Modelling Hydration Kinetics of Cementitous Systems

THÈSE NO 5312 (2012)

PRÉSENTÉE le 16 mars 2012
À LA FACULTÉ DES SCIENCES ET TECHNIQUES DE L'INGÉNIEUR
LABORATOIRE DES MATÉRIAUX DE CONSTRUCTION
PROGRAMME DOCTORAL EN SCIENCE ET GÉNIE DES MATÉRIAUX

ÉCOLE POLYTECHNIQUE FÉDÉRALE DE LAUSANNE

POUR L'OBTENTION DU GRADE DE DOCTEUR ÈS SCIENCES

PAR

Aditya Kumar

acceptée sur proposition du jury:

Prof. P. Muralt, président du jury

Prof. K. Scrivener, directrice de thèse
Dr E. Gartner, rapporteur
Prof. M. Rappaz, rapporteur
Dr J. Thomas, rapporteur

Suisse
2012
 Acknowledgements

I would like to express my gratitude to my supervisor Prof. Karen Scrivener for giving me the
opportunity to work at LMC, for innumerable discussions and insightful comments and for
being a constant source of motivation. I would also like to thank the Doctoral school at EPFL
for accepting me as a Ph.D student and Swiss National Science Foundation (FNS) for proving
financial support for this research.

I would like to thank all my colleagues at LMC for offering plenty of help, advice and
feedback during the course of my Ph.D. Special thanks to Mercedes, Vanessa, Patrick, Amelie,
Caterina, Julien, Alexandra, Elise and Hui for letting me use their hard-earned experimental
results. Thanks to Patrick, Gaurav, Cyrille and Shashank for their criticisms and suggestions.

I would like to express my deepest gratitude to all my friends (Julien S., Julien B., Mathieu,
Hui, Olga, Ruzena, Arnaud, John, Theo, Alexandra, Quang Huy, Christoff, Cedric, Elise,
Alain, Mohamed, Berta, Pavel, Cheng, Amelie, Adrian) at LMC for including me in their
“group” and outside-laboratory activities, which I thoroughly enjoyed being a part of. Thanks
to the Administrative (Sandra, Anne-Sandra, Maude, Marie-Alix) and Technical (Lionel) staff
for helping me with all kinds of stuff. Thanks to everyone in the lab for being very friendly to
me.

Page Number: 2
 Abstract

The modeling platform μic has been extended to study the hydration of cement based
systems. It is shown that the hydration kinetics of alite can be described through the
implementation of two mechanisms: a solution controlled dissolution (SCD) mechanism for
the first two stages of hydration and nucleation and densifying growth (NDG) of products for
the last three stages of hydration. The dissolution rate is varied according to the ratio
between the ion activity product and the equilibrium solubility product. The solution
concentrations are computed directly from the amount of alite dissolved taking into account
the amount of water present and the amount of products formed, with activities and complex
ions formation according to standard thermodynamic rules. Saturation index calculations are
implemented to compute the time of precipitation of C-S-H and portlandite individually. After
the precipitation of portlandite, the rate controlling mechanism is switched to nucleation and
densifying growth. C-S-H grown in a diffuse manner in which the packing density increases
with hydration. The overall heat-evolution profile is obtained by combination of heat evolved
as obtained from the SCD and NDG mechanisms.

The approach is used to describe the influence of particle sizes, thermal treatment, mixing
conditions and additions of alkali hydroxides, alkali sulfates and gypsum on hydration of alite.
Through a study of systems with wide variations in the process parameters, it is shown that
the hydration kinetics of alite can be generically described using the combination of SCD and
NDG mechanisms and a limited number of fit parameters. The results obtained from
simulations were found to be in good agreement with a variety of experimental results. The
various simulations parameters are shown to be strongly correlated to the initial states of the
systems.

A model is presented to describe the hydration of aluminate+gypsum systems. It is shown

that the reaction rate profile can be numerically reproduced by using two mechanisms:
dissolution of the aluminate phase in the first stage and boundary nucleation and growth of
monosulfate phase in the second stage of reaction. It is shown that the effects of particle sizes
and composition on hydration of aluminate can be described using just one fit variable,
isotropic growth rate of monosulfate phase, which correlates well the space available in the
microstructural volume.

Using the combination of mechanisms that describe hydration of alite and aluminate, a model

Page Number: 3
is presented to describe the hydration of multi-phase systems. Preliminary simulations of the
heat-evolution profiles of alite-aluminate-gypsum and commerical cementitous systems are
presented. Good agreement was found between the simulated and measured results.

The current study demonstrates the versatility of μic and shows that the use of mechanistic,
rather than empirical, rules can be an important tool to achieve better understanding of
cement hydration.

Keywords: hydration, thermodynamics, dissolution, nucleation, precipitation, diffuse growth

Page Number: 4
 Résumé

La plateforme de modélisation μic a été étendue pour étudier l'hydratation des systèmes
cimentaires. Il est montré que la cinétique de l'hydratation de l'alite peut être décrite grâce à
la mise en œuvre de deux mécanismes : un mécanisme de dissolution contrôlée par la solution
(SCD) pour les deux premières étapes de l'hydratation et un mécanisme de nucléation,
croissance et densification (NDG) des produits hydratés pour les trois dernières étapes. Le
taux de dissolution est varié en fonction du rapport entre le produit d'activité ionique et le
produit de solubilité à l'équilibre. Les concentrations des solutions sont calculées directement à
partir de la quantité d'alite dissoute en tenant compte de la quantité d'eau présente et la
quantité de produits formés, et en utilisant les activités et les règles usuelles de la
thermodynamique de formation de complexes ioniques. Des calculs d'indice de saturation sont
mis en œuvre pour déterminer individuellement le temps de la précipitation des CSH et de la
portlandite. Après la précipitation de la portlandite, le mécanisme dominant devient la
nucléation, croissance et densification. Les CSH croissent de manière diffuse et leur la
compacité augmente avec l'hydratation. Le profil de la chaleur globale est obtenu par
combinaison de la chaleur obtenue à partir de chacun des mécanismes SCD et NDG.

L'approche est utilisée pour décrire l'influence de la taille des particules, du traitement
thermique, des conditions de mélange, et des ajouts d'hydroxydes alcalins, de sulfates alcalins
et de gypse sur l'hydratation de l'alite. L'étude de nombreux systèmes dont les paramètres
varient fortement montre que la cinétique de l'hydratation de l'alite peut être décrite de façon
générique en utilisant combinant les mécanismes SCD et NDG, et un nombre limité de
paramètres de calage. Les résultats obtenus à partir des simulations ont été jugés en bon
accord avec une variété de résultats expérimentaux. Les différents paramètres des simulations
sont fortement corrélés à l'état initial des systèmes.

Un modèle est présenté pour décrire l'hydratation de systèmes aluminates+gypse. Il est

montré que le profil de vitesse de réaction peut être numériquement reproduit en utilisant
deux mécanismes : dissolution des aluminates dans la première étape, puis nucléation-
croissance des monosulfates dans la deuxième étape de la réaction. Il est montré que les effets
de la taille des particules et de la composition sur l'hydratation des aluminates peuvent être
décrits en utilisant une seule variable, le taux de croissance isotrope des monosulfates, qui est
bien corrélé à l'espace disponible dans la microstructure.

Page Number: 5
En utilisant la combinaison des mécanismes qui décrivent l'hydratation des aluminates et de
l'alite, un modèle est présenté pour décrire l'hydratation des systèmes polyphasés. Des
simulations préliminaires de profils de chaleur de systèmes alite-aluminate-gypse et de
systèmes cimentaires commerciaux sont présentés. Un bon accord a été trouvé entre les
résultats simulés et mesurés.

L'étude actuelle démontre la polyvalence de μic et montre que l'utilisation de règles

mécanistes plutôt qu'empiriques peut être un outil important pour parvenir à une meilleure
compréhension de l'hydratation du ciment.

Mots-clés: hydratation, thermodynamique, dissolution, nucléation, précipitation, croissance

diffuse

Page Number: 6
Table of Contents
Abstract.......................................................................................................................................2
Résumé........................................................................................................................................4
Acknowledgements......................................................................................................................6
Glossary.....................................................................................................................................10
Chapter 1: Introduction.............................................................................................................15
Chapter 2: Cement Hydration- Literature Review....................................................................19
2.0 Cement: Basic Principles.....................................................................................................19
2.1 Alite: Background and Fundamentals.................................................................................21
2.2 Alite: Hydration and Kinetics..............................................................................................21
2.3 Alite Hydration: Dissolution and Induction Period.............................................................22
2.3.1 Metastable barrier hypothesis.......................................................................................23
2.3.2 Nucleation and Growth.................................................................................................26
2.3.3 Solution controlled dissolution......................................................................................26
2.4 Alite Hydration: Acceleration and Deceleration Periods.....................................................30
2.4.1 Nucleation and growth of C-S-H with uniform density................................................30
2.4.1.1 Avrami N&G model: Definition.............................................................................30
2.4.1.2 Boundary N&G model: Definition.........................................................................31
2.4.1.3 Studies based on Avrami and Boundary N&G models using uniform density of C-
S-H....................................................................................................................................32
2.4.2 Diffusion based mechanism beyond the heat-rate peak................................................35
2.4.3 Nucleation and densifying growth of C-S-H ................................................................36
2.5 Alite Hydration: Stage V.....................................................................................................39
2.6 Microstructural modeling of cement hydration kinetics .....................................................40
2.6.1 DuCOM........................................................................................................................40
2.6.2 The Jennings+Johnson model......................................................................................41
2.6.3 HymoStruc model.........................................................................................................41
2.6.4 CemHyd3D model.........................................................................................................42
2.6.5 HydratiCA simulation model........................................................................................44
2.6.6 Integrated Particle Kinetics Model (IPKM).................................................................45
2.6.7 μic modeling platform..................................................................................................46
2.7 Current Study......................................................................................................................50
Chapter 3: Simulation Techniques for Alite Hydration ...........................................................51
3.1 Introduction ........................................................................................................................51
3.2 Basics of µic ........................................................................................................................54
3.3 Solution controlled dissolution (SCD).................................................................................57
3.4 Activity calculations ...........................................................................................................59
3.5 C-S-H precipitation ...........................................................................................................61

Page Number: 7
3.6 Portlandite precipitation ....................................................................................................66
3.7 Calibration of parameters for SCD regime .........................................................................67
3.8 Evolution of calcium and silicate concentrations and C-S-H during SCD regime ..............72
3.9 Nucleation and Densifying Growth (NDG) Regime ...........................................................73
3.10 Sensitivity of the NDG parameters ...................................................................................77
3.11 Transition from SCD to NDG regimes .............................................................................78
3.12 Determination of simulation parameters............................................................................82
3.13 Validation of the SCD mechanism.....................................................................................83
Chapter 4: Studying the Effects of Process Parameters............................................................86
4.1 Effect of particle surface area..............................................................................................86
4.1.1 Materials and methods..................................................................................................86
4.1.2 Simulating the effect of surface area.............................................................................89
4.2 Effect of mixing conditions..................................................................................................95
4.2.1 Materials and methods..................................................................................................95
4.2.2 Simulations: Effects of mixing conditions.....................................................................97
4.3 Effect of thermal treatment of alite...................................................................................110
4.3.1 Materials and methods................................................................................................110
4.3.2 Effect of thermal treatment: Measured heat evolution profiles..................................111
4.3.3 Simulating the effect of thermal treatment of alite....................................................114
4.4 Effect of alkali hydroxide addition.....................................................................................121
4.4.1 Introduction................................................................................................................121
4.4.2 Materials production and compositional information.................................................121
4.4.3 Simulating the effect of particle size distribution on alite hydration.........................123
4.4.4 Effect of alkali hydroxide additions: Experimental results.........................................125
4.4.5 Simulating the influence of the initial solution pH on alite dissolution......................127
4.4.6 Simulating the pre-acceleration period.......................................................................129
4.4.7 Simulating the Acceleration and Deceleration periods...............................................131
4.4.8 Studying the changes in structure and morphology of C-S-H....................................136
4.5 Effect of gypsum addition..................................................................................................139
4.5.1 Materials and Methods...............................................................................................139
4.5.2 Influence of gypsum addition: Measured heat evolution profiles................................139
4.5.3 Simulating the influence of gypsum addition on alite hydration................................141
4.6 Effect of alkali sulfate addition .....................................................................................150
4.6.1 Introduction................................................................................................................150
4.6.2 Materials and methods................................................................................................150
4.6.3 Effect of alkali sulfates addition: Measured heat evolution profiles...........................151
4.6.4 Simulating the evolution of ionic concentrations........................................................153
4.6.5 Simulating the heat evolution profiles........................................................................157
Chapter 5: Perspectives on Modeling Alite Hydration............................................................163

Page Number: 8
5.1 Transition from SCD to NDG mechanism.........................................................................163
5.2 Variation in the NDG parameters.....................................................................................165
5.3 Effect of space-filling on hydration kinetics.......................................................................166
5.4 Densification, phase-arrangement and growth criterion of C-S-H.....................................168
Chapter 6: Modeling Hydration of C3A+Gypsum and Multi-phase Cementitous Systems....171
6.1 Introduction.......................................................................................................................171
6.2 C3A+Gyspum+H systems: Basics ....................................................................................171
6.2.1 Aluminate Hydration: Kinetics of hydration during 1st stage....................................173
6.2.2 Aluminate Hydration: Kinetics of hydration during 2nd stage..................................174
6.3 C3A+H2O+Gypsum systems: Simulations........................................................................175
6.3.1 Preliminary simulations for C3A+H2O+Gypsum systems: Stage 1 .........................176
6.3.2 Preliminary simulations for C3A+H2O+Gypsum systems: Stage 2 .........................179
6.3.3 Conclusions: Preliminary simulations for C3A+H2O+Gypsum systems...................185
6.4 Simulating hydration of multi-phase cementitous systems................................................185
Chapter 7: Conclusions and Perspectives................................................................................194
7.1 Dissolution of alite.............................................................................................................194
7.2 Effects of process parameters of alite hydration................................................................195
7.3 Hydration of C3A+Gypsum systems.................................................................................197
7.4 Perspectives on microstructural modeling ........................................................................197
7.5 Perspectives on thermodynamic modeling.........................................................................198
7.6 Perspectives on simulating multi-phase systems................................................................199

Page Number: 9
 Glossary

Cement chemistry notation

C CaO
S SiO2
H H2O
A Al2O3
F Fe2O3
$ SO3
C3S Tricalcium silicate
C2S Dicalcium silicate
C3A Tricalcium aluminate
C4AF Tetracalcium aluminoferrite
C-S-H Calcium silicate hydrate
CH Calcium Hydroxide (Portlandite)

Page Number: 10
 Abbreviations
AU Arbitrary units
BET Brunauer, Emmett and Teller theory
BSE Back-Scattered Electrons
BWI Bound water index
CV Computational volume
EDS Energy dispersive spectroscopy
FEM Finite elements method
NMR Nuclear magnetic resonance
NDG Nucleation and densifying growth
NG Nucleation and growth
N&G Nucleation and growth
OPC Ordinary Portland cement
PSD Particle size distribution
QENS Quasi-electron neutron scattering
QXRD Quantitative X-ray diffraction analyses
REV Representative elementary volume
SCD Solution controlled dissolution
SEM Scanning electron microscopy
TEM Transmission electron microscopy
TGA Thermo-gravimetric analysis
w/c Water to cement ratio
w/s Water to solid ratio
XRD X-ray diffraction

Page Number: 11
 List of symbols (In the order of appearance in the manuscript)
Symbol Context Description
Vphase Hydration Volume of phase
Q Hydration Heat flow rate
H Hydration Enthalpy
dmphase Hydration Change in mass of phase
t Hydration Time or age
IAP Thermodynamics Ion activity product
Ksp Thermodynamics Solubility product
β Thermodynamics Ratio of IAP to Ksp
SI Thermodynamics Saturation Index
I Thermodynamics Ionic strength
γ Thermodynamics Activity coefficient
aphase Thermodynamics Activity of species “phase”
AT Thermodynamics Debye-Huckel Temperature constant I
BT Thermodynamics Debye-Huckel Temperature constant II
ai Thermodynamics Kielland ion-specific parameter
bi Thermodynamics Kielland ion-specific parameter
T Thermodynamics Temperature
mspecies Thermodynamics Molality of species
zspecies Thermodynamics Unit charge of ionic species
kdiss Alite dissolution Dissolution rate constant
βmax Alite dissolution Critical alite undersaturation constant
βalite Alite dissolution Alite undersaturation (IAPalite/Ksp-alite)
cstep retreat Alite dissolution Step retreat constant
aSSA Alite dissolution Specific surface area of particles
{} AliteDissolution Activity
[] Alite Dissolution Concentration
Vreal NDG Volume of products without impingements

Page Number: 12
Vextended NDG Volume of products with impingements
Af NDG Area fraction covered by products
ρ NDG Density of phase
ρ0 NDG Base density of C-S-H
ρmin NDG Minimum or initial density of Outer C-S-H
ρmax NDG Final density of Outer C-S-H
Idensity NDG Number of nuclei of Outer C-S-H per unit area
Gout NDG Outward Growth rate of the nuclei of the product
Gpar NDG Growth rate of the nuclei along the surface
Irate NDG Rate of nucleation of C-S-H per unit area
kden NDG Rate of densifictation of outer C-S-H
t0 NDG Start time of NDG
Ndensity NDG Nucleation density coefficient (Idensity. Gpar2)
Nrate NDG Nucleation rate coefficient (Irate. Gpar2)
aBV NDG Surface area of particles per unit volume of water initially
dmC3A Ettringite formation Change in mass of C3A
aSA Ettringite formation Surface area of C3A particle
k1 Ettringite formation Rate constant
k2 Ettringite formation Rate constant
r0,i Ettringite formation Initial radius of C3A particle
aSSA Ettringite formation Specific surface area of particles
X Monosulfate formation Fraction of C3A reacted
Vextended Monosulfate formation Extended volume of monosulfate
tr Monosulfate formation Age of product (monosulfate)
tmono Monosulfate formation Start time of monosulfate formation kinetics
Gout Monosulfate formation Outward growth rate
Ndensity Monosulfate formation Nucleation density coefficient (Idensity. Gpar2)
Nrate Monosulfate formation Nucleation rate coefficient (Irate. Gpar2)
aBV Monosulfate formation Area of C3A particles divided by volume of water

Page Number: 13
 falite Multi-phase hydration Fraction of alite in cement
faluminate Multi-phase hydration Fraction of C3A in cement

Page Number: 14
Chapter 1: Introduction
This study presents a modeling approach using the microstructural modeling platform μic.
This platform uses uses the vector approach to represent the geometry of the microstructure.
In the vector approach particles with simple geometry are used to represent the system and at
each step of the simulation the position and the size of the particles are stored. Due to its
flexible design, the users of the platform can define custom materials, particles and reactions,
and control the development of the microstructure by defining laws that define the
mechanisms of the reactions. The versatility of μic allows users to model many different
particulate growth systems not limited to cement hydration.

In this study, the modeling platform μic has been extended to study the effect of various
process parameters that influence the hydration kinetics of cement based systems. The
influence of: particle sizes, mixing conditions and heat treatment at different temperatures
and different durations and additions of: alkali hydroxides, alkali sulfates, and gypsum have
been studied. Comparisons against a variety of experimental results are done for validation. A
model to describe the hydration of aluminate+gypsum systems studying the effects of
composition and particle size distribution is described. Using the mechanisms to describe
hydration kinetics of model alite and aluminate systems, simulations for several multi-phase
systems are also described. The results and discussions on the variable simulation parameters
provide significant insights into the mechanisms that drive the hydration of clinker phases
and highlight the gaps in our understanding of these mechanisms.

The main problem that researchers encounter in their study of cement-based systems is the
confluence of several interactions that occur between the ionic species in the pore-solution and
solid phases during hydration at different rates, at different points in time and for different
durations. Most of our understanding of mechanisms driving cement hydration depend on
interpretation of indirect experimental techniques, and therefore, there exists a knowledge gap
between the values and trends obtained from experiments and their physical meanings. A few
computational techniques have been used in the last few decades to bridge this gap and
provide theories and hypotheses that can explain the mechanisms behind the hydration
process. Numerical models use combinations of fundamental processes and empirical concepts
to simulate systems and processes. Model are devised so as to describe the behavior of smaller
and discrete sub-systems and the interactions between these sub-system units. By integration
of these sub-systems, the behavior of the entire system can be evaluated. Numerical models
can, therefore, can be used as an important technique, which works in a complimentary
capacity to the experiments, in order to further our understanding of cement hydration.

Page Number: 15
Several numerical models have been developed to describe and quantify the hydration of
cement based systems. These models can be broadly classified into two categories:
Microstructure based models and Analytical models. Most of these models are based on
empirical methods and calibrations from experimental results. As a result of this, the
applicability of these models is restricted to validation and prediction of properties of a small
range of systems, the compositions of which are similar to the systems from which the
calibrations have been done. The implementation of fundamental mechanisms from the
ground-up is required to provide a generic and relatively accurate description of the hydration
process. For this, firstly, the model should offer flexibility to the developer to incorporate
custom mechanisms and rules without having to alter the model or program extensively.
Secondly, as calibrations of the input and process variables are mostly done on a trial-and-
error basis, the simulation time and the computational complexity of the model should also be
small. Lastly, the model should offer a wide range of output results so as to provide the user a
range of options to validate the mechanisms and parameters.

It is precisely for these reasons, μic was chosen as the model to study the hydration of
complex systems. μic provides a basic framework on the basis of which new mechanisms and
hypotheses can be programmed and described as “plugins”. Kinetics schemes such as
dissolution, precipitation of phases etc. can be described using mathematical expressions that
can be built upon time-discrete or finite difference schemes. The model offers support for
description of thermodynamic laws within the core, that can serve as a set of binary rules
describing the conditions for precipitation and dissolution. Ultimately, the model provides an
effective means to reconstruct three dimensional (3D) digital microstructure that evolve in a
temporal fashion. These microstructures can be analyzed to calculate different properties. The
output of the simulations constitute several results that can be compared against experimental
results for validation: heat evolution rate, volumes and amounts of phases, ionic
concentrations, porosity etc. The model and its applications will be described in the detail in
the following chapters.

The following chapters discuss the importance of numerical modeling in cement science,
highlight the key areas of research done in the field, discuss the simulation techniques
implemented in this study and present the results that can be obtained by the implementation
of the fundamental principles from different disciplines.

Page Number: 16
Chapter 2 discusses our current understanding of cement hydration and different approaches
used so far to elucidate the mechanisms that drive the hydration of cement-based systems and
evolution of the microstructure. The chapter provides a summary of the work done in the past
several years and discusses several significant strides that have been made both in
experimental techniques and in theoretical models in the intervening years. As the
mechanisms that drive hydration are different during different stages, the chapter has been
divided into sections categorized on the basis of “stage of reaction” to delineate the proposed
mechanisms and hypotheses that pertain specifically to the time-domain in question. The
modeling approaches implemented to study hydration have also been discussed independently
to illustrate the developments made in the past few years and to highlight the gaps in our
understanding or numerical implementation of mechanisms. The limitations of the theoretical
and numerical approaches are also presented that highlight the importance of developing a
new model, that combines the approaches from different disciplines into a unified approach
and can potentially be used to simulate hydration of complex systems without the need of
calibrations or fitting using a large set of empirical parameters.

Chapter 3 describes the simulation techniques implemented in the model. The approaches
used to describe hydration of alite during the early and late ages and their implementation in
the model are outlined. The chapter highlights the problems in some of the widely accepted
mechanisms and propose new hypotheses that are used to overcome these. The assumptions
and approximations made therein are also delineated to demonstrate the need for simplifying
the numerical approach and to outline prospective avenues for more rudimentary research in
the field of hydration of cement-based systems.

Chapter 4 describes the implementation of the model to study the effects of various process
parameters. As such, the effects of surface area of particles, mixing conditions, heat treatment,
and additions/substitutions of alkali hydroxides, alkali sulfates and gypsum on the hydration
of alite/C3S are described in individual sections. The simulation results are compared to
experimental results, isothermal calorimetry in particular. The variation in the simulation
parameters with respect to physical variables are also discussed to highlight the role of the
experimental process parameters towards influencing the hydration of alite/C 3S. Through
several examples, the chapter illustrates the versatility of the model and demonstrates the use
of one unified approach to describe hydration of alite/C 3S under the influence of different
variable process parameters and initial states of the system.

Page Number: 17
Chapter 5 discusses some of the limitations in the approach used to simulate hydration of
alite. Assumptions and approximations made in the model and suggestions and concepts to
further improve these aspects are delineated. Areas which are not covered in this study, such
as hydration of alite in high w/c ratio systems and non-spherical growth criteria of C-S-H, are
discussed with preliminary simulations. Suggestions for potential mechanisms which can be
incorporated in the future are also discussed.

Chapter 6 describes the hydration kinetics of aluminate (C 3A)- gypsum systems. Our current
understanding is highlighted through a short literature review of the recent work in the field
of hydration of model aluminate systems. The effects of particle size distributions and
composition on the reaction rate profiles are discussed using a series of experimental and
simulation results. The implementation of the mechanisms in the model are described with
mathematical formulations and illustrations of the sensitivity of the kinetic parameters. The
variation in the simulation parameters with respect to physical variables, related to the initial
or instantaneous state of the system, are also discussed. The chapter also includes preliminary
simulations for multi-phase systems, which are implemented by unifying concepts and
approaches from simulations of model alite and aluminate systems. The variation in the
kinetic parameters used for the multi-phase systems are discussed.

Chapter 7 presents a brief summary of all approaches used in the model. The variation of all
the simulation parameters across different systems are presented in a unified fashion to
correlate the numerical variables to the physical processes and the initial states of the
respective systems. Perspectives for the future numerical and experimental studies on cement
hydration are also presented.

Page Number: 18
Chapter 2: Cement Hydration- Literature Review
In this chapter, our current understanding of the cement hydration and the approaches used
to simulate it numerically are presented. In the past few years, several aspects of cement
science have been explored and described using a combination of experimental and numerical
approaches. Despite numerous advances made in the recent past, most of the existing models
rely heavily on empirical results, often limiting their applicability. The discussion identifies
the progress made in the recent years towards describing some of the important aspects of
hydration and highlights some of the key areas where a scope for major improvements exist.

2.0 Cement: Basic Principles

Portland cement is the most widely used type of cement. It is produced by burning minerals
that naturally occur on the earth-crust in a kiln at around 1450 0C[1]. The minerals include
primarily lime and clay. After processing at a high temperature followed by cooling, the
minerals fuse to form clinker nodules. The nodular clinker is then mixed with a small
proportion of gypsum and is finely ground to produce cement. The oxides in the clinker
combine to form phases that constitute cement. For example, calcium oxide and silicon
dioxide combine to form a modified form of tricalcium silicate, which is also known as alite
and is the most important phase in cement. The other major phases present in cement are
belite, aluminate and ferrite, as listed in Table 2.1. It should be noted that in cement these
phases are present in their impure form with ionic substitutions in their crystalline structures.

Table 2.1: Typical composition of Ordinary portland cement

Mineral Phase name Abbreviation Typical amount

Tricalcium Silicate Alite C3S 50-70%
Dicalcium Silicate Belite C2S 10-20%
Tricalcium Aluminate Aluminate C3A 5-10%
Tricalcium Aluminoferrite Ferrite C4AF 5-15%

Cement reacts with water in a process called hydration. Hydration leads to decrease in the
amount of the clinker phases and increase in the amounts of product phases. The overall
volume of solid phases increases while the volume of water decreases. The mixture of cement
and water, conventionally called cement paste, decreases in its overall volume and converts
into a stiff solid. The rate of reaction of the individual clinker phases differs from one to

Page Number: 19
another. Alite and aluminate phases react quickly whereas belite and ferrite react slowly and
for longer durations. The overall progress of hydration is generally monitored using the heat
flow as measured by isothermal calorimetry (DSC). Figure (2.1) shows the typical heat
evolution profile of ordinary Portland cement for the first 3 days. The curve can be broadly
divided into five stages.

Figure 2.1: Typical heat evolution curve (measured) of ordinary Portland cement system
against time until approximately 3 days of hydration.

The first stage is marked by a rapid evolution of heat for several minutes. This is generally
attributed to the initial rapid dissolution of alite and gypsum and a rapid hydration of the
aluminate phase [2]. A continuous low evolution of heat is observed in the second stage of the
process. This stage is referred to as the induction period or the dormant period. In stage 3,
the reaction accelerates for a few hours until a peak is reached. In stage 4, the rate of reaction
drops and begins decreasing monotonically that continues during stage 5 until the ultimate
degree of hydration is reached.

During the hydration of cement several reaction products called hydrates are formed that are
responsible for providing its binding properties and strength development. In the following
sections, the hydration of cement, alite and aluminate in particular, the development of its
microstructure and its reaction kinetics will be discussed.

Page Number: 20
2.1 Alite: Background and Fundamentals
Tricalcium silicate ((CaO)3SiO2 or C3S) is the primary constituent of a number of different
types of cement, including Ordinary Portland cement, white cement and oil-well cements [1].
Alite, which is an impure form of C 3S, contains small quantities of magnesium and aluminum
as part of its lattice structure. Reaction between alite and water produces calcium silicate
hydrate (C-S-H) and portlandite (CH or calcium hydroxide) as shown in Equation 2.1.

C3S + (3-x+y)H → CxSHy + (3-x) CH (2.1)

The equation (2.1) is exact for pure systems but the value of (x+y) are not known accurately
[3,4,5,1]. Unlike C-S-H, the structure, morphology and stoichiometry of portlandite are well-
known; it is reported to grow as large clusters or hexagonal platelets in the cement matrix
[6,7,8] depending on pore-solution composition and particle size distribution of cement. The
formation of C-S-H and portlandite is largely responsible for the setting and hardening of the
cement paste in addition to the development of its microstructure and strength [1,9]. For this
reason, the hydration of alite has been widely studied over the past 40 years or so with special
focus on reaction kinetics, phase chemistry and phase equilibria [10,11,12,13,14,15,16,17,18,19].
However, a consensus has not been reached on the qualitative nature of the reaction
mechanisms, especially during the first several hours after mixing. The hydration of alite is
discussed in the next section.

2.2 Alite: Hydration and Kinetics

The hydration of alite progresses through 5 different stages, across which the rate of reaction
varies non-monotonically by several orders of magnitude. Figure (2.2) shows a typical heat
evolution curve as well as the evolution of the calcium concentration for the first 30 hours.
Five main stages can be identified. Stage I is referred to as the initial dissolution period in
which the rate of reaction rapidly slows, leading to period II in which the rate of heat-
evolution is low – known as the induction period. Until the end the period II, the calcium
concentration rises continuously up to a supersaturated value. At the start of the stage III
regime, the calcium concentration begins to drop from the portlandite supersaturated value to
portlandite solubility equilibrium and the heat flow rate rises until a peak is reached. During
stage IV the heat flow rate drops leading to stage V in which the heat flow rate is steady at a
low value and calcium concentration continues to remain at portlandite solubility equilibrium.

Page Number: 21
Figure 2.2: Typical heat evolution curve and calcium concentration evolution in alite-model
systems until approximately 30 hours of hydration. From [20]

It should be noted that the classification of the reaction rate of alite into stages is only a
general one and depends heavily on the initial state of the system, as will be discussed in the
next chapters. The following sections will discuss the theories and mechanisms used to
describe the rate of hydration of alite during different stages. A detailed review of these
mechanisms can also be found in[21,22].

2.3 Alite Hydration: Dissolution and Induction Period

The dissolution of alite is chemically represented as Equation 2.2. This equation assumes
congruent dissociation of alite. Nonat [23] in his study confirmed stoichiometric dissolution of
C3S in very dilute suspensions (w/c = 50000) with a steady increase in Ca and Si
concentrations in a 3:1 ratio.

Page Number: 22
Ca3SiO5 + 3H2O = 3Ca2+ + H2SiO42- + 4OH- (2.2)

Several theories have been proposed to explain the hydration kinetics during the first two
stages. The majority of these studies focus on explaining the onset and the end of the
induction period. Some of the important mechanisms used to explain hydration of alite-based
systems in the first two stages are described below:

2.3.1 Metastable barrier hypothesis

One of the widely accepted theories for the first deceleration period during stage I and the
induction period, is the formation of a protective layer of hydration products on the surface of
reacting grains which limits the dissolution of the anhydrous phases. Stein et al.[24] and
Jennings et al.[25] have proposed the rapid formation of continuous and thin metastable layer
of C-S-H that passivates the surface either by (a) restricting its access to water by providing
full or almost full coverage over the alite surface, or (b) by restricting diffusion of ions away
from the surface of alite. Gartner et al.[9] proposed the formation of a specific form of C-S-H,
called C-S-H(m) that is responsible for passivating the surface of alite during the early ages. A
review of solution composition data from literature done by Gartner and Jennings [11,12]
showed a divergence in the Si versus Ca plots attributing to the presence of two forms of C-S-
H with hydrating alite. The plot shown in Figure (2.3) shows two distinct curves: the upper
curve was interpreted as highly soluble and metastable C-S-H (SI), which acts as a protective
metastable layer inhibiting dissolution of alite causing the onset of the induction period, and
the lower curve was reported as the stable form of C-S-H (SII). It was postulated that the
metastable C-S-H (SI) undergoes transformation to a less-protective form of stable C-S-H
(SII), which marks the end of the induction period and permits rapid dissolution of alite.
However, the presence of such protective layers has never been observed using microscopic
techniques. In addition to this, due to the vast difference between the crystal structures of C 3S
and C-S-H, it is very unlikely for an epitaxial relationship to exist between the hydrating C 3S
particles and the protective membrane as it is observed in metallic systems between the metal
and the passivating oxide thin-film.

