Hybrid Modeling and Optimization of Cs2SiBr6 Perovskite Solar Cells Using DFT, Machine Learning, and SCAPS-1D Simulation

Abstract

In this work, we investigate the structural, electronic, and optical characteristics of the lead-free inorganic double perovskite Cs2SiBr6 using first-principles density functional theory (DFT) calculations. The results indicate that Cs2SiBr6 possesses a direct bandgap of 1.82 eV, exhibits strong visible-light absorption, and shows favorable electronic transition properties, confirming its potential as an effective solar absorber material. To further assess its photovoltaic applicability, SCAPS-1D simulations are performed for different device configurations employing various hole transport layers (HTLs) such as Cu2O, CuSCN, Spiro-OMeTAD, and CBTS within the FTO/SnS2/Cs2SiBr6/HTL/Au structure. Among these designs, the CBTS-based configuration demonstrated the highest power conversion efficiency (PCE) of 19.40%. Device performance is further optimized by systematically adjusting absorber thickness, doping concentration, defect density, and operating temperature. Moreover, five machine learning (ML) regression algorithms are utilized to predict PCE outcomes, where Random Forest and XGBoost exhibited superior predictive accuracy. Overall, this study presents a comprehensive, hybrid methodology integrating DFT modeling, numerical device simulation, and ML-based prediction, offering a robust strategy for the rational design and optimization of high-efficiency, lead-free perovskite solar cells.