Cryst. Res. Technol., 1–16 (2013) / DOI 10.1002/crat.

201200713

1
2
3 Cement hydration: the role of adsorption and crystal growth
4
5
6 G. Artioli*1 and J. W. Bullard2
7 1
Università degli Studi di Padova, Dipartimento di Geoscienze e Centro Interdipartimentale di Ricerca per lo
8 Studio dei Materiali Cementizi e dei Leganti Idraulici (CIRCe), Via Gradenigo, 6-35131, Padova, Italy
2
9 Materials and Construction Research Division National Institute of Standards and Technology 100 Bureau
10 Drive, Gaithersburg, MD, USA
11
12 Received 13 April 2012, revised 15 August 2012, accepted 12 October 2012
13 Published online DD MM YYYY
14
15 Key words Portland cement, nucleation and growth processes, computational modelling.
16
17 Portland cement is a fundamental structural and binding material for industry and society. Its structural and
18 physical properties at different scales show a complexity that can presently be barely managed through ex-
19 perimental and computational methods. State of the art problems and trends in cement studies will be briefly
20 assessed from the point of view of the processes at the molecular scale.
21 
C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

22
23
24 1 Introduction to inorganic binding materials
25
26 Cements based on Portland-type clinkers, mortars (pastes and plasters prepared with fine aggregates), and other
27 inorganic binders form an important class of construction materials: they are all supplied as powders and when
28 mixed with water they form a fluid mass (paste) that can be shaped, moulded, added to other components,
29 or attached to the surface of other materials. The paste then hardens spontaneously at normal environmental
30 conditions.
31 Binding materials are used in buildings with the aim of (a) making structural elements for constructions,
32 (b) increasing the mechanical resistance of the construction by linking the structural and architectural elements,
33 (c) increasing water resistance and protecting masonry surfaces from environmental degradation, and (d) preparing
34 substrates for artwork and decorative purposes. Excluding last century’s binders and adhesives based on organic
35 polymers, the binders used in antiquity are inorganic compounds based on carbonates (calcite, dolomite), sulphates
36 (gypsum) and alumino-silicates (cements).
37 Starting from the properties of classical Roman materials (pozzolanic lime-based mortars) the search for
38 optimal hydraulic binders was actively pursued in several countries in the first half of the 19th century [1,2], the
39 most successful being those of J. Smeaton and J. Parker in Britain, and those of L.J. Vicat in France. The modern
40 cement formulation is commonly traced back to the three-stages process of J. Aspdin (1824), who described
41 his product as Portland cement, because of its similitude with the reputed Portland limestone, favoured among
42 British builders for its quality and durability [1].
43 The clinker is the reactive product formed by cooling a limestone and clay mixture from a high processing
44 temperature (1450 ◦ C) within a kiln. The kiln temperature is high enough to cause partial fusion, with reacted
45 molten material lumping into nodules that are partly crystalline and partly liquid. Upon cooling, the liquid phase
46 crystallizes and the whole material becomes highly reactive with water. The basic four-phase composition of
47 the clinker is reported in Table 1, though the real situation is far more complicated by the existence of several
48 polymorphs for each phase, and by compositional deviation from the indicated stoichiometry. The basic chemical
49 composition of Portland cement is shown in figure 1, where it is evident that the composition of the two major
50 phases (alite and belite) lie on the CaO-SiO2 join as a result of the lime-silica reaction.
51 The main characteristics of modern Portland clinkers are actually derived from the use of rotary kilns in place
52 of the historical shaft kilns, a technical development that also allows continuous production in place of the batch
53 process. The carefully controlled initial formulation of calcareous and argillaceous components, together with
54
∗ Corresponding author: e-mail: gilberto.artioli@unipd.it


C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 2 G. Artioli and J. W. Bullard: Cement hydration

1 Table 1 Main phases of ordinary Portland clinkers [3].

2
Phase composition cement notation wt% range
3
4 alite Ca3 SiO5 = 3CaO·SiO2 = tricalcium silicate C3S 50–70
5 belite Ca2 SiO4 = 2CaO·SiO2 = dicalcium silicate C2S 15–30
6 aluminate Ca3 Al2 O6 = 3CaO·Al2 O3 = tricalcium aluminate C3A 5–10
7 ferrite Ca2 AlFeO5 = tetracalcium aluminoferrite C4AF 5–15
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Fig. 1 Ternary CaO-Al2 O3 -SiO2 diagram showing the composition of major clinker phases and general
24
compositions of commercial cements.
25
26
27 the use of the rotary kiln that provides a long residence time at high temperature and continuous mixing, favors
28 (a) the minimization of the amount of un-reacted lime, (b) the maximization of the alite/belite ratio, and (c) the
29 appropriate crystal size of the mineral components, commonly in the range of 10 μm to 40 μm. Modern clinkers
30 contain about 60% to 65% by mass of C3S, and less than 2% of un-reacted free lime. The standard reaction
31 properties of modern Portland clinker are essentially due to a high alite/belite ratio and to careful control of the
32 grinding of the clinker into a fine powder.
33 1.1 Cement hydration: The nature of the reactions The very reactive Portland cement clinker powder
34 can be mixed with water to produce a final hardened material through a series of complex reactions, the so-called
35 hydration process, involving dissolution of the crystal phases, surface reactions, gel formation, precipitation of
36 new phases, and textural changes [3,4]. The different crystal phases present in the clinker have very different
37 reactivities, C3A having the highest and C2S the lowest. As a matter of fact, C3A is so exothermally reactive
38 that it can cause unwanted rapid setting of the paste (flash set), with concurrent loss of workability. Therefore, a
39 calcium sulphate material (commonly gypsum, but also bassanite and/or soluble anhydrite) is usually interground
40 with the clinker in the amount of about 4% to 8% by mass as a set-controlling agent.
41 The finely powdered mixture of clinker and gypsum is marketed as standard Portland cement (PC, often called
42 ordinary Portland cement), which is then mixed with water. The mass ratio of water to cement, w/c, typically
43 ranges from 0.3 to 0.5, although sufficient water to fully react with cement would require w/c to be about 0.38
44 or greater. When an aggregate material is added to the paste we have mortar (fine aggregate, such as sand with
45 grain size below 1 mm) or concrete (fine and coarse aggregate, such as coarse sands or gravel). They both can be
46 considered cement composites, with the aggregate having the roles of imparting greater stiffness and of reducing
47 the formation of fractures during the shrinkage that accompanies the hydration and hardening. The aggregate
48 may be composed of any type of loose or ground rock, and it is assumed not to react with the cement paste (hence
49 it is inappropriately called the inert component), though this is not necessarily the case.
50 Upon mixing with water the clinker phases react at different rates to produce a collection of hydration product
51 phases such as ettringite (Ca6 Al2 (SO4 )3 (OH)12 ·26H2 O), monosulphoaluminate (Ca4 Al2 (SO4 )O6 ·12H2 O), port-
52 landite (Ca(OH)2 ) and amorphous calcium silicate hydrate (C-S-H), which are the phases forming the interlocked
53 gridwork of the material and producing high mechanical resistance. The completion of the hydration process
54 may take weeks, months, or even years depending on the crystal size, the crystal defects and polymorphism of
the phases, the porosity of the paste, the environmental conditions, etc. With time portlandite may eventually


