A REVIEW OF MACHINE-LEARNING SYNERGY IN HIGH-ENTROPY ALLOYS

Abstract

High-entropy alloys (HEAs) offer remarkable properties in the areas of mechanics, temperature and chemistry. While HEAs have a wide range of possible material compositions, both exploring and identifying them proves difficult for traditional processes. ML methods have proven to be highly effective in speeding up materials research and optimization by exploring the links between different compositions, structures and the related properties. It reviews the interaction between machine learning (ML) and HEA research, pointing out successful applications along the materials creation process. It reviews several ML methods used in developing HEA systems, such as supervised learning for forecasting properties, unsupervised learning for detecting patterns, reinforcement learning for optimizing results and active learning for saving time during experimentation. The review looks at the present obstacles, difficulties and potential paths forward for research involving ML in HEA. Using ML along with physics and experiments greatly speeds up the search for new HEAs useful for many purposes.

Keywords: High-entropy alloys; Machine learning; Materials discovery; Property prediction; Composition structure-property relationships; Materials informatics