21
International RILEM Conference on Materials, Systems and Structures in Civil Engineering
Conference segment on Service Life of Cement-Based Materials and Structures
22-24 August 2016, Technical University of Denmark, Lyngby, Denmark
NANOSCALE SIMULATIONS OF CEMENT FORMATION AND
STRUCTURAL EVOLUTION: A NEW KINETIC APPROACH
Enrico Masoero
(1)
, Igor Shvab
(1)
(1) Newcastle University, Newcastle upon Tyne, United Kingdom
Abstract
The formation and degradation of cementitious materials are largely controlled by the
nanostructural evolution of hydration products. Modelling such evolution across multiple
length and time scales is a great challenge. Here we present a new kinetic approach to
simulate the nucleation, dissolution, and aggregation of cement hydrates nanoparticles. The
approach is based on a Kinetic Monte Carlo algorithm in which the rates of the transitions are
obtained via a new coarse-graining procedure. The rates account for free energy changes due
to both mechanical interactions and chemical reactions. The methodology is able to address
the long timescale of cement formation and captures various possible mechanisms of
nucleation and growth of the hydrates. By coupling chemistry, mechanics, and long
timescales, this work is a first step towards simulating cement hydration and degradation.
1. Introduction
The hardened cement paste is the glue of concrete. It forms upon chemical reaction between
dry cement powder and water [1]. This hydration process leads to the precipitation from ionic
solution of several hydrated phases (HP). The HP progressively fills the space and induces the
liquid-to-solid transition known as setting. In ordinary cement pastes, the main HP is calcium-
silicate-hydrate (C—S—H) and at least 50% of the total hydration reaction takes place during
the first 24 hours after mixing dry cement with water. Setting typically occurs during this
stage, which is known as “early hydration” [2].
There is a considerable scientific and technological interest in controlling the early hydration
and setting of the cement paste. In the last decades, this led to a number of models and
simulation approaches whose target is to predict the early hydration based on the chemistry
and mix design proportions of the paste. These models consider length scales above the m
and use a combination of thermodynamics and chemical kinetics in order to reproduce the
22
International RILEM Conference on Materials, Systems and Structures in Civil Engineering
Conference segment on Service Life of Cement-Based Materials and Structures
22-24 August 2016, Technical University of Denmark, Lyngby, Denmark
microstructural evolution of the paste [3-8]. Their results can fit well the rate of early
hydration from isothermal calorimetry experiments. These models however have the
limitation of considering the HP as homogenous domains, without information on the
underlying texture at sub/micrometre length scales (except for user-defined parameters that
sometimes are employed to mimic the anisotrpic formation of needles and foils [9]). Missing
the sub-micrometre texture is a particularly limiting for the C—S—H hydration product,
whose network of mesopores with size 1-50 nm dictates largely the macroscopic response to
humidity cycles and creep [10-14].
Simulating the formation of HP at the sub-micrometre level is a challenging task. In the last
decade, several authors have shown that aggregating nano-units of 1-10 nm lead to model
structures of the cement HP that display many of the experimentally measured structural
features and mechanical properties [15-23]. Some of these models have also attempted to
describe the process of HP formation by precipitation from solution. However, the timescale
of cement hydration is of the order of 24 hours while nanoparticle simulations are limited to
dynamic processes in the timescale of nano-to-micro seconds or, on the other extreme, to
equilibrium studies in the infinite-time limit. To overcome this limitation, the simulations to
date have either introduced constraints on the mechanisms of particle aggregation, or applied
ad-hoc nonlinear mapping between simulation steps and time [17, 19-21]. None of the
simulations to date can predict the mechanisms of HP nanoparticle formation and aggregation
directly from the chemistry of the aqueous solution.
In this work, we propose a new approach to simulate the formation of cement HP at the
mesoscale of 1-to-500 nm. Our approach is based on Kinetic Monte Carlo (KMC) simulations
of nanoparticle insertion and deletion. The rates of insertion and deletion are calculated using
a new coarse graining scheme that considers both Classical Nucleation Theory (CNT) and
crystal growth theory. First results show that the rates can address the long timescale of
cement hydration and consider the interplay between chemical driving force and mechanical
interactions between the nanoparticles.
2. Methodology
The simulations of HP formation start with an orthogonal box that is empty except for a 30
nm thick layer that represents the surface of a cement grain (see Figure 1). The cement layer
is discretized using 10 nm spherical particles that are fixed, i.e. not displaced nor removed
during the simulations. It is implicitly assumed that the box is filled with aqueous ionic
solution, whose supersaturation with respect to HP formation is known.
After a certain number of KMC steps, a certain number of HP spherical particles will have
formed in the box (see Figure 1). The next KMC step computes first the rates of all possible
particle deletions and then the rates of all possible particle insertions. The insertion is tricky
because there are infinite possible positions for a new particle. In order to manage this, we
create many trial HP particles and allow them to move a bit in order to find a local minimum
of interaction energy with the other existing particles (HP and layer; no interactions between
two trial particles, because they do not exist yet; see Figure 1).