C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org
 Cryst. Res. Technol. (2013) 3

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 Fig. 2 ESEM figure of ettringite prismatic crystals growing on the clinker grains. (Image courtesy of Central
17 Analytical Laboratory, Mapei Spa).
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34 Fig. 3 ESEM figure of C-S-H and a well developed portlandite crystal growing on the clinker grains. (Image
35 courtesy of Central Analytical Laboratory, Mapei Spa).
36
37
38 convert into calcite through a carbonation reaction with atmospheric carbon dioxide, so that abundant calcite is
39 found in old cements. Figures 2 and 3 show environmental scanning electron microscope (ESEM) evidence of
40 the growth of ettringite, C-S-H, and portlandite at different times on the surface of the clinker grains.
41
42
43 2 Properties of cement
44
45 Strength development is related to the degree of hydration and to the speed of the process. When the hydration
46 process is intentionally accelerated by increasing the curing temperature there is usually a corresponding loss
47 in long term compressive strength. However, other factors that increase the hydration rate, such as using finer
48 particle sizes, either do not cause a strength loss or may even increase the ultimate strength. Significant differences
49 are sometimes observed in the difference between modern clinkers and those produced in the early part of the
50 20th century. The modern cement is required to develop high resistance to compression (over 40 MPa) within
51 a few days of its placement, whereas cements used in the early 1900s developed high strengths (in the range
52 20–30 MPa) over much longer times because of their higher belite content and coarser particle sizes.
53 Nowadays only about 35% of the cements industrially produced for the global market are so-called “ordi-
54 nary” Portland cement (PC, in the past also referred to as OPC). The remainder of the materials are special
composite formulations [5,6], in which different components are added to enhance specific physico-chemical

www.crt-journal.org 
C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 4 G. Artioli and J. W. Bullard: Cement hydration

1 Table 2 Main types of common and special cements..

2
Cement type Portland clinker wt% composition properties
3
4 Portland cement (PC) 95–100 Mixed with minor constituents Common cements
5 [CEM Ia ]
6 Special cements [CEM IIa ] 65–94 Mixed with blast furnace slags, Blended cements with enhanced
7 silica fume, natural or durability
8 calcined pozzolans, fly ash,
9 burnt shale, limestone
10 Blast furnace cement 5–64 Mixed with blast furnace slags
[CEM IIIa ]
11
Pozzolanic cement [CEM 45–89 Mixed with pozzolan, silica
12 IVa ] fume, fly ash
13 Composite cement [CEM 20–64 Mixed with blast furnace slag,
14 Va ] pozzolan, fly ash
15 High belite cements 95–100 Belite>alite obtained by fast Slow steady increase in strength,
16 quenching of clinker low heat production
17 Calcium aluminate cement 0 Made by fusion of limestone Rapid increase in strength,
18 (CAC) or high alumina and bauxite superior sulphate resistance,
19 cement (HAC) resistance to high temperature
20 Fiber cement 95–100 Mixed with reinforcing fibers Crack control, reinforcement,
21 (polymeric, steel) reduction of plastic shrinkage,
shotcrete applications
22
Cement-polymer systems variable OPC or HAC mixed with Macro-defect free (MDF)
23 water-soluble polymers materials, high compressive
24 and flexural strength
25 Ultra-rapid hardening 85–90 C11 A7 CaF2 >20 wt% Very rapid development of
26 cements strength, jet cement
27 applications
28 8–12 wt% anhydrite
29 Expansive cements variable Contain 3CA·CaSO4 The expansive component is used
30 (yeelemite), lime, or to compensate OPC shrinkage
31 high-C3A
32 a
Classification (UNI EN 197-1, 2007)
33
34
35 or mechanical properties (Table 2). The main added constituents are (a) latent hydraulic components (granu-
36 lated blast furnace slags, metakaolin, Class C fly ash) having hydraulic properties that need to be activated by
37 portlandite or hydrated lime, (b) pozzolanic aluminosilicate components (Class F fly ash, silica fume, natural
38 pozzolan) that have no hydraulic properties and need to be activated by portlandite, and (c) nonreactive or poorly
39 reactive components that modify the grindability of the clinker or the rheology of the paste, such as limestone
40 powder.
41 Furthermore, the rheological and working properties of cement pastes are now invariably controlled with
42 the use of chemical admixtures (water-soluble organic polymers) that allow good fluidity of the paste with
43 lower water/cement ratio. The net result is also a lower porosity of the paste and a higher mechanical strength:
44 high performance concrete (HPC) and high strength concrete (HSC) are made in this way and they can reach
45 compressive strengths of over 150 MPa. Organic and inorganic additives are used to control specific properties of
46 the mortars and cements, such as air entrainment (for a better resistance to freeze and thaw cycles), acceleration
47 or retardation of the setting (to control the time of progressive increase in mechanical resistance), plasticity and
48 viscosity (plasticizing or super-plasticizing effects to control the workability of the paste), and even resistance to
49 chemical attack. Modern additives are designed for specific actions, but the concept of using additives to change
50 the paste properties has been known for millennia. Ancient Mesoamerican and Chinese briefly mention the use
51 of additives in recipes for lime plasters, and the Egyptians, Minoans, and Romans also used to modify the lime
52 in mortars with a number of organic compounds including Arabic gum, animal glue, fig’s milk, egg’s yolk and
53 many others in the attempt to improve the mechanical or working properties [7]. The modern range of plasticizing
54 additives is mostly confined to polyvinylic alcohols, meta-acrylic polymers, naphthalene, poly-carboxylates, and
ligno-sulphonates, and a few others though this is a very active field of research [8–10].


C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org
 Cryst. Res. Technol. (2013) 5

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39 Fig. 4 The multi-scale nature of cement properties [11]. Adapted with permission from: “Modelling of
40 cement-based systems. The alchemy of cement chemistry” by K. van Breugel, Cement and Concrete Reasearch
41 34, issue 9, (2004) 1661–1670. Copyright
C 2004, Elsevier

42
43 Depending on the nature of the starting phases, the admixture component(s), and the environmental conditions
44 of the hydration reactions, one or more of the occurring chemical reactions can be accelerated or inhibited, because
45 of interference with one or more of the rate limiting steps (adsorption, surface and volume diffusion, dissolution of
46 primary phases, nucleation and growth of secondary phases). Such manipulation can in turn affect the multiscale
47 microstructure of the evolving cement paste (figure 4). The rheology of the cement paste, which is of great
48 importance to manufacturers and building companies, and the mechanical resistance of the hardened cement or
49 concrete, determining its application and durability, are critically controlled by the type and sequence of the
50 chemical reactions occurring during the setting and hardening stages. The complexity of the system is such
51 that the overall monitored evolution of the hydration process, typically performed by calorimetric measurements
52 (figure 5), can be deeply affected by the change of any component. This is what makes every single cement a
53 truly unique material.
54 The changes in the kinetics of the reactions can be caused by the chemistry of the solution, the crystal defects
and polymorphism of the clinker phases, the nature and solubility of the calcium sulphate phases, the presence of

www.crt-journal.org 
C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 6 G. Artioli and J. W. Bullard: Cement hydration

