From prediction to design: Recent advances in machine learning for the study of 2D materials

Abstract

Although data-driven approaches have made significant strides in various scientific fields, there has been a lack of systematic summaries and discussions on their application in 2D materials science. This review comprehensively surveys the multifaceted applications of machine learning (ML) in the study of 2D materials, filling this research gap. We summarize the latest developments in using ML for bandgap prediction, magnetic classification, catalyst material screening, and material synthesis design. Furthermore, we discuss the future directions of ML applications in various domains, providing robust references and guidance for future research in this field. Compared to traditional methods, we particularly emphasize the unique advantages of ML in predicting the bandgap of 2D materials, such as the introduction of advanced feature engineering and algorithms to enhance research efficiency. We also summarize ML algorithms for classifying the magnetism of 2D materials, showing that complex pattern recognition can precisely interpret the correlation between magnetic moments and atomic structures. Additionally, the review outlines how ML algorithms can efficiently sift through large-scale material databases to identify candidates with specific catalytic properties, thereby greatly accelerating the discovery process for new catalysts. ML has become a powerful tool in the field of materials science, promoting the discovery of new materials, improving their properties, and accelerating research across various application domains.