédérale de Lausanne (EPFL), 2008. (LC3) Applied to Thailand, EPFL-Lausanne, 2019, https://doi.org/10.5075/epfl-
[49] G.W. Scherer, J. Zhang, J.J. Thomas, Nucleation and growth models for hydration thesis-9005.
of cement, Cem. Concr. Res. 42 (2012) 982–993. [82] H. Freundlich, Colloid & Capillary Chemistry, Methuen, London, 1926.

15
K. Scrivener, et al. Cement and Concrete Research 124 (2019) 105823

[83] Y. Briki, No Title, Priv. Commun. (n.d.). in model cements, Cem. Concr. Res. 44 (2013) 46–54.
[84] G.W. Scherer, Stress from crystallization of salt, Cem. Concr. Res. 34 (2004) [98] D. Jansen, C. Naber, D. Ectors, Z. Lu, X.-M. Kong, F. Goetz-Neunhoeffer, et al., The
1613–1624, https://doi.org/10.1016/j.cemconres.2003.12.034. early hydration of OPC investigated by in-situ XRD, heat flow calorimetry, pore
[85] H.F. Taylor, C. Famy, K. Scrivener, Delayed ettringite formation, Cem. Concr. Res. water analysis and 1H NMR: learning about adsorbed ions from a complete mass
31 (2001) 683–693, https://doi.org/10.1016/S0008-8846(01)00466-5. balance approach, Cem. Concr. Res. 109 (2018) 230–242, https://doi.org/10.
[86] C. Yu, W. Sun, K. Scrivener, Mechanism of expansion of mortars immersed in 1016/J.CEMCONRES.2018.04.017.
sodium sulfate solutions, Cem. Concr. Res. 43 (2013) 105–111, https://doi.org/ [99] F. Zunino, K.L. Scrivener, The influence of the filler effect in the sulfate require-
10.1016/J.CEMCONRES.2012.10.001. ment of OPC and blended cements, Cem. Concr. Res. (2019) Submitted.
[87] W. Kunther, B. Lothenbach, K.L. Scrivener, On the relevance of volume increase [100] B. Lothenbach, A. Schöler, M. Zajac, M.B. Haha, F. Winnefeld, Effect of SCMs on
for the length changes of mortar bars in sulfate solutions, Cem. Concr. Res. 46 hydration and microstructure of cementitious systems, Int. RILEM Conf. Mater.
(2013) 23–29, https://doi.org/10.1016/J.CEMCONRES.2013.01.002. Syst. Struct. Civ. Eng., Lyngby, Denmark, 2016.
[88] J. Bizzozero, C. Gosselin, K.L. Scrivener, Expansion mechanisms in calcium alu- [101] E. Berodier, K. Scrivener, Understanding the filler effect on the nucleation and
minate and sulfoaluminate systems with calcium sulfate, Cem. Concr. Res. 56 growth of C-S-H, J. Am. Ceram. Soc. 97 (2014) 3764–3773, https://doi.org/10.
(2014) 190–202, https://doi.org/10.1016/J.CEMCONRES.2013.11.011. 1111/jace.13177.
[89] B. Mota-Gasso, Impact of Alkali Salts on the Kinetics and Microstructural [102] X. Ouyang, D.A. Koleva, G. Ye, K. van Breugel, Understanding the adhesion me-
Development of Cementitious Systems (Doctoral Dissertation), Ecole chanisms between C-S-H and fillers, Cem. Concr. Res. 100 (2017) 275–283,
Polytechnique Fédérale de Lausanne (EPFL), 2015. https://doi.org/10.1016/j.cemconres.2017.07.006.
[90] E. Gallucci, K. Scrivener, Crystallisation of calcium hydroxide in early age model [103] S. Pourchet, I. Pochard, F. Brunel, D. Perrey, Chemistry of the calcite/water in-
and ordinary cementitious systems, Cem. Concr. Res. 37 (2007), https://doi.org/ terface: influence of sulfate ions and consequences in terms of cohesion forces,
10.1016/j.cemconres.2007.01.001. Cem. Concr. Res. 52 (2013) 22–30, https://doi.org/10.1016/j.cemconres.2013.04.