1
2
3
4
5
6
7
8
9
10
11
12
13 Fig. 5 Heat evolution during the hydration of a cement paste. The five steps describing the process follow
14 the convention of Gartner et al. [4]. Reprinted with permission from: “Hydration of Portland cement” by E. M.
15 Gartner, J. F. Young, D. A. Damidot, I. Jawed in: Structure and performance of cements-2nd Edition, J. Bensted
16 & P. Barnes Eds., Spon Press, London-NewYork,2002, pp 57–113. Copyright C 2002,Taylor&Francis Books

17 (UK).
18
19
impurities, environmental conditions such as a temperature and humidity, and of course by any chemical additive
20
added to the admixture.
21
22
23 3 Open problems
24
25 Here is a short list of pending problems involving cement systems and needing appropriate understanding at the
26 molecular level.
27 The mechanisms controlling hydration. Given the complexity of the cement systems and of the reactions
28 involved during hydration, it is no surprise that the phenomenological behaviour is mostly known and well
29 described, though a complete interpretation of the mechanisms involved is still largely lacking. The two main
30 models commonly invoked to explain many of the observed features of the hydration process are based on the
31 metastable barrier layer (MBL) or on the slow dissolution step (SDS) assumptions. The MBL model implies the
32 formation of an hydration layer at the surface of the hydrating clinker particles, which slows down the hydration
33 process during the so-called induction (or dormant) period (i.e. period 2 in figure 5), and then becomes unstable
34 for chemical (diffusion) or mechanical (crystallization) reasons, thus allowing the hydration to start again at
35 later times (set period 3 in figure 5) [12,13]. This model has found some support from recent experiments based
36 on nuclear resonance reaction analysis (NRRA, [14,15]). The SDS model on the other hand assumes that the
37 discontinuity in the hydration rate is caused by the contrasting rates of dissolution of the clinker phases and of
38 precipitation of the secondary phases (mainly C-S-H), which cause fluctuations in the saturation of the chemical
39 species in solution [16–18]. The debated topic concerning the mechanism(s) causing the induction period is
40 still fiercely discussed in the cement community. This is certainly an area of poorly understood (albeit critically
41 important) surface processes. Details on the mechanisms and computational demands are discussed below.
42 The control of rheology. Another area of substantial debate is the role of crystal morphology and surface
43 interaction between particles in determining and controlling the rheology of the cement paste [19,20]. The
44 rheology is affected by the micro-texture of the paste, and the micro-textural evolution of the system in turn
45 is affected by the dissolution-crystallization reactions during hydration [21,22]. Again, how the modern super-
46 plasticizing additives operate to change the kinetics of the reactions is empirically observed but hardly modelled
47 at a molecular level [23,24]. The organic-inorganic interaction at the surface of the hydrating particles is a key
48 process that is still poorly investigated, though some modelling at the molecular level has been attempted [25].
49 The atomic and molecular structure of C-S-H. Despite a good deal of evidence in recent years indicating
50 that the C-S-H phases have an essentially tobermoritic or jennitic structure at the nanoscale [26–28], their poor
51 crystallinity and short range character have prevented a satisfactory structural description. How far existing
52 structural models [29–32] can be applied to real systems is an active field of investigation. An intimate structural
53 description of C-S-H is needed to model crystallization and growth of this major phase involved in cement set
54 and hardening, but structural information alone is not sufficient. A number of almost untouched issues include
the structural incorporation of the organic additives [33,34], the role of impurities and disorder in the long-term


C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org
 Cryst. Res. Technol. (2013) 7

1 durability of these phases, and the possible structural role of C-S-H in the stabilization of hazardous or toxic
2 species.
3
4
5 4 The role of adsorption and crystal growth in the hydration process
6
The hydration process starts immediately upon wetting of the clinker by rapid reactions between alite and C3A
7
and water:
8
9 C3S + 3H2 O → 3Ca+2 + H2 SiO−2
4 + 4(OH)
−
(1)
10
11 C3A + 6H2 O → 3Ca+2 + 2Al(OH)−
4 + 4(OH)
−
(2)
12
These reactions have a large exothermic character (see initial peak in stage 1, figure 5), consistent with a large
13
thermodynamic driving force for dissolution. The enthalpy of congruent dissolution of alite is −138 kJ/mol and of
14
C3A is −248 kJ/mol, based on the reactions above [35,36]. Solution chemical analyses have furnished persuasive
15
evidence that both C3S and C3A dissolve congruently and quite rapidly in the first seconds after wetting. In
16
dilute suspensions of C3S, for example, the increase in silicate concentration over the first 30 s suggests that the
17
dissolution rate may be at least 10 μmol · m−2 · s−1 [37]. In fact, Stein [38] calculated a theoretical solubility
18
product for C3S of Ksp ≈ 3, which would imply that C3S should continue to dissolve until reaching equilibrium
19
calcium and silicate concentrations in solution of several hundred mmol/L. In fact, it is well known that C3S
20
and alite dissolution rates decelerate very quickly while the solution is still undersaturated, by about 17 orders of
21
magnitude with respect to the ion activity product of reaction (1) compared to the equilibrium calculation, by the
22
end of Stage 1 [13]. The reason for slow alite dissolution while the solution is still significantly undersaturated
23
is one of the puzzles of hydration. Possible mechanisms (i.e. MBL and SDS model) have been shortly addressed
24
above, each one does indeed show advantages and disadvantages in explaining the experimental evidence in
25
detail. The induction period (Stage 2, figure 5) appears to exist as a distinct stage only when chemical retarders
26
are added, indicating a complex interplay between the C3S/C3A dissolution, the C-S-H precipitation, and the
27
crystallization and growth of intermediate phases in the presence of gypsum (AFt and AFm phases, especially
28
ettringite). Ettringite is known to play a critical role in controlling the rheology of the paste in the early stages
29
of hydration, so that the interaction of the organic additives with the sulphate-containing phases is of extreme
30
interest to develop control of the fluidity [22,23]. When alite starts rehydrating again, at the onset of stage 3
31
(figure 5), the reaction shows a typical autocatalytic behaviour, since the rate of hydration depends on the amount
32
of some hydration product, essentially C-S-H. Experimental evidence, especially from microscopic techniques,
33
indicates that the rate controlling-factor is the heterogeneous nucleation and growth of C-S-H on alite [31,39],
34
and/or the aggregation of the C-S-H nuclei and growing particles [27,40,41]. The process can be greatly
35
accelerated by increasing the nucleation process (for example by seeding [42]) and its kinetics is also remarkably
36
affected by the impurities and defects present in the structure of alite, besides the intrinsic stacking and structural
37
disorder characteristic of the C-S-H sheets. All structural models of C-S-H phases include a remarkable degree
38
of disorder at the nanoscale [29–32], which explains the anisotropic difficult growth of the C-S-H crystals, whose
39
morphology is variously described as frustrated sheets, crumpled foils, and other creative forms. The same
40
structural reason is at the base of the limited crystalline growth of the natural analogues of the calcium silicate
41
hydrates (tobermorites, gyrolite, xonotlite, afwillite, hillebrandite, foshagite, rankinite, kilchoanite, truscottite,
42
Ca-chondrodite, etc. [43]), which are seldom found as good single crystals of reasonable size.
43
It is thus quite obvious that molecular adsorption, crystal nucleation and crystal growth are fundamental pro-
44
cesses acting simultaneously in the different stages of cement hydration, and any small variation in the kinetics of
45
these processes greatly affects the microstructural evolution of the material. The complex transformation kinetics
46
observed between these phases [44] and the colloidal nature of C-S-H [45,46] makes the deep understanding of
47
the real structure and microstructure of cement a substantial challenge, both from the experimental point of view
48
(Section 6) and from the theoretical-computational point of view (Section 7).
49
50
51 5 Experimental challenges
52
53 A multi-scaled material such as cement (figure 4), whose physical and mechanical properties at the macroscale
54 are closely related to its microstructure and nanostructure, can be profitably investigated by different techniques
to obtain complementary information.

