Multivariate analysis: An essential for studying complex glasses

Abstract

Understanding the impact of individual compositional components on the devitrification of complex multicomponent glasses, for example, 10-50+ oxides, typically requires numerous studies to examine each component's impact. Here we apply exploratory data analysis (EDA) to a heterogeneous data set of silicate glasses to determine the cations' individual and interacting effects on the crystallization of nepheline (nominally NaAlSiO 4 ). Our data consisted of 795 simulated high-level nuclear waste glasses composed of, on average, 50 oxide components.We determine the interactions in the heterogeneous data that cause deviations from the behavior found in simplified composition studies. Using both univariate and bivariate EDA techniques, we demonstrate the importance of including calculated structural glass parameters on nepheline's devitrification, including field strength, cation-to-anion radius ratio, and single-bond strength. We also show that studies with simplified glass compositions may fall short in generating knowledge directly transferrable to complex glass compositions. The method used in this study has the potential to inform experimental design for simplified compositions (∼6+ oxides) that can generate knowledge directly transferrable to complex, multivariable compositions. The observations reported here have broad implications for any study attempting to map the physical properties of a complex glass containing numerous cations.