Page Number: 23
Figure 2.3: Si concentration versus Ca concentration collected from extensive literature
search and reported by Jennings[11]. The divergence of the plot to two distinct curves was
interpreted as the formation of the metastable C-S-H (upper curve) and C-S-H (lower curve)
in equilibrium with other phases.

Page Number: 24
Other evidence taken to suggest the metastable barrier hypothesis comes from Nuclear
Magnetic Resonance (NMR) studies conducted by Rodgers et al.[29]. Their study of the
hydration of C3S particles in suspension indicated the presence of silicate hydrate monomers,
which form a protective layer (calculated to be around 40 nm in thickness) and inhibit
dissolution of C3S during stage 1 causing the onset of the induction period. It was shown that
the hydrate monomers later dimerize to stable C-S-H and dissolution of C 3S is re-initiated
marking the end of the induction period. The analyses, however, do not confirm that the
spatial distribution of this layer is uniform and provides full surface coverage of the C 3S
particles, which would be a prerequisite for the rate of dissolution to slow down by two orders
of magnitude and cause the onset of the induction period.

Similar evidence of a metastable product was presented in a recent study by Bellmann et al.
[26], who reported observations of intermediate calcium silicate phase containing silicate
monomers, as also reported in previous studies [27,28,29,30]. The observations were made
using NMR analyses based on 29Si NMR on nano-sized (40 to 200 nm) C 3S particles in
suspension (w/c = 4000) and paste (w/c = 1.50) conditions hydrated for 5 minutes. The
interpretation of their data indicates the formation of a protective layer intermediate silicate
hydrate at very early ages. The authors postulated that this phase undergoes conversion into
stable C-S-H once the solution reaches close to portlandite supersaturation. However, their
study does not demonstrate that the intermediate phase has the requisite passivating
properties to reduce the rate of dissolution by two orders of magnitude. Furthermore,
Bellmann et al. also do not explain the correlation between the saturation of calcium ions in
the pore-solution and end of the induction period which would result in the transformation of
the silicate monomers to stable C-S-H.

The most serious drawback of a “blocking” layer on the C 3S surface, as stated earlier, is the
lack of direct experimental evidence. In order to effectively block hydration or restrict
diffusion of one or more species, the layer has to be (a) uniform, enclosing the surface almost
completely and (b) fairly dense, to block or restrict the diffusion of species. However, surface
analyses conducted on flat C3S surfaces under water using atomic force microscopy [31] and
alite in suspensions using scanning electron microscopy [20] have only revealed the formation
of isolated precipitates of C-S-H as opposed to a continuous layer of a hydrate, which
contradicts the premise for this hypothesis to be true.

Page Number: 25
2.3.2 Nucleation and Growth
The Nucleation and Growth theory postulates that the nature of the induction period is
related to steady formation of embryos of products. The end of the induction period is
attributed to the rapid growth of the pre-existing embryos and formation of new nuclei of
products causing the overall rate to rise again. Studies in cement science supporting this
theory can be divided into two classes: (a) studies claiming that the nucleation of portlandite
is rate determining at early ages [32] (b) studies claiming that the nucleation of C-S-H at
early ages is related the onset and end of the induction period [18,14,31,33].

The nucleation of portlandite was hypothesized as a cause for the induction period since the
peak in the calcium concentration evolution has been observed to coincide with the end of the
induction period, as reported in various studies [1,10]. However experiments of addition of
portlandite to hydrating C3S showed prolongation in the induction period as opposed to
shortening it [34,35]. This behavior can be explained using the solution controlled dissolution
theory, which will be described in the next section. The portlandite nucleation theory also
failed to explain the slow down in the dissolution of alite within the first few minutes.

Other studies have postulated that the steady formation of C-S-H nuclei during the induction
period controls the overall rate [14,18,31,33]. These studies report that there is no true
induction period but only a continuous rise in the rate of formation of C-S-H. According to
these studies, the induction period ends when these nuclei of C-S-H begin to grow rapidly.
However, these studies also fail to explain the decrease in the dissolution rate during stage 1
and the onset of the induction period; C-S-H starts forming minutes after alite begins to
dissolve [20,31] and there is no evidence to suggest that this would lead to the significant
decrease in dissolution rate of alite.

2.3.3 Solution controlled dissolution

The solution controlled dissolution (SCD) hypothesis postulates that the rate of dissolution is
controlled by the solution composition which leads to a regime in which the dissolution rate is
low. Evidence of this comes from the field of Geo-chemistry, in which it has been shown that
the dissolution rates of several naturally occurring minerals do not follow a smooth
exponential type of relationship [36,37,38]. In their study the authors postulated that as
dissolution of several naturally occurring minerals progresses in aqueous solutions, ions are
released from a constantly evolving surface giving rise to formation of three main types of
defects: (a) 2D vacancy islands and impurity defects (b) Etch pit openings at dislocations (c)

Page Number: 26
Step-retreats at pre-existing roughnesses. The evolution of the first two defects occurs at fast
dissolution rates and high undersaturations (far from equilibrium). Step-retreats drive the
dissolution at slow rates close to equilibrium conditions. The evolution of these defects is
shown in Figure (2.4).

(a) (b)
Figure 2.4 (a): a- two-dimensional vacancy islands created due to the nucleation of small pits
at surfaces with or without impurity, b- etch pit formation at dislocations and c- step retreat
taking place at pre-existing roughnesses. The image was adapted from [20].
(b) Schematic representation of evolution of defects on the surface of dissolving mineral as a
function of undersaturation and dissolution rate. The images correspond to dissolving quartz.
On the right extrema of the x-axis, equilibrium with respect to the quartz exists. The image
was adapted from [20].

Principles from the solution controlled dissolution theory have been applied to alite in a
recent study by Juilland in his doctoral study at EPFL and [20]. Using this theory, the
authors devised a set of fairly generic rules to describe dissolution of alite/C 3S as summarized
below:
1. Far from equilibrium and at high undersaturation, two-dimensional vacancy islands
and etch pits open at the surface

Page Number: 27
 2. As dissolution progresses at high rates, undersaturation decreases very rapidly.
3. Beyond a critical minimum level of undersaturation, the driving force is insufficient to
create new etch pits.
4. At this low driving force and low undersaturation close to equilibrium conditions,
dissolution occurs primarily by a retreating step mechanism that is much slower.
Juilland et al.[20] also showed SEM images of unground alite surfaces hydrated in DI-water
and saturated lime solution confirming the importance of the initial undersaturation. Samples
hydrated in DI-water showed significant surface roughening with a prominent presence of etch
pits at early times of hydration (Figure 2.5a), whereas samples hydrating in saturated lime
solution preserved a smooth surface (Figure 2.5b). In DI-water with high initial
undersaturation, the driving force is sufficient to open up etch pits and drive dissolution at
high rates (as shown in Figure 2.5a). This progression of dissolution is different from the one
observed in saturated lime solution, in which the initial undersaturation and, thus, the driving
force for creating etch pits are low. As such, dissolution begins from close to the step-retreat
regime preserving the smoothness of the surface devoid of many etch pits (as shown in Figures
2.5b). Also, in the case of portlandite addition in alite hydration, Juilland et al. explained the
lengthening of the induction period due to the decrease in the initial undersaturation resulting
from rapid dissolution of portlandite. This leads to slower dissolution from the outset and
longer time needed to reach the critical level of portlandite supersaturation.

Figure 2.5: SEM images of alite immersed for (a) 2 minutes in DI-water (b) 2 minutes in
saturated lime solution. From [20].

Page Number: 28
Juilland et al.[20] summarized their study by presenting a schematic progression of dissolution
as a function of undersaturation outlining the two distinct regimes (shown in Figure 2.6).
They proposed that at the end of the induction period, massive precipitation of hydrates takes
place which consumes the ions from the pore-solution. Decrease in ionic-concentration, in
turn, causes the undersaturation to rise again placing the dissolution profile back in the
undersaturation-driven regime. Faster dissolution and a high undersaturation values provide a
driving force for the reaction rate to rise again during the acceleration regime.
These experimental observations provide very strong evidence to suggest that the hydration
kinetics during the first two stages is driven by solution controlled dissolution. The hypothesis
gets corroboration from experimental evidence and draws analogies from well-established
principles in Geo-chemistry. In this study, as will be demonstrated in the following chapters,
numerical evidence has been provided to substantiate this theory and it has been shown that
the same generic description of alite dissolution can be applied to a many alite-based systems
with different initial conditions.

Figure 2.6: Schematic representation of dissolution of alite as a function of alite

undersaturation. The left end of the light grey bar (zone II) represents the transition point of
the dissolution profile from fast to slow regime. This occurs at the onset of the induction
period. The profile is exponential in the fast-dissolution regime (zone I) and constant in the
slow-dissolution regime (zone II). As portlandite precipitates and undersaturation rises
forcing the dissolution back to the high-dissolution regime. The image was adapted from [20].

Page Number: 29
2.4 Alite Hydration: Acceleration and Deceleration Periods
The acceleration period is the third stage of alite hydration occurring right after the end of
the induction period, in which massive precipitation of hydrates takes place and the rate of
reaction increases until a peak is reached. The post-peak regime of the alite hydration rate
profile is the deceleration period. The acceleration regime is generally attributed to nucleation
and growth (N&G) of hydrates [18,15,39], but there is no broad consensus on the mechanisms
driving the hydration kinetics during the deceleration regime.

Several authors have come up with different mechanisms to describe the decrease in the
reaction rate during the deceleration period. These mechanisms can be broadly divided into 3
main categories as summarized below and described in detail in the following sub-sections:
1. N&G of C-S-H with uniform density
2. Transition from N&G to diffusion based mechanism at the peak.
3. Nucleation and densifying growth (NDG) of C-S-H

2.4.1 Nucleation and growth of C-S-H with uniform density

Several authors have used analytical expressions of traditional and modified N&G mechanisms
[15,18,17] (Avrami [40] and Cahn [39] kinetics) to numerically reproduce the acceleration and
deceleration regimes of alite systems. In these models, the density of C-S-H is assumed to be
uniform. In the following sub-sections, the two main forms of N&G models and results
obtained from these models, as reported in different studies, are described.

2.4.1.1 Avrami N&G model: Definition

The Avrami or JMAK (Johnson-Mehl-Avrami-Kolmogorov) N&G equation [40,41] was
developed to describe phase-transformation in metals under isothermal conditions. The
underlying assumptions and the resulting consequence in its mathematical formulations are:
(a) nucleation occurs at spatially random sites (b) the growth of products is linear in any
direction and constant throughout the reaction (c) as a consequence of the assumptions in “a”
and “b”, the geometry of the transformed phase is spherical (d) after a certain time the
volume of the transformed phase decreases exponentially with time due to impingements
between growing product nuclei. The transformed volume without the volume lost due to
impingements is called the “extended volume”, which always increases with time.

Page Number: 30
As these assumptions can be broadly applied to the hydration of alite and formation of C-S-H
as the primary product, the Avrami equation can be derived as:
dX (2.3)
=An k n ( t−t0 )n-1 exp (−[k( t−t0 )n ])
dt

where, A is the normalization constant, t 0 is the time delay between the end of mixing and
start of nucleation and growth of C-S-H, k is the effective rate constant and n is a variable
represented by Equation 2.4.
p (2.4)
n= +q
s

where, p is the growth dimensionality (1 for needles, 2 for sheets and 3 for isotropic growth), s
is the type of rate control mechanism (1 for phase boundary and 2 for diffusion) and q is the
type of nucleation (0 for site saturation, in which the nucleation event occurs a t 0 and 1 for
continued nucleation of products). Depending on the values of p, s and q, the value of n can
range from 0.25 to 4.

2.4.1.2 Boundary N&G model: Definition

Boundary N&G models differ from the standard Avrami equation in terms of their
descriptions of distribution of nucleation sites and growth criteria of the products: (a)
nucleation of products occur preferentially on the defect-sites, primarily grain boundaries on
the surface of the reactant material (b) the growth of products occurs in an isotropic fashion,
both along the lateral and longitudinal directions. The final mathematical formulation, as
modified from the original form given by Cahn [39], for C3S/alite systems was derived by
Thomas [42,18] (shown in Equation 2.5).

Gt
X=1−exp [−2 O BV∫0 (1−exp (−Y e ))] (2.5)

where, G is the isotropic growth rate of the product and Y e is the extended area of the
transformed product phase given by Equation (2.6 and 2.7)

y
e
t− (if t > y/G) (2.6)
Y =π ∫0 G
IB (G 2 (t−τ)2−y 2)d τ

e
Y =0 (if t < y/G) (2.7)

Page Number: 31
where, y is the distance between the untransformed boundary and an imaginary plane, I B is
the density of nuclei, t is the time and τ is the time at which nucleation and growth of the
product begins.

2.4.1.3 Studies based on Avrami and Boundary N&G models using uniform
density of C-S-H
Using the Avrami equation, several authors [15,17,18,46] have attempted to fit alite hydration
profiles, either as the standard N&G mechanism or for the purpose of validation. The Dijon
model, developed by Garrault and Nonat [15,46] for the simulation of hydration of alite based
systems, is based on modified form of Avrami equation. The modification incorporates a non-
isotropic growth rate of the product. In their model, nuclei of C-S-H form on a two-
dimensional surface with periodic-boundary conditions and closed continuous surfaces. At
each step in the simulation, the nuclei grow both parallel and perpendicular to the surface.
The growth is simulated by generation of new C-S-H elements on the boundaries of per-
existing elements. It was observed that good fits with experimental results can be obtained if
the perpendicular growth rate is assumed to be higher than the parallel growth rate. One of
the critical drawbacks of this approach is that the effect of inter-particle interactions is not
taken into account and due to this, the fits can only be obtained with very large variations in
the simulations parameters. With large variations in the parameters, the deceleration regime
can be simulated but the parameters do not correlate well with the physical states of the
systems. Also, the model was calibrated and compared against experimental results with only
slightly different process parameters and was never validated for large variations in
experimental process parameters (such as PSD). The importance of validating mechanisms
against results from systems with very different specific surface areas is discussed below.

Thomas et al.[18] showed that the deceleration regimes of C 3S based systems could not be
reproduced using Avrami kinetics with a uniform density of C-S-H. In their study, the authors
showed that the boundary N&G mechanism using a uniform density of C-S-H could reproduce
the heat-evolution profiles, both in the acceleration and deceleration regimes. In a recent
study, Thomas et al.[43] also showed that the same set of equations could be used to simulate
hydration of C3S in the presence of accelerators, as shown in Figure (2.7). Here it should be
noted that in all simulations presented in [18,43] a parameter “A” was used, which effectively
scales the density of C-S-H to a value that allows the simulations to fit better with the
experimental results. However, like the Dijon model, this model was also not validated for
systems with large differences in process parameters.

Page Number: 32
Figure 2.7: Hydration of C3S in DI-water and in a CaCl2 solution simulated using the
boundary N&G model (from [43]).

The most serious drawbacks in simulating hydration using classical N&G mechanisms, both in
the Avrami and Boundary N&G forms, are highlighted when the specific surface area of
particle of alite changes drastically and the density scaling parameter “A” is not used. Bishnoi
and Scrivener[17] implemented simulations using the classical definition of boundary
nucleation and growth without the use of any scaling parameter in the microstructural model
μic [70] to reproduce hydration profiles of alite-based systems with specific surface areas
ranging from 600 to 3000 cm2/g. The calorimetry profiles for these systems were generated by
M. Costoya during her doctoral study at EPFL [44]. Bishnoi and Scrivener showed
numerically that using the classical definitions of nucleation and growth (both Avrami and
Boundary N&G) and assuming a uniform density of C-S-H, only the cumulative heat
evolution profiles of alite could be reproduced. The deceleration regime, which is accentuated
in the heat rate plot, could not be simulated well (shown in Figure 2.8). Bishnoi and Scrivener
argued that the discrepancy occurred due to the formation of a very low volume of uniformly
dense C-S-H at the time of the main heat-peak (shown in Figure 2.9) which was insufficient
for significant impingement between growing particles leading to a deceleration in the
hydration rate. Low volume of C-S-H at the peak was expected due to low degrees of
hydration at the peak, which was calculated between 6.5% to 14% for the systems with the
lowest and highest specific surface areas respectively. Bishnoi and Scrivener also showed that
the variation in the parameters, growth rate of the product in particular, was huge and did

Page Number: 33
not scale proportionally with the initial specific surface area of the different PSDs. The results
from the work of Bishnoi and Scrivener[17] showed that a N&G based model using a uniform
density of C-S-H cannot explain the post-peak deceleration.

Despite the problems mentioned above, classical N&G models are used prevalently to simulate
hydration of alite based systems. This is primarily because the computation power needed to
execute simulations is low.

Figure 2.8: Measured and numerical reproduction of heat evolution profiles of alite systems
with different PSDs. The names of the PSDs and the median particle diameters of each PSD
are indicated at the top left of each figure. The simulations were implemented using the
classical definition of the nucleation and growth. As can be seen, the mechanism provides
satisfactory fits to the cumulative evolution of the extent of reaction but fails to fit the heat
flow rates. The image was adapted from [17].

Page Number: 34
Figure 2.9: 2D slice for the simulation of alite hydration at the peak. The white rim and the
grey cores represent the product and unreacted alite respectively. The image was adapted
from [17].

2.4.2 Diffusion based mechanism beyond the heat-rate peak

It has been argued that the slow diffusion of ions through the layers of hydrates depositing
over the alite particles leads to deceleration of the rate of reaction [45,46] beyond the heat
rate peak. The authors suggested this from their interpretation of the microstructure, in which
a layer of C-S-H is observed to be providing coverage to the alite particles.

To test this hypothesis, Bishnoi and Scrivener [17] implemented simulations for alite systems
with different PSDs using a mechanism that combined classical N&G and diffusion. The
acceleration part of the heat-flow rates were fit using the N&G mechanism and the
deceleration regime was reproduced using a diffusion based mechanism. In the simulations, the
authors observed that very good fits could be obtained for all systems (shown in Figure 2.10)
but the variation in the simulation parameters, especially the diffusion coefficient of C-S-H
used between different systems differed by over an order of magnitude.
This variation was in stark contradiction to the observations made for the same systems using
microscopy, where no characteristic or structural difference was found between the layers of

Page Number: 35
C-S-H. Peterson and Juenger [47] also reported a large variation in the diffusion coefficient of
C-S-H that was necessary to obtain fits for hydration profile of C 3S in solutions with
additives. In their study, the mechanisms was switched from boundary N&G to diffusion a
few hours after the peak. These studies, thus, strongly suggest that diffusion is not the driving
mechanism right after the peak, though it may become so later on.

Figure 2.10: Measured and numerical reproduction of heat evolution profiles of alite systems
with different PSDs. The names of the PSDs and the median particle diameters of each PSD
are indicated around the top left of each figure. The simulations were implemented using the
N&G mechanism for the acceleration part and diffusion through C-S-H for the deceleration
part. The simulations reproduce the evolution of hydration very well but the variation in the
diffusion coefficient is over an order of magnitude. The image was adapted from [17]

2.4.3 Nucleation and densifying growth of C-S-H

The mechanism of Nucleation and densifying growth (NDG) of C-S-H was proposed in the
study conducted by Bishnoi and Scrivener[17]. The NDG mechanism is based on a N&G
based kinetics combined with a diffuse criterion of growth for C-S-H. The diffuse growth of C-
S-H is the growth of a low density and highly porous product that quickly occupies space in
the microstructure as its packing density increases with time. Bishnoi and Scrivener
summarized the basic aspects of the NDG mechanism as follows:

Page Number: 36
 1. The surface of the particles is only partially covered by the product, and the
availability of product surface controls the growth of the product,
2. C-S-H grows uniformly over its exposed outer surface and
3. The rate of increase in packing fraction of C-S-H is assumed to decrease linearly with
the instantaneous density

Using this mechanism, the authors could obtain very good fits for the acceleration and
deceleration regimes of calorimetry profiles of alite systems with different PSDs (shown in
Figure 2.11). The authors also found that the growth rate of C-S-H (a simulation parameter)
required for the simulations was larger for systems with coarser particles and vice versa
suggesting an opportunistic growth in the presence of space in the microstructural matrix. It
should be noted that the inter-particle spacing is larger for systems with coarser particles and
vice versa and, consequently, the growth rate can also be related to the initial dispersion of
the particles. This variation will be further explored in this study in the upcoming chapters.

The variable density of C-S-H, which is an implicit assumption in the theory proposed by
Bishnoi et al., has also been previously suggested in other studies [48,49]. Powers in this study
proposed a progressive densification of the product [50]. The presence of a loose product can
also indirectly interpreted from results that emerge from NMR and QENS measurements,
which show the presence of excessive fraction of water close to the surface of hydrates [ 51,52].
A loose and low-density product growing as fibres or sheets can also be observed in many
micrographs [53,54], as shown in Figure (2.12).

The hypothesis of space availability driving the hydration of alite at later stages provides
good numerical reproduction of experimental data but remains controversial. In the current
study, this mechanism has been used to simulate the hydration of various alite based systems
and further evidence have been provided to confirm the theory. Several examples using this
theory are included in Chapter 4.

Page Number: 37
Figure 2.11: Measured and numerical reproduction of heat evolution profiles of alite systems
with different PSDs. The names of the PSDs and the median particle diameters of each PSD
are indicated around the top left of each figure. The simulations were implemented using the
diffuse growth criteria of C-S-H combined with boundary nucleation and growth mechanism.
The image was adapted from [17].

Page Number: 38
Figure 2.12: TEM image of OPC showing a relatively loosely-packed C–S–H. From[53]

2.5 Alite Hydration: Stage V

During stage V of alite hydration, the rate of reaction remains low and monotonically
decreases with time. In most early studies [56,55], it was postulated that during this stage the
reaction is controlled by diffusion of species through the layers of hydrates deposited on the
surface of reactant particles. Some recent studies [18,31] have indicated that the nucleation
and growth mechanism can also be used to describe hydration kinetics during stage V. It is
stated that the reduction in the reaction rate occurs due to decrease in the available surface
area for growth, as a result of impingements between solid phases and consumption of the
reactant material. Only a limited number of studies focusing on these stages of hydration have
been found and the reaction mechanism is still not clear. If a diffusion controlled regime is
assumed, the point of transition from the nucleation and growth mechanism to diffusion
controlled kinetics is also not clear. Due to the lack of experimental data, the stage V is not
studied here.

Page Number: 39
2.6 Microstructural modeling of cement hydration kinetics
In the past decades lot of progress has been made towards the development of mathematical
models in the field of cement-based systems. The focus has primarily been on the analytical
models, which simulate hydration kinetics numerically without the use or generation of digital
microstructures. The Avrami and Boundary nucleation and growth based models described in
section (2.4) fall in this category.

A few microstructural models [17,56,57,58,59,65] have been developed in the past to simulate
hydration of cement-based systems. With the rise of the computation power, the capabilities
of models in this category have grown significantly. In the following sub-sections
microstructure based models will be described. The description includes the summaries of
capabilities of the model, their applications and the mechanisms used to describe the
hydration process. The assumptions made in the models are highlighted along with their
individual sets of strengths and limitations.

2.6.1 DuCOM
The model DuCOM was developed by Kondo and Ueda [56] to characterize the hydration of
alite using a single spherical particle and concentric layered growth of hydrates. For the
simulation of hydration kinetics, the authors [56] implemented a series of mechanisms to
describe hydration at different stages. For early ages, a metastable barrier is allowed to form
inhibiting dissolution of alite and leading to the induction period. The layer is then allowed to
disintegrate with time exposing the alite surface and leading to the acceleration period. The
alite particle then shrinks and a layer of hydration products grow as concentric layers into the
capillary pore-space. At the critical value of thickness of the outward product, the kinetics
shifts to diffusion-controlled mechanism as the overall rate decreases asymptotically. Using the
combination of these mechanisms, the results obtained from this model were reported to agree
well with the evolution of degree of hydration with time. However, several fit parameters were
used to obtain these results and the variations in the parameters were not shown to scale in
proportion to the physical parameters.

While DuCOM serves as a simple and effective tool to capture hydration kinetics of simple
systems, the representation of the hydration kinetics is not accurate as the inter-particle
interactions are not taken into account. Also, the influence of available space in the
microstructure on the hydration kinetics cannot be considered within the realms of this
approach. It has already been shown in section (2.5) that the available space in the
microstructure and sizes of the particles strongly influence the hydration rate.

Page Number: 40
2.6.2 The Jennings+Johnson model
The model developed by Jennings and Johnson [57] was the first microstructure based model
developed to simulate hydration of cement based systems. The model was based on vector
approach using spheres to represent cement particles. To simulate the hydration process, the
spheres, representing reactant particles, were allowed to shrink and concentric layers of
products grew on the anhydrous particles. The overlaps between particles were accounted for
by redistribution of products in the capillary-space. Algorithms were implemented to simulate
the mechanical aspects of cement hydration by taking into account the effects of PSDs and
w/c ratios. Aspects related to pore-solution chemistry, nucleation of C-S-H and solid and pore
phase connectivities were proposed but not implemented in the final version. Due to
limitation in the computation power at the time of the development of the model, the
simulations represented the hydration of cement roughly at best. While the use of this model
was limited and lasted only for a short period, it paved the way for development of more
sophisticated models.

2.6.3 HymoStruc model

HymoStruc, an acronym for Hydration, Morphology and Structural development, was a
microstructure based model developed by van Breugel [58] in 1995. The “vector-approach”
based framework of the model is based on the same principles as the Jennings + Johnson
model[57]. Spherical particles are distributed in an “equi-spaced” fashion in a 3D virtual cube
to represent the initial state of the system (Figure 2.13). The remaining volume in the cube
represents water, which is allowed to react with the cement particles based on volumetrically
balanced chemical reactions. As hydration progresses, a single hydration product grows as
concentric layers on top of the shrinking anhydrous particles. The output includes digital
microstructure, heat evolution profiles and phase information. In addition to hydration,
modifications were made later in HymoStruc to compute autogeneous shrinkage and
development of strength based on simulated pore-structure [59].
The two most serious drawbacks of HymoStruc stem from the assumption that there is a
single product and the lack of intrinsic kinetics. Due to this, the model overlooks several
details about individual phases such as density, morphology, phase-assemblage etc. and their
effects on the overall hydration kinetics. The overall hydration rate needs to be calibrated
based on experimental results for each new mix. The other issue with the model is that the
hydration of the particles are simulated on a statistical scale, as opposed to treating the
particles individually based on their sizes. Another drawback in the model with regard to
mechanistic simulations is that the effects of impingements of solid phases, pore-connectivity
and solid-phase connectivity are not calculated explicitly.

Page Number: 41
Figure 2.13: Schematic representation of growth of products in HymoStruc model [58].

2.6.4 CemHyd3D model

CemHyD3D is a 3D microstructural model developed by Bentz and Garboczi[60,61]. The
model is based on a lattice-based approach, in which a 3D cement paste microstructure is
digitized onto a uniform cubic lattice and each volume element of the lattice to assigned to a
phase (porosity, alite, belite etc.). The evolution of the microstructure is simulated through a
large set of rules that are evaluated on the scale of individual pixels with dependencies on
participating materials, temperature, w/c ratio, impingements between solid phases etc. The
underlying mechanisms can be broadly classified into three main categories: (a) dissolution of
solids (b) diffusion of dissolved species according to random walk algorithm (c) nucleation and
growth of products.

The lattice-approach based model implemented in CemHyd3D presents a series of advantages

over the vector-based models. The spatial distribution of phases and random shapes of
particles can be very well reproduced digitally in lattice-based models. Integration of the
approach over a large set of volume elements permits the calculations of development of the
pore-structure, pore-connectivity, permeability, mechanical properties etc. Also, the model can
be used to simulate hydration of multi-phase systems in addition to pure-phase systems. A
wide range of properties can be evaluated including heat flow rate, phase volume fractions,
chemical shrinkage, setting time etc. Moreover, the microstructure at any point in time can
be analyzed to compute eletrical properties[62], transport properties[63] and mechanical
properties[64].

Page Number: 42
CemHyD3D model has been widely used over the years by several researchers to simulate the
development of a realistic microstructures. However, the model lacks the description of
intrinsic kinetics. In order to compensate for this, a large set of empirical rules are used, which
in turn present a number of inherent limitations. One serious drawback of the model is the
lack of a physical time scale parameter. For every cement mix, the time scale needs to be
calibrated with respect to the experimental data that provides a measure of the extent of
hydration with respect to time (generally in hours). This empirical dependency of the
computation time parameter on the physical time parameter renders the presence of
experimental results as a prerequisite. Hence, the model fails to serve as a predictive tool that
would replace experiments.

The time of precipitation of products, AFt and AFm phases in particular, in the model is also
calibrated based on observations from experimental data and, hence, do not necessarily follow
the rules of thermodynamics. Due to this limitation for each new cement mix with a different
composition, the stability of product phases either do not cohere with the experimental
observations or invoke the need of a fresh calibration.

The more serious drawback, perhaps, of the CemHyD3D model is its inherent lattice
resolution limit, which is typically 1 μm, on account of limited computational power available
at the time of development. Consequently, the simulations do not represent particles or
species which are less than 1 μm in size. In reality as several features in the cement
microstructure (gel pores, some capillary pores, several small anhydrous cement particles,
small nuclei of products) are much less than a micron in size.

When conceived CemHyD3D was an important development and its use points towards the
need for a new microstructural model that can make proper use of the currently available
computational power.

Page Number: 43
2.6.5 HydratiCA simulation model
HydratiCA simulation model is in development by Bullard [65] at NIST. Compared to
CemHyD3D, which was also developed at NIST, HydratiCA is based on more fundamental
principles of kinetics of the individual rate processes comprising cement hydration. The model
draws its basic principles from CemHyD3D and simulates hydration using a series of
mechanisms: (a) diffusion of mobile species in the solution (b) dissolution and growth of
mineral phases (c) complexation reactions among species in solution or at solid-solution
interfaces (d) nucleation of new phases. Mass and concentration balances are used to define
rules for the interactions and the overall hydration profiles are simulated using a 3D
microstructure based approach.

Stoichiometric solid phases are represented as a combination of separate chemical species

finely discretized into quanta of concentrations called cells that are arranged onto a regular
cubic lattice. For example, calcium hydroxide is represented as collection of Ca 2+ and OH-
species with concentrations equal to the number of cells they occupy individually. For solid
phases, concentration at a lattice site is directly computed from the fraction of volume it
occupies in the lattice. The digital microstructure generated in the simulations, thus, includes
ionic species, solid phases and water [22].

The reactions in the model are simulated using probabilistic rules that define changes in the
chemistry and structure in small time increments. The reaction progresses with dissolution of
solid phases followed by diffusion of ionic species. Inter-species interactions, both chemical and
eletrostatic in nature, are accounted for as hydration progresses. The input parameters: molar
volumes of phases, diffusive mobility of species, reaction rate constants, equilibrium constants,
activation energies, solubility products of solid phases and enthalpies of formation of different
products – provide necessary definition of the state of the systems and criteria for the
evolution of the microstructure and formation of new products. The differences in the initial
state of the systems reflects in the simulations and results in different evolutions of the
microstructure despite having the same built-in mechanisms. Unlike its predecessor
CemHyD3D, HydratiCA does not require calibrations for new mixes as all calculations are
based on laws of thermodynamics and values from literature, which are given as input.
However, many of the input parameters are not available in the literature or accessible
experimentally so a large lumber of parameters must be fit. For example, the equilibrium rate
constant for the formation of C-S-H (both forward and backward reactions) is required for the
simulation of C3S hydration. As a precise value of this parameters has not been
experimentally determined, the user is required to provide the best estimated value. This can
be critical because it affects the results of the simulation directly.

Page Number: 44
HydratiCA has been used to investigate the plausibility of various possible mechanisms of
hydration of C3S [19] and multi-phase systems[66]. While it has been shown that the model
has the potential to be used to investigate mechanisms driving hydration and validate
hypotheses, the computation power needed to run the simulations presents the biggest hurdle.
The simulations for larger systems require multiple processors for spatial decomposition of the
computational domains. Furthermore, the time required for the simulations is also very large.
For example, simulating the dissolution of C3S (pure) in DI-water for 2 hours requires about 2
weeks of simulation time on a supercomputer that utilizes multiple processors. In smaller
computational systems, the same simulation would require several months to converge. This
limits the applicability of the model. Also, as the simulation times of the model are large, the
choice of input parameters becomes even more critical as re-running the simulations multiple
times to establish precise input values is excessively time-consuming.

While HydratiCA should offer a lot of advantages over several existing models, the
computational requirements make it a tool that is exclusive to the users who have access to
powerful computational systems. Large simulation times that are required by the model make
it infeasible for simulations of systems that are large or for simulations of hydration at later
ages.

2.6.6 Integrated Particle Kinetics Model (IPKM)

IPKM was developed at EPFL by Navi and Pignat [67,68]. The model is based on vector
approach in which three-dimensional microstructure are generated by shrinking the reacting
cores and growing products at concentric layers and new nuclei in the pore-space.

The mechanisms that are included in the model are Avrami nucleation and growth, Phase-
boundary growth and diffusion. The Avrami equation is used for the initial nucleation and
growth regime of the reaction. In the second stage, boundary growth is used, where the rate of
reaction depends on the uncovered surface area of the individual reactant particles. In the
third phase, the rate of reaction is diffusion controlled, where the rate of consumption of the
reactant material is inversely proportional to the thickness of the hydrates deposited on the
surface of reactant particles. The hydration kinetics are simulated at the scale of individual
particles taking into account the effects of impingements, evolution of the exposed surface
area and porosity in the vicinity of the particles. This is a more accurate representation of the
microstructure as compared to other vector-approach based microstructural models like
HymoStruc[58].

Page Number: 45
Using the model, the authors simulated hydration of cement-based systems and obtained good
fits [67]. Simulations were also implemented to post-process the digital microstructure and
characterize pore-sizes and connectivity[69]. However, the model was limited because of the
lack of data-structure algorithms to reduce the computational complexity. As the model
treated hydration at the scale of individual particles, the simulations were computationally
expensive and simulations were limited to ten thousand particles. This was a serious drawback
because a sample of real cement contains millions of particles.

