Performance evaluation of al/ETLs/FASnI3/GO/Au PSCs using numerical and machine learning approaches

Formamidinium tin iodide has emerged as a promising lead-free absorber for perovskite solar cells due to its optimal bandgap (∼1.4 eV) and strong light absorption. In this study, the photovoltaic performance of FASnI 3 -based perovskite solar cells with the configuration Al/ETLs/FASnI 3 /GO/Au was investigated using SCAPS-1D simulations under standard AM1.5G illumination. Five electron transport layers—SnO 2 , TiO 2 , ZnO, PCBM, and IGZO—were systematically analyzed to optimize device efficiency. The SnO 2 -based perovskite solar cell achieved the highest power conversion efficiency of 20.55 %, with an open-circuit voltage of 0.961 V, a short-circuit current density of 25.798 mA/cm 2 , and a fill factor of 82.87 %, attributed to favorable conduction band alignment and enhanced electron mobility. Graphene oxide was employed as the hole transport layer, improving hole extraction, minimizing interfacial recombination, and enhancing device stability. The quantum efficiency approached 100 % across the visible spectrum. Parametric studies identified optimal conditions, including series resistance (Rs = 0 Ω cm 2 ), shunt resistance (Rsh = 10 7 Ω cm 2 ), and near-room temperature operation (300 K). Five machine learning (ML) regression models—K-Nearest Neighbors, Support Vector Regression, Gradient Boosting, Random Forest, and Linear Regression—were trained to predict device efficiency in addition to simulations. The RF model demonstrated the highest accuracy (R 2 = 0.99, RMSE = 0.08), effectively capturing the complex interdependence of device parameters. This combined simulation and ML approach provides a robust framework for the design of stable, high-efficiency, and environmentally sustainable PSCs.