www.crt-journal.org 
C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 8 G. Artioli and J. W. Bullard: Cement hydration

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16 Fig. 6 Nanoscale layered organization of C-S-H phases [31]. Reprinted with permission from: “Tober-
17 morite/jennite and tobermorite/calcium hydroxide-based models for the structure of C-S-H: applicability to
18 hardened pastes of tricalcium silicate, Portland cement, and blends of Portland cement with blast-furnace slag,
19 metakaolin, or silica fume” by I.G. Richardson, Cement and Concrete Research 34, issue 9 (2004) 1733–1777.
20 CopyrightC 2004, Elsevier.

21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36 Fig. 7 X-ray absorption tomographic image of a cement pellet used for waste solidification, showing virtual
37 sections (left), and the pore structure of the pellet extracted through analysis of the reconstructed image (right).
38 (Images courtesy of M. Parisatto @ UNIPD).
39
40
41 Schematically we may identify several techniques that have been successfully employed to visualize (AFM,
42 HR-TEM) or characterize (XRD, MAS-NMR) the structure at the atomic and molecular level. Figure 6 shows
43 the organization of the crystalline C-S-H phases at the nanoscale within the calcium silicate hydrate gel.
44 The structural organization of the crystalline regions are usually derived from the single crystals of natural
45 analogues such as 11 Å or 14 Å tobermorites and jennite [29,30], whereas the local crystallochemical arrangement
46 is typically characterized by high resolution MAS-NMR spectroscopy [47,48]. Quantification of the crystalline
47 phases present in the starting powder and their evolution during hydration is commonly performed by full-profile
48 Rietveld-type analysis of the powder diffraction spectra. In situ X-ray or neutron powder diffraction is probably
49 the optimal technique to follow the phase changes through the process [49,50], though many of the intermediate
50 phases, especially the aluminate hydrates, do not yet have an appropriate description of their crystal structural
51 model [51]. Imaging of the nanostructure of cement is performed by HR-TEM (figure 6) and by AFM [38] though
52 these techniques are intrinsically invasive and generally affect the process and/or the texture of the paste. There
53 are claims that the electron beam may induce formation of crystalline domains in the gel and/or damage in the
54 existing structures [31]. The visualization of the unperturbed microstructure of cement may be obtained through
hard X-ray tomographic techniques (figure 7).


C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org
 Cryst. Res. Technol. (2013) 9

1 By using fluorescence or diffraction signals for the tomographic reconstruction, it is in principle possible to
2 map the three-dimensional distribution of the phases within the sample without any perturbation of the chemical
3 reactions and the microstructure [52]. Present resolution of laboratory tomography is of the order of 0.5 μm,
4 whereas with the advent of the next generation optics, nanoscale resolution should be reached at synchrotron beam
5 lines (http://www.esrf.eu/files/Upgrade/LTS_060706.pdf). The final goal of course is to obtain simultaneously
6 time-, space-, and phase-resolved information. These are needed to reliably assess the spatial evolution of the
7 phases involved in the hydration process, and to fill the gap between theoretical modelling and the interpretation
8 of the macroscopic properties.
9
10
11 6 Progress and challenges in computational modeling
12
13 The mechanisms of C-S-H growth are critical to understanding the development of microstructure and aiding
14 the discovery of ways to control the rate of hydration. The qualitative understanding of the relationship between
15 accelerated hydration and C-S-H growth is reasonable, but the quantitative relationship between and is more
16 difficult to establish. Computational models are used to systematically vary the system parameters and understand
17 their control on the reaction kinetics, on the nature of the products, and on the microstructure of the paste from
18 the nanoscale [53] to the macroscale [54].
19 The earliest mathematical models of cement hydration kinetics focused on the hydration of a single spherical
20 particle with a uniform layer of hydration product formed on its surface and gradually thickening [55,56].
21 Although analytically tractable and giving an expression for the rate of hydration in terms of particle dissolution
22 rates and diffusion rates through the hydrated layer, these models were based on several unwarranted assumptions
23 and were so idealized that they gave little insight into the underlying causes of the induction period or general
24 trends in microstructure development where particle size distributions become important.
25 An alternative and currently popular theoretical approach is to employ phenomenological equations for
26 nucleation and growth of C-S-H similar to those derived by Avrami [57], Johnson and Mehl [58], and Kolmogorov
27 [59] to describe first order phase transformations in melts or single crystals. The main attraction of these models
28 is their simplicity of application and ability to fit experimental calorimetry data in the vicinity of the main
29 exothermic peak. However, there is a growing recognition that these models are based on underlying assumptions
30 that are dubious at best when applied to cement hydration. Among these assumptions are that the thermodynamic
31 driving force for the transformation is constant throughout the process, that the nucleating phase forms at random
32 locations throughout the volume of the parent phase, and that diffusion is not required at any stage of the
33 transformation. Consequently, these models cannot make any account of the induction period, solution chemistry
34 or other microstructural detail, and their fits to experimental calorimetry data break down shortly following the
35 main alite hydration exotherm. Thomas has removed at least the assumption of random nucleation in his recent
36 boundary nucleation and growth model [60] by forcing C-S-H to nucleate only on the surface of C3S particles
37 and has therefore been able to fit calorimetry data to somewhat longer times.
38 Beginning with the seminal work of Jennings and Johnson in 1986 [61], attention turned to modeling more
39 realistic microstructures of cement paste, and these have now become powerful tools for understanding and
40 predicting the development of mechanical and transport properties in these materials. The CEMHYD3D sim-
41 ulation model [62] developed at the National Institute of Standards and Technology (NIST) is the best-known
42 and most widely used of these models. In CEMHYD3D, the 3D microstructure of cement paste is represented
43 on a discretized lattice, capturing realistic particle size and shape distributions, the volume fraction and sur-
44 face area fractions of the initial phases, and the spatial distribution of those phases (figure 8). Changes to the
45 microstructure are simulated through a large set of rules that are evaluated locally at each lattice site, and
46 that depend on the materials involved in the reaction. The resulting hydrated microstructures have a realistic
47 appearance, quantitatively reasonable spatial distribution of phases, and realistic topology of capillary poros-
48 ity down to a minumum pore size of 1 μm. Other properties that can be followed with CEMHYD3D are the
49 heat evolution, chemical shrinkage, phase percolation and setting. The main limitations of CEMHYD3D are
50 that its rules are not formulated to provide an inherent time scale to the changes that occur, and that the in-
51 timate relation between solution chemistry and the thermodynamics and kinetics of the coupled reactions is
52 not modeled. These limitations render CEMHYD3D poorly suited to examine detailed mechanisms of hydra-
53 tion and the role of chemical additives in modifying the reaction rates by selective adsorption or other subtle
54 mechanisms, although it is still a quite powerful tool for tracking general features of hydration that depend on
microstructure.

