Computational optimization of MASnI3 perovskite solar cells using SCAPS-1D simulations and machine learning techniques

Abstract

The lead-free methylammonium tin iodide (MASnI₃) has gained significant interest in perovskite solar cell (PSC) technology due to its ideal band gap and environmentally friendly composition, offering a promising alternative to traditional lead-based perovskites. In this computational work, the SCAPS-1D simulator is utilized to create and assess several PSC structures by merging two electron transport layers and three hole transport layers. Among them, the MASnI₃-based device architecture of Al/FTO/WS₂/MASnI₃/Zn₃P₂/Ni provides superior performance, including efficiency of 32.30%, fill factor (FF) of 87.35%, short-circuit current density (J_sc) of 34.19 mAcm⁻² and open-circuit voltage (V_oc) of 1.08 V. The influence of thickness, carrier density, defect concentration, and carrier lifetime of the MASnI₃ perovskite layer on the photovoltaic performance metrics is investigated meticulously. Furthermore, the effects of interface recombination, operating temperature, and capacitance–voltage (C–V) and capacitance–frequency (C–F) on the performance characteristics of the proposed structure are analyzed. Moreover, we present a comparative analysis conducted among three machine learning algorithms to identify the model that provides the highest accuracy in predicting the efficiency of the specified device. The results demonstrate that the XGBoost model achieves superior performance, yielding an R² value of 0.9999, along with a low mean squared error (MSE) of 0.0092 and mean absolute error (MAE) of 0.051. In addition, the individual influence of several input parameters on the device efficiency has been assessed using the feature importance method. Among the evaluated features, defect density emerged as the most influential parameter, indicating that it plays a significant role in defining the overall performance of the photovoltaic (PV) device.