Due to these limitations, IPKM was not extensively used. However it paved the way for the
development of the modeling platform μic, which draws it basic principles and
implementations from IPKM.

2.6.7 μic modeling platform

The modeling platform μic was developed by Bishnoi and Scrivener [70] in 2009. The model is
based on a vector approach using spherical particles and concentric growth of products as
layers. However, μic provides a new implementation of the vector approach that enables
simulations using millions of particles with a realistic representation of the original PSD.
Built-in libraries are included in the model that allow the computation of hydration and
interactions at the scale of individual particles. Several algorithms from data-structures have
been implemented to expedite the simulations that describe hydration of cement-based
systems in realistic time-scales with a simulation time of a few hours. The benchmark for the
simulation times were established using a single-processor desktop computer.

The model was developed with a view to provide flexibility to the users on account of our
gaps in understanding of the driving mechanisms. As such, the core of model works as a
stand-alone platform that only simulates reactions and generates “resolution-free”
microstructure. “Plugins” are used to describe customized mechanisms and rules in the form of
simple mathematical equations that drive the hydration kinetics, simulate the consumption of
the anhydrous particles and the formation of new products. Different sets of plugins can be
used to describe several variations in the microstructure such as the following:
1. Initial state of system including w/c ratio, PSD, phase-assemblage of the clinker
particles (polyphase systems with a single PSD or collection of monophase systems
with two or more PSDs), addition/substitution by filler particles etc. can be defined by
the user.
2. The spatial arrangement of the phases in the microstructural volume can be

Page Number: 46
 customized. Products can be defined to nucleate as layers on particles or as separate
nuclei in the capillary-pore space or both. Products can also be defined as interspersed
with other products or as a separate phase. This is demonstrated in Figure (2.14).
3. The reactions, reactants, products, and the growth criteria of products can be described
as per the users input. For example: growth of C-S-H can be described as concentric
layers or as fibres. Products can be allowed to grow as fractals, fibers etc.. This is
shown in Figure (2.15).
4. Plugins can be written to describe rules for precipitation of products based on user-
defined thermodynamic rules or manual triggers. The triggers can also be defined to be
based on slower or faster reaction rates. For example, diffusion can be set to begin at
the peak or when the rate of diffusion becomes lower than the rate from classical N&G
based mechanism for products.
5. As finite-difference scheme is built into the core of the model, equations can be
described to evaluate the progression of hydration in customized time-steps, both fine
or coarse. Hydration at very late ages can also be simulated.
6. Plugins can be written to describe hydration as N&G based, diffusion based or other
custom mechanisms or combination of different mechanisms occurring in customized
time domains.
7. Plugins can also be coded to customize the outputs. Numerical outputs can be
generated in the desired units and in file-formats as per the users' need. The
microstructural outputs can also customized to generate a wide range of information as
needed by the user.

Page Number: 47
Figure 2.14: Customized distribution of phases in the μic microstructure. (from [70])

Figure 2.15: Simulation of products (a) as smooth layers and new nuclei in the capillary-pore
space (b) as controlled nucleation of small discrete nuclei on the surface of reactant particle.
(from [22])

Page Number: 48
Using the customized plugins, μic is well suited to investigate various phenomena at the
microstructure level, such as the effect of PSDs, w/c ratio and filler addition. As hydration
progresses and the microstructure evolves, the effect of impingements is taken into account
and is reflected in the spatial distribution of the products (products are not allowed to
nucleate or grow in space that is already occupied). The output of the model includes: 3D
microstructural images, phase volume and mass fractions, heat output and details involving
each reaction. The microstructural images are generated using resolution-free information that
includes the radii and coordinates of each particle that can be utilized to simulate the
evolution of properties such as: porosity, pore-size distribution, permeability, strength etc.

The model μic has been used to describe the hydration of alite studying the effect of particle
sizes[17]. The model was used to test and validate the applicability of several N&G based
equations, as described in the previous sections. Using the flexibility of the model, plugins
were used to describe the hydration of alite using a diffuse growth criteria of C-S-H, which
reproduced the heat evolution profiles of several alite based systems very well, as described in
section 2.4.3. The speed and flexibility of μic allows the simulation of complex systems with
realistic particle sizes and specific mechanisms. The model offers a range of output results that
can be used to test and validate mechanisms and also to post-process and analyze the
mirostructure to derive the development of other properties. In this study, the modeling
platform μic is used to simulate hydration of several alite-based systems.

This study sets out to integrate in the model thermodynamics and solution-phase chemistry,
which determine the rules for dissolution and precipitation of phases. The one other limitation
of the model is the spherical approximation of the particles, which affects the pore structure
and, consequently, the connectivity of the pore-network. Another limitation of the model is a
large number of customizable elements, which on one hand offers flexibility to the users and
on the other hand presents a certain degree of complexity.

Page Number: 49
2.7 Current Study
The current study is based on μic that simulates hydration of cement-based systems.

For the simulation of the early age mechanisms, dissolution of phases is implemented using
our latest understanding from the field of Geo-chemistry and following the theories of solution
controlled dissolution as proposed by Juilland et al.[20]. Thermodynamic laws are
incorporated in the core of model to provide a generic description of the hydration
mechanisms and dissolution/precipitation of phases. The dissolution parameters are calibrated
such that they can be generically used to describe the hydration of alite in different initial
conditions without the need for re-calibration or re-derivation of the simulation parameters.
The calibrations are done using different sets of experiment results from a wide range of
systems to ensure that the final values generically describe the physical meanings they stand
for. The output of the simulations are extended to include solution-phase chemistry to provide
more flexibility to the users for validation against a variety of experimental results.

For the description of hydration during stages III and IV, simulations are implemented using
a modified form of boundary nucleation and growth incorporating the change in the density of
products. The mechanism has been adapted from the study of Bishnoi et al.[17]. The
simulations are implemented and validated against experimental results for several types of
systems to illustrate the ability of the model to generically describe hydration.

While several mechanisms have been incorporated in the model, several algorithms have been
implemented in parallel to reduce the level of computational complexity. The time-step for
simulations is kept small so as to capture the evolution of the microstructure at small time
intervals. At the same time, special care is taken to ensure that the overall simulation time
does not exceed a few hours.

Page Number: 50
Chapter 3: Simulation Techniques for Alite Hydration
This chapter describes the simulation techniques implemented in the model to describe the
hydration of alite. The model has been developed to describe mechanisms at both early and
late ages of hydration. The following sections describe the mechanisms and techniques
employed and their mathematical descriptions in the model. The assumptions and
approximations made as part of the description of the mechanisms are also outlined.

3.1 Introduction
In Chapter 2, several proposed mechanisms pertaining to hydration of alite were delineated.
In this section, concepts which are of specific relevance to development of the model are
described. The heat evolution profile is alite is the primary experiment result that captures
the effects of several process parameters, describes kinetics and is used for comparison against
simulation results. Figure (3.1) shows a typical heat evolution curve for the first 50 hours.
Five main stages can be identified:
1. Dissolution period: During the first minutes, alite dissolves and ions are released into
the pore solution, but the reaction rapidly slows to a low rate.
2. Induction period: During the following hours low activity and low heat evolution are
observed.
3. Acceleration period: The rate of heat evolution increases for several hours until it
reaches a peak
4. Deceleration period: The rate of heat evolution decreases for a few hours
5. Slow hydration period: After the deceleration period, the hydration rate continues
at a low decreasing level.

Page Number: 51
Figure 3.1: Typical heat evolution curve of alite-model systems until approximately 2 days of
hydration.

The mechanisms governing these periods have recently been reviewed by Bullard et al. [71].
Most controversy surrounds the first two periods. The two principal theories are that the
slowdown in the rate of reaction occurs either: due to formation of a metastable barrier
product on the surface of the reacting alite grains; or due to a decreasing rate of alite
dissolution caused by the build-up in concentration of ions in solutions. Arguments for and
against these hypothesis are fully discussed in [71].

In a recent publication [72] the similarities were demonstrated between alite reaction and the
dissolution of many other minerals studied in geochemistry, where the theory of a slow
dissolution step due to the buildup of ions in solution is widely accepted. The dissolution rate
of most natural minerals does not follow a smooth relationship with respect to the saturation
state of the solution [73,74]. Far from equilibrium, high rates of dissolution are enabled by
etch pit opening. Whereas at high concentrations, the transition to the slow dissolution regime
occurs at the levels of undersaturation still several orders of magnitude more dilute than the
equilibrium concentration. In this regime, the driving force from undersaturation is insufficient
to activate the etch pit opening and dissolution occurs by a retreating step mechanism that is
much slower. We refer hereafter to this mechanism as solution controlled dissolution – SCD.

Page Number: 52
As described in Chapter 2, periods 3 and 4 are now generally accepted to be controlled by the
nucleation and growth of the main hydrate phase C-S-H. The kinetics in this period have been
successfully reproduced by an analytical Boundary Nucleation and Growth algorithm [75] and
numerically in the μic modelling platform [76]. In the later case it was found that period 4 was
only well captured across a range of particle sizes, when C-S-H was presumed to grow out
rapidly with a low packing density until the products growing from adjacent grains impinged.
Concurrently this diffuse product densifies as a function of its distance from the underlying
grain. This mechanism is subsequently referred to as nucleation and densifying growth –
NDG.

As discussed in references [71,75,77]and demonstrated in [74], the weight of evidence indicates

that diffusion through the C-S-H layer is NOT the rate controlling mechanism in period 4,
although it may become so during period 5.

Several numerical models[78,75,79,80,81,82,83,84,85] have been developed to describe and

quantify the hydration of alite. These are discussed in a recent detailed review [71]. The
modeling platform µic [78] is unique in being able to rapidly simulate the microstructure using
a representative volume of around 100 µm in dimension, taking account of particles down to
below 0.1 µm (around 2-10 million particles in the simulation), based on physical and
chemical reaction mechanisms. Other models, either simplify simulation by dealing with a
single particle [80]; have a resolution limited by the pixel size of around 1μm [81]; implement
kinetics purely by external empirical calibration with overall reaction rates measured by
experiment [83,79,75]; require intensive computational power and protracted time-lengths to
simulate digital microstructure and reaction rates [84].

In this chapter the implementation of two reaction mechanisms is described: - solution

controlled dissolution (SCD) for stages 1 and 2 and nucleation and densifying growth (NDG)
for stages 3 and 4 – can be used to capture the reaction kinetics for alite. During the SCD
regime the composition of the solution phase is calculated at each step of the hydration
process and thermodynamic calculations are used to determine when the different phases
precipitate. This regime ends with the precipitation of calcium hydroxide which corresponds
well with the onset of the acceleration period. The main heat evolution peak is well captured
by the nucleation and densifying growth mechanism as first shown by Bishnoi and
Scrivener[74]. There is transition period between the two numerical regimes the reasons for
which are discussed later.

Page Number: 53
3.2 Basics of µic
The modeling platform µic [78] is used in this study, a detailed description of which can be
found in Chapter 2. This is a microstructural modeling framework, which uses a vector
approach to generate a resolution free microstructure. That is to say particles are represented
geometrically as spheres with their respective positions and radii stored as vectors. Reactions
and mechanisms are implemented as “plugins”, which are user definable. Reactions,
consumption of reactants and formation of products are carried out whilst monitoring the
volumetric quantities of each phase. The model is flexible and can be used to study the effects
of various processes and mechanisms in hydration of cement-based systems.

Particle size distribution plays an important role in determining the rate of reaction
[74,85,86]. Finer particles with higher specific surface area react rapidly compared to cement
with coarser particles having low specific surface area. In order to capture the effect of particle
size distribution it is crucial to have a representation of the real particle size gradation in the
model.

Initially, in the simulations, alite particles are randomly dispersed in the microstructure
without any overlaps. Spheres are used to represent the alite particles using the real PSD as
obtained from laser granulometry. Product phases which form as a result of hydration and
their arrangement in the microstructure are also defined by the user. As the microstructure
evolves with the growth of products, the effects of impingements between hydration products
and clinker phases are taken into account to evaluate the overall rate of reaction.

In all the simulations presented in this and the following chapters, the arrangement of
hydrates is as shown in Figure (3.2). Inner product C-S-H forms in the space left by the
dissolving alite grains was assumed to have a uniform density of 2.10 g/cm 31. The density
values used for alite, water and portlandite are 3.15 g/cm 3, 1.0 g/ml and 2.24 g/cm3
respectively. Outer C-S-H forms as a layer initially with low packing density. This packing
density increases over time from a low value to 2.10 g/cm 3, as described below and in [78]. It
should be noted that the change in density of the outer C-S-H only indicates its packing
density. As such, the H20/Si and Ca/Si ratios of outer C-S-H are the same as for inner C-S-H.
Portlandite nucleates in the original water filled space. This arrangement of hydrates in the

1 We are aware of the extensive debate about the density of C-S-H, which depends on the water content. We
consider the value of 2.1 for inner or “high density” C-S-H to be reasonable for a composition of C 1.7SH4 as
expected in standard systems. However changes in this value will only have a minor impact on the rest of
the paper.

Page Number: 54
microstructure follows the widely accepted view of microstructural development from
microscopical observations [87,88,89,90].

As alite dissolves and hydrates, C-S-H (inner and outer products) and portlandite (CH) are
formed according to volumes given in Equation 3.1, where the H/S ratio assumed for C-S-H is
4.0.

1. 0⋅VC S +1.318⋅V H =1.00⋅Vinner C−S−H +0.495⋅V outer C−S−H +0.593⋅V CH (3.1)

3

The heat released under isothermal conditions is computed by using the standard enthalpy of
reactions [89,90], as given in Table 1. At each step, the amount of each hydrated alite is
calculated and used to compute the heat that would be released under isothermal conditions
at T = 293.15 K. The cumulative heat released is then computed by the addition of heat
released in individual steps. The rate of heat evolution is computed by differentiating the rate
of cumulative heat released with time.

Figure 3.2: Distribution of phases in different cementitious systems as used in the model.

Table 3.1: Standard enthalpies for alite reactions.

Reaction Heat Released (kJ/mol)

C3S hydration -120 [1,33]
C-S-H precipitation 30 [1,33]
Portlandite precipitation 20 [1,33]

Page Number: 55
Up until the precipitation of portlandite (stages 1 and 2, Figure 3.1), the rate controlling
mechanism is the solution controlled dissolution (SCD) of alite. The rate of dissolution is
determined by the solution concentration following the hypothesis advances by Juilland et al.
[2], as described below. Subsequently the main heat evolution profile is simulated by
nucleation and densifying growth (NDG), this fit to the main heat evolution peak and so
there is a virtual starting point for the NDG regime at t0, as will be described later.

As described earlier, the simulations are implemented with two different kinetic mechanisms:
SCD and NDG. The heat evolution during the SCD phase is computed using the set of
Equations (3.2 – 3.5). It should be noted that negative sign indicates decrease in the mass of
alite as dissolution progresses.

dm alite (before C-S-H precipitates) (3.2)

Q1=− ⋅Halite dissolution
dt

dm alite (after C-S-H precipitates) (3.3)

Q2=− ⋅Halite dissolution+m C−S−H (dt)⋅H C−S−Hformation
dt

Q3=Q 2+m CH (dt)⋅HCHformation (after C-S-H and CH precipitate) (3.4)

where, Q, H and m are the heat flow rate, enthalpy and mass respectively.

Likewise, the heat evolution during the NDG regime is calculated using Equation 3.5.

dm alite (3.5)
Q2=− ⋅Halite hydration
dt

where, Halite hydration is fixed at -484 J/g (-110.51 kJ/mol).

Page Number: 56
3.3 Solution controlled dissolution (SCD)
The dissolution of alite plays a key role in the evolution of the concentration of ions in the
pore solution. It has been shown by Bullard et al. [91] that the concentration in solution
become uniform throughout the overall volume very fast. Furthermore, any concentration
gradients only exist on the scale of the distances between particles, which are very small (a
few μm). Therefore, in the model the solution concentrations are considered to be uniform
throughout the simulation volume.

The dissolution of alite is described by Equation (3.6). In this study “[]” and “{}” represent
concentrations and activities respectively. It should be noted that the ionic species used in
Equation (3.6) are “elemental” and their respective speciations will depend on the overall pH
of the pore-solution.
C3S + 3H2O ----- > 3Ca2+(aq) + 4OH-(aq) + H2SiO42-(aq) (3.6)

Juilland et al. [72] proposed that the rate of dissolution of alite varies as a function of the
undersaturation of alite. At high degrees of undersaturation, i.e. on, and shortly after the
addition of water, the dissolution is rapid, with the creation of etch pits. As ions enter
solution, the undersaturation of the solution decreases until a critical degree of
undersaturation is reached, where there is no longer enough energy to generate etch pits and
the rate of dissolution slows down dramatically. Beyond this point dissolution is slow
corresponding to dissolution by step retreat from existing etch pits. The dependence of
dissolution rate on undersaturation close to this transition is controversial as some claim it is
almost a step change [92], but here we assume that the rate of dissolution varies linearly with
the parameter βalite, which is the ion activity product of the ions in solution divided by the
equilibrium solubility products (Equation 3.7), as this fits best with data report in the
literature [24,25,26]. While βalite represents the degree of undersaturation, log β alite is the
Saturation Index (SIalite).

IAPalite ( t)
β alite ( t )= (3.7)
K sp alite

where, IAP alite ( t)={Ca 2+ }3.0⋅{ H2 SiO42- }⋅{ OH - }4.0 is the ion activity product for alite and,
K spalite (t )=1.14⋅10 is the equilibrium solubility product for alite.
−4

The value for the solubility product of alite is a matter of discussion; here we use the value
given by Bellmann et al. [93]. This value was selected (as opposed to other values, i.e., Stein

Page Number: 57
et al. [94]) as this series of calculations assumes H 2SiO42- as the primary charged silicate
species in solution. It should be noted that the value of both Ion Activity Product and
Solubility product of alite will change depending on the choice of primary charged silicate
species.

Overall the rate of dissolution of alite (µmol. m-2. s-1) is described mathematically by Equation
(3.8). This equation was modified from the form as originally described by Damidot et al. [95]
in the light of hypotheses advanced by Juilland et al. [72], a brief description of which is
provided above. It describes the dissolution regimes of fast dissolution facilitated by etch pit
formation for undersaturation below the critical value of β max (for βalite <= βmax given by
Equation 3.8a) and slow dissolution driven by step retreats beyond the critical
undersaturation (for βalite > βmax given by Equation 3.8b). The negative sign in the equations
indicates loss in the mass of alite as a function of time as dissolution progresses.

−dm alite ( t)
=( −k diss⋅aSSA⋅( β max−βalite ( t)) ) −a SSA⋅c step retreat (3.8a)
dt

for βalite(t) <= βmax , and

dm alite (t )
=−aSSA⋅cstep retreat (3.8b)
dt

for βalite(t) > βmax

where, dm (mol. m-2) is the amount of alite dissolved, βmax is the critical degree of
undersaturation of alite, kdiss (mol. kg. m-4. s-1) is a dissolution rate constant, a SSA (m2. kg-1) is
the specific surface area of particles divided by the volume fraction of the water in the
representative volume (to take in account the w/c ratio), and c step retreat (mol. kg. m-4. s-1) is the
step retreat constant. In the simulations, βalite(t) is calculated at each step based on the
amount of alite dissolved and taking into account the total volume of water present in the
representative volume. As such, the rate of dissolution of alite varies as a linear function of
undersaturation (β) when the undersaturation is high. When the undersaturation approaches

Page Number: 58
a critical undersaturation value, given by β max, dissolution continues at a constant value driven
by step retreats (given by aSSA. cstep ). The equation consists of three parameters which
retreat

need to be calibrated: kdiss, βmax and cstep retreat

. The calibrations of these parameters are
discussed in section (3.7).

With the dissolution equation implemented in the present form, the effects of surface area of
alite particles and differences in the initial states of the mixing solution can be described
without the need for any modification in the format of the equation. The description of the
effects of these process parameters is further discussed in the upcoming sections and chapters.

3.4 Activity calculations

Solubility calculations were implemented to compute the time of precipitation of C-S-H and
portlandite in the model. The molality of all species in the model were calculated using the
concentration of ionic species as obtained from simulation of dissolution of alite and taking
into account the volume of water present in the representative volume 2. Equation (3.9)
describes the activity of species in a solution. The expression for calculating the ionic strength
is given by Equation (3.10).
a i =m i⋅γi (3.9)

I=0.5 ∑ m i⋅zi
2
(3.10)
where γi is the activity coefficient, zi is the unit charge and m i (moles. Kg-1) is the molality of
the species i. For the calculations of the activity coefficients, Truesdell-Jones form of [96]
extended Debye-Hückel equation shown by Equation (3.11) was used. This equation is
applicable for electrolytes having an ionic strength up to 2.0 mol/L, which is accepted as a
good upper-bound for cement-based systems.

−A T z2i √ I
log  γi= +b i⋅I (3.11)
1+BT a i √ I

where,

A T=( 2 .74⋅10−6 ) T2+(−7 .60⋅10−4 ) T+ ( 0 .4916 ) (3.12a)

2 Molality (mi) = moles of species i( in moles) / Mass of water (in Kg). Water is assumed to have a density of
1 g/ml here.

Page Number: 59
B T=( 1 .62⋅10−4 ) T+ ( 0 .2799 ) (3.12b)

T (K) is the temperature, I (moles. Kg -1) is the ionic strength and, a i (Å) and bi (Å) are
Kielland ion-specific parameters.
Values of AT and BT were calculated by using a temperature value of 293.15 K in
Equations(3.12a - 3.12b). The values of ai and bi for different ionic species used in the
simulations are shown in Table 3.2.

Table 3.2: Ion specific parameters for ionic species. All values have been adapted from [26]

Ionic Species ai (Å) bi (Å)

Ca2+ 4.86 0.15
OH- 10.65 0.21
3
H2SiO42- 1/BT 0.30/AT

The expression of Saturation Index (SI) for any phase in solution is given by Equation (3.13)

IAP
SI=log
( )
K sp
(3.13)

where IAP is the ion activity product and Ksp is the solubility product. A negative value of SI
implies undersaturation, a positive value indicates supersaturation and SI = 0 implies
equilibrium with respect to the pore solution. A product is expected to precipitate when it is
supersaturated with respect to the pore solution.

In order to implement the thermodynamic calculations with precision, the time-resolution

should be fine. As such, the time step for set at 0.01 seconds. Figure 3.3 shows schematically
the steps followed in the simulations controlled by SCD.

3 For H2SiO42-, Davies form of equation was used. The ion-specific parameters were modified to incorporate the
effect of temperature.

Page Number: 60
Figure 3.3: Flow-chart depicting the steps followed for simulation of the SCD regime in the
model.

3.5 C-S-H precipitation

There are several thermodynamically equivalent (but kinetically inequivalent) expressions for
computing the ion activity product of C-S-H. Rothstein et al. in their study [97], reported
Equation 3.14a to describe the solubility of C-S-H. This equation assumes a fixed Ca/Si ratio
of 1.7 and does not consider contributions of ions such as CaOH +. Glasser [98] proposed
Equation 3.14b, which assumes a Ca/Si ratio of 1.75 and takes into account the contributions
of CaOH+. In this equation, y represents the ratio of concentration of CaOH + ions in the
solution to the sum of the Ca 2+ and CaOH+ concentration. At the high pHs occurring in
hydration the formation of CaOH+ is important and cannot be ignored (Figure 3.4), therefore
we used a generalized form of Equation (3.14b); rewritten as Equation (3.15) to account for
variability of the Ca/Si ratio.

Page Number: 61
IAPCSH−Rothstein ={Ca 2 + }1.7⋅{ H2 SiO42- }⋅{ OH - }1.4 (3.14a)

2+ (1.75−y / 2 ) + y /2 2- - (1.5−y / 2 )
IAP CSH−Glasser ={Ca } ⋅{CaOH } ⋅{ H 2 SiO4 }⋅{OH } (3.14b)

Ca Ca
(3.15)
( −y /2 ) (2 −y /2−2 )
2+ + y/2 2- -
IAPCSH ={Ca } Si
⋅{CaOH } ⋅{ H2 SiO4 }⋅{ OH } Si

To calculate the activity of the CaOH+ species, the law of mass-action was used (Equation
3.16, [99,100]). Using this equation, the concentration of CaOH+ can be computed using the
set of equations described by Equation 3.16-3.18, where γCa2+ and γCaOH+ are the activity
coefficients of CaOH+ and Ca2+ (using the simulated Ionic Strength) respectively.

2+ + + -12.78
Ca +H2 O=CaOH +H where , log k=10 (3.16)

[Ca2+ ]+[CaOH+ ]=[Ca Total ] (3.17)

γCaOH+⋅[CaOH+ ] 10−12.78 (3.18)

2+ 2+
=
γCa ⋅[Ca ] 10-pH

Page Number: 62
Figure 3.4: Ratio of activities of CaOH+ and Ca2+ as a function of pH.

The calculation of the saturation index of C-S-H also requires knowledge of its solubility
product. Various values are reported in the literature corresponding to CaO-SiO 2-H20 systems
with specific Ca/Si ratios. Values as a function of Ca/Si ratio can be obtained by
consideration of the CaO-SiO2 solubility curve published and adapted by several authors
[94,76] (Figure 3.5) and the widely published relationship [89,101,102,103,104,105] between the
Ca/Si ratio of C-S-H as a function of Ca2+ (Figure 3.6). Mathematical fits to these two sets of
data, shown in the respective figures, were used to obtain the solubility product of C-S-H as a
function of Ca/Si, as shown in Figure (3.7). The solubility product values reported for Jennite
(Ksp = 10-11.85) [88] and Tobermorite (Ksp = 10-8.0) [106] correspond to Ca/Si ratios of 1.5 and
0.833 respectively are also shown in Figure (3.7). This plot indicates that C-S-H is somewhat
more soluble than these reference phases as expected for a disordered phase. The line of
congruent dissolution of alite (Ca:Si = 3:1) is also shown in Figure (3.5) and from the point of
intersection of this congruent dissolution line and the CaO-SiO 2 solubility line, the Ca/Si ratio
of the first product formed can be calculated, which is also indicated on Figures (3.6 and 3.7).
However, this first C-S-H is predicted to have a low Ca/Si ratio, which would require silicate
chain length more than the dimeric species typically seen at such early stages of hydration
[107]. In any case the Ca/Si ratio of the C-S-H quickly evolves to around 1.7 in around 20
minutes, so small differences between predicted and actual composition will not significantly
affect the simulations. The saturation index of C-S-H was calculated at each step from these
data. Precipitation was assumed to occur at the first step where SI C-S-H exceeds 0.

Page Number: 63
Subsequently calcium and silicate ions were removed from the solution to keep SI C-S-H
= 0.
The Ca/Si ratio of the product at each step was calculated according to the instantaneous
calcium concentration. At the point of C-S-H precipitation, the Ca/Si was calculated as 0.84.
It should be noted here, that in the simulations it was approximated that all of Si
concentration in the pore solution are considered as H 2SiO42-. We are aware that the
speciation of silica differs from this, including H 2SiO42-, H3SiO4- and H4SiO40, but at present
this can only be calculated iteratively using a system of linear equations and so has not yet
been implemented in the computation for reasons of speed.

Figure 3.5: Solubility curve for C-S-H taken from data of Brown et al. [76]. The dotted line
shows the calibration of the SiO2-CaO solubility data as implemented in the model. The solid
line shows the path of solution concentration for congruent dissolution of alite which will
intersect the curve at the point indicated by the arrow. From this point and Figure (3.5) by
using the equivalent total Ca concentration, the composition of the first C-S-H to precipitate
can be calculated.

Page Number: 64
Figure 3.6: Evolution of Ca/Si ratio as a function of Ca 2+ concentration. The calibration of
Ca/Si ratio was done using data points (shown as solid symbols) extracted from several
authors [89,93,92,91]. The point at which first product forms is also indicated.

Page Number: 65
Figure 3.7: Simulated plot of log (K sp-C-S-H) as a function of Ca/Si ratio. The data points
correspond to tobermorite and jennite [88,87]. The point at which first product forms is also
indicated.

3.6 Portlandite precipitation

To calculate the Saturation Index of portlandite, Equations (3.19 - 3.20) were used. K sp at
293.15 K for portlandite was fixed at 10-5.18, as reported in the PHREEQC thermodynamic
database of Parkhurst [1]. The computation of activities of Ca 2+ and OH- was implemented in
the same way as described in sections (3.4 and 3.5). In the model, portlandite is allowed to
precipitate when SIPortlandite reaches a value of 0.40 for the first time, which corresponds to the
maximum widely reported in the literature[76,73].

Page Number: 66
 IAPPortlandite
SIPortlandite =log
( K s p Portlandite ) (3.19)

where,

IAPPortlandite ={Ca 2+ }⋅{OH - }2 (3.20)

3.7 Calibration of parameters for SCD regime

As discussed above, the equation controlling the dissolution is described by Equation (3.8).
This equation contains 3 parameters:
• kdiss: the dissolution rate coefficient in the faster dissolution regime, before the onset of
step retreat.
• cstep : equals the assumed constant rate of dissolution in the step retreat regime
retreat

when multiplied by the specific surface area of the particles

• βmax: the critical degree of undersaturation of alite for onset of the step retreat regime.

These parameters were calibrated iteratively by trying to match the first part of the heat
evolution (up to the end of the induction period) from the simulations to the experimental
data from isothermal calorimetry. Figure (3.8a) shows the sensitivity of the output
simulation curves during the first few hours to the values of these coefficients. All three
values must be defined independently to obtain the correct shape of the experimental heat
evolution curves. The value of kdiss influences mainly the steepness of the slowdown in
dissolution rate, while the parameter cstep retreat affects the minimum heat evolution during the
induction period. The value of βmax changes the point of transition from the undersaturation to
the step-retreat regime and, consequently, the time spent in the step-retreat regime. At higher
values of βmax, the dissolution kinetics stay in the fast dissolution regime longer and, hence, a
shorter time in the step retreat regime is required to reach portlandite supersaturation. The
reverse happens when the value of β max is lower, as shown in Figure (3.8d). The values of all
three dissolution parameters affect the time at which portlandite precipitates. These best
values obtained are shown in Table 3.3 are kept constant in all subsequent simulations.

Page Number: 67
Table 3.3: Calibrated values of dissolution parameters at the temperature of 293.15 K

Dissolution parameters Values

kdiss 24.0 (10-6 mol. kg. m-4. s-1)
βmax 10-22 (Unitless)

cstep retreat 4.6 (10-29 mol. kg. m-4. s-1)

The best value of βmax was found to be 10-22.0. As discussed by Julliand et al.[72] it is seen that
the onset of slow dissolution takes place when IAP alite is still many orders of magnitude away
from the hypothetical equilibrium solubility for alite 4 between 3 and 1.14 x 10-4 [86,82]. This is
depicted in Figure (3.9), which shows log (IAP alite/Ksp-alite)5 with respect to time. Once
portlandite precipitates (when SIPortlandite reaches a value of 0.40), the concentrations of calcium
and hydroxyl ions decrease causing a corresponding decrease in log (IAP alite/Ksp-alite) back to
the high dissolution regime. Therefore one reason for the acceleration period is due to the
transition of the dissolution kinetics back to the high dissolution regime.

The relationship between the dissolution rate and the undersaturation of alite (log(βalite))is
shown in Figure (3.10). The dissolution profile of alite comprises of two regimes:
Undersaturation below βmax in which the rate varies linearly with the difference (βmax - βalite(t))
or logarithmically with the Saturation Index; and undersaturation above βmax corresponding to
the step retreat regime, where rate of dissolution is constant equal to (aSSA.cstep retreat). As can be
seen, after the precipitation of portlandite the rate of dissolution increases in magnitude and
transitions back to the high dissolution regime. The form of this curve is the same as that
proposed by Juilland (Fig 20 in [72])6. The solid symbols on this curve are the experimental
data points from Damidot [77]. It is interesting to note the excellent agreement between the
values obtained from this study of alite synthesized by Costoya and the value from a totally
independent alite (or C3S) synthesis by Damidot.

4 NB it is not possible to precisely measure the equilibrium solubility for alite, due to the precipitation of the
hydrate phases. Values between 3 and 10-4 are reported.
5 log(IAPalite/Ksp-alite) = log(βalite)
6 The values for alite undersaturation in the original paper were not calculated. The values are simple
projections which represent the decrease in undersaturation from high values to very low values as dissolution
progresses.

Page Number: 68
 (a) (b) (c)
Figure 3.8: Sensitivity of the dissolution parameters as indicated at the top of each figure. In
all simulations w/c = 0.4 and a SSA = 2000 cm2/g (Alite C2). The values were calibrated (see
Table 3.3) so as to provide the best fit with the measured heat evolution profile during the
dissolution and induction periods. As can be seen, values lower or higher than the calibrated
values show deviation from the measured profile.

Figure 3.8(d): Evolution of CaTotal concentration as the value of βmax deviates from the
calibrated value of 10-22.0.

Page Number: 69
Figure 3.9: Simulation evolution of log (IAPalite/Ksp-alite) as a function of time. The black
dotted line represents the critical undersaturation value (log (βmax). The specific surface area
and w/c ratio are indicated at the top of the figure. Once Portlandite precipitates at the
point indicated by the circle symbol, log (IAPalite/Ksp-alite) decreases again (shown
schematically by the arrow) to the high dissolution regime until portlandite saturation
equilibrium is reached. The value of log (IAP alite/Ksp-alite) at the point of portlandite solubility
equilibrium (indicated by the grey dotted line) was calculated as -22.17.