www.crt-journal.org 
C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 10 G. Artioli and J. W. Bullard: Cement hydration

1
2
3
4
5
6
7
8
9
10
11
12
13 Fig. 8 (Left) SEM false color segmented image of a PC (sample 152 from the Cement and Concrete Reference
14 Laboratory Proficiency Sample program). (Right) 2D slice from a 3D reconstruction of the same cement with
15 a slightly lower packing fraction of particles. The phases are colored as follows: brown = C3S, blue = C2S,
16 gray = C3A, white = C4AF, yellow = gypsum, red = arcanite, green = CaCO3 , cyan = quartz.
17
18
19 To address these limitations, a kinetic cellular automaton model has been developed at NIST to simulate
20 the chemical kinetics of coupled reactions among solids in heterogeneous aqueous suspensions [63–65]. This
21 model has been applied recently to examine microstructural development and solution chemistry during C3S
22 hydration [66,67]. In the absence of unequivocal evidence supporting either the MBL hypothesis or the SDS
23 hypothesis of the induction period, the model was used to examine the plausibility and implications of each
24 hypothesis. The results showed that, no matter which of the two hypotheses is chosen, only heterogeneous
25 nucleation of C-S-H on C3S surfaces can account for experimental observations of the evolution of the solution
26 composition and degree of hydration in Stages 1 through 3. Those simulations also differentiated between
27 heterogeneous nucleation and growth of C-S-H, and indicated that nucleation occurs in a fairly short burst over
28 only a few minutes, either at the end of Stage 0 when assuming the SDS hypothesis or near the beginning of
29 Stage 3 when assuming the MBL hypothesis. The narrow window of C-S-H nucleation occurs in the simulations
30 because nucleation consumes calcium and especially silicates from solution, thus lowering the saturation index of
31 C-S-H enough that the thermodynamic driving force for continued precipitation becomes less than the energetic
32 barrier to nucleation.
33 These basic insights about the timing, location, and rates of nucleation and growth of C-S-H still provide no
34 clear way to differentiate between the MBL and SDS hypotheses, and yet a better knowledge of the trigger for
35 ending the induction period could have important consequences for designing chemical additives that control
36 setting time. Therefore, subsequent uses of the model have investigated whether one hypothesis or the other makes
37 distinguishable predictions about the influence of different mineral phases on the kinetics of C3S hydration. The
38 investigation focused on the role of portlandite because (1) it is the only other major product of C3S hydration
39 other than C-S-H, (2) portlandite is often observed to first precipitate near the end of the induction period, and
40 (3) once formed, portlandite acts as a buffer controlling the calcium concentration in solution. In fact, the close
41 correlation between portlandite formation and the end of the induction period has led some researchers to propose
42 that portlandite precipitation actually causes the end of the induction period. Motivated by this, the model was
43 used to test both the MBL and SDS hypotheses in simulations where portlandite was artificially prevented from
44 forming.
45 The simulation results revealed a significant difference in behaviors depending on which hypothetical mech-
46 anism governs the induction period. Figure 9 shows the predicted degree of hydration (i.e. mass fraction of C3S
47 consumed) as a function of time for three different scenarios. In the first simulation (open circles in figure 9), the
48 SDS hypothesis was used to simulate normal hydration of C3S in water without any restriction on portlandite
49 precipitation. In the second simulation, (open squares in figure 9), the SDS hypothesis was used to simulate
50 hydration of the same system except that portlandite was not allowed to precipitate. Finally, a third simulation
51 (open diamonds in figure 9) was identical to the second simulation except that the MBL hypothesis was used in
52 place of the SDS hypothesis. The figure clearly shows a distinct difference in the effect of suppressing portlandite.
53 When the SDS mechanism is operative, suppressing portlandite has a strong retarding effect on hydration; the
54 induction period is prolonged by about 10 hours before a weak accelerating period begins. However, when the
MBL mechanism is operative, there is no measurable retarding effect. The length of the induction period is


C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org
 Cryst. Res. Technol. (2013) 11

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17 Fig. 9 Simulated rates of hydration of C3S as a function of time using NIST’s kinetic cellular automaton
18 model, testing the influence of the MBL and SDS hypothesis on changing the kinetics when the precipitation
19 of portlandite is suppressed
20
21
22 virtually the same as in the control simulation, although the acceleration period which follows is shorter and a
23 period of steady but slower hydration ensues.
24 The underlying reason for the differences in retarding effect is the higher calcium concentrations in solution
25 when portlandite is not able to precipitate. Under the SDS hypothesis, much higher calcium concentrations force
26 the silicate concentration to lower values, on the order of 0.1 μmol/L to maintain a near-equilibrium condition
27 with C3S. The lower silicate concentrations result in a lower driving force for C-S-H growth. Therefore, neither
28 the dissolution rate of C3S nor the growth rate of C-S-H is as great when portlandite cannot form and the induction
29 period is extended. However, the situation is much different under the MBL hypothesis. When a metastable layer
30 of hydrate protects highly soluble C3S from dissolving, this layer is destabilized by the formation of C-S-H,
31 which happens early in the induction period. The destruction of the metastable layer causes renewed dissolution
32 of C3S, and the higher calcium concentrations in solution due to the absence of portlandite has virtually no effect
33 on the rate of dissolution of C3S because its solubility product is so high. Similarly, the silicate concentration is
34 not reduced as much because it continues to be supplied to solution by C3S dissolution, and C-S-H can continue
35 to grow. Therefore, neither the induction period nor the beginning of the acceleration period is noticeably affected
36 by suppressing portlandite when the MBL mechanism is operative. Incidentally, we remark that the apparent
37 reason for the shortening of the acceleration period under the MBL hypothesis is a continual re-formation and
38 dissolution of the metastable hydrate at the C3S surfaces, which provides partial protection and decrease of the
39 C3S dissolution rate. We cannot yet confirm that this latter effect is realistic, although Gartner has proposed a
40 similar effect in an earlier review [68].
41 Another area, directly related to adsorption and crystal growth phenomena in cement, where these kinds
42 of models can provide useful insight and guidance is in understanding the influence of calcium sulfate in
43 controlling the rate of hydration of Portland cement and pozzolanic cement (defined in Table 2). Since the
44 seminal study by Tenoutasse in 1968 [69] it has been recognized that C3S hydration is strongly retarded in
45 Portland cements where the sulfate content is insufficient to promote complete conversion of C3A to ettringite.
46 Likewise, the presence of calcium sulfate in solution is known to mildly retard the hydration of pure C3S
47 pastes [70]. However, a compelling mechanistic explanation of these admixture effects has not been given.
48 Indeed, these admixture effects seem unusual at first because sulfate species do not participate in the reactions for
49 simple C3S dissolution or growth of C-S-H or portlandite. A key to understanding these admixture effects may be
50 to understand whether selective adsorption of ions occurs at reactive surface sites of clinker minerals or hydration
51 products.
52 Recently, nanoscopic surface techniques such as atomic force microscopy (AFM) or vertical scanning inter-
53 ferometry (VSI), along with molecular-scale kinetic Monte Carlo or molecular dynamics simulations have been
54 used to elucidate the mechanisms of interaction between adsorbates and mineral surfaces [71], and these same
techniques may eventually be applied to sulfate interactions with cement. Even so, the kinetic cellular automaton

www.crt-journal.org 
C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 12 G. Artioli and J. W. Bullard: Cement hydration

