Buffering Capacity of pH in Backfill

Gravel backfills may help retard the migration of hyperalkaline cement pore fluids in repository designs containing large amounts of cement and concrete (e.g. the proposed Swedish SFL 3-5 repository) by acting as a 'sacrificial' reactive barrier. This behaviour relies on the reaction of silicate and aluminosilicate minerals in the backfill through hydroxyl ion-catalysed mineral dissolution reactions and the associated precipitation of hydroxyl ion bearing solids, such as calcium silicate hydrates. Recent work published by SKB on potential reaction mechanisms and assessment of the potential performance of such backfills has been reviewed, and scoping calculations to assess likely backfill performance have been carried out. The approaches and methodologies employed by SKB to model the potential for buffering of hydroxyl ions released by cementitious engineered barriers by reaction with a surrounding gravel barrier are considered to have a number of deficiencies. In particular, mass balance calculations used a non-conservative estimate of the amount of hydroxyl ions which may be consumed by precipitation of CSH minerals. More conservative choices of the composition of CSH suggest that complete reaction of at least 36 % of the gravel barrier (as currently designed), and possibly much more, would be required to contain the release of all hydroxyl ions contained within cement in SFL 3-5. Also, SKB's scoping calculations overestimate the amount of quartz/SiO2 likely to be present in the gravel backfill, thus leading to an overestimation of the likely amount of retardation of hydroxyl ion migration through the backfill. Moreover, SKB's calculations assumed that the reactive surface area of particles in the gravel backfill does not change with time and that the rate-limiting step of release of silica remains detachment of silicate ions from the mineral surfaces. However, evidence from SKB's own experiments suggests that the surface area available for dissolution will decrease with time due to the coating of surfaces by precipitates and it is likely that the rate limiting step for release of Si (and hence OH- consumption) ultimately becomes diffusion through an alteration layer of precipitates surrounding rock fragments. Modelling of quartz-water reactions at the particle-scale carried out in this study using different plausible models for the dissolution of quartz has revealed that the total amount of buffering of the cement water provided by diffusion into, and reaction in, the particle pore space is small and is not enough to satisfy the amounts of buffering claimed by SKB. In every model case, the interior of the particle becomes shielded from the cement water by a layer of CSH that precipitates in the pore space adjacent to the surface, thus removing the interior of the particle from further reaction