Page Number: 70
Figure 3.10: Dissolution profiles versus undersaturation for alite (as obtained from
simulations for w/c = 0.5 and aSSA = 4265 cm2/g). The round symbols represent measured
values from [25]. The point of portlandite precipitation exists at the right extrema of the
plot, which also marks the end of the simulated SCD regime. As portlandite precipitates,
ions are consumed and the dissolution kinetics transition back to the undersaturation regime.
The decrease is shown schematically by the arrow and the point of portlandite solubility
equilibrium (calculated as log (IAPalite/Ksp-alite) equal to -22.17) is indicated by the circle.

Page Number: 71
3.8 Evolution of calcium and silicate concentrations and C-S-H during SCD
regime
The evolution of the solutions during the SCD regime was examined. Brown [76] provides the
most detailed experimental solution data at a w/c ratio close to those used practically in
pastes (w/c=0.7). Figure (3.11) shows the simulated evolution of total Si, total Ca and Ca2+
concentration with time compared to experimental data reported by Brown[76].

(a) (b)
Figure 3.11: Evolution of concentration (a) Total Calcium and Ca 2+ (b) Total Si . In the
model it is assumed that all of Si concentration exists as H 2SiO42-. Once portlandite
precipitates, CaTotal and Ca2+ decrease and SiTotal increases until portlandite solubility
equilibrium is reached (shown by arrows). The levels of Ca 2+, CaTotal and SiTotal at portlandite
solubility equilibrium are indicated by dotted lines. The data points have been extracted from
[76].

Figures (3.12) shows, the evolution of the Ca/Si ratio of the C-S-H and the total amount of
C-S-H formed up to the end of the induction period. It should also be noted that the rate of
dissolution of alite is assumed to be unmitigated by the physical precipitation of C-S-H during
the early ages. As such, the change in the surface area of particles is also not taken into
account in the model during the SCD regime. This, however, has little or no impact on alite
dissolution because the amount of C-S-H formed until the end of induction period (shown in

Page Number: 72
Figure 3.12b) is very small. This amount of C-S-H is also not critical if spread thinly on the
surface of alite.

(a) (b)
Figure 3.12: Simulated time dependent evolution of (a) amount of C-S-H until the end of the
induction period (b) Ca/Si ratio of C-S-H

3.9 Nucleation and Densifying Growth (NDG) Regime

A modified form of boundary nucleation and densifying growth mechanism proposed by
Bishnoi and Scrivener [74] is used to capture the main heat evolution peak. This is shown
schematically in Figure (3.13a). Portlandite, outer C-S-H and inner C-S-H are the products
which continue to form as the alite dissolves. The average growth of the outer C-S-H phase
around the alite grains is represented as a product of low packing density and it forms as the
outer layer on alite grains. However, in order to speed up calculations, the equations were
modified so that the interaction on the scale of individual particles need not be calculated. In
order to do this, the concept of extended volume (V extended), used often in nucleation and
growth, will be used. Vextended represents the expected volume of a set of growing nuclei if the
impingements between these nuclei is neglected. Usually, the extended volume is converted to
the real volume using Equation (3.21a or 3.21b), where all volume units are shown in (μm-3):

V real=1−e
−V extended
(3.21a)

Page Number: 73
or,
Vs , t−V s , t=0
V real (t)= (3.21b)
1−Vs , t =0

where Vs , t is the total volume of all solids at time “t” respectively.

These equations hold for systems with multiple products and takes into account the volume
occupied by other products such as portlandite. For these systems, it can be assumed that the
probability, and therefore the fraction of the extended volume lost to overlaps, increases as
the real volume available for growth reduces, as shown in Equation (3.22).
dVreal dVextended (3.22)
= (1−Vreal )
dt dt

If a densifying growth of C-S-H is assumed, the mass of alite reacting in a period δt can be
written as:
ρ(t r+dt r)−ρ( tr )
−dmalite =
1 ρ(t)
k ρ0 [(dV extended, CSH (t+dt)−dV extended, CSH (t))(1−V real )+V real ,CSH⋅( ρ0 )
]
(3.23)
where k is the ratio of mass of alite reacting to the mass of C-S-H produced. The equation
now takes into account both the space occupied by other phases and the change of density of
C-S-H which is already present in the system.
G out t
Where: Extended Volume ( V extended ) =∫0 a BV⋅( 1−exp (−Af ) ) dy (3.24)

Where Af can be written as:

[ y2
2

Gout
2
r
2
t3r y 2 tr 2y 3
A f =π Idensity⋅Gpar ⋅( t − 2 )+Irate⋅Gpar ⋅( − 2 + 3 )
3 Gout 3Gout ] (3.25)

where, Gout (μm. hour-1) and Gpar are the outward and parallel growth rates of outer C-S-H, t r
(hour) is the age of the product, Irate (μm-2.hour-1) is the rate of nucleation per unit area of the
untransformed boundary and Idensity (μm-2) is the nucleation density of the product (number of
nuclei of product at the time = 0). Equation 3.25 can also be expressed using N rate and Ndensity
which are respectively the products of the nucleation rate and the nucleation density with the

Page Number: 74
square of the parallel growth rate (IrateGpar2 and IdensityGpar2). In this model, the parameters
Ndensity and Nrate are referred to as coefficients of nucleation density and nucleation rate
respectively. It should be stated here that the use of N density and Nrate in this study is simply to
convolute the Irate and Idensity parameters with Gpar to eliminate Gpar as the variable parameter.
The dependence of the density of C-S-H on time can be written as below:

(3.26)
(( )
−k den⋅t r
ρ( tr )=ρmax −( ρmax −ρmin )⋅exp
ρmax−ρmin )

where,
(tr = 0) & (y = 0) when (t < t0) (3.27)
and (tr = t – t0) when (t >=t0 ) (3.28)

where ρmax (gm. cm-3) is the final packing density of outer C-S-H, ρmin (gm. cm-3) is the initial
packing density of outer C-S-H, k den (gm. cm-3.hour-1) is the rate of densification of outer C-S-
H, aBV is the total surface area of alite powder per unit volume (μm -1) and t0 is the start time
parameter. In order to compute the values of aBV, the total specific surface area of the
particles aSSA (cm2/g), as measured from experimental techniques, were used as inputs and the
dimensions were transformed to (μm-1) using the density value of alite of 3.15 and the initial
w/c ratio.

While generating the heat evolution curves it was observed that a unique combination of the
fit parameters was not possible and similar curves could be obtained by modifying a
combination of the NDG parameters. The sensitivity of the parameters and their influence on
the kinetics are discussed in section 3.10.

The NDG kinetics presented above can be implemented using two different versions: local and
global. In the local version of the model all calculations are implemented on the scale of
individual particles whereas in the global version all calculations are implemented using the
overall surface area of the particles and overall volume of water. In this study, all simulations
were run using the global version on account of computation time, which is less than 10
minutes as opposed to the local version, which takes about 6 hours. The differences between
the local and global version are highlighted in Figure (3.13b)

Page Number: 75
Figure 3.13a: Diffuse growth of C-S-H as the microstructural volume as it densifies. Single
particle (top) and between particles (bottom)

Local version Global version (used in this study)

Figure 3.13b: Steps followed in the simulations for NDG kinetics using the local and the
global versions.

Page Number: 76
3.10 Sensitivity of the NDG parameters
In order to illustrate the sensitivity of the NDG parameters, simulations were implemented
with a base value of the parameters as shown in Table (3.4). The parameters were then varied
individually while preserving the value of the other parameters to show the influence of the
variable on the overall kinetics. The simulated plots with variations of the parameters G par,
kden, ρmin, t0, Gout and Irate are shown in Figure (3.14). As can be seen, each parameter affects
the heat evolution profile in a unique fashion.

Table 3.4: Base values and short description of the parameters used for the simulations shown
in Figure (3.13).
Parameters and Base Values
Nucleation density ( Idensity) = 8.0 (µm-2)
Outward Growth rate (Gout) = 0.215 ( µm. hour-1)
Parallel Growth rate (Gpar) = 0.05 (µm. hour-1)
Nucleation rate (Irate) = 2.04 (µm-2 . hour-1)
Initial C-S-H density ( ρmin ) = 0.148 (gm. cm-3)
Final C-S-H density ( ρmax ) = 2.10 (gm. cm-3)
Rate of densification ( kden ) = 0.0035 (gm. cm-3.
hour-1)
Base density of C-S-H (ρ0) = 2.10 (gm. cm-3)
Starting time (t0) = 0.0
Description
Number of nuclei of C-S-H per unit C3S
surface area at (time = t0)
Outward Growth rate of the nuclei of
the product
Growth rate of the nuclei along the
surface
Rate of nucleation of C-S-H per unit C3S
surface area
Initial packing density of outer C-S-H.
Final packing density of outer C-S-H.
Rate of densification of outer C-S-H
Density of Inner C-S-H
Time at which NDG kinetics begins

Page Number: 77
Figure 3.14: Sensitivity of the parameters ρmin, Gout, kden, Gpar, Irate and t0 in the NDG regime.
The variable parameter is indicated at the top of each figure.

3.11 Transition from SCD to NDG regimes

In the model, the SCD regime is terminated at the point of portlandie precipitation. The
kinetics beyond the end of the induction period are simulated using NDG kinetics. But this
leaves a gap between the end of the simulated induction period and the acceleration period
(shown in Figure 3.15). The sequence of steps that are followed to implement this transition
are described below and shown schematically in Figure (3.16).
1. Until the end of the induction period, SCD mechanism is used. Heat-flow rate during
this period is calculated using only SCD kinetics.
2. NDG kinetics are also implemented in parallel and independent of the SCD kinetics.
The parameters for the NDG kinetics also include the start time parameter (t 0) which

Page Number: 78
 is determined to best fit the measured calorimetry curve. The heat-flow rate obtained
from NDG kinetics is calculated and stored but not used to compute the overall
hydration rate until the SCD regime is terminated.
3. SCD regime is terminated when portlandite reaches a supersaturation value of 0.40.
As stated earlier, until this point the overall heat-flow rate is equal to the heat-flow
rate obtained from SCD kinetics.
4. Once portlandite precipitates, calcium and hydroxyl ions are simulated to decrease
from portlandite supersaturation value of 0.40 to portlandite solubility equilibrium
value of 0.00. This decrease in the concentrations is implemented in a single step (a
progressive decrease is not implemented). The concentrations of calcium and hydroxyl
ions at portlandite solubility equilibrium are calculated by simulating simultaneous
decrease of calcium and hydroxyl ions in a 1:2 ratio from their respective
supersaturated values (values at SIPortlandite = 0.40) until SIPortlandite decreases to 0.0.
5. Due to the decrease in concentrations, undersaturation of alite transitions back to the
high dissolution regime and the dissolution rate (which is calculated from SCD
kinetics) rises to high value (shown in Figure 3.15). The dissolution rate at this point
is computed in the model by calculating the undersaturation of alite with respect to
the pore-solution at portlandite solubility equilibrium (calcium, hydroxyl and silicate
ions at SIPortlandite = 0.0).
6. Since the NDG kinetics are implemented independently, the heat-flow rate calculated
from the NDG kinetics remains unaltered at the point of portlandite precipitation. As
can be seen in Figure (3.15), at this point the heat-flow rate from NDG kinetics is
lower that the heat-flow rate from SCD kinetics. NDG kinetics is allowed to continue
until the heat-flow rate equals the value of heat-flow rate at the end of the induction
period.
7. From the point of portlandite precipitation to the point at which the heat-flow rates
from NDG kinetics and SCD kinetics at the end of induction period equalize, the
overall heat-flow rate is not calculated. This can be seen as the gap in Figure (3.15).
8. Once (Heat-flow rate from NDG kinetics = Heat-flow rate from SCD kinetics at the
end of the simulated induction period), the overall heat-flow rate is calculated from
NDG kinetics.

Page Number: 79
Figure 3.15: Simulated evolution of heat flow using SCD, NDG and combinations of SCD
and NDG mechanisms. As can be seen, once portlandite precipitates, dissolution transitions
to the high-dissolution regime rendering NDG as the slower and rate determining process.
The switch from SCD to NDG mechanism is made when the heat flow rates from both
regimes equalize.

Page Number: 80
Figure 3.16: Flow-chart depicting the steps followed in the simulations to compute the
overall heat-flow using the combination of SCD and NDG kinetics.

This scheme of transition between the SCD and NDG regimes was chosen based on the idea
of the slower physical process being the rate determining process. If portlandite does not
precipitate, the undersaturation and, therefore, the dissolution rate would gradually decrease
to a near-zero value. Precipitation of portlandite causes the consumption of ions and places
the undersaturation and the dissolution rate in the “faster-dissolution regime” as can be seen
in Figures (3.10) and (3.15), while the rate of precipitation of hydrates (from NDG) remains
the slower process. Therefore, being the slower of the two simultaneous processes, the rate of
precipitation becomes the rate determining process. As a summary of these points, it could
also be said that beyond the end of the induction period, the systems comes to a state of
“dynamic equilibrium” where the rate of dissolution is regulated (supply) by the nucleation
and densifying growth of products (demand). The same theory is implemented in the
simulations and NDG kinetics, which is slower than the dissolution rate after the termination
of the induction period, is used to simulate the overall rate for stages III and onwards.

Page Number: 81
3.12 Determination of simulation parameters
Determination of simulations parameters for any system is conducted separately for the SCD
and the NDG regimes.

For the SCD regime, the calibrated dissolution parameters are used in addition to the input
parameters, which include: aSSA, initial pH, initial Ionic strength and initial calcium
concentration. The SCD parameters are then manually varied, if necessary, to obtain best fit
to the measured calorimetry profile. The SCD parameters which best match the length of the
induction period in the measured plot are set as the final values.

In the NDG regime, the number of the variable parameters is 6 and each of these parameters
affects the simulated heat-flow profile uniquely, as shown in Figure (3.14). For this regime,
the parameters for the reference system are initially set at the values used for a previous
system with similar aSSA value and the deviation is evaluated. Thereafter, the parameter or
parameters that best corresponds towards minimizing the deviation (as shown in Figure 3.14)
are perturbed until the difference between the measured and simulated plots are reduced to
10% for the majority of the hydration regime, as shown in Figure (3.17). The parameters are
then fine-tuned to obtain simulation profiles which remain within 5% difference bound with
respect to the measured profiles, as depicted in Figure (3.18).

Figure 3.17: Difference (in %) between the simulated and the measured calorimetry profiles
as a function of time for 3 NDG simulation variables indicated at the top. The grey region
represents the 10% difference bound.

Page Number: 82
Figure 3.18: Difference (in %) between the simulated and the measured calorimetry profiles
as a function of time as the NDG parameter G out varies. The grey region represents the 5%
difference bound.

3.13 Validation of the SCD mechanism

It has been stated earlier that in the simulation of our experimental data the dissolution
parameters remains the same for all conditions and the kinetics are determined only by the
build-up of ions in solution. In addition it was seen in section 3.8 that the data of Brown for
alite hydration could also be well captured by the same parameters. To explore this further,
the impact of the composition of the starting solution was studied by simulation and
compared to the data of Brown [108] for staring solution of saturated DI-water, calcium
hydroxide, sodium hydroxide and sodium chloride. For all systems, the input parameters
included : initial Ionic strength, initial pH, initial OH - concentration and initial calcium
concentration, listed in Table (3.5). The results of the simulations are shown in Figure (3.19).
As previously discussed in section 3.8, it should be noted that the Calcium concentrations
provided by Brown correspond to the total Calcium concentration NOT Ca 2+. In the
measured and simulated plots, the points of portlandite precipitation are marked by the
decrease in the CaTotal concentration and the circular symbols respectively.

Page Number: 83
Table 3.5: Input parameters for simulations shown in Figure (3.19) for systems from [39]. The
parameters represent the initial state of the system. The dissolution parameters used for these
simulations are shown in Table 3.3. The w/c for all systems is 2.0.
Solution pH Ionic Strength OH- concentration Ca concentration
(mol. L-1) (mol. L-1) (mol. L-1)
DI-water 7.00 0.0000 0.0000 0.0000
NaCl solution 6.98 0.0408 0.0000 0.0000
Ca(OH)2 solution 12.45 0.06333 0.0422 0.0211
NaOH solution 12.68 0.0408 0.0408 0.000

Page Number: 84
Figure 3.19: Simulated and measured evolutions of CaTotal concentration and Ca2+ in alite
systems with different initial states of pore solution. Once portlandite precipitates (marked by
circular symbols), the Ca2+ concentration and the SIPortlandite values decrease (schematically
shown by arrows) until portlandite solubility equilibrium is reached (indicated by the dotted
lines) data from [73].

Page Number: 85
Chapter 4: Studying the Effects of Process Parameters
In this chapter, the effects of various experimental process parameters on the hydration of
alite are described. The following sections discuss the influence of particle surface area, mixing
conditions, thermal treatment and additions of alkali hydroxides, alkali sulfates and gypsum
on the hydration of alite. Using the model and simulation techniques described in Chapter 2,
heat evolution curves of different alite-based systems are simulated to explore the way in
which the hydration mechanisms are affected by different process parameters.

4.1 Effect of particle surface area

In this section, the effect of particle surface area on the hydration of alite is discussed. Particle
size distribution (PSD) plays an important role in determining the rate of reaction
[109,110,111,112]. Finer particles with higher specific surface area react rapidly compared to
cement with coarser particles having low specific surface area.

4.1.1 Materials and methods

In order to capture the effect of particles surface area, a group of simulations [113] was run for
the different PSDs of alite investigated by Costoya [109]. This consists of 6 particle size
distributions from the same synthesis of alite described in Table 4.1 and Figure (4.1). The
PSD of the original alite powder, Alite-C2, is close to that of cement. The other narrow PSDs
were chosen to cover the range of particle sizes present in cement. The water to cement ratio
was fixed at 0.40.

The measured heat evolution profiles for all systems as obtained from isothermal calorimetry
are shown in Figure (4.2). Figure (4.3) shows the early age heat flow for the same systems. As
can be seen, in systems with lower specific surface area of particles (a) the heat evolved
during the induction period is lower (b) the induction period is longer causing a delay in the
start of the acceleration period (c) the upslope of the acceleration regime is lower (d) peak at
the maximum heat flow is lower.

Page Number: 86
Table 4.1: Median size, Specific surface area calculated from PSD measurements obtained
from Laser diffraction method, and density values of all alite gradations used in this section.
From [109].

PSD Dmodal (µm) Specific Surface Density (g/cm3)

Area (cm2/g)
Alite C2 33 2000 3.15
Alite-Batch B 32 2100 3.14
PSD 2 38 1212 3.15
PSD 3 18 2393 3.17
PSD 5 13 3059 3.15
C3S 6 3952 3.15

Figure 4.1: Measured particle size distributions of all alite systems presented in this section.
Other details about the PSDs can be found in Table 4.1. From [109].

Page Number: 87
Figure 4.2: Measured heat evolution profiles of all alite systems presented in this section.
Other details about the PSDs can be found in Table (4.1).

Figure 4.3: Measured early age heat evolution profiles of all alite systems listed in Table
(4.1). From [109].

Page Number: 88
4.1.2 Simulating the effect of surface area
Simulations for these systems were implemented using the model as described in Chapter 3.
The dissolution parameters for all simulations were set at their calibrated values without any
alteration (listed in Chapter 3, Table 3.3). The NDG parameters were varied to obtain the
best fits and are listed in Table 4.2.

Table 4.2: Values of the NDG parameters used in the simulations for systems listed in Table
4.1. The units of each parameter are the same as described in section 3.9 (Chapter 3).
PSDs aSSA ρmin Irate Idensity Gpar Gout kden t0
Alite C2 2000 0.105 0.800 8.00 0.0500 0.260 0.0015 0.0
PSD 3 2393 0.097 1.680 8.00 0.0500 0.238 0.0020 0.0
PSD 5 3059 0.148 2.040 8.00 0.0500 0.215 0.0035 0.0
Alite-Batch B 2100 0.105 1.200 8.00 0.0500 0.250 0.0020 0.0
PSD 2 1212 0.073 0.615 8.00 0.0403 0.335 0.0010 0.0

C3S 3952 0.290 0.850 8.00 0.0968 0.210 0.0050 2.0

The main parameter from the different PSDs which varies in the simulations is a SSA, which
has a direct impact on the rate of dissolution as described by Equation 3.8 (Chapter 3). The
measured and simulated heat evolution profiles at early ages are shown in Figure (4.4). Figure
(4.5) shows the calorimetry curves and the simulations up to 30 hours of hydration. It can be
seen that across the range of PSDs the simulation is able to well capture the slowdown in
reaction rate and the induction period with the same parameters for the dissolution kinetics.

Figure (4.6) shows the simulated evolution of Ca 2+ concentration and SIPortlandite in the pore
solution. As can be seen, the times at which SIPortlandite exceeds 0.40, the value set for critical
supersaturation, for the first time changes as PSD changes and these times are approximately
the same as the times at which the rate of reaction begins to accelerate in the measured
calorimetry plots.

Page Number: 89
Figure 4.4: Simulated and Measured heat flow rates for alite systems with different PSDs for
the first 5 hours. w/c ratio for these systems is 0.40 and the PSDs are indicated at the top of
each figure. In each plot, the solid black and dotted lines represent heat flow rates obtained
using the SCD and NDG mechanisms respectively. The termination points of the solid black
lines represents the end of the simulated induction period and precipitation of portlandite.
The full plots shown in Figure (4.5) are obtained by combining the heat flow rates obtained
from the two mechanisms.

Page Number: 90
Figure 4.5: Measured and simulated rate of heat released for alite systems up to 50 hours of
hydration. w/c ratio of all systems is 0.40. For the simulations, the calibrated values of SCD
parameters were used (shown in Table 3.3, Chapter 3) and the G out parameter of the NDG
kinetics was varied in addition to minor variations in the other parameters (shown in Table
4.2).

Page Number: 91
 (a) (b)
Figure 4.6: Evolution of (a) Ca 2+ concentration (b) SIPortlandite. Once portlandite precipitates,
the Ca2+ concentration SIPortlandite values decrease (schematically shown by arrows) until
portlandite solubility equilibrium is reached (indicated by the dotted lines). In all systems,
w/c = 0.40.

The NDG parameters with the largest variations in the simulations are G out, which represents
the outward growth rate of C-S-H, and I rate, which is the rate of nucleation of C-S-H per unit
area. The variation in the rest of the parameters: ρ min and kden are 3.97 and 3.5 times
respectively in comparison to their lowest values. These parameters (ρmin and kden) increase
proportionally with respect to the specific surface area (Figure 4.7a and 4.7b) but the exact
nature of this relationship is not understood. It must also be stated here that the need to vary
four of the NDG parameters is unique to systems in which the surface area of alite particle
changes drastically. The number of variable parameters is reduced to 2 when simulations are
implemented for the same PSD while studying the variation of other process parameters such
as: effect of mixing conditions, effect of alkali salts, effect of heat treatment on alite etc. as
will be demonstrated in other sections.

The values of Irate were compared to the specific surface area of the particles as shown in
Figure (4.7c). It was observed that Irate increases with increasing specific surface area of the
particles. It appears that smaller particles have higher nucleation rates, even when normalized

Page Number: 92
with respect to the surface area. The exact reasons for this are unclear, although it may be
related to the fact that smaller particles have undergone more “grinding action” and so may
have a more highly damaged surface and curvature. It is also seen that C 3S has a lower
nucleation rate, again for reasons unclear at this stage.

(a) (b) (c)

Figure 4.7: Variation of the (a) ρmin (b) kden (c) nucleation rate (Irate) in the model with respect
to the specific surface area of the PSD.

The values of Gout were also compared to the average distance between closest neighboring
particles and the specific surface area of the particles as shown in Figure (4.8). The
calculation of the distance between closest neighbors was done using As can be seen, G out
decreases as the specific surface area of the particles increases (Figure 4.8a). Interestingly this
corresponds to an increase in direct proportion to the distance between the particles (Figure
4.8b). The variation of Gout with respect to the initial inter-particle space suggests an
opportunistic growth of the hydrates according to the available space in the microstructure,
the description of the exact mechanism is not known. It may, however, be speculated that this
variation is due to slight concentration gradients in the immediate vicinity of particles.

Page Number: 93
 (a) (b)
Figure 4.8: Variation of the Outward growth rate (G out) in the model with respect to (a)
specific surface area of the PSD (b) Average distance between closest neighboring particles
(dclosest neighbours). The value of the dclosest neighbours was calculated at time = 0 by (1) Identifying the
closest neighbours for each particle through iterations over the entire system (2) Calculating
the distance between all closest neighbours in a set of 2 (3) Averaging the value of distance
over the number of calculations in step 2.

Page Number: 94
4.2 Effect of mixing conditions
Mixing is a crucial step for any concrete application. Poor mixing results in large
inhomogeneities, which reduces the workability, leads to a poor packing of cement particles,
increases overall porosity and reduces strength. High mixing rates, on the other hand, reduce
the length of the induction period and enhance the hydration process [114,115].

4.2.1 Materials and methods

To study the effects of mixing conditions, an experimental dataset generated by Juilland [115]
was used which provides the heat evolution profiles of alite hydrating in solutions prepared at
varying mixing speeds. As part of the variation in the mixing conditions, rates ranging from
Hand Mixing to 2000 rpm were used and the heat flow was monitored using Isothermal
Calorimetry. For convenience, in the subsequent parts, hand mixing will be referred to as a
mixing rate of 0 rpm.

To highlight the effects of mixing on early age hydration and the evolution of the surface
defects during the dissolution of alite, Juilland used two kinds of solution: DI-water and
saturated lime solution. This was primarily done to alter the initial state of the pore-solution
and so the dissolution of alite began at different levels of undersaturation; high in the case of
DI-water and low in the case of saturated lime solution. To keep the complexity of the
systems to a minimum, for all experimental measurements the same w/c ratio (0.40) was
used. The PSD of the alite used for hydration in DI-water and saturated lime solution have
specific surface area values of 1300 and 1900 cm2/g respectively.

Figure (4.9a) shows the heat evolution profiles of alite hydrating in DI-water prepared at 6
different mixing rates, ranging from hand mixing to 2000 rpm. As can be seen, the heat
evolution profiles are significantly altered as the mixing rate changes. The kinetic effects
observed in the calorimetry measurements that vary directly with the speed of mixing are (a)
slow down of the dissolution rate during stage 1 is slightly altered (b) heat flow during the
induction period is higher (c) length of the induction period is shorter (d) the upslope of the
acceleration period is higher (e) the duration of the main peak occurs is shorter.

Figure (4.9b) shows the heat evolution profiles of alite hydrating in saturated lime solution
prepared at 4 mixing rates ranging from hand mixing to 1600 rpm. As can be seen, the mixing
conditions do not significantly affect the heat flow during the first two stages. The changes in

Page Number: 95
the latter stages, however, are more pronounced and are summarized here: (a) upslope of the
acceleration regime is higher at higher mixing rates (b) intensity of the main heat-peak is
enhanced when the speed of mixing is high. These effects are similar to the ones observed for
alite hydration in DI-water.

(a)

Page Number: 96
 (b)
Figure 4.9: Measured heat evolution profiles of alite hydration in (a)DI-water (b) Saturated
lime solution (left)up to 30 hours (right) up to 5 hours. From[115].

4.2.2 Simulations: Effects of mixing conditions

Simulations were implemented to study the effects of mixing and highlight the underlying
mechanisms that drive the change in the hydration profiles of alite. As such, the methodology
and simulation techniques were kept the same as for the simulations described in section 4.1
and Chapter 3.

Simulations for alite hydration in DI-water under different mixing conditions are shown in
Figure 4.10 ((a) for early ages and (b) for the hydration up to 30 hours) with comparisons
against the measured rate of heat evolution.

To capture the effects of mixing it was necessary to vary the values of the SCD parameters
kdiss and cstep retreat. Amongst the NDG parameters, the nucleation density coefficient (N density)
was varied within the widest range with small variations in the t 0 and ρmin parameters. As
described in section 3.9 of Chapter 3, N density is equal to (Idensity. Gpar2), where Idensity and Gpar are
the nucleation density and parallel growth rate respectively. The SCD and NDG parameters
used are listed in Table 4.3.

Page Number: 97
Table 4.3: SCD and NDG parameters used for the simulations of alite hydrating in DI-water
at different mixing conditions. The units of parameters are the same as described in section
3.9. The other input parameters used for these simulations were w/c ratio and a SSA equal to
0.40 and 1300 cm2/g respectively.
NDG parameters SCD parameters
Mixing ρmin NRate Ndensity Gout kden t0 kdiss cstep βmax
speed retreat

Hand 0.255 0.006 0.0110 0.215 0.0025 0.7 8.0 3.2 10-22
500 0.255 0.006 0.0115 0.215 0.0025 0.7 10.0 3.3 10-22
915 0.300 0.006 0.0170 0.215 0.0025 0.6 14.0 3.5 10-22
1350 0.310 0.006 0.0210 0.215 0.0025 0.40 20.0 3.8 10-22
1600 0.340 0.006 0.0240 0.215 0.0025 0.40 24.0 4.6 10-22
2000 0.360 0.006 0.0270 0.215 0.0025 0.40 26.0 4.8 10-22

Page Number: 98
Figure 4.10(a): Measured and simulated early age heat evolution profiles for alite systems in
DI-water at different mixing rates. The mixing rates are indicated on top of each plot.
Parameters shown in Table 4.3 were used in the simulation in addition to the input
parameters which included aSSA and w/c ratio equal to 1300 cm2/g and 0.40 respectively.

Page Number: 99
Figure 4.10(b): Measured and simulated rate of heat evolution for alite systems in DI-water at
different mixing rates. The mixing rates are indicated on top of each plot. Parameters shown
in Table 4.3 were used in the simulation in addition to the input parameters which included
aSSA and w/c ratio equal to 1300 cm2/g and 0.40 respectively.

Page Number: 100

Simulations for the alite hydration in saturated lime solution at different mixing conditions
were also implemented and the comparisons against experimental results are shown in Figure
4.11 ((a) for early ages and (b) for hydration up to 30 hours). To simulate the initial
conditions of saturated lime, the values of initial calcium concentration of 22 mmol. L -1 and
pH of 12.45 were used as inputs.

Similar to simulations described earlier, for the simulations of the heat evolution profiles of
alite hydrating in saturated lime solution, variations in SCD and NDG parameters were
needed (shown in Table 4.4).

Table 4.4: SCD and NDG parameters used for the simulations of alite hydrating in saturated
lime solution at different mixing conditions. The units of the parameters are the same as
described in section 3.9. The other input parameters used for these simulations were w/c
ratio, aSSA, initial calcium concentration and initial pH equal to 0.40, 1900 cm 2/g, 22 mmol. L-1
and 12.45 respectively.
NDG parameters SCD parameters
Mixing ρmin NRate Ndensity Gout kden t0 kdiss cstep βmax
speed retreat

Hand 0.167 0.006 0.014 0.237 0.0025 -1.0 24.0 3.80 10-22
500 0.170 0.006 0.015 0.237 0.0025 -1.3 24.0 4.25 10-22
800 0.175 0.006 0.017 0.237 0.0025 -1.3 24.0 4.45 10-22
1600 0.208 0.006 0.022 0.237 0.0025 -0.7 24.0 4.60 10-22

Page Number: 101

Figure 4.11(a): Measured and simulated rate of heat evolution for alite systems in saturated
lime solution at different mixing rates. The mixing rates are indicated on top of each plot.
The w/c ratio for all systems is 0.40.Parameters shown in Table 4.4 were used in the
simulation in addition to the input parameters which included a SSA, w/c ratio, initial pH and
initial calcium concentration equal to 1300 cm2/g, 0.40, 12.45 and 22 mmol/L respectively.

Page Number: 102

Figure 4.11(b): Measured and simulated rate of heat evolution for alite systems in saturated
lime solution at different mixing rates. The mixing rates are indicated on top of each plot.
The w/c ratio for all systems is 0.40. Parameters shown in Table 4.4 were used in the
simulation in addition to the input parameters which included a SSA, w/c ratio, initial pH and
initial calcium concentration equal to 1300 cm2/g, 0.40, 12.45 and 22 mmol/L respectively.

Page Number: 103

Figure (4.12) shows the variation of the SCD parameters, k diss and cstep retreat used in the
simulations against the mixing rates. For the simulations in DI-water it was observed that the
value of both these parameters increase with mixing speed.

(a) (b)
Figure 4.12: Variation in the SCD parameters (a) k diss and (b) cstep retreat. The round and
rectangular symbols represent the values used for alite systems in DI-water and saturated lime
solution respectively. For saturated lime solution, kdiss was not varied but the points are shown
in the plot to show the relative difference with respect to values for DI-water at equivalent
mixing rates.

The increase in the kdiss parameter captures the increase in dissolution rate. This increase
would correspnd to the effect the removal of the ions from the surface by the mixing process.
With a higher value of k diss, faster dissolution kinetics are observed. It should be noted that
the dissolution equation as implemented in the model does not explicitly represent the process
of removal of ions but provides a simple mathematical description of the dissolution kinetics.

The increase in the cstep retreat with the mixing rate implies a higher density of steps at higher
mixing rates. It could be hypothesized that at higher mixing rates, more etch pits open up
during the undersaturation regime. These pits later evolve to steps.

Page Number: 104

Figure 4.13: Simulated alite dissolution rate against alite undersaturation for the DI-water
and saturated lime solution systems. In this figure, the dotted line marks the split between
the two dissolution regimes: Undersaturation and Step-retreat regime. For the alite systems
with saturated lime solution, the dissolution regime begins very close to the step-retreat
regime. Once portlandite precipitates, the concentrations decrease causing an increase in alite
undersaturation and alite dissolution rate (shown schematically by dotted arrow) until
portlandite solubility equilibrium is reached (shown by circle symbol).

In the simulations with saturated lime solution, the SCD parameter k diss was kept constant
(Figure 4.12 and Table 4.4). As can be seen in Figure (4.13), in the presence of additional
CaTotal ions and a higher initial pH value, the initial undersaturation of alite is much lower as
compared to DI-water. As a result of this, dissolution begins from very close to the step
retreat regime. This is reflected in these simulations, which are relatively insensitive to the
value of kdiss chosen, as the system spends so little time in this regime. However, as the system
moves to the step retreat regime, the mixing rate imparts a small influence on the dissolution
rate. This is reflected in the c step retreat parameter, which was observed to rise slightly with the
mixing rate. This variation in cstep retreat was also found for hydration of alite in DI-water.

