Autonomous Closed-Loop Discovery and Reproducible Manufacturing for High-Performance Perovskite Solar Cells

Autonomous experimentation is emerging as a powerful paradigm for accelerating materials discovery and improving reproducibility in complex energy materials systems. In this talk, I will present our recent work on an autonomous closed-loop framework that integrates interpretable machine learning, quantum-mechanical modeling, and automated device manufacturing for perovskite solar cells. On the discovery side, we combined active learning with density functional theory calculations to screen a large molecular design space and establish quantitative structure-property-performance relationships for interfacial passivation molecules. On the fabrication side, we developed an automated manufacturing platform that uses Bayesian optimization and symbolic regression in a feedback loop to continuously refine processing conditions and device performance. Using this integrated framework, we identified a new passivation molecule, 5-(aminomethyl)nicotinonitrile hydroiodide (5ANI), and autonomously optimized its implementation in perovskite solar cells. This approach enabled 0.05 cm² devices with a power conversion efficiency of 27.22% and 21.4 cm² mini-modules with an efficiency of 23.49%, while also achieving substantially improved operational stability and nearly fivefold better fabrication reproducibility compared with manual processing. More broadly, this work demonstrates how AI-driven closed-loop systems can bridge materials discovery and manufacturing, offering a general strategy for reproducible, scalable, and interpretable scientific discovery in photovoltaics and beyond.