Challenges for 3D metal halide perovskites and the ameliorating role of machine learning

Metal halide perovskite thin films remain promising materials for LEDs and solar cell materials. Their appeal stems from room temperature solution processing and compositional flexibility. But challenges lie in operational stability, lack of understanding of ion migration and issues with scale-up are inhibiting reliable commercialization. We have studied the interplay between speciation, crystallization kinetics and defect formation on thin film production. This search is complicated by managing the high dimensionality of different compositions and processing options that is best attacked by a machine learning approach. We use a Bayesian optimization approach to sift between these options, aided by a “front-end” feature engineering code base. We will cover our insights into using Bayesian optimization for materials and processing discovery and its future outlook.