The evolution of calcium ions in solution in the two cases is shown in Figure (4.14). For the
DI-water systems it is clear that the first part of the curve is steep, corresponding to the fast
dissolution regime with the formation of etch pits. This part of the curve is clearly very

Page Number: 105

sensitive to the speed of mixing; higher slope implying faster dissolution at higher mixing
rates. Then there is a transition to a much less steep slope, corresponding to the step retreat
regime, which is nevertheless faster. For the systems mixed with saturated calcium hydroxide
solution, the first steep part is short and the systems enter very quickly into the slower step
retreat regime. Beyond this transition point, the small variation in the step retreat parameter
results only in a minor variation in the dissolution rate and, consequently, a small variation in
the evolution of the calcium concentration.

(a) (b)
2+
Figure 4.14: Simulated evolution of Ca concentration in (a) DI-water (b) Saturated lime
solution. Once portlandite precipitates, the concentrations decrease (indicated schematically
by arrows) from portlandite supersaturation value (shown by the thin grey line) towards
portlandite solubility equilibrium (shown as the thick dotted black line).

From these experiments and simulations, it is clear that the overall rate of dissolution up to
the end of the induction period is limited by two different processes. At the start, when mixed
with pure water the anhydrous phase quickly dissolves with the formation of etch pits at high
undersaturation, the concentration at the interface quickly approaches the critical
concentration at which the surface reactions will slow down. However, when the initial
solution is already saturated with lime, the systems enter quickly into the of step retreat
regime and is not significantly affected by mixing (see Figures 4.9b and 4.13).

Page Number: 106

From the end of the induction period the hydration kinetics are controlled by nucleation and
densifying growth of C-S-H. To capture the effect of mixing speed in the simulations, it was
necessary to change the parameter pertaining to the number of nuclei of C-S-H per unit
surface area, Ndensity, along with minor variations in the ρ min and t0 parameters. The variation in
the ρmin parameter and its correlation to a physical process is not well understood at this stage.
The t0 parameter for the DI-water systems increased with decreasing mixing rates suggesting
a delay in the start of the nucleation and growth of C-S-H in mixes prepared at low mixing
rates. Longer induction periods and delayed starts of NDG regimes in systems prepared at low
mixing rates can also be observed in the measured and simulated heat-evolution profiles
(Figures 4.9 and 4.10) and simulated calcium concentration profiles (Figure 4.14). For the
saturated lime solutions, negative values of t0 were used, indicating a very early precipitation
of products. It is likely that due to the presence of calcium ions and high pH initially, the
precipitation of products occurs during the mixing process.

Figure (4.15) shows the variation of the N density parameter for the systems mixed with DI-
water and those mixed with saturated lime solutions. In contrast to the regime prior to the
induction period, the effects of mixing speed are the same in the two systems and it is clear
that mixing speed significantly increases the number of nuclei of hydrates in the system. The
increase in nucleation density for both DI-water and saturated lime solutions is proportional
and almost linear with respect to the mixing rate.

The increase in nucleation density of hydrates is likely to be a purely mechanical effect. C-S-H
precipitates (mainly on the surface of the alite grains), very quickly after the addition of
water. During mixing these “nuclei” are dislodged from the surface forming extra nucleation
sites away from the surfaces of the alite particles. The microstructures of the different systems
(Figures 4.16) support this hypothesis as it is clear that there is much more growth of C-S-H
between the cement grains when the mixing speed is increased.

It was shown that the model can reproduce the influence of the mixing conditions on
hydration kinetics by varying the simulation parameters (a) k diss and cstep retreat, which are
related to the ease at which species detach from the dissolving surface of alite, and (d) N density,
which is related to the number of C-S-H nuclei per unit surface area of the particles.
Experimental results confirm the increase in the nucleation density of hydrates as a higher
packing density of C-S-H in the microstructural matrix is observed for samples prepared at
higher mixing rates.

Page Number: 107

Figure 4.15: Variation of the Ndensity parameter with respect to the mixing rate for systems
presented in this section.

Page Number: 108

Figure 4.16: Micrographs of polished section of alite samples at (left) maximum heat-flow
(right) 7 days of hydration (a) hand mixing (16 hours of hydration), (b) at 915 rpm (13.5
hours of hydration) and (c) at 1600 rpm (12.5 hours of hydration) From[115].

Page Number: 109

4.3 Effect of thermal treatment of alite
In order to the study the role of defects, the effect of heat treatment on alite hydration was
investigated. Numerical simulations were performed to describe the underlying mechanisms
that drive the hydration up to the first 50 hours and elicit the impact of heat treatment. The
geochemical theories of dissolution used in this study emphasize the importance of
crystallographic defects.

4.3.1 Materials and methods

The synthesis and heat treatment of C 3S and alite were carried out by A. Bazzoni [116]. Pure
tricalcium silicate (Ca3SiO5, C3S) was synthesized in the laboratory from a 3:1 stoichiometric
mixture of calcium carbonate (CaCO 3) and high purity quartz powder (SiO 2). Powders were
homogenized for 24 hours in water and the blend was dried at 100° for 24 hours. The dried
mass was then pressed into pellets and burned at 1600° during 10 hours. The pellets were
quenched in the air after heating and finally grounded into powder. The specific surface area
as calculated from laser diffraction was 4000 cm2/g.

Three different batches of alite were synthesized with the same composition but different
specific surface areas: 2000 cm2/g, 2100 cm2/g and 3100 cm2/g. These are referred to as
batches A, B and C respectively in the order of increasing specific surface area values. The
PSDs as measured using Laser diffraction are shown in Figure (4.17).

Heat treatment of alite and C3S were carried out at two different temperatures: 6500C and
8500C and for 3 different durations: 10 minutes, 1 hour and 6 hours. Quenching in air was
used as the cooling method post heat-treatment.

Page Number: 110

Figure 4.17: Particle size distributions of C3S and alite described in section 4.4.

4.3.2 Effect of thermal treatment: Measured heat evolution profiles

Heat evolution profiles were experimentally measured for C3S (Figure 4.18) and alite
hydrating in DI-water (Figure 4.19)[116].

As can be seen, heat treatment of alite and C 3S influences the hydration kinetics, particularly
in stages I and II. The effects as a function of the duration of treatment and temperature of
treatment are identical, as summarized below:
1. Heat flow during stage 1 remains unaltered
2. Heat flow rate during the induction period is lowered
3. Length of induction period is lengthened
4. Acceleration and deceleration profiles are altered. In some cases, a higher heat-peak is
observed and in some cases the heat-peak intensity is lowered.

Page Number: 111

 (a) (b)
Figure 4.18: Measured heat evolution profiles of C3S hydrating in DI-water (a) up to 40 hours
(b) up to 6 hours. From [116]

Page Number: 112

Figure 4.19: Measured heat evolution profiles of alite hydrating in DI-water (left) up to 40
hours (right) early ages. The batches are indicated at the top of each figure.

Page Number: 113

4.3.3 Simulating the effect of thermal treatment of alite
Numerical simulations were implemented to describe the hydration of heat-treated alite. The
methodology used was the same as described in the previous sections and in Chapter 3. In
order to capture the hydration profiles numerically, firstly the best NDG parameters were
determined for the reference systems while fixed SCD parameters were used. For the heat-
treated samples, the step-retreat parameter (cstep retreat) was varied to obtain best fits to the
heat evolution profiles until the end of the induction period. To reproduce the nucleation and
growth regime, the nucleation density parameter (N density) was varied in the widest range in
addition to variations in the ρ min and t0 parameters. These variations in the NDG parameters
were made with respect to the reference sample from the same type of system or batch. Table
4.5 lists the SCD and NDG parameters used for the simulations, which are compared against
the experimental results (Figures 4.20 for early ages and 4.21 for hydration up to 50 hours).
The variation of the parameters is discussed later.

As shown in Table (4.5), in order to fit the early age hydration profiles of heat treated alite
and C3S, the cstep retreat parameter was varied from the calibrated value (shown in Table 3.3,
Chapter 3) while preserving the values of the other two SCD parameters. When plotted
against the duration of heat treatment, a progressive decrease in c step retreat was observed (shown
in Figure 4.22a). For the same duration of treatment, a higher decrease in c step retreat parameter
with respect to the temperature of treatment was observed. Also, for the same durations and
temperatures of treatment, the values of c step retreat for C3S and alite were found to be similar.
The trend suggests a decrease in the density of step-retreat defects on the surface of alite and
C3S as the duration and time of heat treatment increases. It should be noted here that a
decrease in the density of step-retreat defects would imply a decrease in the density of etch-
pits during stage I for hydration. However, in the simulations the k diss parameter was not
varied because the simulation results were found to best fit the measured curves with the
calibrated value. This is expected because the fist stage of hydration lasts only for a fraction
of an hour and the kinetics during this regime are controlled by the solution composition
(undersaturation), which is the same for all systems presented in this section.

Page Number: 114

Table 4.5: SCD and NDG parameters used to simulate hydration of C 3S and alite described in
section 4.3. The units of each parameter are the same as described in section 3.9 (Chapter 3).
w/c ratio for all systems are 0.40. For the reference sample of each batch, calibrated
dissolution parameters (Table 3.3) were used and the best NDG parameters were determined.
The other samples from the same batch were then simulated by varying c step retreat and Ndensity
parameters, in addition to variations in the ρ min parameter, all with respect to the reference
sample.

SCD
NDG parameters
parameters
Batch Temp. Time ρmin NRate Ndensity Gout kden t0 kdiss cstep βmax
retreat

C3S - - 0.690 0.020 0.100 0.175 0.005 2.3 24.0 4.60 10-22
C3S 650 10 min 0.685 0.020 0.110 0.175 0.005 0.8 24.0 2.50 10-22
C3S 650 1 hour 0.715 0.020 0.130 0.175 0.005 2.0 24.0 2.00 10-22
C3S 650 6 hours 0.715 0.020 0.160 0.175 0.005 4.5 24.0 1.20 10-22
C3S 850 1 hour 0.710 0.020 0.150 0.175 0.005 3.0 24.0 1.60 10-22
Alite-A - - 0.075 0.003 0.026 0.25 0.0015 0.3 24.0 4.60 10-22
Alite-A 850 1 hour 0.087 0.003 0.018 0.25 0.0015 9.5 24.0 0.80 10-22
Alite-A 850 6 hours 0.080 0.003 0.026 0.25 0.0015 10.2 24.0 0.78 10-22
Alite-B - - 0.100 0.003 0.020 0.25 0.0022 0.0 24.0 4.60 10-22
Alite-B 650 10 min 0.145 0.003 0.015 0.25 0.0022 1.5 24.0 2.50 10-22
Alite-B 650 1 hour 0.136 0.003 0.013 0.25 0.0022 1.7 24.0 2.0 10-22
Alite-B 650 6 hours 0.185 0.003 0.025 0.25 0.0022 3.5 24.0 1.50 10-22
Alite-C - - 0.250 0.003 0.020 0.22 0.0057 1.3 24.0 4.60 10-22
Alite-C 650 1 hour 0.380 0.003 0.009 0.22 0.0057 1.0 24.0 1.60 10-22

Page Number: 115

 C3S

Alite Batch A

Alite Batch B

Alite Batch C

Figure 4.20: Early age measured and simulated heat evolution profiles of alite and C 3S
hydrating in DI-water.

Page Number: 116

 C3S

Alite Batch A

Alite Batch B

Alite Batch C

Figure 4.21: Measured and simulated heat evolution profiles of alite and C 3S hydrating in DI-
water.

Page Number: 117

The variation in the step retreat parameter with respect to the treatment duration is in
accord with the analyses made by Juilland et al.[115,117]. In their study, the authors
proposed that the defects on the surface of alite decrease due to the heat treatment during the
high-temperature processes of grain recovery. During hydration, a lesser number of defects
initially results into lesser number of step-retreats during the induction period, hence lowers
the dissolution rate and the heat flow rate during the dormant period. Consequently, the
increase in the calcium and hydroxyl concentrations also occurs at a slower rate and the
induction period is lengthened. The increase in the length of the induction period can also be
observed in the simulated concentration profiles, as shown in Figure (4.22b). As can be seen,
due to lower values of parameter c step retreat, the increase in calcium concentration is at a slower
rate in the heat-treated samples which manifests as (a) a longer duration to reach portlandite
supersaturation (b) longer induction period

(a) (b)
Figure 4.22: (a) Variation in c step retreat as a function of duration of heat treatment (b) Calcium
concentration profiles of heat treated C 3S samples. As portlandite precipitates, the calcium
concentration drops (indicated by the arrow) from the supersaturated value of 0.40 (shown by
the dotted line) to portlandite solubility equilibrium (shown by the solid line).

Page Number: 118

As shown in Table 4.5, the NDG parameters were varied to fit the N&G regimes of the heat
evolution profiles. Minor variations were observed in ρmin parameter with respect to the
reference system of each batch. The variation in the t 0 parameter indicates an increase in the
time at which nucleation and growth of C-S-H begins with respect to the temperature and
duration of treatment. The values of t 0 interestingly showed a decrease proportional to the
value of the cstep retreat parameter used for the corresponding system in the SCD regime (Figure
4.23a). From this it could be interpreted that the nucleation and growth of C-S-H is delayed
or “slower” or a combination of both when the defect-density on the surface of alite is lower
and the induction period is longer.

The other variable NDG parameter was N density, which is related to the number of C-S-H
nuclei per unit surface area of the anhydrous alite particles. When plotted against the
duration of heat treatment, no distinct trend was found (shown in Figure 4.23). In the case of
C3S, Ndensity showed an increase whereas in the case of alite, the value of N density remained fairly
constant with respect to increasing durations of treatment. On the whole, this variation
suggests an alteration in the nucleation criterion and growth of C-S-H due to the heat-
treatment process but with no straight forward relationship with the duration or temperature
of treatment. It should also be noted that the values of N density are higher for C 3S as compared
to the values for alite. This difference can be attributed to the high reactivity of C 3S, as a
result of its high specific surface area.

It is shown that the impact of heat treatment on the hydration of C 3S and alite can be
explained by the model. Due to a lower value of step-retreat density on the surface of alite
particles, dissolution occurs slowly and the time to reach portlandite supersaturation
increases, thus lengthening the induction period. The conclusions from this study provide
insights into the role of surface defects on the dissolution of alite and corroborate the
hypothesis proposed by Juilland et al. [117].

Page Number: 119

 (a) (b)
Figure 4.23: Variation in NDG parameters (a) t0 as a function of cstep retreat parameter used for
the corresponding system in the simulations (b) Ndensity as a function of duration of heat
treatment. Values for all systems presented in section 4.3 are included.

Page Number: 120

4.4 Effect of alkali hydroxide addition
Addition of alkali hydroxides (NaOH, KOH) to the mixing water increases its pH. For alite
(impure tricalcium silicate; MIII-Ca3SiO5) hydration, this pH increase accelerates the rate of
alite hydration and reduces the duration of the induction and deceleration regimes. These
effects have not been quantitatively linked to parameters which control the progress of alite
hydration. This section looks at alite hydration in solutions of varying compositions and
alkalinities (0.1M, 0.2M and 0.5M NaOH and KOH) by measurements of their heat release
behavior and analysis of the solid/liquid phases. The numerical modeling platform, μic, is used
to study the impact of the global solution chemistry and reaction product formation
parameters on alite hydration[118].

4.4.1 Introduction
Past research indicates that alkalis enhance hydration of alite reactions (assessed using
isothermal calorimetry), leading to an increase in the rate and magnitude of heat release
during the acceleration regime (Stage III), a shorter time to the main (2 nd) heat peak and
rapid deceleration beyond the main heat peak [119,120,121,122]. It has been reported that
while alkali additions enhance the degree of reaction at short times scales (first few days), at
longer time scales (28 days) the degree of hydration is similar [118].

Acceleration by alkalis has been attributed to the rate of dissolution and precipitation of the
reactant/product phases in relation to the chemistry of the contacting solution (i.e., the Ca 2+,
Na+, K+, Si4+, OH- abundance) and the common-ion effect [123,124,125,126,127] or alterations
in the composition/morphology of the hydration products [128,129,130,131]. However most of
the studies conducted in the past have been qualitative in nature.

4.4.2 Materials production and compositional information

Two batches of alite (Batches A and B of monoclinic MIII-Ca 3SiO5 [132]) were synthesized as
described by Costoya [109]. The minerals were mixed in excess distilled water and the slurry
further agitated for 24 hours in a ball mill with Zirconium spheres. The mixture was then
oven dried at 100°C for 5 days to evaporate residual water. The dry mineral mixture was
ground and pelletized. The pellets were burnt in a muffle-furnace as per the burning
procedure described by Costoya [109]. The burnt pellets were ground in a ball-mill for 12
hours and sieved. The unburnt free-lime (CaO) content of both batches was around 0.8%
using quantitative X-ray diffraction (QXRD). The particle size distributions (PSD) are shown
in Figure (4.24). The specific surface area of Batch A and Batch B was determined to be 2760
cm2/g and 1923 cm2/g respectively; assuming spherical particle geometry. Batch A was used
for the calorimetry and thermogravimetric analyses (TGA) and batch B for the pore solution

Page Number: 121

extraction. Microscopy and Energy Dispersive Spectroscopy (EDS) analyses were done on
both batches. As shown below, the two batches hydrate identically when their slightly
different PSDs are taken into account. Alite pastes were prepared at a constant water-to-
solids ratio (w/s) of 0.35. The solids were combined with de-ionized water or the appropriate
salt (NaOH, KOH; AR-grade) solution and mixed using a high-shear immersion mixer for 2
minutes at a mixing speed of 500 rpm. Here, care was taken to minimize exposure of the
mixtures to the air to limit carbonation. The compositions of the mixing solutions used are
detailed in Table 4.6.

Figure 4.24: Particle size distribution of the two batches of alite as measured using Laser
diffraction. Data from C. Gianocca[118]

Table 4.6: The compositions of mixing water utilized in the evaluations and their calculated
pH.
Concentration pH of NaOH salt solution pH of KOH salt solution
0.1 M 13.10 13.08
0.2 M 13.34 13.33
0.5 M 13.71 13.68

Page Number: 122

4.4.3 Simulating the effect of particle size distribution on alite hydration
Simulations were implemented to capture the hydration kinetics of alite, with different
particle size distributions (PSDs), hydrating in DI-water. Comparisons against experimental
results are shown in Figures (4.25) (early age hydration) and (4.26) (up to 50 hours).
Calibrated dissolution parameters shown in Table 3.3 (Chapter 3) were used. The heat release
during the NDG regime is matched to the experimental response by varying the NDG
parameters, as shown in Table 4.7. The variation of the NDG parameters is similar to the
variation used to describe the effect of particle sizes on alite hydration in a different study [25]
and section 4.1. Upon matching the dissolution rate and NDG regime parameters, the
simulations results for Batches A and B clearly indicate that the difference in reaction kinetics
of the two batches is strictly a “specific surface area” effect. This is an important point as it
enables the simulations to discriminate aspects related the physical properties of the particles
from chemical compositional considerations of the mineral phases or the mixing water.

Table 4.7: NDG parameters used to simulate hydration of alite from batches A and B in DI-
water. The units of each parameter are the same as described in 3.9 (Chapter 3). In addition
to this, the input parameters were w/c ratio (0.35 for both batches) and a SSA (1923 and 2760
cm2/g for batches B and A respectively).

PSDs ρmin Nrate Ndensity Gout kden t0

Batch A 0.280 0.004 0.035 0.230 0.003 -0.3
Batch B 0.305 0.004 0.025 0.225 0.003 -0.3

Page Number: 123

Figure 4.25: Measured and simulated early age heat evolution profiles of alite from batches A
and B. Data from C. Gianocca[118]

(a) (b) (c)

Figure 4.26: Measured and simulated heat evolution profiles of (a) Batch A (b) Batch B
hydrating in DI-water. (c) Degree of hydration computed from the calorimetry curves. The
w/c and aSSA are indicated at the top of each figure. As can be seen, the kinetics of reaction
are different as the PSD of alite changes and they can be captured by using calibrated
dissolution parameters and slight variations in the NDG parameters.

Page Number: 124

4.4.4 Effect of alkali hydroxide additions: Experimental results

Figure (4.27) shows the measured heat evolution profiles of alite hydrating in mixing solutions
with different compositions. Figure (4.27b) shows the early age heat evolution profiles of alite
hydrating in varying compositions of mixing water. Here, it is clearly noted that the
dissolution (heat release) response of alite varies according to the composition of the mixing
water. This results in heat-evolution profiles (Figure 4.27a) to produce a combination of
effects including:
• Alterations in hydration kinetics during stages I and II
• An increase the rate of heat release during the acceleration regime
• Increase in the intensity of the 2nd (main) heat peak
• Rapid deceleration

As can be seen in Figure (4.27c), alkali additions enhance the degree of reaction at short times
scales (up to 15 hours or so) but at longer time scales (beyond 18 hours), the degree of
hydration is lower as compared to the reference. However, at high levels of alkali additions
(between 0.2M and 0.5M), the degree of hydration at later ages remains higher than the
reference. These differences in the degree of hydration between the systems were found to be
in agreement with the portlandite contents evaluated using Thermal Gravimetric Analyses
(TGA) performed at 30 hours (Figure 4.28).

Here it is critical to note that inspite of similar mixing water parameters (pH, ionic strength),
isomolar compositions of NaOH and KOH induce slight differences in the alite hydration (heat
release) response. A combination of these effects ensures that alite systems containing NaOH
or KOH at early ages hydrate faster, wherein the increase in rate is proportional to the initial
concentration of salt in the mixing solution.

Page Number: 125

 (a) (b) (c)
Figure 4.27: Measured heat evolution profiles (a) up to 50 hours (b) up to 3 hours (c) Degree
of hydration of alite hydrating in salt solutions of different concentration. Data from C.
Gianocca[118].

Figure 4.28: Portlandite content, as calculated from TGA measurements at 30 hours of

hydration, against degree of hydration as calculated from calorimetry curves, for the samples
presented in this section. The red bars represent a 5% error bound.

Page Number: 126

4.4.5 Simulating the influence of the initial solution pH on alite dissolution
The addition of alkali hydroxides increases the initial pH and ionic strength of the mixing
water. These effects alter the dissolution of alite as may be inferred using the implementation
of the SCD mechanism used to simulate the following:

• Dissolution-Undersaturation master-curves (shown in Figure 4.29) which at the left-

extrema tends towards DI-water and towards the right-extrema indicates portlandite
precipitation
• Concentration profiles of the major ionic species and pH (shown in Figure 4.30). The
supression of the Ca2+ concentration and elevation of pH can be seen. A sharp
elevation in the total Si concentration is also observed.

Here, it is noted, that an increase the mixing solution pH and initial ionic strength, results in
a rightward shift of the “initial undersaturation level” and a faster transition of dissolution
kinetics from the fast “undersaturation regime” to the slow “step-retreat regime”. In addition
to this, it can also be observed that there is a reduction in the level of Ca 2+ required to
invoke portlandite precipitation, which corresponds to the termination of the induction period.
As a combination of these effects, it can be said that:

• Due to low initial undersaturation, the system transitions from the fast to slow
dissolution regime rapidly. Consequently, the kinetics remain in the slow-dissolution
regime for longer duration rendering the release of Ca2+ concentration at a slower rate.
• Precipitation of portlandite is invoked at lower level of Ca 2+ concentration. This is
because at high pH values, the concentration and, hence, the activity of OH - ions is
higher as compared to the reference system. The activity of calcium ions to precipitate
SI

portlandite is given by {Ca }= 2+ √

10 ⋅K spPortlandite
Portlandite

. For fixed values of SIPortlandite

-
{OH }
(0.40 in the model) and Ksp, Portlandite (10-5.18), it can be seen that as the activity of OH -
increases, the activity of Ca2+ ions required to precipitate portlandite decreases.
• As a combination of these two effects, the length of the induction period is altered as
can be seen in the measured heat-evolution profiles Figure (4.28).

It should be noted that all the simulation profiles were obtained using calibrated SCD
parameters shown in Table 3.3 (Chapter 3).

Page Number: 127

Figure 4.29: Rate of dissolution versus undersaturation of alite for mixing solutions with
different initial pH values and alkali concentrations. The arrow indicates the initial
undersaturation. The w/c ratio and specific surface area of the particles are indicated at the
top of the figure.

(a) (b) (c)

2+
Figure 4.30: Simulated evolution of (a) Ca (b) pH (c) Total Si concentration for alite
hydrating in solutions with two different initial states of pore-solution. The circle-symbols
represent the points at which portlandite precipitates. The dotted lines represent the
portlandite solubility equilibrium.

Page Number: 128

4.4.6 Simulating the pre-acceleration period
Simulations were implemented to capture the evolution of concentrations in the systems
described above using the input parameters as shown in Table (4.8). Figure (4.31) shows the
evolution of Ca2+, pH and total Si as obtained from simulations for systems with NaOH salts
in the mixing solution. Measured values at different times are also shown in the plots. As can
be seen, portlandite precipitates when SI Portlandite reaches a value of 0.40 (indicated by circular
symbols ) at lower levels of Ca 2+ and higher levels of SiTotal when the initial pH of the mixing
solution is higher. This is expected as the elevated pH values and ion-activities cause a
reduction in the critical supersaturation-barrier for portlandite precipitation, as described in
the previous section.

Table 4.8: Input parameters for simulations of systems described in this study. The
parameters represent the initial state of the system. The dissolution parameters used for these
simulations are shown in Table 3.3. w/c ratio for all systems is 0.35. The values of a SSA used
for batches A and B are 2760 and 1923 cm2/g respectively.
Solution pH Ionic Strength Alkali concentration OH- concentration
(mol. L-1) (mol. L-1) (mol. L-1)
DI-water 7.00 0.0 0.0 0.0
0.1 M NaOH 13.10 0.1 0.1 0.1
0.1 M KOH 13.10 0.1 0.1 0.1
0.2 M NaOH 13.35 0.2 0.2 0.2
0.2 M KOH 13.35 0.2 0.2 0.2
0.5 M NaOH 13.69 0.5 0.5 0.5
0.5 M KOH 13.69 0.5 0.5 0.5

Page Number: 129

 (a) (b) (c)
Figure 4.31: Measured and Simulated evolutions of (a) Ca 2+ (b) pH and (c) Total Si. The type
and concentration of the salt used to prepare the salt solution are indicated as legends. The
circle-symbols indicate the point at which portlandite precipitates and the arrows show the
decrease in the concentrations from portlandite supersaturation (SI Portlandite = 0.40) to
portlandite solubility equilibrium (SIPortlandite = 0.0, indicated by the dotted lines). The w/c
ratio and specific surface area of the particles are shown at the top of the figures.

Using the calibrated dissolution parameters (shown in Table 3.3, Chapter 3) and input
parameters shown in Table (4.8), early age hydration profiles were simulated for alite (Batch
A) hydrating in mixing solutions with different compositions. Comparisons against
experimental results are shown in Figure (4.32). The time of portlandite precipitation
indicated in Figure (4.31) and the end of the induction period in Figure (4.32) are identical as
they are results from the same simulations. As can be seen, the model captured reasonably
well the onset of the induction period. However, the time at which portlandite precipitates is
not captured well. It should be noted here that for isomolar concentrations of NaOH and
KOH, as the input and dissolution parameters are the same, the simulated dissolution profiles
are also identical.

Page Number: 130

 (a) (b) (c)
Figure 4.32: Measured and Simulated heat evolution profiles for alite hydration in (a) 0.1 M
(b) 0.2 M (c) 0.5 M alkali hydroxide solution. The w/c ratio and a SSA are indicated at the top
of each figure.

4.4.7 Simulating the Acceleration and Deceleration periods

Simulations for the post-Induction regime for all systems were carried out using NDG kinetics,
as described in the previous sections and (section (3.8) Chapter 3). With the exception of two
parameters (Ndensity and ρmin), the simulations for all the systems were implemented using
identical values of NDG parameters (Table 4.9). The results obtained from the simulations
compared against the experimental measurements are shown in Figure (4.33).

Page Number: 131

Table 4.9: NDG parameters used to simulate hydration of alite from batch A in mixing
solutions with different compositions. The units of each parameter are the same as described
in section 3.9 (Chapter 3). w/c and aSSA for all systems are 0.35 and 2760 cm2/g respectively.

Mixing ρmin Nrate Ndensity Gout kden t0

Solution
DI-Water 0.280 0.004 0.035 0.23 0.003 -0.3
0.1 M NaOH 0.245 0.004 0.045 0.23 0.003 -0.3
0.1 M KOH 0.265 0.004 0.045 0.23 0.003 -0.3
0.2 M NaOH 0.250 0.004 0.075 0.23 0.003 -0.3
0.2 M KOH 0.265 0.004 0.075 0.23 0.003 -0.3
0.5 M NaOH 0.295 0.004 0.150 0.23 0.003 -0.3
0.5 M KOH 0.350 0.004 0.170 0.23 0.003 -0.3

As discussed earlier, simulations for all systems were implemented using calibrated dissolution
parameters. However, for the simulations of the NDG regime, variations in the Ndensity and ρmin
parameters were required. As can be seen in Table (4.9), the variation in the ρ min parameter is
small (from 0.245 to 0.350) and the exact correlation of this variable with respect to a
physical process or parameter is not known at this stage. A constant negative value of the t 0
parameter (-0.30) was used for all systems and this suggests that the precipitation of hydrates
occurs very early in these systems. It should be noted that the value of t 0 was set at -0.3 as
opposed to 0.0 to improve the fit.

Amongst the NDG parameters, the widest variation in found in the N density parameter. The
values of this parameter when plotted against the initial ionic strength of the mixing solution,
an increase in direct proportion (shown in Figure 4.34) was observed. This trend suggests an
increase in the number of C-S-H nuclei per unit surface area of anhydrous alite leading to a
denser matrix of hydrates in systems with higher initial pH.

Page Number: 132

Figure 4.33: Measured and Simulated profiles of alite hydrating in salt solutions with different
concentrations. The type and concentration of the salts are indicated at the top of each figure.
For all simulations w/c = 0.35 and aSSA = 2760 cm2/g.

Page Number: 133

Figure 4.34: Initial ionic strength of the alkali hydroxides plotted against values of nucleation
density coefficient (Ndensity) used in the simulations. As can be seen, for both NaOH and
KOH, the variations in Ndensity is proportional to the initial ionic strength.

In order to corroborate the increase in C-S-H nucleation density, SEM images obtained at the
age of 30 hours (Figure 4.35) were analyzed. With the limited resolution of this technique, the
differences in the packing of hydrates between the images are not distinguishable. However,
through close observation it can be seen that the density of the C-S-H matrix is higher for the
0.2 M NaOH mixing solution, indicating a higher number of C-S-H nuclei as compared to the
reference sample resulting in denser packing of the hydrates around the alite grains. It should
be noted that a higher nucleation density C-S-H leads to faster hydration of alite at early
ages, as shown in Figure (4.27c) and the denser packing of hydrates observed in Figure (4.35)
could simply be a result of enhanced hydration.

It has been hypothesized that the growth of C-S-H is kinetically limited by the very low
concentration of silicate in solution and as this concentration rises facilitating the onset of
rapid growth, as previously proposed in [133]. As can be seen in Figures (4.30c) and (4.31c),
the rise in Si concentration is more pronounced when the initial pH is higher. This could
result in a enhanced nucleation and growth of the C-S-H phase. However, this hypothesis is
fairly speculative at this stage.

Page Number: 134

 (a) (b)
Figure 4.35: SEM images of alite at the age of 30 hours for (a) hydrating in DI-water (b)
hydrating in 0.2 M NaOH solution. The magnification for the images at the top and bottom
are 1200x and 8000x respectively.

Page Number: 135

4.4.8 Studying the changes in structure and morphology of C-S-H
Additional experiments were done to examine the impact of alkali salts on the structure and
morphology of C-S-H. Figure (4.36) shows the results obtained from EDS analyses of alite
samples (both batches) hydrated for 30 hours in mixing solutions of different compositions. It
was observed that the scatter of the Ca/Si ratio for all samples remained within the range of
1.7, thus indicating no influence of the alkalis on the stoichiometry of C-S-H.

Figure 4.36: EDS analyses for alite samples hydrated in mixing solutions of different
compositions at the age of 30 hours. Data from C. Patapy[113].

For further verification, C/S and H/S ratios for C-S-H were calculated using data from
calorimetry and QXRD (for reacted SiO2) and Thermogravimetric analyses (for C-S-H bound
water content), as shown in Figure (4.37), for alite samples at the age of 30 hours. It was
observed that the differences in the C/S and H/S ratios were minor confirming no major
change in the C-S-H stoichiometry.

Page Number: 136

Figure 4.37: C/S (left axis) and H/S (right axis) ratio calculations for alite samples hydrated
in mixing solutions of different compositions at the age of 30 hours. XRD, TGA and
calorimetry analyses were used to compute the ratios. For clarity, the average values are
plotted as vertical bars with the minimum (blue line) and maximum(red line) values as error
bars.

To examine the morphological changes in C-S-H, fracture surface images were taken at 11 and
20 hours (shown in Figure 4.38). From the images it was concluded that the morphology of C-
S-H was not affected significantly by the addition of alkali salts in the mixing solution.

It has been shown that hydration profiles of alite in mixing solutions with different
compositions can be captured by the use of calibrated dissolution parameters and varying two
of the NDG parameters. The variation in the N density parameter in the simulations, the
experimental results from calorimetry and SEM, showed a progressive increase in the
nucleation density of C-S-H as a function of the initial ionic strength of the pore-solution. It
was also shown that the addition of alkalis in the mixing solution does not influence the
structure and morphology of the hydrates significantly. However, as can be seen in Figure
(4.33), the matches between the simulation and measured heat flow profiles during the
deceleration regimes are not good. This could be a result of the assumption that C-S-H only
grows as layers on alite particles, while in reality C-S-H could also nucleate in the pore-space
thus affecting the impingements and evolution of the exposed surface area.

