ML-integrated predictive modeling and performance evaluation of inorganic Ba3SbI3 perovskite solar cell using SCAPS-1D

The transition to sustainable energy sources has heightened interest in perovskite solar cells (PSCs), yet commercialization remains challenged by the toxicity of lead-based compounds and their vulnerability to environmental degradation. In response, this study presents a comprehensive computational framework for designing and optimizing a lead-free, inorganic PSC that employs Ba3SbI3 as the absorber material. Implementing SCAPS-1D simulations, we systematically evaluate the influence of device architecture, carrier transport layers (CTLs), defect densities, doping concentrations, and operational conditions on key photovoltaic metrics. When different device layouts were compared, the Al/FTO/SnS2/Ba3SbI3/CBTS/Au structure stood out as the most promising. It yielded a simulated power conversion efficiency (PCE) of 33.31%, accompanied by an open-circuit voltage (Voc) of 1.2068 V, short-circuit current density (Jsc) of 32.17 mA/cm2 and fill factor (FF) of 85.81%, outperforming the reference model without a hole transport layer (HTL) achieved PCE = 21.55%, Voc = 0.9089 V, Jsc = 27.83 mA/cm2, FF = 85.19%. Beyond parametric optimization, we incorporate machine learning (ML) via Random Forest Regression (RFR) to enhance predictive modeling capabilities. The model demonstrates high fidelity (R² > 0.97) and offers insight into variable importance, identifying absorber defect density, doping level, and operational temperature as primary performance drivers. Our hybrid approach not only accelerates the design of high-efficiency Ba3SbI3-based PSCs but also underscores the viability of environmentally benign materials for next-generation photovoltaics. The integration of physics-based simulation with data-driven analysis provides a scalable methodology for future studies in sustainable solar energy technology.