A comparative study of the modelling of cement hydration and cement–rock laboratory experiments

The use of cement and concrete as fracture grouting or as tunnel seals in a geological disposal facility for radioactive wastes creates potential issues concerning chemical reactivity. From a long-term safety perspective, it is desirable to be able model these interactions and changes quantitatively. The ‘Long-term Cement Studies’ (LCS) project was formulated with an emphasis on in situ field experiments with more realistic boundary conditions and longer time scales compared with former experiments. As part of the project programme, a modelling inter-comparison has been conducted, involving the modelling of two experiments describing cement hydration on one hand and cement–rock reaction on the other, with teams representing the NDA (UK), Posiva (Finland), and JAEA (Japan).This modelling exercise showed that the dominant reaction pathways in the two experiments are fairly well understood and are consistent between the different modelling teams, although significant differences existed amongst the precise parameterisation (e.g. reactive surface areas, dependences of rate upon pH, types of secondary minerals), and in some instances, processes (e.g. partition of alkali elements between solids and liquid during cement hydration; kinetic models of cement hydration). It was not conclusive if certain processes such as surface complexation (preferred by some modellers, but not by others) played a role in the cement–rock experiment or not. These processes appear to be more relevant at early times in the experiment and the evolution at longer timescales was not affected. The observed permeability profile with time could not be matched. The fact that no secondary minerals could be observed and that the precipitated mass calculated during the simulations is minor might suggest that the permeability reduction does not have a chemical origin, although a small amount of precipitates at pore throats could have a large impact on permeability.The modelling exercises showed that there is an interest in keeping the numerical models as simple as possible and trying to obtain a reasonable fit with a minimum of processes, minerals and parameters. However, up-scaling processes and model parameterisation to the timescales appropriate to repository safety assessment are of considerable concern. Future modelling exercises of this type should focus on a suitable natural or industrial analogue that might aid assessing mineral–fluid reactions at these longer timescales.

► A modelling intercomparison has shown that the dominant reaction pathways are well understood.
► However, significant differences in model parameterisation produced similar results.
► The modelling showed that there is benefit in keeping the numerical models as simple as possible.