Abstract

Perovskites have attracted widespread attention in various applications such as optoelectronic devices, catalysis, and flexible electronic devices due to their excellent optoelectronic properties and cost-effective fabrication. Despite promising prospects, obstacles in their commercial applications necessitate addressing the shortcomings of existing perovskites on one hand and discovering novel and high-performance perovskites on the other hand. With the significant progress in theoretical methods and computational ability, various material simulation methods and machine learning (ML) techniques can be utilized to efficiently design novel perovskite materials and guide the experimental synthesis. In this report, I will describe my recent works on data-driven material discovery and synthesis for perovskite optoelectronic materials. In the first, I will introduce a ML-aided interpretable approach for photovoltaic properties of two-dimensional/three-dimensional (2D/3D) perovskite heterojunctions. Physicochemical insights, ML techniques, and causal inference are integrated to explore the key factors influencing formation, stability, and photovoltaic performance of 2D/3D perovskite heterojunctions. In the second, I will introduce the ML-accelerated experimental synthesis of 2D silver/bismuth perovskites. Compared to the traditional approach based on chemical intuition, our strategy have increased the success rate of the synthesis feasibility by a factor of four.