23
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
Figure 1: (a) Layer of particles representing a cement grain surface, in red, and hydration
phase HP particles formed near it, in blue. (b) Same configuration showing also the trial
particle for KMC insertion in yellow.
The rate of particle insertion is coarse grained from the molecular scale following the classical
theories of Classical Nucleation Theory (CNT) and crystal growth [24]:
(1)
Details on the derivation of Eq. 1 and on the corresponding equation for the deletion rate R
del
are given in ref. [25] (notice that the chemical conditions considered for the simulations in
this paper will be such that the number of dissolution events will be negligible alongside their
effect on the rate, so the results that we will show will be reproducible even without
implementing a deletion step at all). In Eq. 1, C
cn
is the concentration of possible particle
nucleation sites in solution (e.g. the concentration of ions), V
box
is the volume of the
simulation box, and M is the number of trial particles assumed as uniformly distributed in
V
box
. In the curly bracket, the first term is the characteristic time to form a critical nucleus as
per CNT, while the second integral term is the time to grow the radius of the critical nucleus
by single-molecule growth reactions until reaching the wanted particle size R
part
(here
diameter = 10 nm, thus R
part
= 5 nm).
Z is the Zeldovich factor, expressing the probability that
a critical nucleus will indeed start growing rather than dissolving back. G
CNT
is the
difference in free energy between critical nucleus and solution. For spherical particles:
(2)
R
nucl
is the radius of the critical nucleus, given by the condition d G
CNT
/ dR = 0, k
B
is the
Boltzmann constant,
T is the temperature in Kelvin degrees, and is
the interfacial energy
between cement HP and water. U
nucl
is the change in total interaction energy in the system
in case the critical nucleus appears; its relationship to the nucleus size depends on the type of
interaction potential employed. Here we use a pairwise interaction potential U
ij
(r) that
depends only on the distance r between the particles and has been shown to capture well the
mechanical properties of cement HP at the 500 nm mesoscale [13, 18, 23]:
24
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
(3)
(R) is the energy strength, which is assumed to scale as the volume of the particle. A
different hypothesis on the scaling of is discussed in ref. [25], in relation to the prediction of
the rate of early cement hydration.
Going back to Equation 1, a is the linear size of a molecule of cement HP (here we assume a
= 0.645 nm, thus
a
3
= 0.267 nm
3
, as for C—S—H [26]). r
0
*
is a kinetic constant (rate per unit
surface) that contains the activity coefficient of the activated complex and the standard free
energy barrier of the C—S—H formation reaction [24].
1rx
and U
1rx
are the change of
particle surface and interaction energy caused by one molecular reaction of HP product
forming on the surface of a growing nanoparticle. Each molecular reaction changes the
particle volume by a
3
and consequently the radius R; both
1rx
and U
1rx
therefore depend
on the current R. The rate of particle deletion can be obtained in a very similar manner as R
in
and contains neither the supersaturation term nor the sampling-related prefactor before the
curly brackets. Deletion will be considered in the simulation but the rate expression is omitted
here for brevity and because we will only consider values of beta that are sufficiently high for
deletion to be negligible.
To simulate the formation of cement HP, we consider = 86.7 mJ/m
2
[26]. The interpretation
of experiments based on CNT indicates that the size of the critical nucleus in cement hydrates
is as small as one single molecule [27]. This means that the CNT term in Equation 1 is small
compared to the crystal growth term (the integral one) and we will neglect it. For cement
hydration, both C
cn
and r
0
*
are unknown, hence the two will be combined and treated as a
single effective parameters that is used to fit the right timescale (actually, if one decides to
identify C
cn
with the concentration of calcium or silicon ions in solution, then literature data
could be used to set a value for it, e.g. [28]). It is important to notice that differently from the
existing simulations, the only scaling of time here is linear. This means that any nonlinearity
in the rate can only be a true reflection of the formation mechanism and is not imposed ad-
hoc.
The particles of cement hydrates that we insert during the simulations are monodisperse even
though just after insertion, a small random change of the diameter by 5% is imposed to avoid
crystallization. It is known in the cement literature that realistic details of the heterogeneities
at the sub-micrometer scale can only be captured using polydiperse nanoparticles [16, 23],
and there are experiments that suggest that anisotropic shapes of the particles are more
realistic [29]. However, spherical and monodisperse particles are sufficient to discuss the
qualitative kinetics of collective precipitation-aggregation mechanisms, as shown in the
literature [19] and in line with the objective of this work.
As a final note for this section, it is interesting to note that the coarse grained rate expression
in Equation 1 combines some terms that are related to the chemical kinetics of the process
(e.g. r
0
*
) with other terms that are related to the thermodynamics. Among the latters, there are
terms that are related to the chemistry of the system (e.g. ) and terms that are related to the