Data analysis of compositional distribution and glass transition temperature of low-melting phosphate glass using big data

Abstract

Regression analysis was performed on the data of low-melting phosphate glasses extracted from a glass database. The data were categorized based on duplication, and the most commonly used components for each temperature range were extracted. The relationship between metal-oxide-based compositions and cation-based compositions was examined. Even though various compositions were included in the database, it was found that the average and standard deviation values of divalent or tetravalent cations in cation-based compositions were approximately 2/3 of those in oxide-based compositions. Multiple regression analysis suggested that cation-based analysis is more suitable than oxide-based analysis. Predictions using cations are expected to become more important in future structure-driven data analyses of glasses.