Machine Learning‐Assisted Screening of Small Molecules for Surface Modification of 2D Perovskites

ABSTRACT In recent years, inspired by the substantial reductions in non‐radiative losses and improvements in efficiency and stability achieved through surface modification of 3D systems, research into the surface engineering of 2D and quasi‐2D systems has progressed rapidly. This study systematically explores the integration of small‐molecule surface engineering within the 2D perovskite framework and employs data‐driven methodologies to expedite the screening process. We have constructed a library of small molecules composed solely of carbon, hydrogen, oxygen, and nitrogen elements and have utilized first‐principles calculations to determine the interface adsorption energy (E _ads ) and charge transfer (Δq) following adsorption on the 2D perovskite surface. Furthermore, we have deployed an interpretable machine learning model to identify critical descriptors of these small molecules. The findings indicate that molecular polarity and hydrophobicity collaboratively influence E _ads and Δq, which in turn impact device performance. We selected three small molecules, urea, 5‐amino‐1‐pentanol hydroiodide, and tris(hydroxymethyl)aminomethane (Tris), for device‐level experimental validation. The devices doped with these small molecules exhibited markedly enhanced photodetector performance compared to their undoped counterparts. These outcomes underscore the practical significance of rapid screening methodologies for the development of high‐performance photodetectors, perovskite light‐emitting diodes, and solar cells.