3.1 Numerical simulation

The microstructures of the six cement pastes were simulated using the CEMHYD3D program. This model operates by a sequence of steps : dissolution, random-walker diffusion of the mobile agents, and reaction between colliding pixels. One complete sequence is called a cycle of hydration. The simulated microstructures were representative of the real microstructures, in that the hydration cycle number was fixed in such a way that the simulated CH volume fraction was chosen to be as close as possible to that found in experiment (Kamali 2003). First, simulations of 5000 hydration cycles were performed. Then, as shown in Figure 2, a cycle number was chosen so that the numerical CH content evolution agreed within two or three significant figures with the experimental CH content value. Table 5 presents the number of cycles identified for each cement paste and a comparison between simulated CH content values and those measured experimentally. In paste number 6, the 5000 cycles of hydration used, as well as model inaccuracies in simulating the effect of silica fume, made it not possible to obtain such close agreement with the CH content. A unit cell 100 µm in size was used for all the simulations.

Figure 2: Evolution of the principal hydrate volume fractions with the number of hydration cycles (for portland cement paste at w/c=0.4).

Once the 3D microstructures of unleached hydrated cement pastes were built, the leaching of CH, AFm and AFt were simulated. Figure 3 presents 2D images obtained from 3D numerical microstructures of Portland cement paste with w/c = 0.4, before hydration, after hydration and after CH dissolution. It is clearly shown that the dissolution of CH (shown in white) increases the porosity (shown in black).

Figure 3. 2D images obtained from 3D microstructures of portland cement paste with w/c=0.4 before hydration, after hydration and after CH dissolution. The images are 100 µm in width.