Application of portland cement-based materials to radioactive waste immobilization

Abstract

The immobilization potential of cement may be either physical or chemical. Physically it may act as a barrier and chemically as a selective binder for radwaste species. The latter aspect is believed to assume increasing importance at long ages. The nature of cement and modified cement matrices are explored to explain their chemical binding potential. It is shown that Portland cements are very similar worldwide but supplementary materials such as fly ash and slag are less well-specified. The main immobilizing potential of cement systems comes from their high internal pH allowing precipitation of many nuclides as hydroxides. In slag systems, the low redox potential can also reduce the solubility of some nuclides. In order to establish a predictive capability to determine the time-dependence of the internal state of cements, it has proven necessary to develop physico-chemical models of cement performance. The state of modelling is reviewed. Significant advances have been made in recent years, but there remain several problems, one of which is validation. The present database still requires a considerable input with respect to specific radwaste-cement interactions which can only come from experimentation. Experimental work can also help in verification of the existing database. Findings of experiments on uranium, iodine, and strontium interactions with cement are presented. Iodine is shown to be immobilized by lattice incorporation into crystalline cement phases and by ‘sorption’ into C-S-H gel. Uranium VI however, can be precipitated as low solubility hydrated Ca salts. Much remains to be determined. The impact of higher temperatures is singled out as an important unknown factor affecting the performance of cements. It is known that elevated temperatures cause crystallization of C-S-H gel, thereby lowering the pH at which it buffers. In systems open to the environment, the course and speed of reaction of groundwaters with cement components are only known qualitatively. However, models can be modified to incorporate these interactions, provided the appropriate database is available.