POLARIS: perovskite optimization using LLM-assisted refinement and intelligent screening

Key Points

• The aim is to optimize 2D halide perovskites using large language model (LLM) techniques.
• Utilized LLM-assisted mining to convert literature into a spacer-cation dataset.
• Applied GNN-GP models to predict structural and optoelectronic properties with uncertainty.
• Implemented clustering methods to categorize chemical spacer families.
• Identified significant patterns in the structural and optoelectronic properties of perovskites.
• Generated a comprehensive dataset from 200 studies that facilitates further materials research.