Executive Summary : | Borosilicate-based glass matrices are widely used as vitrification matrices for long-lived and high-activity radioactive waste. These glasses are sent to underground repositories to isolate them from the ecosystem. However, the aqueous dissolution of these glasses is inevitable, and has been extensively studied for the past 50 years. The process involves hydration, hydrolysis of the ionic-covalent network, and exchange between alkali or alkaline-earth ions and protons in solution. This results in the formation of a silica-rich hydrated and porous layer called a gel layer. The limiting mechanisms controlling the properties of this gel layer are still unclear, and there is limited research on the effect of irradiation on the gel layer. This proposal aims to investigate the gel layer using atomistic simulations, experiments, and machine learning. An information extraction framework using Natural Language Processing (NLP) will be developed to extract information on the dissolution behavior and chemical durability of glasses. A knowledge base will be developed to include glass compositions, processing and testing conditions, and chemical durability in terms of dissolution. A machine learning model will be developed to predict the durability of glasses. Molecular dynamics simulations of select glass compositions will be performed to understand the mechanism governing durability. The project aims to understand the composition-structure-durability relationship in nuclear waste glasses using a holistic approach. |