Machine Learning-Integrated Modeling of Thermal Properties and Relaxation Dynamics in Metallic Glasses

Abstract

Metallic glasses are a promising class of materials celebrated for their exceptional thermal and mechanical properties. However, accurately predicting and understanding the melting temperature (T_m) and glass transition temperature (T_g) remains a significant challenge. In this study, we present a comprehensive approach that integrates machine learning (ML) models with theoretical methods to predict and analyze these key thermal properties in metallic glasses. Our ML models using distributional data derived from elemental composition-based features obtain high accuracy while minimizing data preprocessing complexity. Furthermore, we explore the correlation between T_m and T_g to elucidate their dependence on alloy composition and thermodynamic behavior. When the T_g value of metallic glasses is known, further analysis using the Elastically Collective Nonlinear Langevin Equation (ECNLE) theory provides a deeper understanding of structural relaxation dynamics. This integrated framework establishes a quantitative description consistent with experimental data and previous works and paves the way for the efficient design and discovery of advanced materials with tailored thermal properties.