Large Language Model-Assisted Additive Selection for Synergistic Defect and Crystallization Control in Efficient...

Precursor additives are crucial for enhancing the efficiency and stability of perovskite solar cells. However, their traditional selection of additives primarily relies on empirical trial-and-error approaches, which are time-consuming and inefficient. Herein, we utilize Perovskite-R1, a large language model, to rapidly identify an efficient additive: ethyl 2-aminopropanoate hydrochloride (EAH). This additive simultaneously passivates defects and regulates crystallization through the coordination of its -CO and -NH₃ ⁺ groups with the uncoordinated Pb²⁺ and I^- ions in the perovskite. These interactions significantly improve charge-carrier transport and suppress nonradiative recombination, leading to a champion power conversion efficiency (PCE) of 22.58%. Furthermore, the EAH-modified device exhibits excellent long-term stability, maintaining 95.1% and 94.1% of its initial PCE after 1368 h of storage in N₂ and 1272 h of thermal aging at 65°C, respectively. This study highlights the potential of integrating artificial intelligence with materials design to accelerate the discovery of high-performance, stable, and sustainable perovskite optoelectronic materials.