Page Number: 137

 (a) (b)
Figure 4.38: Fracture surface images of C-S-H at the ages of 11 hours (top) and 24 hours
(bottom) for (a) DI-water (b) 0.5 M NaOH. Batch B alite samples were used for this analysis.
The images were taken at magnifications of 12000x and 30000x for the ages of 11 and 24
hours respectively. The scales are indicated at the bottom right of each image. As can be seen,
the morphology of C-S-H is not affected by the addition of alkalis in the mixing solution.

Page Number: 138

4.5 Effect of gypsum addition
In commercial cements, gypsum is typically added to control the hydration rate of the
aluminate phase and prevent flash setting. It has been shown in other studies that the
addition of gypsum also affects the hydration of alite [134] and influences the long term
strength [135]. In this section, the effect of gypsum addition on the hydration of alite and C 3S
are described. Numerical simulations are implemented to outline the mechanisms that drive
the changes induced due to the addition of gypsum.

4.5.1 Materials and Methods

To study the influence of gypsum addition on hydration profiles, the experimental data
generated by A. Quennoz [114] was used. In her study, both C 3S and alite were used as the
main reactant clinker phases with different levels of replacement by gypsum. The protocols
followed for the synthesis of C 3S and alite are the same as described in section (4.3) and
here[118,114]. As such, the PSD generated for C 3S and alite were measured to have specific
surface area values of 3800 cm2/g and 2500 cm2/g respectively.

In order to monitor the kinetics, heat evolution profiles were generated for C 3S and alite
hydrating in DI-water with three different substitution levels of gypsum: 2%, 5% and 10%.
For comparison, the profiles for reference systems were also obtained and are described in the
next section.

4.5.2 Influence of gypsum addition: Measured heat evolution profiles

Heat evolution profiles were experimentally measured for C3S and alite hydrating in DI-water
(Figure 4.39).
As can be seen, gypsum addition affects hydration kinetics at both early and later ages. The
points below summarize the changes caused due to the addition of gypsum as interpreted from
the heat flow profiles:
1. Initial slow down of the dissolution rate remains unaltered.
2. Heat flow during the induction period is low
3. Length of induction period increases
4. Hydration, during the acceleration and deceleration regimes, is enhanced to a higher
degree in alite (significant increase in the main heat-peak) and to a lower extent in the
case of C3S (minor increase in the main heat-peak).

Page Number: 139

Figure 4.39: Measured heat evolution profiles of C3S and alite hydrating in DI-water (left) up
to later ages (right) early ages.

Page Number: 140

4.5.3 Simulating the influence of gypsum addition on alite hydration
Numerical simulations were implemented to reproduce the heat evolution profiles of hydrating
alite and C3S with different levels of gypsum substitutions. The input parameters included
initial calcium and sulfate concentrations of 14.0 mmol. L -1, in addition to the aSSA and w/s
values. Initial concentrations of Ca 2+ and SO42- were determined as per the solubility of
gypsum which is reported as 2.40 grams. L-1[136]. It should be noted that this value
corresponds to 2.406 grams. L-1 of dissolved gypsum, a value very close to gypsum's solubility
limit of 2.40 grams. L-1. As this value of dissolved gypsum (14 mmol. L-1) amounts to less than
2% substitution of gypsum (calculated as 29.04 mmol. L -1 for w/s = 0.40 ), it can be stated
that for all substitution levels (2, 5 and 10%), solid (undissolved) gypsum still exists and the
solution is saturated with respect to gypsum.

For simulations of the early age hydration profiles, calibrated SCD parameters (Table 3.3)
were used for the reference samples. For the samples with gypsum substitution, the critical
undersaturation parameter (βmax) was varied with respect to the calibrated value. The reason
for choosing βmax as the variable parameter is described later. Three of the NDG parameters,
Ndensity, t0 and ρmin, were also varied with respect to the best-fit values determined for the
reference samples to fit the heat flow rates for the post-induction period regime. The values
ultimately determined to best fit the hydration profiles, shown in Figure 4.40 ((a) for early
ages and (b) for full hydration plots up to 30 hours), are enumerated in Table (4.10).

Page Number: 141

Table 4.10: SCD and NDG parameters used to simulate hydration of C3S and alite with
gypsum substitutions described in section 4.5. The units of each parameter are the same as
described in section 3.8 (Chapter 3). w/s ratio for all systems is 0.40. For the reference sample
of each batch, calibrated dissolution parameters (Table 3.3) were used and the best NDG
parameters were determined. The other samples from the same batch were then simulated by
varying βmax and Ndensity parameters, in addition to variations in the ρ min parameter, all with
respect to the reference sample.

NDG parameters SCD parameters

Batch ρmin NRate Ndensity Gout kden t0 kdiss cstep βmax
retreat

C3S 0.790 0.020 0.090 0.16 0.005 1.6 24.0 4.6 10-22.00
C3S +2%G 0.830 0.020 0.160 0.16 0.005 3.6 24.0 4.6 10-22.43
C3S +5%G 0.690 0.020 0.240 0.16 0.005 4.0 24.0 4.6 10-22.43
C3S +10%G 0.690 0.020 0.240 0.16 0.005 4.0 24.0 4.6 10-22.43
Alite 0.130 0.003 0.031 0.235 0.002 0.0 24.0 4.6 10-22.00
Alite + 2%G 0.255 0.003 0.071 0.235 0.002 1.6 24.0 4.6 10-22.35
Alite + 5%G 0.275 0.003 0.120 0.235 0.002 2.0 24.0 4.6 10-22.35
Alite + 10%G 0.263 0.003 0.095 0.235 0.002 2.0 24.0 4.6 10-22.35

Page Number: 142

Figure 4.40a: Measured and simulated early age heat evolution profiles of C 3S and alite with
gypsum substitutions. The batches and levels of gypsum substitution are indicated at the top
of each figure. Simulations were implemented using parameters as listed in Table (4.10). The
w/s ratio for all systems is 0.40.

Page Number: 143

Figure 4.40b: Measured and simulated heat evolution profiles of C 3S and alite with gypsum
substitutions. The batches and levels of gypsum substitution are indicated at the top of each
figure. Simulations were implemented using parameters as listed in Table (4.10). The w/s
ratio for all systems is 0.40.

As shown in Table (4.10), for the simulations, two of the SCD parameters (k diss and cstep retreat)
were kept at their calibrated values and the parameter βmax was varied by the same magnitude
(see Table 4.10) irrespective of gypsum substitution so as to best fit the heat flow rate until
the end of the induction period. Since the amount of dissolved gypsum is the same in all
substituted systems, as stated earlier, the initial state of pore-solution is also the same and,
hence, the variation needed in βmax were also identical. Consequently, the simulated SCD
regimes for all systems also are identical (see Figure 4.40a).

Due to this variation in βmax, in addition to the change in the initial concentrations of calcium
and sulfate, the evolution of concentration of the ionic species differ from the reference. Figure
(4.41a) shows the evolution of the calcium, silicate and sulfate concentrations for the reference
and gypsum substituted samples. As can be seen in the gypsum substituted samples, calcium

Page Number: 144

rises and sulfate falls from the initial values of 14 mmol. L -1. The fall in sulfate concentration
was simulated to maintain gypsum solubility equilibrium. The concentration profiles (Figure
4.41a) show that the point at which portlandite precipitates (when SI Portlandite reaches a value
of 0.40) and the simulated induction period ends. As can be seen, portlandite precipitation is
delayed in systems with gypsum additions. Once portlandite precipitates (marked by the
circular symbols), calcium concentration drops sharply (indicated by arrows) to the
portlandite solubility equilibrium value (shown by dotted lines). The lengthening of the
induction period can also be observed in the measured (Figure 4.39), and simulated heat
evolution profiles (Figure 4.41b). It should be noted that by decreasing the step-retreat
parameter (cstep retreat), the heat-flow during the induction period can be reproduced but the
time of portlandite precipitation cannot be captured (shown in Figure 4.41b). Due to this
reason, the parameter βmax was varied instead of cstep retreat.

The simulated pore-solution concentrations presented in this section are in good agreement
with experimental data reported by Brown et al.[113]. In their study, the authors presented
pore-solution concentrations of total calcium and sulfate ions in solution when C 3S is hydrated
in a 0.0158 M CaSO4 solution, which showed (a) increase in the level of calcium concentration
at the point of portlandite precipitation (b) lengthening of the induction period as compared
to C3S hydrating in DI-water and saturated lime solution (shown in Figures 4.41c).

It can be hypothesized that in the presence of sulfate ions in the pore-solution, the evolution
of defects on the surface is altered; either due to sorption of sulfate ions on alite surface or
inhibition of defect sites on alite surface. As a result of this, the critical undersaturation
(given by βmax ), below which it is not possible to create new etch pits, is altered. However,
this hypothesis is fairly speculative at this stage and requires further investigation.

Page Number: 145

 (a)
Figure 4.41(a): Simulated concentration profiles of reference and 2% gypsum substituted C 3S
hydrating in DI-water (w/s = 0.40). The changes in concentrations from supersaturated
value(SIPortlandite = 0.40) of portlandite to solubility equilibrium (SI Portlandite = 0.00) is shown by
the arrow. Due to the presence of calcium initially and change in βmax, the induction period is
longer in the gypsum substituted samples.

Page Number: 146

 (b) (c)
Figure 4.41(b): Simulated heat evolution profiles of C 3S + 2% gypsum system using calibrated
and fitted values of βmax. As can be seen, due to the decrease in βmax from the calibrated value
(10-22.00), the induction period is lengthened and fits the measured plot. Simulations with
variations in cstep retreat parameter are also shown (c) Measured pore-solution concentrations for
C3S hydrating in mixing solution of different compositions (indicated in the legends) adapted
from [113]. The dotted and solid lines represent points of portlandite precipitation for C 3S
hydrating in CaSO4 solution and DI-water/saturated lime solution respectively.

As shown in Table (4.10), amongst the NDG parameters, N density, t0 and ρmin were varied. The
values of the t0 parameter suggest an increase in direct proportion with respect to the increase
in gypsum addition. This variation indicates a delay in the start of nucleation and growth of
C-S-H in the presence of gypsum, as can also be seen in Figure (4.40a) which indicate a delay
in the time of portlandite precipitation (and, hence, the end of the induction period) in
systems with gypsum additions.

Amongst the other NDG parameters, the widest variation was found in N density. This
parameter is related to the nucleation density of C-S-H that form as layers on the alite
particles. When plotted against the replacement percentage of gypsum, a trend emerged
(shown in Figure 4.42) which suggests that the nucleation density of C-S-H increases up to an
optimum level of substitution by gypsum. Beyond this level, the nucleation density decreases

Page Number: 147

again or stabilizes at the same value. This observation from the simulations is in agreement
with studies that have reported optimum gypsum contents in terms of compressive strength
development[112]. A higher value of nucleation density until the optimum would imply a
denser packing of hydrates in addition to enhanced hydration and, hence, greater compressive
strength. It should also be noted that the values of N density are higher for C3S as compared to
the values for alite. This difference can be attributed to the high reactivity of C 3S, as a result
of its high specific surface area.

From the trend observed in Figure (4.42), it appears that the hydration and strength
development are enhanced up to an optimum level of gypsum replacement (5% or more in our
study). Replacements beyond the optimum level lead to decrease in the hydration rate and,
hence, the development of strength.

Figure 4.42: Variation in Ndensity with respect to levels of gypsum replacement. Values for
both C3S and alite systems are included.

Page Number: 148

In this section, it was shown that the model can be used to explain the influence of gypsum
addition on alite/C3S hydration. It was shown that the critical undersaturation parameter
(βmax) needed to be varied to capture the early age hydration profiles of the substituted
systems. The variation of this parameter was the same across all levels of substitutions,
implying that the amount of gypsum dissolved initially in the system was the primary cause
driving the alteration in the early age hydration profile with respect to the reference. Through
the variation in the NDG parameters, it can be stated that the nucleation density of C-S-H
increases up to an optimum level of gypsum replacement and then decreases.

Page Number: 149

4.6 Effect of alkali sulfate addition
The addition of alkali sulfates (Na 2SO4, K2SO4) to the mixing water elevates its initial ionic
strength in proportion to the level of addition. For alite (impure tricalcium silicate; MIII-
Ca3SiO5) hydration, this initial ionic strength elevation alters the rate of alite dissolution and
enhances the overall hydration. These effects have not been quantitatively linked to
parameters which control the progress of alite hydration. This section looks at alite hydration
in solutions of varying compositions (0.1 M, 0.2 M and 0.5 M Na 2SO4 and K2SO4) in context of
their heat release behavior and analysis of the solid/liquid phases. The numerical modeling
platform, μic, is used to simulate, describe and discriminate the impact of the global solution
chemistry and reaction product formation parameters on alite hydration.

4.6.1 Introduction
Past research indicates that alkali sulfate salts enhance the hydration of alite by a faster
acceleration, higher heat-peak (main peak) and faster deceleration[111,110]. Similar to
addition of alkali hydroxide salts, alkali sulfate salts are reported to enhance the degree of
reaction at short times scales (first few hours), at longer time scales (3 days or so) the degree
of hydration is noted to be similar or decreased with respect to the reference system.

4.6.2 Materials and methods

For the study of the influence of alkali sulfate salt additions on alite hydration, the same
materials described in section (4.4) were used. Two batches of alite were synthesized namely
Batches A and B with specific surface area values of 2760 cm 2/g and 1923 cm2/g respectively.
The PSD of both batches are shown in Figure (4.24). Batch A was used for the calorimetry
and Thermo Gravimetric Analyses (TGA) and batch B for the pore solution extraction and
Quantitative X-ray diffraction analyses (QXRD). Microscopy (SEM) was done on both
batches. It has already been shown in section (4.4) that the two batches of alite hydrate
identically when their different specific surface area values are taken into account.

Alite pastes were prepared at a constant water-to-solids ratio (w/s) of 0.35. The solids were
combined with de-ionized water or the appropriate salt (Na 2SO4, K2SO4; AR-grade) solution
and mixed using a high-shear immersion mixer for 2 minutes at a mixing speed of 500 rpm.
Here, care was taken to minimize exposure of the mixtures to the air to limit the potential for
carbonation. The compositions of “mixing water” (solutions) used are detailed in Table (4.11).

Page Number: 150

Table 4.11: The compositions of mixing water utilized in the evaluations and their calculated
ionic-strengths

Concentration Ionic strength of Na2SO4/K2SO4 salt solution

0.1 M 0.3 M
0.2 M 0.6 M
0.5 M 1.5 M

4.6.3 Effect of alkali sulfates addition: Measured heat evolution profiles

Figure (4.43) shows the measured heat evolution profiles of alite hydrating in mixing solutions
with different compositions. It can be clearly observed that the hydration kinetics are altered
as a function of the dosage of salt addition. The changes with respect to increasing values of
the initial concentration of alkali salt in the mixing solution are summarized below:
1. Dissolution of alite during stage 1 is steeper: Rate lowers to low values within short
duration.
2. The heat flow during the induction period is lowered
3. The time at which induction period ends is altered
4. The rate of hydration during stage III is enhanced, as shown by steeper upslopes
during acceleration, higher heat peaks, faster decelerations.
5. The hydration of alite at later ages (beyond 15 hour or so) is also enhanced for initial
alkali sulfate concentration above 0.1 M, as can be seen in Figure (4.43c). As the
degree of hydration profiles (Figure 4.43c) are parallel at later ages, it could be
hypothesized that the enhanced degree of hydration at later ages is a result of
“carrying-over” of enhanced hydration at early ages. Concurrent to the degree of
hydration calculations, an increase in portlandite content proportional to initial alkali
sulfate content was computed from the Thermal Gravimetric Analyses (TGA)
performed at 30 hours (Figure 4.43d). Compared to the reference system, portlandite
contents at 30 hours are higher in systems with alkali sulfate concentrations of or
above 0.2 M, as is also reflected in the degree of hydration plots (Figure 4.43c).

In addition to these effects, it should also be noted that for the same dosage of salt, there are
differences in the hydration profiles between the systems with Na 2SO4 and K2SO4 salts
additions. Minor differences in overall degree of hydration can be observed in systems with
the same concentration of salt but of different types (K+ or Na+) as shown in Figure (4.43c).

Page Number: 151

 (a) (b) (c)
Figure 4.43: Measured heat evolution profiles (a) up to 50 hours (b) up to 5 hours, and (c)
degree of hydration of alite hydrating in salt solutions of different concentration. The types
and concentrations of the salts are indicated in the legends. The w/c ratio and a SSA are
indicated at the top of each figure.

Figure 4.43(d): Portlandite content, as calculated from TGA measurements at 30 hours of

hydration, against degree of hydration, as calculated from calorimetry curves, for the samples
presented in this section. The red bars represent a 5% error bound.

Page Number: 152

4.6.4 Simulating the evolution of ionic concentrations
Simulations were implemented to capture the evolution of concentrations in the systems
described above using the input parameters as shown in Table (4.12) and SCD parameters,
shown in Table (4.13). As can be seen, only one of SCD parameter (βmax) was varied while the
other two were kept at their calibrated values.

In addition to this, the measured sulfate concentration (using pore-solution analyses) as a

function of time was also used as an input in the simulations. Further analysis of the pore-
solution measurements showed a rise in pH proportional to the decrease in the sulfate
concentration. In order to check for possible precipitation of gypsum, the solid phase
compositions were studied using “continuous” and “discrete” QXRD analyses but the results
showed no precipitation of gypsum either at early or later ages. Following these findings, in
the simulations decrease in sulfate concentration and the corresponding rise in hydroxyl ions
were implemented using calibration plots as shown in Figure (4.44). The plots are based on an
exponential decrease of the sulfate ions from the initial value. As the initial sulfate
concentration changes (in a different system), the same equation can be used to simulate the
decrease in sulfate species and the corresponding increase in pH.

As the initial ionic-strength of the solution changes, the corresponding alteration in the
dissolution profile also changes the evolution of concentration of the ions. Figure (4.45) shows
the evolution of Ca2+, pH and SiTotal as obtained from simulations for systems with Na 2SO4
salts in the mixing solution. Measured values at different times are also shown in the plots. In
the simulated plots, the points at which the Ca concentrations reach a maximum and Si
reaches a minimum, represent the first precipitation of portlandite (shown by the circular
symbol) and the end of the simulated induction period. As the decrease in sulfate species
continues, the pH continues to rise as can be seen in the simulated and measured plots.
Beyond this, the concentrations stabilize at the simulated levels of portlandite solubility
equilibrium, indicated using dotted lines in each plot.

Page Number: 153

Table 4.12: Input parameters for simulations of systems described in this study. The
parameters represent the initial state of the system. The dissolution parameters used for these
simulations are shown in Table 3.3, Chapter 3. w/c ratio for all systems is 0.35. The values of
aSSA used for batches A and B are 2760 and 1923 cm2/g respectively.
Solution pH Ionic Strength Alkali concentration SO42- concentration
(mol. L-1) (mol. L-1) (mol. L-1)
DI-water 7.00 0.0 0.0 0.0
0.1 M Na2SO4 7.00 0.3 0.2 0.1
0.1 M K2SO4 7.00 0.3 0.2 0.2
0.2 M Na2SO4 7.00 0.6 0.40 0.2
0.2 M K2SO4 7.00 0.6 0.40 0.2
0.5 M Na2SO4 7.00 1.5 1.0 0.5
0.5 M K2SO4 7.00 1.5 1.0 0.5

Page Number: 154

Table 4.13: SCD and NDG parameters used for the simulations of alite hydrating in saturated
lime solution at different mixing conditions. The units of the parameters are the same as
described in section 3.9, Chapter 3. The other input parameters used for these simulations are
shown in Table 4.12. The w/c ratio for all systems is 0.35 and a SSA value for batches A and B
are 2760 and 1923 cm2/g respectively.

NDG parameters SCD parameters

Mixing ρmin NRate Ndensity Gout kden t0 kdiss cstep βmax
Solution retreat

DI-water 0.280 0.004 0.035 0.23 0.003 -0.3 24.0 4.60 10-22

0.1 M 0.255 0.004 0.055 0.23 0.003 -1.2 24.0 4.60 10-22.1
Na2SO4
0.1 M 0.240 0.004 0.055 0.23 0.003 -1.2 24.0 4.60 10-22.1
K2SO4
0.2 M 0.420 0.004 0.160 0.23 0.003 1.0 24.0 4.60 10-22.30
Na2SO4
0.2 M 0.420 0.004 0.160 0.23 0.003 1.0 24.0 4.60 10-22.30
Na2SO4
0.5 M 0.580 0.004 0.220 0.23 0.003 2.0 24.0 4.60 10-22.45
Na2SO4
0.5 M 0.470 0.004 0.220 0.23 0.003 1.0 24.0 4.60 10-22.45
K2SO4

Page Number: 155

Figure 4.44: Measured (symbols) and calibrated concentrations (lines) of SO 42- species(left)
and the corresponding rise on OH - concentration (right) as a function of time. As sulfate
concentration drops in the pore-solution, OH- ions are released to maintain electro-neutrality.

(a) (b) (c)

2+
Figure 4.45: Measured and Simulated evolution of (a) Ca (b) pH and © Total Si. The type
and concentration of the salt used to prepare the salt solution are indicated as legends. The
circle-symbols indicate the point at which portlandite precipitates and the arrows show the
decrease in the concentrations from portlandite supersaturation (SI Portlandite = 0.40) to
portlandite solubility equilibrium (SIPortlandite = 0.0, indicated by the dotted lines).

Page Number: 156

4.6.5 Simulating the heat evolution profiles
Numerical simulations were implemented to reproduce the heat evolution profiles of alite
hydrating in mixing solution with different initial concentrations of alkali sulfate salts. The
input parameters and simulations techniques used are the same as described in section (4.6.4).

Figure (4.46) shows the comparison between the simulated and measured heat evolution
profiles at early ages. As can be seen, the dissolution profiles and the nature of the induction
period (initiation point, length and termination point) changes with respect to the dosage of
salt addition in the mixing solution.

(a) (b) (c)

Figure 4.46: Measured and Simulated early age heat evolution profiles for alite hydrating in
mixing solutions with alkali sulfate additions of concentrations (a) 0.1 M (b) 0.2 M © 0.5
M. The type and concentration of the salt used to prepare the salt solution are indicated as
legends. The w/c ratio and aSSA are shown at the top of the figures. For the simulations, SCD
parameters shown in Table 4.13 and input parameters shown in Table 4.12 were used.

Figure (4.47) shows the full heat evolution profiles for all systems. Reasonably good fits were
obtained for all systems using the same simulation techniques as used for other systems. For
the simulations, the nucleation density parameter (N density) was varied in addition to
adjustments in ρmin and t0 parameters.

Page Number: 157

Figure 4.47: Measured and simulated early age heat evolution profiles for alite hydrating in
mixing solutions with alkali sulfate additions of concentrations different concentrations. The
type and concentration of the salt used to prepare the salt solution are indicated as legends.
The w/c ratio and aSSA are shown at the top of the figures. For the simulations, SCD
parameters shown in Table 4.13 and input parameters shown in Table 4.12 were used.

Page Number: 158

As shown in Table (4.13), for the simulation of systems described in this section, the SCD
parameter βmax pertaining to the critical undersaturation of alite was varied. The other SCD
parameters kdiss and cstep retreat were kept at their calibrated values (Table 3.3, Chapter 3). The
values of βmax when plotted against the initial ionic-strength of the alkali sulfate in the mixing
solution exhibited an almost linear relationship, as shown in Figure (4.48). This variation in
βmax is similar to the one described in section 4.5 for gypsum additions in alite systems.
Drawing analogies from these systems, it can be hypothesized that in the presence of sulfate
ions in the pore-solution, the evolution of defects on the surface is altered; either due to
sorption of sulfate ions on alite surface or inhibition of defect sites on alite surface. As a result
of this, there in an alteration in the critical undersaturation of alite, which determines the
threshold energy below which new etch pits cannot be created. This hypothesis is fairly
speculative at this stage and needs further investigation.

Figure 4.48: Variation in the SCD parameter βmax (log scale) with respect to the initial ionic-
strength of the alkali sulfate addition in the mixing solution. As can be seen, the critical
undersaturation of alite decreases progressively as the initial ionic-strength increases.

As shown in Table (4.13), for the simulations of the NDG regimes of the systems presented in
this section, the Ndensity parameter needed to be varied in addition to minor variations in ρmin
and t0. The variation of the ρmin parameter and its correlation to a physical variable is not
understood at this point. The variation in the t 0 parameter, on the other hand, reflects the

Page Number: 159

length of the induction period. For the reference and 0.1 M alkali sulfate systems, the negative
values of t0 indicate short induction period and rapid formation of products, as can be seen in
Figure (4.43b). For systems with higher initial concentrations of alkali sulfates, positive values
of t0 had to be used and this suggests that the acceleration regime is delayed, as can also be
seen in Figure (4.43b).

Figure (4.49) shows the variation of the N density parameter against the initial ionic-strength of
the alkali sulfate salt. As can be seen, the nucleation density of hydrates increases
proportionally with respect to the initial dosage of alkali sulfate in the pore-solution. This
trend is similar to the trend described for systems with alkali hydroxide salts addition in the
mixing solution in section 4.4 (Figure 4.34). The increase in the nucleation density was also
visually observed in SEM images (shown in Figure 4.50) obtained for samples with different
initial salt dosages at the age of 30 hours. As can be seen in Figure (4.50), the packing of the
product is denser in the samples with salt additions as compared to the reference indicating a
greater nucleation density of the product.

Figure 4.49: Variation in the NDG parameter Ndensity with respect to the initial ionic-strength
of the alkali sulfate addition in the mixing solution. As can be seen, the nucleation density of
hydrates increases progressively as the initial ionic-strength increases.

Page Number: 160

(a)

Figure 4.50: SEM images of alite at the age of

30 hours for (a) hydrating in DI-water (b)
hydrating in 0.5 M Na2SO4 solution (c)
hydrating in 0.5 M K2SO4 solution. The scales
for each image are indicated at the bottom
right.

(b)

(c)

Page Number: 161

In this section, it was shown that hydration profiles of alite in mixing solutions with different
compositions can be captured by varying one SCD parameter and two of the NDG
parameters. The variation of the βmax parameter with respect to the initial ionic-strength of
the alkali sulfate salts suggests a decrease in the critical energy required to create etch pits.
The variation in the Ndensity parameter in the simulations and the experimental results from
calorimetry and SEM showed a progressive increase in the nucleation density of C-S-H as a
function of the initial ionic strength of the pore-solution.

Page Number: 162

Chapter 5: Perspectives on Modeling Alite Hydration
Several examples describing the influence of different process parameters on hydration of alite
were discussed in Chapter 4. In this chapter, the limitations of the approaches and
perspectives for improvements are discussed.

5.1 Transition from SCD to NDG mechanism

In the model, the SCD regime is terminated at the point of portlandite precipitation. The
kinetics beyond the end of the induction period are simulated using NDG kinetics. But this
leaves a gap between the end of the simulated induction period and the acceleration period
(shown in Figure 3.14). The sequence of steps that are followed to implement this transition
are described in section (3.11) of Chapter 3.

At the point of portlandite precipitaiton (end of the simulated induction period), calcium and
hydroxyl ions are consumed in a single time-step to let the saturation index of portlandite
drop from 0.40 to 0.0. However, in experimental results [137] it is observed that this drop in
concentrations is progressive and continues for a few hours, as can be seen in Figure (5.1). If
the consumption of calcium and hydroxyl ions is simulated using a progressive decrease from
portlandite supersaturation to portlandite solubility equilibrium, the transition between the
two regimes can be modeled without a gap. In order to illustrate this a simple fit function,
shown in Equation (5.1) according to observations of pore-solution concentration data from
literature [137] and experiments (Figures 4.24 and 4.45) is used.

r Ca ( t−tprecipitaton ) (5.1)
Ca( t)=Caeq +(Ca SS−Ca eq ). exp( )
(Ca SS−Ca eq )

where, Ca(t) is the concentration of total Ca, Ca eq is the total calcium concentration at
portlandite solubility equilibrium, CaSS is the value of calcium concentration at portlandite
supersaturation of 0.40, t is the time, tprecipitation is the time at which portlandite precipitates
and rCa is a fit constant that depends on a SSA and w/c ratio. Figure (5.2) shows the simulated
evolution of calcium concentrations and the resulting heat-evolution profiles. The heat evolved
is calculated using Equation (3.4), as described in Chapter 3, which takes into account the
increase in dissolution of alite (exothermic) due to increase in undersaturation and heat
consumed as portlandite precipitates (endothermic) and ions are consumed. Using this
criterion, the bump at the end of the induction period can be reproduced (shown in Figure
5.2b).

Page Number: 163

Figure 5.1: (a) Measured calcium concentration evolutions of C 3S hydrating in solutions with
different compositions. As can be seen, the drop in calcium concentration is progressive and
continues for a few hours. Image taken from [206]

Page Number: 164

 (a) (b)
Figure 5.2: (a) Calcium concentration evolution (b) Heat evolution rates using sudden and
progressive decrease in the concentrations of calcium.

5.2 Variation in the NDG parameters

In the simulations presented in Chapter 4, it was shown that hydration of both C 3S and alite
with and without additives can be numerically reproduced. As described in sections (4.1),
(4.3) and (4.5), the values of NDG parameters: minimum density of outer C-S-H (ρmin) and
the nucleation density/rate (Ndensity/Nrate) were found to be higher for C 3S as compared to alite.
It was shown in section (3.10) of Chapter 3 that increasing the value of ρmin resulted in the
scaling of the simulated heat-flow rate to higher values. It was also shown that by increasing
the values of Ndensity or Nrate, the hydration during the stages 3 and 4 were enhanced with faster
acceleration, higher main heat-peak and faster deceleration. As C 3S-based systems hydrate
faster than alite-based systems, the values of the NDG parameters ρmin and Nrate/Ndensity were
required to be set to higher values in the simulations of the NDG regime. These high values of
the parameters in the model and enhanced hydration observed in the measured heat-evolution
profiles can be attributed to high intrinsic reactivity of C 3S. In the future, the variation in
these parameters and mechanisms leading to differences in intrinsic reactivities need to be
further investigated.

Page Number: 165

5.3 Effect of space-filling on hydration kinetics
The hydration kinetics for C-S-H, as used in the model, assumes a diffuse growth of product.
The hypothesis of space availability driving the hydration kinetics of alite at later stages
provides good reproduction of experimental data for systems with realistic w/c ratios, as
described through several examples in Chapter 4. In the future the influence of variation in
w/c ratio on hydration can be explored.

Figure (5.3) shows the heat-evolution profiles of a alite (90%)-aluminate 7 (10%) system with
3.5% of gypsum for different w/c ratios. As can be seen, at higher w/c ratios, the first
acceleration continues until a higher peak (marked by A 1) is reached. Similar behavior can
also be observed for the growth of the monosulfate phase (marked by A 2). This difference in
the reaction rate profiles can be explained as a manifestation of difference in the available
space in the different systems. In the presence of extra space in the high w/c ratio systems,
the reaction rate continues for longer durations resulting in a higher and delayed main peaks.
As the available space is still higher at the time of formation of monosulfate, the second peak
(A2) is also distinctly more intense in the high w/c ratio systems.

Figure 5.3: Measured heat evolution profiles of alite (90%)- aluminate (10%) system with
3.5% of gypsum at different w/c ratios. The w/c ratios are indicated in the legends. from
[138]

7 aluminate = C3A

Page Number: 166

Simulations of alite-water systems at different w/c ratios are shown in Figure (5.4). The same
set of parameters, as listed in Table (5.1), are used for all simulations. As the w/c ratio
increases, the available space in the microstructure increases and, consequently, the
acceleration regime continues until a more intense main peak is reached. A higher initial water
content also leads to a higher degree of hydration at later ages, as reported previously in other
studies [139,140,141,142] and shown in Figure (5.4b). It should be noted these simulations
were implemented in the model without any validation from experimental data on account of
their unavailability. Interestingly, the behavior observed from the simulations is analogous to
the one observed in the measured heat-evolution profiles of multi-phase systems shown in
Figure (5.3).

Table 5.1: NDG parameters used for simulations of alite hydration at different w/c ratios.
The simulated results are shown in Figure (5.4).

Batch ρmin Nrate Ndensity Gout kden t0

Alite 0.280 0.004 0.035 0.23 0.003 0.0

(a) (b)
Figure 5.4: Simulated heat evolution profiles of alite in DI-water at different w/c ratios. All
NDG parameters for these systems, except for w/c ratio, were the same and are shown in
Table (5.1).

Page Number: 167

However, in the case of ordinary portland cement, which includes other reactant clinker
phases (such as aluminate, belite and ferrite) and alkali salts, the influence of w/c ratio on
hydration kinetics is reported to be different from the above description. Figure (5.5) shows
measured heat-evolution profiles (unpublished data) of ordinary portland cement hydrating in
DI-water at different w/c ratios. As can be seen, at higher w/c ratios, the heat-evolution
profiles are broader and the main heat-peak is less intense. On the whole, this behavior is the
reverse of what is observed in the simulated alite (Figure 5.4) and measured alite-aluminate-
gypsum heat-evolution profiles (Figure 5.3). The reversal in the trends could be an effect of
presence of alkali salts in the ordinary portland cement but this hypothesis has not been
verified at this stage. Mechanisms driving hydration of cement-based systems, both with and
without alkalis, at higher water contents could be studied and implemented in the model in
the future.

