25
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
mechanical interactions between particles (e.g. U
1rx
). The proposed scheme therefore
combines chemistry and mechanics over the large timescales of chemical kinetics.
3. Results
The KMC approach presented here has been used to study the effect of the interaction
potential and initial system configuration on the mechanism of growth-by-aggregation of the
cement HP. Here we are interested only in the qualitative trends of rates as a function of time,
and how these relate to the mechanisms of precipitation that are triggered. Converting the
qualitative results into quantitative rates would require a model calibration to define the r
0
*
terms in Equation 1; this is left to another manuscript currently under review [21] The results
for three scenarios are shown in Figure 2:
a)
very large in Equation 2, taken from ref. [18] where it was shown that such a large
captures well the mechanics of the C—S—H; in this first case the interaction between
cement surface particles and cement HP particles is taken to be as strong as the
interaction between two HP particles;
b)
same as the previous scenario, but this time the between cement surface and HP
particles is smaller than the that between two HP particles (25% of it to be precise)
except for a small region in the middle of the cement surface where a full 100% of the
HP-HP interaction is maintained; this small region serves as a preferential site for
nucleation of the mesoscale HP domain;
c)
a 3 orders of magnitude smaller , taken from atomic force microscopy experiments
[30].
Figure 2: Results from our simulations with different initial conditions (a) strong
interactions between nanoparticles and homogeneous cement surface; (b) strong interaction
and surface with one preferential site for HP particle formation; (c) weak interactions.
300 nm
26
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
The snapshots in Figure 2 show that the first scenario leads to a layered growth of the HP
domain, formed by aggregated nanoparticles. For this scenario we kept constant in time and
very large, in order to avoid particle deletion. The layered growth results into a constant rate,
very different from what is expected during early hydration (see simulation and experimental
values in Figure). Notice that the size of the HP domain simulated here is in the same order as
that observed experimentally during early hydration, i.e. ~500 nm, so it makes sense to aim
for a rate curve that is similar to the one measured experimentally from calorimetry).
The second scenario, with the favourable nucleation site, gives a much better result, with a
roughly hemispherical domain growing radially. The increasing surface of the hemispherical
domain is at the origin of the acceleration in the rate curve. Indeed, in the KMC algorithm the
total rate is the sum of all particle insertion rates, and an increasing surface of the HP domain
implies that more trial nuclei can find a favourable spot that leads to a high rate. When the HP
domain crosses the periodic boundary on the cement surface, the lateral impingement causes
the rate peak, similar to the calorimetry experiments. The continuous decrease of rate after the
peak is due to the fact that, for this simulation, we took a supersaturation that decreases with
time, as known for the cement solution and as quantified in ref. [31].
Finally, the third scenario with much weaker strength of the interactions causes the
mechanisms of HP formation to change from heterogeneous on the cement surface to
homogeneous in the bulk. This also leads to a rate curve that agrees qualitatively with the
experiments. In this case however the peak occurs when the space gets filled, which implies a
quantitative dependence of the area under the rate curve (number of HP particles inserted) and
the amount of water in the cement mix (viz. the size of the simulation box perpendicular to
the cement surface). Such dependence would contradict the experiments [8], but the problem
can be avoided assuming that the homogeneous nucleation is limited to a small region near
the surface of the cement grain, as proposed in ref. [20]. This constraint might indeed
originate from a gradient of supersaturation in the simulation box, sustained by a diffusive
process of the ions in solution, which move from the cement surface toward the bulk solution.
To explore this possibility, future work should couple the simulations presented here with the
type or reaction-transport simulations mentioned in the introduction of this manuscript [7,8].
4. Conclusion
We have presented a new approach to simulate the precipitation of nanoparticle of cement
hydrates from aqueous solution and their aggregation to form mesoscale domains. The
approach is based on new coarse-grained expressions for the rate of particle insertion and
deletion, which are derived from the theories of classical nucleation and crystal growth. The
rate expression involve one kinetic parameter r
0
*
which has physical meaning but is often hard
to measure. r
0
*
determines
a linear scaling of time, which means that even if it is not
quantified, all the nonlinearities in the rate of HP formation obtained from the simulations are
still a true reflection of the predicted mesoscale mechanism only. Another important feature
of our coarse-grained rates is that they combine chemistry and the mechanical interactions
between nanoparticles, creating new opportunities to study technologically important
phenomena where chemistry and mechanics play together, e.g. crystallization pressure and