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 Fig. 10 Time dependence of solid phases during a kinetic cellular automaton simulation of hydration of the
19 C3A-gypsum-portlandite system.
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34 Fig. 11 Evolution of the solution composition for the simulation of hydration of the C3A-gypsum-portlandite
35 system depicted in Fig. 10.
36
37
38 approach can be used now to begin analyzing some possible mechanisms by which sulfates alter hydration
39 kinetics.
40 A random particulate microstructure in water was created with C3A, gypsum, and portlandite particles, with
41 the initial volume fractions being 0.14, 0.06, and 0.06, respectively. The portlandite was added because its
42 influence in buffering the calcium concentration in cement provided solution compositions that are similar to
43 those in low-alkali Portland cement pastes. The only reactions that were assumed were the congruent, reversible
44 dissolution/growth of gypsum, portlandite, C3A, ettringite, calcium monosulphoaluminate, and the hexagonal
45 C4AH19 phase. In addition, the solution speciation reactions for formation of CaOH+ and CaSO4 0 complexes
46 were assumed. Stoichiometries and equilibrium constants for these reactions were taken from standard ther-
47 mochemical databases [35,36]. Finally, the reversible adsorption of CaSO4 0 on active dissolution sites of C3A
48 was also assumed, as suggested by Minard [72]. This adsorption reaction is assumed to slow the rate of C3A
49 dissolution in proportion to the surface concentration of the adsorbate, and is the only plausible mechanism that
50 has been proposed to explain the rapid reduction in dissolution rates of C3A that occurs at solution concentrations
51 far below the saturation point. Nevertheless, direct evidence for adsorption reaction is lacking and in the absence
52 of experimental data, we have assumed that it involves the single-step attachment of CaSO4 0 on the surface, with
53 an equilibrium constant of 10−4 .
54 Figure 10 shows the simulated changes in the volume fraction of all the solid phases as a function of time,
and figure 11 shows the change in solution composition for the same system. The dissolution of C3A is initially


C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org
 Cryst. Res. Technol. (2013) 13

1 very rapid, consistent with its high solubility, but decreases within about 10 min to a steady value as indicated
2 by the constant rate of decrease in the C3A volume fraction. Likewise, gypsum also dissolves fairly rapidly at
3 the beginning but slows to a steady state value. Ettringite, [Ca3 Al(OH)6 ·12H2 O]2 ·(SO4 )3 ·2H2 O, a very sparingly
4 soluble mineral, nucleates and grows rapidly almost as soon as C3A and gypsum begin dissolving, but once again
5 reaches a steady state value for several hours afterward. Between 10 min and 6 h, the kinetics of the system is
6 controlled by the dissolution of C3A, which is slowed by the initial adsorption of CaSO4 0 . Ettringite grows at
7 a steady rate governed by the rate at which aluminate ions are made available in solution by the dissociation of
8 C3A, and gypsum dissolves at a rate that keeps it in equilibrium with the solution as sulfate ions are consumed by
9 ettringite formation. During this time interval, the solution composition is nearly constant (figure 11). However,
10 this kinetic behavior begins to change once all of the gypsum is consumed at about 6 h. When this happens,
11 ettringite continues to grow for about another hour, although its rate steadily decreases because the continual
12 reduction in sulfate concentration reduces the driving force for ettringite growth (see right panel of figure 11).
13 In addition, when the ettringite has removed most of the remaining sulfates from solution, sulfates adsorbed on
14 C3A spontaneously desorb by Le Chatelier’s principle. When this happens, the dissolution rate of C3A sharply
15 increases shortly after 7 h have elapsed. The renewed dissolution rapidly increases the concentrations of calcium,
16 aluminates, and hydroxyls in solution (see right panel of figure 11). C4AH19 nucleates and rapidly grows in this
17 solution, causing the concentrations of these species to be reduced somewhat.
18 As already described, the general trends in the dissolution of C3A shown in figure 10 are typically observed
19 experimentally in the C3A-gypsum system [68]. It had long been thought that the initial decrease in C3A
20 dissolution rate was caused by the formation of a layer of ettringite around the C3A particles which, like the
21 hypothetical metastable hydrate in the C3S system, partially sealed the particles off from the solution. The renewal
22 of C3A dissolution at later times was thought to be due to a transformation of the ettringite layer to a more porous
23 calcium mono-sulfoaluminate, the latter phase being thermodynamically preferred to ettringite once sulfates have
24 been removed from solution. However, an abundance of microscopic evidence suggests that ettringite forms as
25 rather shortened needles nearby, but not on, the C3A particle surfaces, and it is difficult to see how these needles
26 could form a protective layer. More importantly, recent experimental work has shown that the transformation
27 from ettringite to calcium mono-sulfoaluminate happens quite slowly, perhaps over a few days time [72]. The
28 current simulations are consistent with this observation, for even though calcium monosulfoaluminate is able to
29 form in the simulations, it is never observed to do so for as long as three hours after the sulfates have been removed
30 from the solution by ettringite. In any event, the transformation from ettringite to calcium monosulfoaluminate
31 is unlikely to occur rapidly enough to account for the rapid change in C3A dissolution rates at later times.
32 In cement technology jargon, one would say that the system just discussed was “undersulfated” because the
33 gypsum was consumed before all the C3A reacted. A basic stoichiometric analysis suggests that a 3:1 molar ratio
34 of gypsum:C3A (or, equivalently, a 5:2 volume ratio) is required to allow complete conversion to ettringite. In
35 contrast, the system simulated here, which was formulated to mimic one reported by Gartner [68], had only a 3:7
36 volume ratio.
37 In the C3S-C3A-gypsum system, which is similar to Portland cement, an undersulfated condition causes the
38 formation of hexagonal aluminate hydrates like C4AH19, just as seen in these simulations, but it also causes
39 the acceleration period of C3S dissolution to be both delayed and substantially weakened [69], The result is
40 a severe retardation and delay of setting in undersulfated Portland cements. The reason for this effect is not
41 clear. It is known that aluminates can substitute for silicates in the “drierketten” silicate chains if the aluminate
42 concentration in solution is sufficiently high. It has been speculated that an undersulfated condition causes an
43 elevation in aluminate concentrations, similar to those observed at later times in figure 11, and that the aluminate
44 substitution in C-S-H causes a significant reduction in its growth rate [73]. This argument seems plausible,
45 although direct experimental evidence of a significant dependence of the growth rate of C-S-H on its composition
46 has not been obtained. Another proposal is that the hexagonal aluminate hydrates, which form so rapidly and
47 abundantly in the absence of sulfates, effectively fill the pore space to such an extent that there is no room left for
48 C-S-H to grow [74]. Because there is strong evidence that C-S-H growth controls the rate of hydration of C3S,
49 as discussed earlier, this argument also seems plausible. Again, however, direct experimental evidence of space
50 filling by an aluminate hydrate in undersulfated cement paste is sparse.
51 The undersulfation problem in Portland cement is so well known that it is common practice to determine
52 the “optimum” sulfate level for a Portland cement clinker. The optimum sulfate level is determined not just
53 by the aluminate stoichiometry, but also seems to be influenced by other factors such as the surface area
54 (fineness) of the cement particles, as well as the distribution and polymorphic form of the clinker phases.
Therefore, the determination of optimum sulfate levels is often made by trial and error experimentation, using

www.crt-journal.org 
C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 14 G. Artioli and J. W. Bullard: Cement hydration

