The development of lead-free and environmentally benign absorbers is a critical step toward sustainable solar technologies. This study investigates the optoelectronic behavior and performance optimization of beryllium silicon diphosphide (BeSiP2) based thin-film solar cells using SCAPS-1D simulation coupled with machine learning (ML) analysis. Various electron transport layers (ETLs) PCBM, SnO2, WS2, TiO2 and hole transport layers (HTLs) V2O5, CuI, CuSCN, NiO were examined to determine the optimal device configuration. The optimized FTO/PCBM/BeSiP2/V2O5/Cu–C device architecture achieved a high-power conversion efficiency (PCE) of 28.12%, with open circuit voltage (Voc) = 1.13 V, Short Circuit Current Density (Jsc) = 29 mA cm⁻², and Fill Factor (FF) = 88%. Numerical analysis indicates that thin transport layers (50–100 nm), moderate doping concentrations (1017–1018 cm−3), and low interface defect densities (≤ 1014 cm−3) are critical in minimizing recombination losses and achieving this efficiency. To complement the numerical analysis, an ML framework utilizing eight regression algorithms was implemented to predict key photovoltaic parameters. Gradient Boosting (GB) achieved the highest accuracy (R2 > 0.98), identifying shunt resistance (Rsh) and irradiance as dominant factors influencing device metrics. The hybrid SCAPS-ML approach effectively bridges physics-based and data-driven insights, enabling rapid, interpretable optimization of lead-free chalcopyrite solar cells.
Ali et al. (2026) studied this question.