(a) (b)
Figure 5.5: Measured heat evolution profiles of ordinary portland cement in DI-water at
different w/c ratios (a) from E. Berodier (b) from H. Chen

5.4 Densification, phase-arrangement and growth criterion of C-S-H

In the model, the densification of C-S-H is implemented using a rate that is directly
proportional to the instantaneous density (shown in Equation 5.2). This low-density product
occupies space in the microstructure constricting free microstructural volume for products to
grow further. In addition to this effect, the impingements between solid phases and

Page Number: 168

consumption of reactant material result in the deceleration of the overall rate.

dρ ρ −ρ (5.2)
=k den⋅ max
dt ρmax −ρmin
n
dρ
dt
=k den⋅
( ρmax −ρ
ρmax−ρmin ) (5.3)

As stated in Chapter 3, the dependency of the rate of densification on the instantaneous

density is empirical. Changing the correlation between these two variables, as shown in
Equation 5.3, results in a change in the overall hydration kinetics. Figure (5.6) shows the
simulated evolution of heat using different values of n in Equation (5.3) and using the same
set of NDG parameters. As can be seen, the hydration kinetics are altered significantly,
especially at later ages when available space in the microstructural volume becomes a limiting
factor. It is important to study the densification process and formulate an equation that
describes the process in a more fundamental fashion.

Figure 5.6: Simulated heat evolution profiles of alite in DI-water using different values of n in
Equation (6.2). In all simulations, the same NDG parameters were used.

Page Number: 169

In the model the formation of C-S-H was implemented as smooth concentric layers on the
surface of alite particles. A different arrangement or distribution of C-S-H in the
microstructural volume will affect the impingements between particles at later stages and, in
turn, affect the overall hydration kinetics. For example, if C-S-H is also allowed to nucleate in
the pore-space in addition to forming concentric layers in the surface of alite, the overall
hydration kinetics is expected to be different, as reported in a previous study by Thomas et
al.[143]. In the future, different arrangements of the C-S-H phase should also be studied to
further understand the role of impingements of solid phase towards driving the hydration
kinetics beyond 24 hours or so.

In all simulations presented in this study, C-S-H grows as concentric rings. However, other
growth criteria such as fibrous growth and branching dendrite-growth (demonstrated in
Figure 5.7) are expected to affect the impingements between solid phases and evolution of the
exposed surface area with time. Consequently, the reaction rates are subject to change. In the
future, the growth criteria of C-S-H should also be studied to highlight the influence of
morphology of C-S-H on hydration kinetics. As our knowledge of structure and morphology of
C-S-H is improving, “plugins” can be developed in the model to replicate the observations and
implement them to study the hydration kinetics.

(a) (b) (c)

Figure 5.7: Different growth criteria of product simulated using the model (a) Branching
cuboid growth (b) Dendrite type growth using spherical particles (c) Fibrous growth

Page Number: 170

Chapter 6: Modeling Hydration of C3A+Gypsum and Multi-phase
Cementitous Systems

6.1 Introduction
In this chapter, simulations for model C 3A (aluminate) + gypsum (C$H2) systems are
described. The chapter also includes preliminary simulations for multi-phase systems, which
are implemented by unifying concepts and approaches from simulations of model alite and
aluminate systems. Perspectives and suggestions for simulating hydration of the model
aluminate systems are discussed.

6.2 C3A+Gyspum+H systems: Basics

In commercial cements, aluminate (C 3A) is the phase other than alite that affects the
hydration kinetics and evolution of the microstructure at early ages. C 3A reacts very rapidly
in the presence of water and forms products that are poorly crystallized forms of aluminum
hydroxide (AFm phases), generally described as C 2AH8 and C4AH13 [144]. The products are
metastable and with time transform to hydrogarnet (C 3AH6) [144]. The rate of formation of
these products is very high and causes flash setting of cement paste, which is undesirable as
it reduces workability. In order to control the rate of reaction, Gypsum (CaSO 4.2H20) is added
to cement. In the presence of Gypsum, C3A reacts in two different stages:

C3 A+3 C ̄S H 2+26 H=C 6 A ̄S3 H 32 (6.1)

2C3 A+C 6 A S̄3 H32+4H=3C4 A ̄S H 12 (6.2)

Equation (6.1) represents the reaction during the first stage where the product that forms is
ettringite. The formation of ettringite begins in the first few minutes and continues until
gypsum is exhausted [145,146,147]. Beyond this, ettringite is consumed and formation of
monosulfate begins. As such, the duration of formation of ettringite depends on the gypsum
content. This sequence of reactions and formation of products is very well illustrated in a
recent work by A. Quennoz [145] conducted on C3A+gypsum systems and shown in Figures
(6.1), which shows the heat evolution profiles, and Figure (6.2), which shows the phase
assemblage.

Page Number: 171

Figure 6.1: Heat evolution profiles of C 3A+gypsum systems with different compositions. The
specific surface of the particles and w/c ratio are indicated at the top of the figure. The first
peak (left extrema of the plot) occurs in all systems initially due to the dissolution of gypsum
and C3A. Ettringite continues to form at a low and a non-zero rate until gypsum is
exhausted. The second peak corresponding to the formation of monosulfate occurs at later
times when the gypsum content is higher. (from [145]).

Page Number: 172

Figure 6.2: Evolution of phase assemblage during the first hours of hydration measured by
in-situ XRD. As can be seen, the formation of monosulfate beings after gypsum is exhausted.
A corresponding decrease in the amount of ettringite can also be observed. The heat-
evolution profile is shown by the dotted black line. (From [145]).

6.2.1 Aluminate Hydration: Kinetics of hydration during 1st stage

As described earlier, during the hydration of the C 3A+gypsum systems, ettringite is the main
product that forms. The duration of ettringite formation depends on the amount of gypsum in
the system. It has been shown in studies [146,147] that within minutes calcium and sulfate
ions reach gypsum solubility limit and continue to remain at that level until gypsum is
exhausted. As ettringite continues to form, gypsum dissolves to replenish the consumption of
calcium and sulfate ions. After the exhaustion of solid gypsum, the remaining sulfate ions in
the pore-solution are consumed to form more ettringite. After sulfate ions are also exhausted
rapidly, the formation of monosulfate beings.

One question during the hydration of these systems in the first stage is the reason for the
rapid slow down of the initial reaction, as seen in the left extrema of Figure (6.1). Minard et

Page Number: 173

al.[146] attributed the early deceleration of C3A+gypsum+H reaction to the adsorption of
sulfate ions on the surface of C3A particles. The authors corroborated their theory by
reporting that the deceleration was very rapid in a C 3A+hemihydrate system, as hemihydrate
is more soluble than gypsum. The authors suggested that before the start of the second stage
of the reaction, the sulfate ions rapidly desorb from the C 3A surface re-initiating rapid
dissolution of the C3A, which is reflected as the second peak. This also eliminates the barrier
theory, as reported in previous studies[148], that suggests that the formation of a barrier of
ettringite crystals as a cause for rapid slow down. It can be hypothesized that a process
similar to the solution controlled dissolution of C 3S (section 2.3.3, Chapter 2) could occur in
the reaction of C3A with calcium sulfate, whereby sulfate ions adsorb at defect sites and
inhibit the formation of etch pits, so slowing down the rate of dissolution. However, this
remains to be investigated.

6.2.2 Aluminate Hydration: Kinetics of hydration during 2nd stage

During the second stage of the hydration of C 3A+gypsum systems, the heat rate rises again
until a peak is reached followed by a rapid slow-down of hydration (shown in Figure 6.1). It
is observed in a recent study[145] that the exothermic peaks become broader and less intense
(lower heat peaks) with increasing gypsum content, as shown in Figure (6.3). Quennoz
postulated two mechanisms that cause this:
1. The specific surface area of C3A particles decreases to a lower level for higher gypsum
contents, as more C3A is consumed due to prolonged ettringite formation. As a result
of this, the renewed dissolution that begins at the start of stage 2, occurs at a slower
rate, thus resulting in a broader peak.
2. For higher gypsum contents, prolonged formation of ettringite lowers the space
availability for new products to form and grow. Due to limited space available in
systems with more gypsum, and hence more ettringite, the rate of growth of
monosulfate is slower resulting in broader peaks.

Page Number: 174

Figure 6.3: Second stage reaction of C 3A+gypsum systems with different gypsum contents
(indicated in the legends as G). As can be seen, the peaks are broader when the gypsum
content increases. (from [145])

6.3 C3A+H2O+Gypsum systems: Simulations

A model to simulate C3A+Gypsum+H2O is currently under development as part of
collaborative project between Laboratory of construction materials, EPFL (by A. Kumar, A.K
Mohamed, K. L. Scrivener), Switzerland and Indian Institute of Technology Delhi (S.
Bishnoi), India.

The two stages of hydration of C 3A are numerically handled by two separate mechanisms in
the model. The following sections describe the implementation of the two mechanisms in the
model.

Page Number: 175

6.3.1 Preliminary simulations for C3A+H2O+Gypsum systems: Stage 1
For the first stage of C 3A hydration, a simple rate law as described by Equation (6.3) is used.
This equation uses the instantaneous surface area of the C 3A particles, which decreases with
time (Equation 6.4). These equations do not take into account the heat released from
dissolution of gypsum and C3A in the first few minutes. As such, only the formation of
ettringite through consumption of C3A is reproduced.

dmC A 3
(6.3)
=−k 2⋅a SA (t)
dt

where,

a SA=∑ i 4 π(r 0, i−k 1 t)

2 (6.4)

where, k1(cm. hour-1) and k2(g. hour-1. cm-2) are rate constants and r0,i (cm) is the initial radius
of the particle i. Of these parameters, k1 and k2 are calibration parameters and r0,i is used as
an input from the particle size distribution (size versus cumulative volume percentage). The
other input parameters for these simulations are composition of the C 3A+Gypsum system and
w/s ratio. The heat of hydration for this stage is set at 866 kJ/g [148]. It should be noted that
these simulations are carried out analytically without the use or generation of digital
microstructures.

Figure (6.4) shows the simulation results for the first stage for reaction for 3 different batches:
A (aSSA = 3050 cm2/g), B (aSSA = 4360 cm2/g) and C (aSSA = 4600 cm2/g). The set of
parameters used for these simulations are listed in Table 6.1.

Page Number: 176

Figure 6.4: Measured and simulated heat evolution profiles of C 3A+gypsum+H systems with
different compositions and PSDs. Only the first stage of reaction is shown. The batches
(PSDs) and compositions are indicated at the top of each figure.

Page Number: 177

Table 6.1: Parameters used for simulations shown in Figure 6.4.

Batch C3A (%) Gypsum (%) k1 k2

(cm. hour-1) (g. hour-1. cm-2)
A 93 7 0.06 3.69. 10-7
A 88 12 0.06 3.69. 10-7
A 80 20 0.06 3.69. 10-7
A 73 27 0.06 3.69. 10-7
B 90 10 0.125 7.59. 10-7
B 80 20 0.125 7.59. 10-7
B 65 35 0.125 7.59. 10-7
C 90 10 0.125 7.59. 10-7
C 80 20 0.125 7.59. 10-7
C 70 30 0.125 7.59. 10-7
C 60 40 0.125 7.59. 10-7

As can be seen in Figure (6.4), the results from the simulations provide satisfactory
reproduction of the heat-evolution profiles. The simulation results fit the measured hydration
profiles very well at later stages but poorly before approximately 5 hours. This discrepancy in
the simulations at early ages could be a result of lack of descriptions of gypsum and aluminate
dissolution, which affect the hydration profiles for the first few hours.

The values of k1 and k2, as listed in Table (6.1), show that the same rate law can be
generically used to describe the hydration of C 3A+gypsum systems in the first stage. For
batches B and C, irrespective of the composition, the parameters were identical and could be
used as calibrated constants. For batch A, the values of k 1 and k2 were both halved compared
to batches B and C. This difference in the values could be a result of changes in the intrinsic
reactivities between the batches or large difference in the specific surface area values between
batches A and B/C. In order to evaluate the precise calibrated values of k1 and k2, simulations
for more systems are intended to be implemented in the future.

Page Number: 178

6.3.2 Preliminary simulations for C3A+H2O+Gypsum systems: Stage 2
The second stage reaction involves the formation of monosulfate phase and starts when
gypsum is exhausted in the system. For the 2 nd reaction classical boundary nucleation and
growth mechanism, shown in Equations (6.5 – 6.10), are used to simulate the growth of
monosulfate.

dmC A X (t+dt)−X( t) (6.5)

Rate ( 3
)=
dt dt

Fraction reacted(X)=1−exp (−V extended (t)) (6.6)

G out t
Extended Volume ( V extended )=∫0 a BV⋅( 1−exp (−A f ) ) dy ` (6.7)

Where Af can be written as:

[
A f =π Ndensity⋅(t2r −
y2
2
Gout
)+N rate⋅(
t3r y 2 tr 2y 3
− +
3 G 2out 3G3out
)
] (6.8)

where y = 0 and tr = 0 (for t < tmonosulfate formation) (6.9)

and tr = t – tmonosulfate formation (for t > tmonosulfate formation ) (6.10)

where Ndensity (hour-2) and Nrate (hour-3) are coefficients of nucleation density and nucleation rate
respectively. Ndensity and Nrate can also be written as (Idensity. Gpar2) and (Irate. Gpar2), where Idensity
is the number of nuclei of product per unit area, I rate is the rate of nucleation of product per
unit area and Gpar is the parallel growth rate. The parameter a BV is the boundary area per
unit volume of water, the value of which is used as an input while taking into account the
area of C3A particles at the start of monosulfate formation and amount of water (space) left
in the system. This set of equations include 4 NG parameters (N rate, Gout, tmonosulfate formation, and
Ndensity) that influence the hydration profiles uniquely. The sensitivity of these parameters with
respect to baseline values (shown in Table 6.2) are illustrated in Figure (6.5).

Page Number: 179

Table 6.2: Baseline values of NG parameters used to evaluate the sensitivity of individual
parameters (shown in Figure 6.5).

Ndensity Nrate Gout tmonosulfate formation

(hour-2) (hour-3) (μm. hour-1) (hour)
0.0 5.0 0.26 0.0

Figure 6.5: Sensitivity of boundary NG parameters: N rate, Ndensity, tmonosulfate formation and Gout. Only
the parameters listed on the top of each figure were changed while keeping the other
parameters at their baseline values as listed in Table (6.2).

Using this mechanism, simulations were implemented for reaction in the second stage. As
such, two of the NG parameters: G out and tmonosulfate formation were varied while the others were
retained at identical values. The value of N density was set at 0.0 and Nrate was set at 5.0,
implying that monosulfate continues to nucleate as hydration progresses. The comparison of
simulation results against experimental results are shown in Figure (6.6). The boundary NG
parameters used are listed in Table (6.3).

Page Number: 180

Table 6.3: Values of NG parameters used for simulations of the second stage reaction of
C3A+gypsum+H2O systems shown in Figure (6.6).

Batch C3A Gypsum Ndensity Nrate Gout tmonosulfate formation

(%) (%) (hour-2) (hour-3) (μm. hour-1) (hour)
A 93 7 0.0 5.0 0.43 1.1
A 88 12 0.0 5.0 0.31 3.6
A 80 20 0.0 5.0 0.25 20.2
A 73 27 0.0 5.0 0.157 46.0
B 90 10 0.0 5.0 0.86 1.50
B 80 20 0.0 5.0 0.57 8.05
B 65 35 0.0 5.0 0.30 35.6
C 90 10 0.0 5.0 0.91 1.38
C 80 20 0.0 5.0 0.42 4.10
C 70 30 0.0 5.0 0.41 18.45
C 60 40 0.0 5.0 0.096 50.0

Page Number: 181

Figure 6.6: Simulated and measured heat-evolution profiles of C 3A+gypsum+H systems for
the second stage of reaction. The parameters used for these simulations are listed in Table 6.3.

As can be seen in Figure (6.6), the overall hydration kinetics can be captured reasonably well
using the boundary NG kinetics. The acceleration and the main heat-peak can be well
reproduced but the deceleration for some systems does not fit the experimental values very
well. This could be due to a change in the driving mechanism from standard boundary NG to
some other mechanism at high volumes of monosulfate around the C 3A grains. Modifications
in the boundary NG kinetics also need to be incorporated if the monosulfate phase densifies
with time, like C-S-H. These concepts will be further examined and implemented in the model
in the future.

Page Number: 182

Amongst the boundary NG parameters, the nucleation density and nucleation rate coefficients
were kept the same for all systems and the parameters G out and tmonosulfate formation were varied.
For each batch, as the gypsum content increases, it was observed that t monosulfate formation
increased. This is expected as more gypsum implies formation of ettringite for longer
durations and delayed starts of the second stage reaction, which is reflected in the t monosulfate
formation. In order to verify this, the values of tmonosulfate formation were compared against values of

time at which gypsum exhausts in the systems calculated using both experimental and
simulated heat evolution profiles (Figure 6.7). These calculations were carried out using the
scheme described below:
1. Using the measured/simulated calorimetry profile to compute the degree of hydration
until the second stage of reaction
2. Evaluation of amount of ettringite formed from the degree of hydration calculations
3. Calculating the corresponding gypsum consumption to form ettringite
4. Computing the time to exhaust gypsum, considering the initial amount in the system
As can be seen, the simulation time values at which monosulfate begins to form corresponds
very well to the time at which there is no solid gypsum in the system. Thereby, it can be said
that ettringite is the only product that forms in the first stage of reaction and monosulfate
begins forming when gypsum is all consumed.

Page Number: 183

Figure 6.7: Time to exhaust gypsum calculated from the measured (circular symbols) and
simulated (diamond shaped symbols) heat-evolution profiles versus values of the t monosulfate
formation parameters used in the simulations. The symbols represent the data points and the

dotted line represents the ideal trend.

Larger amounts of ettringite forms in systems with higher gypsum contents. This leads to
space-constraint for monosulfate to grow and this is reflected in the G out parameter, which is
lower for systems with higher gypsum contents and vice-versa. In order to further corroborate
this, porosity (or space in the microstructural volume) left in the systems at the time of AFm
peaks were calculated and plotted against the growth rate values (shown in Figure 6.8). As
stated earlier, it can be seen that Gout increases as space available in the system increases.
This trend is similar to the trend observed for alite-based systems, as described in section 4.1,
Chapter 4. It should also be noted that while batches B and C follow the same trend, batch A
has lower values. Deviation in batch A compared to batches B/C were also observed in the
first stage of reaction, as described in section 6.3.1. This could be due to differences in
intrinsic reactivities or large difference in specific surface values between batches A and B/C.

Page Number: 184

 (a) (b)
Figure 6.8: Variation of the G out parameter against (a) porosity at the time of the AFm peak
(b) time of AFm peak

6.3.3 Conclusions: Preliminary simulations for C3A+H2O+Gypsum systems

In the previous sections it was shown that the basic analytical model can be used to simulate
the hydration kinetics of C 3A+H2O+Gypsum systems. It was shown that calibrated
dissolution parameters can be used to describe hydration during the first stage. The second
stage reaction involving the formation of monosulfate can be simulated using classical
boundary NG reaction and using two variable parameters G out and tmonosulfate formation were varied.
The values of the Gout parameter decreased with increasing gypsum content indicating a
strong correlation between availability of space and growth of products.

6.4 Simulating hydration of multi-phase cementitous systems

In Chapter 3, the mechanisms for simulating alite hydration kinetics were described. In the
previous sections, simulation techniques for C3A hydration are delineated. Combining these
two mechanisms, preliminary simulations were implemented to reproduce the heat-evolution
profiles of alite-aluminate-gypsum and commercial cement-based systems. For the simulations,
the hydrations of alite and aluminate phases were treated separately and the heat evolved

Page Number: 185

from their respective reactions were added to obtain an overall heat evolution profile. Using
this overall heat-evolution profile, the heat-flow rates were simulated.

For validation, heat-evolution profiles of three alite-aluminate-gypsum based systems from the
work of A. Quennoz and commercial systems from the study of V. Kocaba were used
[145,149]. The three alite-aluminate-gypsum systems vary in compositions and initial phase
assemblage: Batch A (Polyphase; Alite(90%)-Aluminate(8%)-Gypsum(2%)), Batch B
(Monophase; Alite(90%)-Aluminate(6.5%)-Gypsum(3.5%)) and Batch C (Polyphase;
Alite(88%)-Aluminate(7.8%)-Gypsum(4.2%)). The three commercial systems are:
• White cement (Alite(68.1%)-Belite(23.6%)-Aluminate(6.3%)-Ferrite(1.9%)-
Gypsum(3.5%))[207],
• Ordinary Portland Cement B (Alite(62.6%)-Belite(17.6%)-Aluminate(6.1%)-
Ferrite(7.5%)-Gypsum(4.4%))[207], and
• Ordinary Portland Cement C (Alite(68.2%)-Belite(13.5%)-Aluminate(3.6%)-
Ferrite(8.8%)-Gypsum(3.7%)))[207].

For the simulations, the hydration of the alite was simulated using the combination of SCD
and NDG mechanisms as described in Chapters 3 and 4. The hydration of C 3A was simulated
using the mechanisms described in previous sections. The hydration of belite and ferrite was
not taken in account assuming that these phase do not hydrate to a large extent in the first 3
days[149]. The parameters used for simulations are listed in Tables (6.4) (aluminate hydration
kinetics) and Table (6.5) (alite hydration kinetics). In addition to these parameters, w/c ratio
and aSSA are used as input parameters for the simulations. For the hydration of alite and
aluminate, the aSSA values are computed using the equations below:

a SSA,alite =f alite aSSA (6.11)

a SSA,aluminate=f aluminate a SSA (6.12)

where, f represent the mass fraction of the respective phases. The results obtained from
simulations are shown with the measured heat-evolution profiles in Figure (6.9). Figure (6.9a)
shows the heat-evolution rates as obtained from alite and aluminate hydration separately and
Figure (6.9b) shows the overall heat-flow rate obtained by combining the heat evolved from
the hydration of alite and aluminate phases. As can be seen, the simulation results are in good
agreement with the experimental results.

Page Number: 186

Table 6.4: Stage 1 and 2 kinetic parameters used to simulate aluminate hydration kinetics.
The units of all parameters are the same as described section (6.3).

Batch k1 k2 Nrate Ndensity tmonosulfate Gout

formation

Alite-aluminate-gypsum systems
A 0.125 7.59. 10-7 1.6 0.0 1.93 0.70
B 0.125 7.59. 10-7 1.6 0.0 14.00 0.04
C 0.125 7.59. 10-7 1.6 0.0 11.00 0.02
Commercial systems
White Cement 0.125 7.59. 10-7 1.6 0.0 25.00 0.0030
OPC B 0.125 7.59. 10-7 1.6 0.0 11.00 0.0065
OPC C 0.125 7.59. 10-7 1.6 0.0 23.00 0.0050

Table 6.5: SCD and NDG parameters used to simulate alite hydration kinetics. The units of
all parameter are the same as described in examples in Chapter 4.

NDG parameters SCD parameters

Batch ρmin NRate Ndensity Gout kden t0 kdiss cstep βmax
retreat

Alite-aluminate-gypsum systems
A 0.135 0.01 0.035 0.150 0.0030 2.0 24.0 4.6 10-22.00
B 0.135 0.01 0.035 0.250 0.0030 0.0 24.0 4.6 10-22.00
C 0.135 0.01 0.035 0.350 0.0030 0.0 24.0 4.6 10-22.00
Commercial systems
White cement 0.260 0.05 0.100 0.160 0.0011 -1.70 24.0 4.6 10-22.00
OPC B 0.270 0.05 0.100 0.120 0.0011 0.70 24.0 4.6 10-22.00
OPC C 0.270 0.05 0.100 0.092 0.0011 -1.70 24.0 4.6 10-22.00

Page Number: 187

Page Number: 188
Page Number: 189
 (a) (b)
Figure 6.9: (a) Measured and separate simulated heat evolution profiles obtained from alite
and aluminate hydration (b) Measured and simulated overall heat evolution profiles for multi-
phase systems. The compositions of the systems are indicated at the top of each figure.

Page Number: 190

As can be seen in Table (6.4), for stage 1 of C 3A reaction, the values of both k1 and k2 are
identical and equal to the values used for model C 3A+gypsum systems. For the stage 2
reaction, the outward growth rate (Gout) and start time parameter (t monosulfate formation) were
varied. Like in the case of model systems, it can be observed that as gypsum content
increases, tmonosulfate formation increases. This is expected because higher gypsum content leads to
formation of ettringite until longer durations and, hence, a delayed start of the second stage
reaction. The values of Gout decrease with increasing gypsum content in a trend similar to that
was identified for model C3A systems. This decrease can be attributed to space constraint in
the microstructural volume, as shown in Figure (6.10). Higher gypsum leads to formation of
more ettringite and more C-S-H (as alite reacts for a longer duration). As these products
occupy more space in the microstructure, the growth rate is lower on account of less available
space. It should be noted that for similar values of porosity, the values of the growth rate for
multi-phase systems are less in magnitude as compared to model systems (B and C) but
similar to values used for batch A. Interestingly, the aluminate used in the alite-aluminate-
gypsum systems was from batch A and, from this it could be said that the growth rate of
monosulfate in multi-phase systems follows that same trend as in model systems.

(a) (b)
Figure 6.10: Variation of the monosulfate growth rate against porosity at the time of AFm
peak for (a) multi-phase systems (b) all model and multi-phase systems presented in this
chapter.

Page Number: 191

Amongst the kinetic parameters used for alite hydration, G out and t0 were varied separately for
alite-aluminate-gyspum and commercial systems (see Table 6.5). The value of t 0 is higher for
the undersulfated system (Batch A), which suggests a delayed start of the nucleation and
growth of C-S-H. This is expected because in this system, the formation of monosulfate begins
at very early ages, which would consume calcium and hydroxyl ions from the pore-solution
and delay the precipitation of portlandite and the start of the nucleation and growth of C-S-
H. For all other systems, which are properly sulfated, nucleation and growth of C-S-H starts
early.

The values of the C-S-H growth rate (G out) parameter ranged from 0.092 to 0.35. When
plotted against the surface area of particles (using the area directly from the PSD
measurements and Equation 6.11), a progressive decrease in G out was observed (Figure 6.11).
The same trend was observed in the alite/C 3S model systems (also shown in Figure 6.11).
This suggests that the speed of growth of C-S-H is higher when the particles are coarser and
the initial dispersion between the particles is more, as described in section 4.1, Chapter 4. It
should also be noted that values of the C-S-H growth rate are lower in the multi-phase
systems as compared to the model systems. Lower growth rates in the multi-phase systems
were also observed for the monosulfate phase, as described earlier. This suggests that the
hydration of both alite and aluminate phases are inhibited due to interactions with each
other.

Page Number: 192

Figure 6.11: Variation of the C-S-H growth rate against surface area of particles. G out values
for both model (section 4.1, Chapter 4) and multi-phase systems (this chapter) are included.

Page Number: 193

Chapter 7: Conclusions and Perspectives
In the current study it was shown that numerical models can provide important information
about cement hydration. It was also shown that a unified approach combining dissolution,
thermodynamics and nucleation & growth mechanisms can be consistently used to explain
hydration of a wide variety of cement-based systems. It was shown that with a limited
number of fit parameters, the influence of several additives on hydration of alite can be
numerically reproduced. It was also shown that through the integration of mechanisms from
hydration of alite and aluminate phase, hydration of multi-phase systems can be described.
The variations in the fit parameters for all systems were shown to have strong correlations to
the initial states of the systems.

In the following sections some of the important numerical and scientific advancements made
in this study are described. Perspectives for simulating hydration in complex systems are also
discussed.

7.1 Dissolution of alite

In the previous chapters, it was shown that dissolution of alite can be numerically described
using a solution controlled dissolution (SCD) mechanism. The mechanism is based on
principles that are used to describe dissolution of several naturally occurring minerals. Some
of the key concepts that can be drawn from the implementation of dissolution of alite in the
model are summarized below:
• Dissolution of alite is a two-regime phenomenon. In the first regime, the rate of
dissolution is controlled by the undersaturation of the pore-solution with respect to
alite. Beyond a critical value of undersaturation, dissolution of alite is controlled by
retreating steps on the surface of alite.
• During the induction period, which is the second dissolution regime, portlandite does
not precipitate as the supersaturation of portlandite is still below the critical
supersaturation. In the model, the value of supersaturation for portlandite is set at
0.40. This value was chosen from observations of pore-solution composition. The
induction period is terminated when portlandite precipitates.

Page Number: 194

7.2 Effects of process parameters of alite hydration
Through several examples described in Chapter 4, it was shown that the hydration of alite
can be generically described using the numerical implementation described in Chapter 3. It
was shown, using experimental and simulation results, that the initial state of the system and
changes in experimental process parameters significantly influence hydration kinetics. The
influence of various process parameters on the hydration of alite/C3S are summarized below:
1. Coarser particles hydrate slower as compared to finer particles on account of difference
in the specific surface areas. The changes in dissolution kinetics for different particle
sizes were captured without any modifications in the calibrated SCD variables. For the
simulations of the NDG regime, variations in the outward Growth rate (Gout) and
nucleation rate (Irate) parameters were required in addition to minor variations in 3
other parameters. Irate showed a progressive increase with respect to the specific surface
area indicating a higher number of nucleation sites. G out was shown to have a linear
relationship with increasing inter-particle spacing initially. This variation suggests an
opportunistic growth criteria of the hydrates, in which the growth is limited by space
constraint.
2. Mixing conditions dramatically influence the hydration of alite in DI-water. Dissolution
is faster, induction period is shorter and hydration is enhanced when mixing is
performed at high rates. It was also shown that the model can reproduce the influence
of the mixing conditions on hydration kinetics by proportionally increasing the values
of the simulation parameters (a) k diss and cstep retreat, which are related to the ease at
which species detach from the dissolving surface of alite, and (d) N density, which is
related to the number of C-S-H nuclei per unit surface area of the particles.
Experimental results confirmed the increase in the nucleation density of hydrates as a
higher packing density of C-S-H in the microstructural matrix was observed for
samples prepared at higher mixing rates.
3. Mixing conditions have little or no effect on the dissolution of alite in saturated lime
solution. Hydration during the NDG regime is enhanced at high mixing rates. In the
presence of calcium ions and high initial pH, undersaturation of the pore-solution with
respect to alite is very low initially and, consequently, dissolution begins from very
close to the step-retreat regime. As such, only slight variations in the c step retreat
parameter in addition to calibrated values of the other two SCD parameters were used
to simulate the SCD regime. The enhanced hydration during the NDG regime was
numerically reproduced by increasing the nucleation density of hydrates in proportion
to the mixing rates. SEM images confirmed an increase in the packing density of

Page Number: 195

 hydrates in the matrix.
4. Thermal treatment at high temperatures and longer durations decreases the defect
density and induce a series of changes in the hydration profiles of alite/C 3S: low heat
evolution, slower dissolution and longer duration of the induction period. This behavior
was numerically reproduced in the model by proportionally decreasing the c step retreat
parameter with respect to the duration and temperature of treatment. Contrary to the
dissolution regime, the NDG regime showed little or no change with the heat treatment
procedure.
5. Addition of alkali hydroxides in the mixing solution shortens induction periods and
enhances hydration at early ages. Precipitation of portlandite takes place at lower
levels of calcium concentration suggesting that the supersaturation-barrier is reduced at
high initial pH of the mixing solution. In the model, this behavior was captured by
providing the initial pH and initial alkali concentrations as inputs and using calibrated
SCD parameters. For the NDG regime, the variation in kinetic parameters suggested
an increase in the nucleation density of C-S-H in the presence of NaOH/KOH salts in
the mixing solution. Microscopy images also revealed a denser packing of hydrates in
systems with a high initial concentration of hydroxide salts.
6. Partial replacement of alite or C3S by gypsum lengthens the induction period. Through
the variation in SCD parameter (βmax), that was used to simulate the dissolution
profiles, it was hypothesized that the presence of sulfate ions in the pore-solution
inhibits the defect-sites and reduces the dissolution rate. In the presence of gypsum, the
overall hydration is enhanced, significantly in the case of alite and slightly in the case
of C3S in the presence of gypsum. In the simulations, the increase in t 0 parameter
confirmed that the induction periods are longer and nucleation and growth of hydrates
begin later in the gypsum substituted samples. The values of the nucleation density
parameter increased up to an optimum (5%) level of gypsum substitution and then
decreased or stabilized.
7. The addition of alkali sulfates in the mixing solution has an effect analogous to the
effects observed in systems with gypsum additions. Lower heat flow and longer
induction periods were observed for mixing solutions with higher dosages of alkali
sulfate salts. Similar to gypsum additions, these effects were numerically reproduced by
varying the critical undersaturation parameter (βmax) in linear proportion to the dosage
of salt. The enhancement in hydration, which was also observed in the measured
calorimetry profiles, was reproduced by proportionally increasing the nucleation density
of hydrates. An increase in the nucleation density of hydrates was also evidenced by
observations from SEM images.

Page Number: 196

7.3 Hydration of C3A+Gypsum systems
In Chapter 6 preliminary hydration kinetics of C 3A+Gypsum systems were discussed.
Experimental results and simulations showed that the overall reaction takes place in two
stages: Ettringite formation in the first stage and Monosulfate formation in the second stage.
Some of the important findings from the simulations are summarized below:
• Reaction during stage 1 can be numerically reproduced using calibration kinetics
parameters across systems with different PSDs or w/c ratios.
• Reaction during stage 2 can be simulated using boundary nucleation and growth of the
monosulfate phase. As such, all kinetic parameters can be calibrated and retained at
constant values except for the outward Growth rate (G out) parameter. It was shown
that the growth rate increase proportionally with respect to the available space in the
microstructure. This variation is similar to the trend observed in alite-based systems,
thus highlighting the importance of space-constraint in hydration kinetics.

