A critical review of the machine learning guided design of metallic glasses for superior glass-forming ability

Abstract

The discovery of novel metallic glasses (MGs) with high glass-forming ability (GFA) has been an important area of active research for years in materials science and engineering. Unfortunately, the traditional approach based on trial-and-error methods is inefficient, time consuming and costly. Therefore, machine learning (ML) has recently drawn significant research interest as an alternative approach for the development of MGs. In this review, we discuss the current progress regarding the ML guided design of MGs from a variety of perspectives, including the GFA database, data representation, ML algorithms and numerical evaluation. Furthermore, we consider the challenges facing this field, including the scarcity and quality of GFA data, the development of physics informed data descriptors, the selection of appropriate algorithms and the necessity for experimental validation. We also briefly discuss possible solutions to tackle these challenges.