Smart Exploration of Perovskite Photovoltaics: From AI Driven Discovery to Autonomous Laboratories

ABSTRACT Metal halide perovskite solar cells have been developed as a front runner in next generation photovoltaic technology. However, such rapid development is hampered by complex relationships among material compositions, process routes, device architectures, and environmental conditions. The fabrication system is strongly coupled with many variables, which implies that existing empirical optimization methods are less capable to explore design space in high dimensions. Additionally, perovskite materials have sensitive intrinsic properties to environmental factors hence facilitating reproducible optimization across the composition space is notoriously difficult. Now, in this coupled system, artificial intelligence and machine learning have driven progress toward the understanding of the nonlinear composition–process–property relationships. But their predictive reliability is limited by patchy data and low reproducibility in manual experiments. Therefore, controlling lab environments and generating reproducible data with help from autonomous laboratories is paramount to accelerate predictive design. This review highlights recent progress in data driven materials discovery and closed loop experiments in perovskite research. We also report that AI guided process optimization is capable of identifying reproducible manufacturing windows. In parallel, robotic manipulation, high throughput synthesis with in situ monitoring and autonomous characterization are enabling systematic exploration of the multi‐dimensional process space.