From Formability to Bandgap: Machine Learning Accelerates the Discovery and Application of Perovskite Materials

Perovskite materials are considered promising candidates for applications in solar cells, photodetectors, catalysts, and light-emitting diodes, owing to their exceptional physicochemical and structural properties. Recently, the integration of machine learning into perovskite research has revolutionized the discovery and optimization process by overcoming the limitations of traditional trial-and-error methods and computationally intensive first-principles calculations. This review examines the role of machine learning in predicting perovskite properties and advancing their practical applications. First, the representative literature and the development trend of machine learning in perovskite materials in recent years were organized and analyzed. Second, the workflow of machine learning for perovskite materials was delineated, accompanied by a brief introduction to the fundamental algorithms. Third, by analyzing the structure and composition of perovskite materials, the role of machine learning in accelerating the discovery of perovskites, particularly in predicting formability and bandgap, is detailed. Finally, four practical applications of machine learning on perovskite materials were presented, along with an innovative proposal of the potential challenges and future directions of machine learning in the field of perovskite materials. Overall, this review aims to provide comprehensive insights and practical guidance for perovskite research, fostering the further development of machine learning-accelerated discovery and application of perovskite materials.