7.4 Perspectives on microstructural modeling

The model μic is currently limited to spherical particles. Products can only grow either as
smooth layers or “solid” nuclei in the pore-space. Consequently, it is not possible to represent
phases that have intrinsic porosity, like C-S-H, or to represent phases that have a plate or
needle shaped geometry, like ettringite and portlandite. These limitations affect the post-
process simulations like the computation of solid and pore-phase connectivities, transport of
water and ionic-species etc. While it is possible to numerically include other shapes, the
simulation-times and computational power increase drastically. It is therefore important to
develop algorithms that can enable fast computations of interactions between non-spherical
shapes.

Page Number: 197

7.5 Perspectives on thermodynamic modeling
Thermodynamics is integrated in the model to compute activities and saturation index values
for several phases. However, the model does not use an exhaustive database that is used in
models such as PHREEQC and GEMS. A thermodynamics-based model would be useful for
the following reasons:
• Calculation of thermodynamic parameters such as activity and saturation index values
for several species that exist in the cement-based systems such as AFt and AFm
phases, carbonate containing species etc.
• Describe the stability or instability criteria for phases depending upon the state of
pore-solution and solid phases present in the system. For example, thermodynamic
calculations can be used to evaluate the time of monosulfate formation in
C3A+gypsum and multi-phase systems. This would eliminate the necessity of using
tmonosulfate formation as a variable parameter.
• Possibility to simulate chemical interactions between clinker phases and hydrates with
supplementary cementitous materials such as Fly ash, metakaolin, slag etc.
• Possibility to compute speciation of bulk ionic species such as Si (to H 4SiO40, H3SiO4-,
H2SiO42-), Al (to Al3+ and Al(OH)4-) etc.
• Study the effect of temperature on the solubility of phases
In the future, the two models can be integrated such that the numerical and microstructural
computations are carried out by μic and the thermodynamic calculations are handled by
PHREEQC/GEMS. In order to do so, the thermodynamic model can be integrated as a
“plugin” which would get inputs (bulk concentrations of all species) from μic and would return
the thermodynamic values as outputs (activities and saturation index values), as shown
schematically in Figure (7.1). With this integration, it would then be possible to simulate
hydration of multi-phase systems more precisely.

Page Number: 198

Figure 7.1: Schematic flow-chart of integration of a thermodynamic model with μic.

7.6 Perspectives on simulating multi-phase systems

In Chapter 6, some preliminary simulations for multi-phase systems were discussed. Using the
principles from alite and aluminate hydration and integrating the mechanisms, it was shown
that the hydration alite-aluminate-gypsum systems and commerical systems can be
numerically reproduced.

In order to describe the temporal progress of hydration of more complex systems, it is also
important to incorporate hydration of the belite phase. As such, it is critical to study
hydration of belite both individually and in the presence of alite. Combining mechanisms and
concepts from these individual phases, a model can be developed to simulate hydration of
commercial and more-complex systems.

Page Number: 199

References

Page Number: 200

1 H.F.W. Taylor, Cement Chemistry, Thomas Telford, 1997
2 H.N. Stein, Thermodynamic considerations on the hydration mechanisms of Ca 3SiO5 and Ca3Al2O3, Cement
and Concrete Research 2, 167-177 (1972)
3 S. Diamond, C/S mole ratio of C-S-H gel in a mature C 3S paste as determined by EDXA, Cement and
Concrete Research 6, 413-416 (1976)
4 J.A Gard and H.F.W Taylor, Calcium silicate hydrate (II) ("C-S-H(II)"), Cement and Concrete Research 6,
667-677 (1976)
5 D.L Rayment and A.J Majumdar, The composition of the C-S-H phases in portland cement pastes, Cement
and Concrete Research 12, 753-764 (1982)
6 J.F Young and H. S. Tong, Microstructure and strength development of beta- dicalcium silicate pastes with
and without admixtures, Cement and Concrete Research 7 627-636 (1977)
7 M. Daimon, S. Ueda and R. Kondo, Morphological study on hydration of tricalcium silicate, Cement and
Concrete Research 1, 391-401 (1971)
8 E. Gallucci and K. Scrivener, Crystallisation of calcium hydroxide in early age model and ordinary
cementitious systems, Cement and Concrete Research 37, 492-501 (2007)
9 E. M. Gartner and J. M. Gaidis, Hydration Mechanisms, I, Materials Science of Concrete 1 (95–125) 1989.
10 P. W. Brown, E. Franz, G. Frohnsdorff, and H. F. W. Taylor, Analyses of the Aqueous Phase During Early
C3S Hydration, Cement and Concrete Research, 14, 257–62 (1984).
11 H. M. Jennings, Aqueous Solubility Relationships for Two Types of Calcium Silicate Hydrate, Journal of
American Ceramic Society, 69 [8] 618 (1986).
12 E. M. Gartner and H. M. Jennings, Thermodynamics of Calcium Silicate Hydrates and their Solutions,
Journal of American Ceramic Society, 80 [10] 743–9 (1987).
13 P. W. Brown, Phase Equilibria and Cement Hydration, Materials Science of Concrete, Vol. 1, (73-93) 1989.
14 S. Garrault-Gauffinet and A. Nonat, Experimental Investigation of Calcium Silicate Hydrate (C–S–H)
Nucleation, Journal of Crystal Growth, 200, 565–74 (1999)
15 S. Garrault, A. Nonat, Hydrated layer formation on tricalcium and dicalcium silicate surfaces: experimental
study and numerical simulations, Langmuir 17 (2001) 8131–8138.
16 D. Damidot, A. Nonat, and P. Barret, Kinetics of Tricalcium Silicate Hydration in Diluted Suspensions by
Microcalorimetric Measurements, Journal of American Ceramic Society, 73 [11], 3319–22 (1990).
17 S. Bishnoi, K. L. Scrivener, μic : Studying nucleation and growth kinetics of alite hydration using
μic, Cement and Concrete Research 39 (2009) 849-860.
18 J.J Thomas, A New Approach to Modeling the Nucleation and Growth Kinetics of Tricalcium Silicate
Hydration, Journal of American Ceramic Society 90 (2007) 3282-3288
19 J. W Bullard, A determination of hydration mechanisms for tricalcium silicate using a Kinetic Cellular
Automaton Model, Journal of American Ceramic Society 91 (2008) 2088-2097
20 P. Juilland, E. Gallucci, R. Flatt, K. L. Scrivener, Dissolution theory applied to the induction period in alite
hydration, Cement and Concrete Research 40 (2010) 831-844.
21 J. W. Bullard, H. M. Jennings, R. A. Livingston, A. Nonat, G. W. Scherer, J. S. Schweitzer, K. L. Scrivener,
J. J. Thomas, Mechanisms of cement hydration, Cement and Concrete Research 41 (2010) 1208-1223.
22 J.J. Thomas, J.J. Biernacki, J. W. Bullard, S. Bishnoi, J. S. Dolado, G. W. Scherer, A. Luttge, Modeling and
simulation of cement hydration kinetics and microstructure development, Accepted for publication in Cement
and Concrete Research 41 (2011) 1257-1278
23 A. Nonat, Modelling hydration and setting of cement, Ceramics 92 (2005) 247–257.
24 H.N. Stein, J.M. Stevels, Influence of silica on the hydration of 3CaO, SiO 2, Journal of Applied Chemistry 14
(1964) 338–346.
25 H.M. Jennings, P.L. Pratt, An experimental argument for the existence of a protective membrane
surrounding portland cement during the induction period, Cement and Concrete Research 9 (1979) 501–506.
26 F. Bellmann, D. Damidot, B. Möser, J. Skibsted, Improved evidence for the existence of an intermediate
 phase during hydration of tricalcium silicate, Cement and Concrete Research 40 (2010) 875–884.
27 H.F.W. Taylor, Cement Chemistry, 2nd Edition. Thomas Telford, London, 1997.
28 N.J. Clayden, C.M. Dobson, C.J. Hayes, S.A. Rodger, Hydration of tricalcium silicate followed by solid-state
29Si NMR spectroscopy, Journal of Chemical Society of Chemistry, Communication 21 (1984) 1396–1397.
29 S.A. Rodger, G.W. Groves, N.J. Clayden, C.M. Dobson, Hydration of tricalcium silicate followed by 29Si
NMR with cross polarization, Journal of American Ceramic Society 71 (2) (1988) 91–96.
30 R.A. Livingston, J.S. Schweitzer, C. Rolfs, H.W. Becker, S. Kubsky, Characterization of the induction period
in tricalcium silicate hydration by nuclear resonance reaction analysis, J. Mater. Res. 16 (3) (2001) 687–693.
31 S. Garrault, E. Finot, E. Lesniewska, A. Nonat, Study of C–S–H growth on C 3S surface during its early
hydration, Materials and Structures 38 (2005) 435–442.
32 J.F. Young, H.S. Tong, R.L. Berger, Compositions of solutions in contact with hydrating tricalcium silicate
pastes, Journal of the American Ceramic Society 60 (5–6) (1977) 193–198.
33 I. Odler, H. Dörr, Early hydration of tricalcium silicate II. The induction period, Cement and Concrete
Research 9 (1979) 277–284.
34 P.W Brown, The effect of inorganic salts on tricalcium silicate hydration, Cement and Concrete Research 16 (
1985) 17-22
35 D. Damidot, A. Nonant, P. Barret, Kinetics of tricalcium silicate hydration in diluted suspensions by
microcalorimetric measurements, Journal of American Ceramic Society 73 (2011) 19-22.
36 A.C. Lasaga, A. Luttge, Variation of crystal dissolution rate based on a dissolution step wave model, Science
291 (2001) 2400–2404.
37 P.M. Dove, N. Han, J.J.D. Yoreo, Mechanisms of classical crystal growth theory explain quartz and silicate
dissolution behavior, Proceedings of National Academics and Science U.S.A. 102 (43) (2005) 15357–15362.
38 A.C. Lasaga, A.E. Blum, Surface chemistry, etch its and mineral water reactions, Geochimica et
Cosmochimica Acta 50 (1986) 2363–2379.
39 J.W. Cahn, The kinetics of grain boundary nucleated reactions, Acta Metallurgica 4 (1956) 449–459.
40 M. Avrami, Kinetics of phase change. I, Journal of Chemical Physics 7 (1939) 1103–1112.
41 J. W. Christian, The Theory of Transformations in Metals and Alloys, Part 1, 3rd edition, Pergamon Press,
Oxford, 2002
42 J. J. Thomas and H. M. Jennings, ‘‘Effects of D 2O and mixing on the early hydration kinetics of tricalcium
Silicate,’’ Chem. Mater., 11, 1907–14 (1999).
43 J.J. Thomas, A.J. Allen, H.M. Jennings, Hydration kinetics and microstructure development of normal and
CaCl2-accelerated tricalcium silicate (C3S) pastes, Journal of Physical Chemistry C 113 (2009) 19836–19844.
44 M. Costoya, Kinetics and microstructural investigation on the hydration of tricalcium silicate, Doctoral
Thesis, École Polytechnique Fédérale de Lausanne, Switzerland, 2008
45 R. Kondo, S. Ueda, Kinetics of hydration of cements, Proceedings of the 5th international symposium on
chemistry of cement, Tokyo, 1968, pp. 203–248.
46 S. Garrault, T. Behr, A. Nonat, Formation of the C–S–H layer during early hydration of tricalcium silicate
grains with different sizes, Journal of Physical Chemistry 110 (2006) 270–275.
47 V.K. Peterson, M.C.G. Juenger, Hydration of tricalcium silicate: effects of CaCl 2 and sucrose on reaction
kinetics and product formation, Chem. Mater. 18 (2006) 5798–5804.
48 H.M. Jennings, A model for the microstructure of calcium silicate hydrate in cement paste, Cement and
Concrete Research 30 (2000) 101–116.
49 P.D. Tennis, H.M. Jennings, A model for two types of calcium silicate hydrate in the microstructure of
Portland cement pastes, Cement and Concrete Research 30 (2000) 855–863.
50 T.C. Powers, Physical properties of cement paste, Proceedings of the Fourth International Symposium on
Chemistry of Cement, Washington, USA, 1960, pp. 577–613.
51 P.J. McDonald, J. Mitchell, M. Mulheron, L. Monteilhet, J.-P. Korb, Two-dimensional correlation relaxation
studies of cement pastes, Magnetic Resonance Imaging 25 (2007) 470–473.
52 W.P. Halperin, Jyh-Yuar Jehngb, Yi-Qiao Song, Application of spin–spin relaxation to measurement of
surface area and pore size distributions in a hydrating cement paste, Magnetic Resonance Imaging 12 (1994)
169–173.
53 I.G. Richardson, The nature of C–S–H in hardened cements, Cement and Concrete Research 29 (1999) 1131–
1147.
54 P.C. Mathur, Study of cementitious materials using transmission electron microscopy, Doctoral Thesis, École
Polytechnique Fédérale de Lausanne, 2007.
55 J.M Pommersheim and J.R Clifton J.R., Mathematical modeling of tricalcium silicate hydration. II.
Hydration sub-models and the effect of model parameters, Cement and Concrete Research, 12 (1982) 765-772
56 K. Maekawa, T. Ishida and T. Kishi, Multi-scale modeling of concrete performance - integrated material and
structural mechanics, Journal of Advanced Concrete Technology 1 (2003) 91-126
57 H.M. Jennings, S.K. Johnson, Simulation of microstructure development during
the hydration of a cement compound, Journal of American Ceramic Society 69 (1986) 790–795.
58 K. van Breugel, Numerical simulation of hydration and microstructural development in hardening cement
paste (II): applications, Cement and Concrete Research 25 (1995) 522–530.
59 G. Ye, K. van Breugel, A.L.A. Fraaij, Three-dimensional microstructure analysis of numerically simulated
cementitious materials, Cement and Concrete Research 33 (2003) 215–222.
60 D.P. Bentz, E.J. Garboczi, A digitized simulation model for microstructural development, Ceram. Trans. 16
(1991) 211–226.
61 D.P. Bentz, Three-dimensional computer simulation of cement hydration and microstructure development,
Journal of American Ceramic Society 80 (1997) 3–21.
62 D.P. Bentz, O.M. Jensen, A.M. Coats, F.P. Glasser, Influence of silica fume on diffusivity in cement-based
materials I. Experimental and computer modeling studies on cement pastes, Cem. Concr. Res. 30 (2000) 953–
962.
63 J.M. Torrents, T.O. Mason, E.J. Garboczi, Impedance spectra of fiber-reinforced cement-based composites: a
modeling approach, Cem. Concr. Res. 30 (2000) 585–592.
64 K.A. Snyder, J.W. Bullard, Effect of Continued Hydration on the Transport Properties of Cracks through
Portland Cement Pastes in a Saturated Environment: A Microstructural Model Study, NIST IR 7265, U.S.
Department of Commerce, Washington, DC, 2005.
65 J.W. Bullard, A three-dimensional microstructural model of reactions and transport in aqueous mineral
systems, Modelling and Simulations in Materials Science and Engineering 15 (2007) 711–738.
66 J.W. Bullard, E. Enjolras, W.L. George, S.G. Satterfield, J.E. Terrill, A parallel reaction-transport model
applied to cement hydration and microstructure development, Modeling and Simulations in Materials Science
and Engineering 15 (2010) 711–738.
67 P. Navi, and C. Pignat, Simulation of cement hydration and the connectivity of the capillary pore space,
Advanced Cement Based Materials, 4 (1996) 58-67
68 C. Pignat, Simulation numérique de l'hydration du silicate tricalcique, charactérisation de la structure
poreuse et de la perméabilité, Doctoral Thesis, École Polytechnique Fédérale de Lausanne, 2003
69 P. Navi and C. Pignat, Three-dimensional characterization of the pore structure of a simulated cement paste,
Cement and Concrete Research, 29 (1999) 507-514
70 S. Bishnoi, K.L. Scrivener, μic: a new platform for modelling the hydration of cements, Cement and Concrete
Research 39 (2009) 266–274.
71 J. W. Bullard, H. M. Jennings, R. A. Livingston, A. Nonat, G. W. Scherer, J. S. Schweitzer, K. L. Scrivener,
J. J. Thomas, Mechanisms of cement hydration, Cement and Concrete Research (2010),
doi:10.1016/j.cemconres.2010.09.011
72 P. Juilland, E. Gallucci, R. Flatt, K. L. Scrivener, Dissolution theory applied to the induction period in alite
hydration, Cement and Concrete Research 40 (2010) 831-844.
73 A.C. Lasaga, A. Lüttge, Variation of crystal dissolution rate based on a dissolution stepwave model, Science
 291 (2001) 2400–2404.
74 A. Lüttge, Crystal dissolution kinetics and Gibbs free energy, Journal of Electron Spectroscopy and Related
Phenomena 150 (2006) 248–259
75 J.J Thomas, A New Approach to Modeling the Nucleation and Growth Kinetics of Tricalcium Silicate
Hydration, Journal of American Ceramic Society 90 (2007) 3282-3288
76 S. Bishnoi, K. L. Scrivener, μic : Studying nucleation and growth kinetics of alite hydration using μic, Cement
and Concrete Research 39 (2009) 849-860.
77 J.J Thomas, J.J Biernacki, J W Bullard, S. Bishnoi, J S Dolado, G. W Scherer, A. Luttge, Modeling and
simulation of cement hydration kinetics and microstructure development, Accepted for publication in Cement
and Concrete Research, 2010
78 S. Bishnoi, K. L. Scrivener, μic : A new platform for modelling the hydration of cements, Cement and
Concrete Research 39 (2009) 266-274.
79 S. Garrault, A. Nonat, Hydrated layer formation on tricalcium and dicalcium silicate surfaces: experimental
study and numerical simulations, Langmuir 17 (2001) 8131–8138.
80 R. Kondo, S. Ueda, Kinetics of hydration of cements, Proceedings of the 5th international symposium on
chemistry of cement, Tokyo, 1968, 203–248.
81 D.P Bentz, Modelling cement microstructure : pixels, particles, and property prediction, Materials
and Structures 32 (1999) 187-195.
82 J. W Bullard, A determination of hydration mechanisms for tricalcium silicate using a Kinetic Cellular
Automaton Model, Journal of American Ceramic Society 91 (2008) 2088-2097
83 K. van Breugel, Numerical Simulation of hydration and microstructure development in hardening cement-
based materials (1) : Theory, Cement and Concrete Research 25 (1995) 319-331.
84 J.W. Bullard, A Three-Dimensional microstructural model of reactions and transport in aqueous mineral
systems, Modelling and Simulation in Materials Science and Engineering 15 (2007 ) 711-738
85 T. Knudsen, Modelling hydration of portland cement : the effect of particle size distribution, Proceedings of
the Engineering Foundation Conference, 1983.
86 M. M. Costoya, Effect of particle size on the hydration kinetics and microstructural development of
tricalcium silicate, Doctoral Thesis, Ecole Polytechnique Federale de Lausanne, Switzerland 2008
87 H.M. Jennings, L.J. Parrott, Microstructural analysis of hardened alite paste, Part II: Microscopy and
reaction products, Journal of Materials Science 21 (1986) 4053–4059.
88 K. Scrivener, The Microstructure of Concrete, Materials Science of Concrete I, J.P. Skalny. Ed (American
Ceramic Society, Columbus, Ohio) 1989, 127-161.
89 H. F. W. Taylor, Cement Chemistry, 1st edition, Thomas Telford Publishing, 1990.
90 I. Odler, Chapter 6: Hydration, setting and hardening of Portland cement, in LEA’s chemistry of cement and
concrete, Peter C.Hewlett (ed), Arnold, London, (2001) 241-298
91 J. W Bullard, A determination of hydration mechanisms for tricalcium silicate using a Kinetic Cellular
Automaton Model, Journal of American Ceramic Society 91 (2008) 2088-2097
92 M. S. Beig, A. Luttge, Albite dissolution kinetics as a function of distance from equilibrium: Implications for
natural feldspar weathering, Geochimica et Cosmochimica Acta 70 (2006), 1402- 1420
93 F. Bellmann, D. Damidot, B. Moser, J. Skibsted, Improved evidence for the existence of an intermediate
phase during hydration of tricalcium silicate, Cement and Concrete Research 40 (2010)
94 H.N. Stein, Thermodynamic considerations on the hydration mechanisms of Ca 3SiO5 and Ca3Al2O6, Cement
and Concrete Research 2 (1972) 167–177.
95 D. Damidot, F. Bellmann, B. Möser, T. Sowoidnich, Investigation of the early dissolution behaviour of C3S,
12th International Congress on the Chemistry of Cement, Montreal, Canada, 2007, 8.-13.7., paper W1-06.5.
96 Truesdell, AH, and Jones, BF. WATEQ, a computer program for calculating chemical equilibria of natural
waters, Journal of Research, U.S. Geological Survey 2 (1974) 233-274
97 D. Rothstein, J.J. Thomas, B. J. Christensen, H. M. Jennings, Solubility behavior of Ca-, S-, Al-, and Si-
 bearing solid phases in Portland cement pore solutions as a function of hydration time, Cement and Concrete
Research 32 (2002) 1663-1671.
98 F.P. Glasser, E.E. Lachowski, D.E. Macphee, Compositional model for calcium silicate hydrate (C-S-H) gels,
their solubilities, and free energies of formation, Journal of American Ceramic Society 70 (1987) 481 – 485.
99 B. Lothenbach, F. Winnefeld, Thermodynamic modelling of the hydration of Portland cement, Cement and
Concrete Research 36 (2006) 209 – 226
100D.L. Parkhurst, C.A.J. Appelo, PHREEQC 2: a Computer Program for Speciation, Batch-reaction,
One-dimensional Transport and Inverse Geochemical Calculations, USGS, Denver 2001.
101P. W. Brown, E. Franz, G. Frohnsdorff, and H. F. W. Taylor, Analyses of the aqueous phase during early
C3S hydration, Cement and Concrete Research 14 (1984) 257-262
102S.A. Greenberg, T.N. Chang, Investigations of the colloidal hydrated calcium silicates II. Solubility
relationships in the calcium oxide–silica–water system at 25 °C, The Journal of Physical Chemistry 69 (1965)
182–188.
103H.M. Jennings, Aqueous solubility relationships for two types of calcium silicate hydrate, Journal of
American Ceramic Society 69 (1986) 614 – 618.
104K. Fuji, W. Kondo, Heterogeneous equilibrium of calcium silicate hydrate in water at 30 °C, Journal of
Chemical Society, Dalton Transaction 2 (1981) 645–651.
105A. Atkinson, J.A. Hearne, C.F. Knights, Aqueous Chemistry and Thermodynamic Modelling of CaO – SiO 2 –
H2O Gels, AERE R 12548, UKAEA, 1987.
106B. Lothenbach, T. Matschei, G. Mőschner, F.P. Glasser, Thermodynamic modelling of the effect of
temperature on the hydration and porosity of Portland cement, Cement and Concrete Research, submitted
for publication.
107J.J Chen, J.J Thomas, H.F.W. Taylor, H.M. Jennings, Solubility and structure of calcium-silicate- hydrate,
Cement and Concrete Research 34 (2004) 1499-1519.
108P.W Brown, The effect of inorganic salts on tricalcium silicate hydration, Cement and Concrete Research 16
( 1985) 17-22
109M. M. Costoya, Effect of particle size on the hydration kinetics and microstructural development of
tricalcium silicate, Doctoral Thesis, Ecole Polytechnique Federale de Lausanne, Switzerland 2008
110T. Knudsen, Modelling hydration of portland cement : the effect of particle size distribution, Proceedings of
the Engineering Foundation Conference, 1983.
111K. G. McCurdy, Suspension hydration of C3S at constant pH. I. Variation of particle size and C 3S content,
Cement and Concrete Research 3 (1973) 247-262
112S. Bishnoi, K. L. Scrivener, μic : Studying nucleation and growth kinetics of alite hydration using μic,
Cement and Concrete Research 39 (2009) 849-860.
113A. Kumar, S. Bishnoi, K.L. Scrivener, Modelling early age hydration kinetics of alite, Accepted for
publication in Cement and Concrete Research 2011
114J.J. Thomas and H.M. Jennings, Effect of D 2O and mixing on the early hydration kinetics of tricalcium
silicate, Chemistry Materials, 11 (1999) 1907 - 1914
115P. Juilland, A. Kumar, E. Gallucci, R. Flatt, K.L Scrivener, Effect of mixing on the early hydration of
cementitous systems, Accepted for publication in Cement and Concrete Research
116A. Bazzoni, M.Cantoni, K. Scrivener, Early hydration of tricalcium silicate, 31 st Cement and Concrete Science
Conference, U.K
117P. Juilland, E. Gallucci, R. Flatt, K. L. Scrivener, Dissolution theory applied to the induction period in alite
hydration, Cement and Concrete Research 40 (2010) 831-844.
118A. Kumar, G. Sant, C. Patapy, C. Gianocca, K. L. Scrivener, An Experimental and Microstructural
Simulation Based Interpretation of Alite Hydration I: The Influence of Sodium and Potassium Hydroxides, In
progress for submission in Journal of American Ceramic Society
119W. Lerch, PCA Bulletin 12, (1946)
120I. Jawed, and J. Skalny, Alkalis in Cement: A Review, Cement and Concrete Research 8 (1978)37-52
121D.D Double, New developments in understanding the chemistry of cement hydration, Philosophical
Transactions of the Royal Society of London A 310 (1983) 53-66
122G.C. Edwards and R. L. Angstadt, The effect of some soluble inorganic admixture on the early hydration of
portland cement, Journal of Applied Chemistry 16 (1966) 166-168
123J.J. Thomas, and H.M Jennings, Free-energy based model of chemical equlibria in the CaO-SiO 2-H2O system,
Journal of the American Ceramic Society, 81(3), (1998) 606-612
124Thomas, N. L., and Double, D. D., ‘Calcium and silicon concentrations in solution during the early hydration
of portland cement and tricalcium silicate’, Cement and Concrete Research, 11, pp. 675-687, (1981)
125P.W. Brown, C.L. Harner, and E.J. Prosen, The effect of inorganic salts on tricalcium silicate hydration’,
Cement and Concrete Research 16 (1985) 17-22, (1985)
126J.F. Young, H.S. Tong, and R. L. Berger, Compositions of solutions in contact with hydrating tricalcium
silicate pastes, Journal of the American Ceramic Society, 60(5-6), (1977) 193-198, (1977)
127R. Kondo, M. Daimon, E. Sakai, H. Ushiyama, Influence of inorganic salts on the hydration of tricalcium
silicate, Journal of Applied Chemical Biotechnology 27 (1977) 191-197, (1977)
128D. E. Macphee, K. Luke, F. P. Glasser, and E.E. Lachowski, Solubility and aging of calcium silicate hydrates
in alakline solutions at 25C, Journal of the American Ceramic Society, 72(4), pp. 646-654, (1989)
129K. Suzuki, T. Nishikawa, H. Ikenaga, and S. Ito, Effect of NaCL or NaOH on the formation of C-S-H,
Cement and Concrete Research, (1986) 333-340, (1986)
130H. Stade, On the reaction of C-S-H with alkali hydroxides, Cement and Concrete Research, 19, (1989) 802-
810
131S. J. Way, and A. Shayan, Early hydration of a portland cement in water and sodium hydroxide solutions:
compositions of solutions and nature of solid phases, Cement and Concrete Research 19 (1989) 759-769
132A. G De La Torre, S. Bruque, J. Campo, M. A. G. Aranda, The superstructure of C3S from synchroton and
neutron powder diffraction and its role in quantitative phase analyses, Cement and Concrete Research 32
(2002) 1347-1356
133J.W. Bullard, R.J. Flatt, New Insights into the Effect of Calcium Hydroxide Precipitation on the Kinetics of
Tricalcium Silicate Hydration, Journal of American Ceramic Society 93 (2010) 1894-1903.
134A. Quennoz, Doctoral Thesis, PhD dissertation, École Polytechnique Fédérale de Lausanne, Lausanne,
Switzerland, 2011.
135E. M. Gartner, Structure and Performance of cements, 2nd edition, Chapter 3. Spon Press ed. 2002.
136Truesdell, AH, and Jones, BF. WATEQ, a computer program for calculating chemical equilibria of natural
waters, Journal of Research, U.S. Geological Survey 2 (1974) 233-274
137P. W. Brown, E. Franz, G. Frohnsdorff, and H. F. W. Taylor, Analyses of the aqueous phase during early
C3S hydration, Cement and Concrete Research 14 (1984) 257-262
138Alexandra, Doctoral Thesis, Hydration of C3A with calcium sulfate alone and in the presence of calcium
silicate, Ecole Polytechnique Federale de Lausanne, 2011
139R. Berliner, M. Popovici, K. W. Herwig, M. Berliner, H. M. Jennings and J. J. Thomas, Quasielastic Neutron
Scattering Study of the Effect of Water-to-Cement Ratio on the Hydration Kinetics of Tricalcium Silicate ,
Cement and Concrete Research, 28 (1998) 231-243.
140D.P Bentz, Influence of water-to-cement ratio on hydration kinetics: Simple models based on spatial
considerations, Cement and Concrete Research, 36(2) (2006) 238-244.
141J. Bensted, Early hydration of portland cement – effects of water/cement ratio, Cement and Concrete
Research, 13 (1983) 493-498.
142D. Damidot, A. Nonant, P. Barret, Kinetics of tricalcium silicate hydration in diluted suspensions by
microcalorimetric measurements, Journal of American Ceramic Society 73 (2011) 19-22.
143J.J. Thomas, H.M. Jennings, J.J. Chen, Influence of nucleation seeding on the hydration
mechanisms of tricalcium silicate and cement, J. Phys. Chem. C, 113, 4327-4334 (2009).
144W.A. Corstanje, H.N. Stein, J.M. Stevels, Hydration reactions in pastes C 3S + C3A + CaSO4.2aq + H2O at
25 °C, Cement and Concrete Research 3 (1973) 791–806
145A. Quennoz, Hydration of C 3A with calcium sulfate alone and in the presence of calcium silicate, Doctoral
Thesis, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland, 2011.
146H. Minard, S. Garrault, L. Regnaud, A. Nonant, Mechanisms and parameters controlling the tricalcium
aluminate reactivity in the presence of gypsum, Cement and Concrete Research 37 (2007) 1418-1426.
147S. Pourchet, L. Regnaud, J.P. Perez, A. Nonat, Cement and Concrete Research 39 (2009) 989-996
148H. F. W. Taylor, Cement Chemistry, 1st edition, Thomas Telford Publishing, 1990.
149V. Kocaba, Development and evaluation of methods to follow microstructural development of cementitous
systems including slags, Doctoral Thesis, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland,
2011.
 Aditya Kumar

Graduate Research Assistant 10401 Wilshire Blvd., Los Angeles, CA-90024

Swiss Federal Institute of Technology (EPFL) Mobile: +1-310-433-1827
Materials Science and Engineering (EDMX) Email: adityaku@ucla.edu

Education
Swiss Federal Institute of Technology, Lausanne, Switzerland (2008 - 2012)
• Doctor of Philosophy in Materials Science and Engineering
• Graduate Advisor: Professor Karen L. Scrivener
• Dissertation Title: Modeling hydration kinetics of cementitous systems

Georgia Institute of Technology, Atlanta, U.S.A (May 2007- July 2008)

• Masters Project in Department of Materials Science and Engineering

Indian Institute of Technology, Kanpur, India (2003-2007)

• Bachelor of Science in Materials Science and Engineering

Research Experience
Swiss Federal Institute of Technology, Lausanne, Switzerland (2012)
• Ph.D student in Materials Science and Engineering
• Studying the influence of various process parameters on the hydration of alite : salt-
solutions, inert fillers, particle sizes, mixing and water content
• Simulation of hydration of aluminate-gypsum systems studying the effect of composition
particle sizes and water content
• Development of a 3D microstructural model to study the microstructural evolution of
hydrating cement paste

Georgia Institute of Technology, Atlanta, U.S.A (June 2007 – July 2008)

• Masters Project in Packaging Research Center in Department of Materials Science and
Engineering

Georgia Institute of Technology, Atlanta, U.S.A (May 2006 – July 2006)

• Summer Internship in Packaging Research Center in Department of Materials Science
and Engineering
Indian Institute of Technology, Kanpur, India (August 2006 – May 2007)
• Undergraduate research assistant in Materials Science and Engineering
• Determination of optimum parameters for plasma nitriding of mild steel to achieve
optimum hardness and corrosion-resistance properties

Skills
Computing
• Programming in C++
• Programming in Java
• Linux, MacOSX and Windows
Laboratory Techniques
• Electron Microscopy
• X-Ray diffraction
• Dynamic Vapor Sorption
• Isothermal Conduction Calorimetry
• Thermo-gravimatric analyses

Publications
• A. Kumar, S.Bishnoi and K. L. Scrivener, Modelling early age hydration kinetics of alite,
Accepted for publication in Cement and Concrete Research, 2010
• P. Juilland, A. Kumar, E. Gallucci, and K.L. Scrivener, Effect of mixing on early hydration
of cementitous systems, Accepted for publication in Cement and Concrete Research, 2011
• A. Kumar and K. L. Scrivener, Modelling hydration kinetics of multi-phase cementitous
systems, CONMOD 2010, Lausanne, Switzerland
• A. Kumar and K. L. Scrivener, Modelling hydration kinetics of multi-phase cementitous
systems, ICCC 2011, Madrid, Spain
• A. Kumar and K. L. Scrivener, Modelling hydration kinetics of multi-component
cementitous systems, Cement and Concrete Science 2010, Birmingham, U.K
• A. Kumar, G. Sant and K. L. Scrivener, Modelling water vapor isotherms of cementitous
materials using a microstructural model, CONMOD 2010, Lausanne, Switzerland
• A. Kumar, G. Sant, C. Patapy, C. Giannocca and K.L Scrivener, Modelling the effect of
sulfate salts addition on alite hydration, In progress for submission in Cement and Concrete
Research
• A. Kumar, G. Sant, C. Patapy, C. Giannocca and K.L Scrivener, Modelling the effect of
alkali hydroxides addition on alite hydration, In progress for submission in Cement and
Concrete research