Modelling Melt Viscosity For Nuclear Waste Glass

This work forms part of a Collaborative Awards in Science and Engineering industrial studentship (iCASE), jointly funded by the Engineering and Physical Sciences Research Council and the National Nuclear Laboratory. The aim is to develop improved models for calculating viscosities of vitreous nuclear waste melts, particularly with respect to the variation in temperature and composition. Both in vitro and in situ experimentation on nuclear material is complicated by radioactivity and its associated expenses, so computational modelling is the principal means we use to study these industrially important glasses. The problem is approached with both top-down and bottom-up methods. From a more fundamental perspective, beginning in Section 7 Molecular Dynamics techniques are used to simulate glass melts at atomic resolution. An audit of literature forcefields, using a systematic methodology for particulate systems generation, involved calculation of structural and diffusive properties to reveal the advantages and disadvantages of contemporary sodium-borosilicate models. After developing an improved glass model, from Section 11 different methods of viscosity computation were trialled to determine that most appropriate for the conditions of the nuclear glass melters. In Section 14 the Inoue2 SBN forcefield was combined with the Green-Kubo technique, using simulated runtimes more than double those of previous literature work. The analyses produced qualitative agreement in compositional and temperature trends, as well as order-of-magnitude quantitative agreement between experimental and computational viscosity results for ternary nuclear glass frits. Complimentary top-down approaches were also used, with rotary viscometry experimentation employed in Section 4 to gather temperature-composition-viscometry data for nuclear waste glasses. These data were used with different fitting algorithms in Section 15 to compare the efficacy of theoretical descriptions for glass viscosity, described in Section 3. A combination of fitting techniques assembles in Section 17 an interpolative second-order model for which the maximum discrepancy between prediction and experiment is 17% of the absolute viscosity.