Machine learning-guided humidity-induced degradation analysis of Cs2CuSbCl6 sustainable perovskite solar cell

In this study, we present a machine learning (ML)-predicted framework to evaluate both the performance and environmental stability of a novel lead-free perovskite solar cell with the structure FTO/AZnO/Cs2CuSbCl6/MoO3. Our approach integrates theoretical calculations, device simulations, and ML to predict device behaviour under varied conditions. Density Functional Theory (DFT) calculations were employed to determine intrinsic electronic properties, yielding bandgaps of 3.481 eV for FTO, 3.330 eV for AZnO, 1.7 eV for Cs2CuSbCl6, and 3.170 eV for MoO3. Optical analysis confirmed that Cs2CuSbCl6 exhibits strong UV absorption with minimal losses, positioning it as an effective absorber. Subsequently, SCAPS-1D simulations optimized photovoltaic parameters, achieving an open-circuit voltage (Voc) of 1.58 V, a short-circuit current density (Jsc) of 21.82 mA/cm2, a fill factor (FF) of 91.12 %, and a power conversion efficiency (PCE) of 31.50 %. Nearly 100 % quantum efficiency in the visible spectrum further demonstrates efficient charge carrier extraction. To address moisture-induced degradation, we developed a linear regression ML model trained on both experimental and simulated data, which achieved an R2 of 0.989. The model reliably predicts that increased humidity and prolonged exposure significantly reduce PCE. This integrated methodology of DFT, device simulation, and ML offers a powerful tool for designing robust, high-performance perovskite photovoltaics with enhanced environmental resilience.