The use of ChatGPT to generate experimentally testable hypotheses for improving the surface passivation of perovskite solar cells

Hypothesis formulation is a creative process fundamental to scientific exploration, and there is an increasing interest in using generative AI to augment humans in this regard. This study explores the potential of large language models, particularly ChatGPT, to generate viable hypotheses within experimental materials science. Focusing on the subdomain of surface passivation of hybrid perovskites, our aim is to identify untested molecules capable of reducing surface recombination and thereby boost the efficiency of perovskite solar cells. By utilizing ChatGPT to brainstorm ideas and formulate new hypotheses, we identify polyallylamine (PAA) as a potential surface modifier, which is a molecule we would not have considered otherwise. Subsequent experimental investigations demonstrate PAA to indeed be effective in decreasing recombination rates and improving device efficiency. This demonstrates a potent synergy between human expertise and AI capabilities also in more intellectual and intuitive facets of the scientific process, such as hypothesis generation.