Efficiently Screening Organic Ligands by Machine Learning for Stabilizing 2D/3D Perovskites

Abstract

Currently, the commercial photovoltaic applications of non-toxic tin-halide perovskites are still fundamentally restricted by their intrinsic phase instability. One of the most effective strategies for stabilizing these perovskites involves using organic ligands for passivation or intercalation. However, the selection of suitable ligands typically necessitates extensive and time-consuming experimental trials, often accompanied by inevitable errors. Herein, a significant step toward efficiently screening potential ligand spacers is made by integrating machine learning (ML) with high-throughput first-principles computation (HTFPC). Ligands selected by the proposed ML model can remarkably stabilize the 2D capping layer, thus facilitating the exploration of 2D/3D perovskites tandem solar cells with robust stability. Through first-principles computation combined with experimental stability tests under humid conditions, the predictions generated by ML were validated, and two target organic ligands have been identified as the most effective spacers for passivating interfaces to enhance the stability of 2D perovskites. The present work offers theoretical insights and practical guidance for improving the power conversion efficiency (PCE) and stability of perovskite solar cells.