Machine learning-guided design of Cs₂SnBr₆-based solar cells: a DFT and SCAPS-1D analysis with N-doped TiO₂ HTL

Abstract This work presents a novel perovskite solar cell (PSC) architecture—FTO/ TiO 2 /Cs 2 SnBr 6 /N-doped TiO 2 /Au—designed to enhance efficiency, and stability. A key innovation is the use of emerging N-doped TiO 2 as the hole transport layer (HTL), offering superior environmental stability, a wide bandgap for better energy alignment, and low-cost processing. Combined with the stable and optically efficient Cs 2 SnBr 6 absorber, this configuration overcomes common HTL-related challenges. The optimized device achieves a remarkable power conversion efficiency (PCE) of 38.70%, with an open-circuit Voltage (V oc ) of 1.29 V, short-circuit current density (J sc) of 33.34 mA cm −2 , and fill factor of 90.21% under standard illumination at 300 K. A machine learning model trained on a dataset predicted PCE degradation due to effect of relative humidity with high coefficient of determination (R 2 = 0.987), enabling performance forecasting across environmental conditions. The work sets a new benchmark for AI-driven material design in photovoltaics, showcasing a stable, efficient, and scalable PSC platform.