Rational design of high-entropy ceramics based on machine learning-A critical review

High-entropy materials provide a versatile platform for the rational design of novel candidates with exotic performances. Recently, it has been demonstrated that high-entropy ceramics (HECs), depending on their compositions, show great application potential because of their superior structural and functional properties. However, the immense phase space behind HECs significantly hinders the efficient design and exploitation of high-performance HECs through traditional trial-and-error experiments and expensive ab-initio calculations. Machine learning (ML), on the other hand, has become a popular approach to accelerate the discovery of HECs and screen HECs with exceptional properties. In this article, we review the recent progress of ML applications in discovering and designing novel HECs, including carbides, nitrides, borides, and oxides. We thoroughly discuss different ingredients that are involved in ML applications in HECs, including data collection, feature engineering, model refinement, and prediction performance improvement. We finally provide an outlook on the challenges and development directions of future ML models for HEC predictions.

Automated summary

High-entropy ceramics (HECs) have shown great potential due to their superior structural and functional properties, but the vast phase space behind HECs makes it difficult to design high-performance HECs through traditional methods. Machine learning (ML) has become a popular approach to accelerate the discovery and design of HECs with exceptional properties. This article reviews the recent progress of ML applications in discovering and designing novel HECs and discusses the challenges and future development of ML models for HEC predictions.