Machine Learning–ECNLE Integrated Framework for Predicting the Molecular Relaxation Dynamics of Metallic Glasses

Abstract

We combine machine learning and theoretical modelling to predict and analyze the thermal and molecular dynamics properties of metallic glasses. First, we use machine learning models to estimate the glass transition temperature (Tg) from alloy compositions. Our approach uses the most minimal input features to date, simplifying the data processing workflow while still achieving high accuracy. Then, the machine learning predicted Tg values are input into the Elastically Collective Nonlinear Langevin Equation (ECNLE) theory to calculate the structural relaxation time as a function of temperature. Our theoretical calculations show excellent agreement with previously reported experimental data. Overall, this study provides a quantitative framework in agreement with experiments and prior works. This approach would pave the way for the discovery and design of metallic glass materials with engineered thermal properties.