AI-Guided Closed-Loop Discovery of Photostable Light-Harvesting Molecules*

AI-guided closed-loop experimentation has recently emerged as a promising method to optimize objective functions, but the potential of this traditionally black-box approach to reveal new chemical knowledge for functional materials is not yet clear. In this talk, I will discuss a closed-loop approach combining automated synthesis, photophysical characterization, and AI-guided prediction methods to identify organic light-harvesting molecules with optimized photostability. A Bayesian optimization framework is used to efficiently guide the search through a large molecular space using key physicochemical descriptors while maintaining a customizable tradeoff between exploitative and explorative sampling. Candidate molecules suggested by the AI framework are then prepared via automated synthesis using a "Lego-like" molecular building block approach based on Suzuki cross-coupling, and the photophysical properties are characterized. Our results show that high-energy regions of the triplet state manifold are key to controlling molecular photostability in solution across a diverse chemical library of light-harvesting donor-bridge-acceptor oligomers. Remarkably, this insight emerged after automated modular synthesis and experimental characterization of only ~1.5% of the total chemical space of 2,200 oligomers. Supervised learning models considering millions of combinations of 100+ physics-based descriptors showed that high energy triplet states most strongly correlate with photostability, while excluding more commonly considered predictors such as the lowest energy triplet state. The physics-informed model for photostability was further confirmed and strengthened using an explicit experimental test set. Broadly, this work shows that interfacing physics-based modeling with closed-loop discovery campaigns unimpeded by synthesis bottlenecks can rapidly illuminate fundamental chemical insights and guide more rational pursuit of frontier molecular functions.