Unlocking a self-driving research workflow for perovskite photovoltaics

Abstract

Perovskite solar cells (PSCs) have attained high power-conversion efficiencies in recent years. However, as the core materials of PSCs, metal halide perovskites (MHPs) still require the field to overcome the relatively low stability and processing reproducibility. In this context, artificial intelligence (AI) has been examined as a transformative tool for chemical-space exploration and experiment automation. This review presents a self-driving research workflow for MHP studies. We integrate computation and automatic experiments to realize this workflow, achieving a closed-cycle design from automated platform data outputs to theoretical models. In computational design, generative AI and discriminative AI are used to explore the vast MHP chemical space. In automatic experiments, recent innovations in hardware and the integration of experimental data streams are discussed. Global optimization incorporates experimental data into the overall workflow to achieve self-iteration. This proposed virtual workflow aims to provide a robust framework for self-driven research to accelerate the development of MHPs and PSCs.