Understanding cementitious backfill interactions with groundwater components

Abstract

Listen

Translate article

Abstract Many concepts for the disposal of intermediate-level radioactive waste specify the emplacement of waste packages in underground vaults backfilled with cementitious material. The cementitious backfill performs a number of functions including maintaining alkaline conditions and providing a high sorption capacity for radionuclides. In the UK, the Nirex Reference Vault Backfill (NRVB) has been developed for such a role. To improve our understanding of its long-term performance, a detailed programme was undertaken to investigate interactions of the NRVB with selected groundwater components using sodium sulphate and magnesium chloride solutions. Experiments using the selected solutions were performed under conditions of constant flow, over timescales of up to 860 days. Elemental composition and pH of the egressing eluent were determined throughout the experiments. At the end of the experiments, detailed solid analyses were performed on samples from various locations in the NRVB using a wide range of techniques. In the sodium sulphate experiment, sulphate attack resulted in the formation of ettringite, thaumasite and gypsum. For the magnesium chloride case, the results were dependent on the concentration of the magnesium in the ingressing solution. At higher concentration, precipitation of brucite at the upstream surface of the NRVB led to clogging and termination of the experiment after about one month. At lower concentration, no such clogging occurred, despite almost all the ingressing magnesium being removed by the NRVB over timescales of more than two years. For both systems, high pH values of the eluent were maintained over the full periods studied; for the sodium sulphate experiment the final pH was 12.7 (after 180 sample volumes) and for the low concentration magnesium chloride case pH 11.5 (after 749 sample volumes). A “top-down” modelling approach was applied to interpret the results, involving solution equilibration with the NRVB represented by a simple set of mineral phases. Further processes were modelled only where they improved the overall fits. The resulting models were considered to be simple while maintaining consistency with the major features of the experimental data sets. This has helped to provide mechanistic interpretations of the evolution of NRVB, based on relatively simple sets of secondary mineral phases, early kinetic control of mineral dissolution and diffusion-restricted release of portlandite at longer times.