(403e) Taming the Complexity of a Messy Solution Process to Find Better Materials for Solar Cells: A Bayesian Optimization-Guided Route to Better Bouillabaisse

The nucleation of hybrid organic-inorganic perovskites (HOIPs) in solution occurs by a molecular self-assembly process that remains poorly understood and under-investigated. Despite progress to date, the lack of understanding of processing-structure-property relationships in HOIPs inhibits the community from solving several key challenges that could, for example, mitigate issues currently surrounding scale-up of these photovoltaic (PV) devices or their stability. There is considerable current interest in having a fundamental understanding of the interactions that take place between the constituents of the medium from which HOIPs will ultimately nucleate and grow. This puts a premium on understanding molecular-scale interactions between the choices of solvent(s), lead salt, and organic and/or inorganic cations that will form a particular HOIP crystal. We tackle this complexity in two ways: Firstly, a detailed quantum mechanical level study of the interactions between the species in the process, our "bouillabaisse," and secondly, using Bayesian optimization techniques to vastly accelerate the search for the combinations of chemical species in the "soup" that lead to a desired performance outcome, e.g., best binding energy between chosen species. This study includes investigating "solvent engineering," namely, assessing the interaction of lead salts and ancillary cationic species as we change processing solvents (both "bath" solvents and "anti-solvents") and determine the impact of solvent choice on complexation and formation of perovskite crystal building blocks. We will show how the solvating strength of the bath solvents, while helpful in solubilizing the lead, hinders the formation of perovskite-generating nuclei and prohibits precipitation of crystallites in solution. We observe that coordination of the solvent with the Pb atom causes a change in Pb-halide bond length and the shift in electron density prohibits complexation with the cation. Our ab initio calculations show that the binding energy between the two major building blocks of the final perovskite crystal structure can be increased using an anti-solvent to tune the overall relative dielectric of the solvent. As the relative dielectric of the solution decreases, the distance between lead and halide ions, and that between halide and cation also (favorably) decrease. This provided an important new way to tune the crystallization outcome.