Innovative computational framework for Sr3SbCl3 absorber optimization: DFT, SCAPS-1D, and machine learning perspectives

Abstract

Recent progress in solar technology has drawn attention to novel inorganic cubic perovskites like Sr₃SbCl₃, due to their excellent physical properties and solar cell potential, as supported by machine learning (ML) insights. This study investigates the optoelectronic properties of Sr₃SbCl₃ using FP-DFT, revealing a bandgap of 1.908 eV at the Γ point. The partial density of states (PDOS) analysis identifies the atomic contributions, while dielectric and absorption studies confirm strong light absorption within the 1.5–3.5 eV range, extending into the visible spectrum. These findings demonstrate that Sr₃SbCl₃ exhibits semiconducting behavior, making it a promising candidate for solar absorption applications. Photovoltaic performance was evaluated using SCAPS-1D simulations with different ETLs: CdS, PCBM, SnS₂, and ZnO. Key parameters, including absorber/ETL thickness, bulk defects, and interface defect density at the ETL/Sr₃SbCl₃ junction, were optimized. Maximum efficiencies of 16.23%, 14.11%, 16.63%, and 15.65% were achieved with CdS, PCBM, SnS₂, and ZnO, respectively. To enhance device prediction, a K-nearest neighbors regressor (KNR) ML model was trained on 2187 SCAPS-1D-generated data points, incorporating absorber thickness, defect densities, ETL (SnS₂) properties, and interface defects. KNR outperformed CatBoost and ridge regression, yielding an RMSE of 0.225, MAE of 0.180, MSE of 0.051, and R² of 0.961. Absorber thickness emerged as the most critical parameter, with a 56.30% importance value. These findings demonstrate Sr₃SbCl₃'s excellent optoelectronic and photovoltaic properties, reinforcing its potential for high-efficiency, next-generation solar cell applications.