[91] B. Mota, T. Matschei, K. Scrivener, The influence of sodium salts and gypsum on 002.
alite hydration, Cem. Concr. Res. 75 (2015) 53–65, https://doi.org/10.1016/j. [104] P.J. Scales, S.B. Johnson, T.W. Healy, P.C. Kapur, Shear yield stress of partially
cemconres.2015.04.015. flocculated colloidal suspensions, AICHE J. 44 (1998) 538–544, https://doi.org/
[92] E. Pustovgar, R.K. Mishra, M. Palacios, J.B. d'Espinose de Lacaillerie, T. Matschei, 10.1002/aic.690440305.
A.S. Andreev, et al., Influence of aluminates on the hydration kinetics of tricalcium [105] A. Schöler, B. Lothenbach, F. Winnefeld, M. Ben Haha, M. Zajac, H.-M. Ludwig,
silicate, Cem. Concr. Res. 100 (2017) 245–262, https://doi.org/10.1016/j. Early hydration of SCM-blended Portland cements: a pore solution and isothermal
cemconres.2017.06.006. calorimetry study, Cem. Concr. Res. 93 (2017) 71–82, https://doi.org/10.1016/J.
[93] M.P.E. Pustovgar, J.-B. d'Espinose de Lacaillerie, T. Matschei, N. Ruffray, R. Verel, CEMCONRES.2016.11.013.
R.J. Flatt, New insights into the retarding effect of aluminates on C3S hydration, [106] E. Berodier, Impact of the Supplementary Cementitious Materials on the Kinetics
10th ACI/RILEM Int. Conf. Cem. Mater. Altern. Bind. Sustain. Concr, 2017, pp. and Microstructural Development of Cement Hydration (Doctoral Dissertation),
14.1–14.12. Ecole Polytechnique Fédérale de Lausanne (EPFL), 2015.
[94] R.J. Myers, G. Geng, E.D. Rodriguez, P. da Rosa, A.P. Kirchheim, P.J.M. Monteiro, [107] P. Termkhajornkit, R. Barbarulo, Modeling the coupled effects of temperature and
Solution chemistry of cubic and orthorhombic tricalcium aluminate hydration, fineness of Portland cement on the hydration kinetics in cement paste, Cem. Concr.
Cem. Concr. Res. 100 (2017) 176–185, https://doi.org/10.1016/J.CEMCONRES. Res. 42 (2012) 526–538, https://doi.org/10.1016/J.CEMCONRES.2011.11.016.
2017.06.008. [108] D.P. Bentz, CEMHYD3D: A three-dimensional cement hydration and micro-
[95] G. Geng, R.J. Myers, Y.-S. Yu, D.A. Shapiro, R. Winarski, P.E. Levitz, et al., structure development modelling package, NIST Intern. Rep, 2000 https://ci.
Synchrotron X-ray nanotomographic and spectromicroscopic study of the tri- [109] A.C.A. Muller, K.L. Scrivener, A.M. Gajewicz, P.J. McDonald, Densification of C-S-
calcium aluminate hydration in the presence of gypsum, Cem. Concr. Res. 111 H measured by 1H NMR relaxometry, J. Phys. Chem. C 117 (2013) 403–412,
(2018) 130–137, https://doi.org/10.1016/J.CEMCONRES.2018.06.002. https://doi.org/10.1021/jp3102964.
[96] H. Minard, S. Garrault, L. Regnaud, A. Nonat, Mechanisms and parameters con- [110] J. Jung, W. Nishima, M. Daniels, G. Bascom, C. Kobayashi, A. Adedoyin, et al.,
trolling the tricalcium aluminate reactivity in the presence of gypsum, Cem. Concr. Scaling molecular dynamics beyond 100,000 processor cores for large-scale bio-
Res. 37 (2007) 1418–1426, https://doi.org/10.1016/J.CEMCONRES.2007.06. physical simulations, J. Comput. Chem. 40 (2019) jcc.25840, , https://doi.org/10.
001. 1002/jcc.25840.
[97] A. Quennoz, K. Scrivener, Interactions between alite and C3A-gypsum hydrations

16

You might also like