1 isothermal calorimetry to detect the timing and strength of the main C3S hydration peak, and then inter-
2 grinding this concentration of gypsum to make the Portland cement. Unfortunately, this same sulfate level is
3 unlikely to be optimal for the same Portland cement with partial replacement by fly ash or other pozzolanic
4 materials, because these latter materials themselves often contain calcium aluminates in the form of crystals
5 or reactive glassy phases. Under-sulfated conditions in these blended cements often leads to what concrete
6 technologists call “incompatibility”, which is a failure to set. Because of its importance in effectively utiliz-
7 ing sustainable concrete materials and because of the cumbersome trial-and-error experimental methods that
8 are required to determine optimum sulfate levels in these materials, one major objective of this kind of com-
9 puter modeling research is to provide a tool that can help predict the optimum sulfate level for any cement
10 composition.
11
12
13 7 Summary
14
15 The hydration of cement paste is responsible for the development of essentially every engineering property of
16 concrete, including, but not limited to, setting and hardening, temperature rise in massive concrete structures,
17 autogenous shrinkage that can lead to cracking, and resistance to leaching, sulfate attack and other degradation
18 phenomena. This fact provides a high motivation to understand the fundamental mechanisms of cement hydration
19 with the aim of controlling the properties of interest in a given situation. Progress in this area has been hindered
20 by the significant complexity of the hydration process, from the points of view of chemistry, kinetics, and
21 microstructure development. This chapter has discussed various aspects of these complexities from the point of
22 view of the impact of adsorption and crystal growth phenomena. Recent advances in experimental techniques [75]
23 and computational modeling [76] are beginning to provide a more basic understanding of these processes in cement
24 paste, but there are still a large number of unanswered questions that will require sustained interdisciplinary
25 research efforts well into the future.
26
27
28
29 References
30 [1] R. G. Blezard, in: The History of Calcareous Cements, edited by P. C. Hewlett, Lea’s Chemistry of Cement and Concrete,
31 4th Edition (Butterworth-Heimann, Oxford, 1998), pp. 1–23.
32 [2] A. Bentur, Cementitious materials – Nine millennia and a new century: Past, present, and future. American Society of
33 Civil Engineers, 150th Anniversary Paper. J. Mater. Civ. Eng., JAN/FEB 2002, pp. 2–22.
34 [3] H. F. W. Taylor, Cement chemistry. 2nd Edition. (Thomas Telford, London, 1997).
35 [4] E. M. Gartner, J. F. Young, D. A. Damidot, and I. Jawed, Hydration of Portland cement, in: J. Bensted, P. Barnes (Eds.),
36 Structure and Performance of Cements, 2nd ed. (Spon Press, London – New York, 2002), pp. 57–113.
37 [5] I. Odler (ed.), Special Inorganic Cements (Spon Press, London – New York, 2000).
38 [6] A. K. Chatterjee, in Special Cements, edited by J. Bensted, P. Barnes, Structure and Performance of Cements, 2nd ed.
39 (Spon Press, London – New York, 2002), pp. 186–236.
[7] G. Artioli, Scientific methods and cultural heritage (Oxford University Press, Oxford, 2010).
40
[8] V. S. Ramachandran, V. M. Malhotra, C. Jolicoeur, and N. Spiratos, Superplasticizers: properties and applications in
41 concrete (CANMET, Materials Technology Laboratory, Ottawa, 1998).
42 [9] R. Rixom and N. Mailvaganam, Chemical Admixtures for Concrete, 3rd ed. (E & FN Spon, London, 1999).
43 [10] S. Chandra, in: Properties of Concrete with Mineral and Chemical Admixtures, edited by J. Bensted, P. Barnes, Structure
44 and Performance of Cements, 2nd ed. (Spon Press, London – New York, 2002), pp. 140–185.
45 [11] K. Van Breugel, Cem. Concr. Res. 34, 1661 (2004).
46 [12] H. M. Jennings and P. L. Pratt, Cem. Concr. Res. 9, 501 (1979).
47 [13] E. M. Gartner and H. M. Jennings, J. Am. Ceram. Soc. 80, 743 (1987).
48 [14] R. A. Livingston, J. S. Schweitzer, C. Rolfs, H. W. Becker, and S. Kubsky, J. Mater. Res. 16, 687 (2001).
49 [15] J. S. Schweitzer, R. A. Livingston, C. Rolfs, H. W. Becker, S. Kubsky, T. Spillane, M. Castellote, and P. G. de Viedma,
Nucl. Instrum. Meth. Phys. Res. B 241, 441 (2005).
50
[16] S. Garrault and A. Nonat, Langmuir 17, 8131 (2001).
51 [17] S. Garrault-Gauffinet and A. Nonat, J. Cryst. Growth 200, 565 (1999).
52 [18] S. Garrault, E. Finot, E. Lesniewska, and A. Nonat, Mater. Structures 38, 435 (2005).
53 [19] A. Nonat (ed.) Hydration and Setting. Why Does Cement Set? An Interdisciplinary Approach, Proc. 2nd International
54 RILEM Symposium (PRO13, RILEM Publications, Cachan, 2000).
[20] R. J. Flatt, Mat. Struct. 37, 289 (2004).


C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org
 Cryst. Res. Technol. (2013) 15

