Testing the Grambow glass dissolution model by comparing it with Monte Carlo simulation results

We have developed a Monte Carlo (simulation) model to describe glass dissolution. This model uses similar parameters to those used in the Grambow model, which has been the main glass dissolution model during the last decade. While the Grambow model is macroscopic, ours is microscopic. This assures that the value of our parameters is time independent and allows changes in the gel structure (not just concentrations) to be described. We relate our parameters to those of the Grambow model and test whether the basic assumptions of the Grambow model are consistent with our simulation results. In these simulations, as the solution evolves towards silica saturation, the silica concentration in solution can become higher than the (final) silica saturation concentration. A possible explanation for this behavior, which is observed experimentally as well, is that the silica saturation concentration is not constant as a function of time. Initially, the silica saturation concentration is the glass saturation concentration. The glass saturation concentration is higher than the final silica saturation concentration, which corresponds to saturation of the gel. In the Grambow model, the initial dissolution rate (at the gel/water interface) and the diffusion coefficient of silica in the gel are not constant either. The existence of a final dissolution rate depends on the silica content of the glass and on the protection provided by the gel layer. In the simulations, a protective gel is mainly formed by the adsorption (precipitation) of dissolved silica particles from the solution in contact with the gel. This makes us expect that (1) the gel would be non protective for dynamic tests (with a low silica content in solution) and (2) the gel would be very protective in static tests with high surface to volume ratios.