Challenges, Opportunities, and Prospects in Metal Halide Perovskites from Theoretical and Machine Learning Perspectives

Abstract Metal halide perovskite (MHP) is a promising next generation energy material for various applications, such as solar cells, light emitting diodes, lasers, sensors, and transistors. MHPs show excellent mechanical, dielectric, photovoltaic, photoluminescence, and electronic properties, and such intriguing physical and chemical properties have drawn attention recently. However, there exists a chasm between the successful applications of MHPs and theoretical understandings. The difficulty arises from the intrinsic properties of MHPs, including structural disorder, ionic interactions, nonadiabatic effects, and composition diversity. Machine learning (ML) approaches have shown great promise as a tool to overcome the theoretical obstacles in many fields of science. In this perspective, the pending theoretical challenges from experiments are overviewed and promising ML approaches, including ab initio ML potentials, materials design/optimization models, and data mining strategies are proposed. Possible roles and pipelines of ML frameworks are highlighted to close the gap between experiment and theory in MHPs.