1 [21] F. Cella, T. Cerulli, D. Salvioni, and S. Stella, Morphological variation of cement paste microstructure due to the use of
2 admixtures. Proceedings of the 23th International Conference on Cement Microscopy, pp. 117–136, (2001).
3 [22] C. Rößler, A. Eberhardt, H. Kucerova, and B. Moser, Cem. Concr. Res. 38, 897 (2008).
4 [23] C. Rößler and J. Stark, in: The Influence of Superplasticizers on the Microstructure Development in Normal Portland
5 Cement and C3S, edited by P. C. Nkinamubanzi, Seventh CANMET/ACI International Conference on Superplasticizers
and other Chemical Admixtures in Concrete in Berlin, Supplementary papers, pp. 17–32, (2003).
6
[24] E. Sakai, K. Yamada, and A. Ohta, J. Adv. Concr. Techn. 1, 16 (2003).
7 [25] J. L. W. Griffin, P. V. Coveney, A. Whiting, and R. Davey, J. Chem. Soc. Perkin Trans. 2, 1973 (1999).
8 [26] X. Zhang, W. Chang, T. Zhang, and C. K. Ong, J. Am. Ceram. Soc. 83, 2600 (2000).
9 [27] A. J. Allen, J. J. Thomas, and H. M. Jennings, Nature Materials 6, 311 (2007).
10 [28] J. Skibsted and C. Hall, Cem. Concr. Res. 38, 205 (2008).
11 [29] E. Bonaccorsi, S. Merlino, and A. R. Kampf, J. Am. Ceram. Soc. 88, 505 (2005).
12 [30] E. Bonaccorsi, S. Merlino, and H. F. W. Taylor, Cem. Concr. Res., 34, H. F. W. Taylor Commemorative Issue, 1481
13 (2004).
14 [31] I. G. Richardson, Cem. Concr. Res. 34, H. F. W. Taylor Commemorative Issue, 1733 (2004), See also I. G. Richardson,
15 Electron Microscopy of Cements, in: Structure and performance of cements, edited by J. Bensted, P. Barnes, 2nd ed.
16 (Spon Press, London – New York), pp. 500–556, (2002).
[32] I. G. Richardson, Cem. Concr. Res. 38, 137 (2008).
17
[33] A. Popova, G. Geoffroy, M.-F. Renou-Gonnord, P. Faucon, and E. Gartner, J. Am. Ceram. Soc. 83, 2556 (2000).
18 [34] J. J. Beaudoin, H. Dramé, L. Raki, and R. Alizadeh, Mater. Struc. 42, 1003 (2009).
19 [35] D. L. Parkhurst, User’s guide to PHREEQC-a computer program for speciation reaction-path, advective-transport, and
20 geochemical calculations, Water-Resources Investigations Report 95–4227, U.S. Geological Survey, (1995).
21 [36] W. Hummel, U. Berner, E. Curti, F. J. Pearson, and T. Thoenen, Nagra / PSI Chemical Thermodynamic Data Base
22 01/01 (Universal Publishers, Parkland, Florida, 2002).
23 [37] S. Garrault and A. Nonat, Langmuir 17, 8131 (2001).
24 [38] H. N. Stein, Cem. Concr. Res. 2, 167 (1972).
25 [39] S. Gauffinet, E. Finot, R. Lesniewska, and A. Nonat, Earth Planet. Sci. 327, 231 (1998).
26 [40] A. J. Allen, R. C. Oberthur, D. Pearson, P. Schofield, and C. R. Wilding, Phil. Mag. B 56, 263 (1987).
[41] H. M. Jennings, Cem. Concr. Res. 30, 101 (2000).
27
[42] J. J. Thomas, H. M. Jennings, and J. J. Chen, J. Phys. Chem. C 113, 4327 (2009).
28 [43] N. Meller, K. Kyritsis, and C. Hall, Cem. Concr. Res. 39, 45 (2009).
29 [44] J. R. Houston, R. S. Maxwell, and S. A. Carroll, Geochem. Trans. 10, 1 (2009).
30 [45] A. Nonat, Cem. Concr. Res. 34, 1521 (2004).
31 [46] H. M. Jennings, Cem. Concr. Res. 38, 275 (2008).
32 [47] X. Cong and R. J. Kirkpatrick, Advn. Cem. Bas. Mater. 3, 133 (1996).
33 [48] F. Brunet, Ph. Bertani, Th. Charpentier, A. Nonat, and J. Virlet, J. Phys. Chem. B 108, 15494 (2004).
34 [49] A. N. Christensen, N. V. Y. Scarlett, I. C. Madsen, T. R. Jensen, and J. C. Hanson, Dalton Trans. 1529 (2003).
35 [50] M. Merlini, G. Artioli, T. Cerulli, F. Cella, and A. Bravo, Cem. Concr. Res. 38, 477 (2008).
36 [51] T. R. Jensen and A. N. Christensen, J. C. Cem. Concr. Res. 35, 2300 (2005).
37 [52] G. Artioli, T. Cerulli, G. Cruciani, M. C. Dalconi, G. Ferrari, M. Parisatto, A. Rack, and R. Tucoulou, Anal. Bioanal.
Chem. 397, 2131 (2010).
38
[53] H. Manzano, A. Ayuela, and J. S. Dolado, J. Computer-Aided Mater. Des. 14, 45 (2007).
39 [54] S. Bishnoi and K. L. Scrivener, Cem. Concr. Res. 39, 266 (2009).
40 [55] R. Kondo, and S. Ueda, in: Kinetics and Mechanisms of the Hydration of Cements, edited by Proceedings of the Fifth
41 International Symposium on the Chemistry of Cement, Vol. II (Tokyo, Japan, 1968), pp. 203–248.
42 [56] J. M. Pommersheim and J. R. Clifton, Cem. Concr. Res. 9, 765 (1979).
43 [57] M. Avrami, J. Chem. Phys. 7, 1103 (1939).
44 [58] W. A. Johnson and R. F. Mehl, Trans. Am. Inst. Min. (Metall.) Eng. 135, 416 (1939).
45 [59] A. N. Kolmogorov, Bull. Acad. Sci. USSR Phys. Ser. 1, 255 (1937).
46 [60] J. J. Thomas, J. Am. Ceram. Soc. 90, 3282 (2007).
47 [61] H. M. Jennings and S. K. Johnson, J. Am. Ceram. Soc. 69, 790 (1986).
[62] D. P. Bentz, J. Am. Ceram. Soc. 80, 3 (1997).
48
[63] J. W. Bullard, Modelling Simul. Mater. Sci. Eng. 15, 711 (2007).
49 [64] J. W. Bullard, J. Phys. Chem. A 111, 2084 (2007).
50 [65] J. W. Bullard, E. Enjolras, W. L. George, S. G. Satterfield, and J. E. Terrill, Modelling Simul. Mater. Sci. Eng. 18,
51 025007 (2010).
52 [66] J. W. Bullard, J. Am. Ceram. Soc. 91, 2088 (2008).
53 [67] J. W. Bullard and R. J. Flatt, J. Am. Ceram. Soc. 93, 1894 (2010).
54 [68] E. M. Gartner, J. M. Gaidis, in: Hydration Mechanisms, edited by J. Skalny, Materials Science of Concrete, Amer.
Ceramic Soc. (Westerville, OH, 1989), pp. 95–125.

www.crt-journal.org 
C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
 16 G. Artioli and J. W. Bullard: Cement hydration

1 [69] N. Tenoutasse, in: The Hydration Mechanism of C3A and C3S ins the Presence of Calcium Chloride and Calcium
2 Sulfate, edited by Proceedings of the Fifth International Symposium on the Chemistry of Cement Vol. II (Tokyo, Japan,
3 1968), pp. 372–378.
4 [70] P. W. Brown, C. L. Harner, and E. J. Prosen, Cem. Concr. Res. 16, 17 (1985).
5 [71] J. Cama, L. Zhang, G. De Giudici, J. M. Soler, R. S. Arvidson, and A. Luttge, Geochim. Cosmochim. Acta 73, A187
(2009).
6
[72] H. Minard, S. Garrault, L. Regnaud, and A. Nonat, Cem. Concr. Res. 37, 1418 (2007).
7 [73] S. Garrault, H. Minard, and A. Nonat, in: Hydration of Silicate Phase and Mechanical Evolution in “Alite-Tricalcium
8 Aluminate-gypsum” Complex System, edited by J. Beaudoin, Proceedings of the 12th International Congress on the
9 Chemistry of Cement (Montreal, Canada, 2007), pp. TH6–08.1.
10 [74] K. Scrivener, personal communication, (2009).
11 [75] J. W. Bullard, H. M. Jennings, R. A. Livingston, A. Nonat, G. W. Scherer, J. S. Schweitzer, K. L. Scrivener, and J. J.
12 Thomas, Cem. Concr. Res. 41, 1208 (2011).
13 [76] J. J. Thomas, J. J. Biernacki, J. W. Bullard, S. Bishnoi, J. S. Dolado, G. W. Scherer, and A. Luttge, Cem. Concr. Res.
14 41, 1257 (2011).
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54


C 2013 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.crt-journal.org