Explainable machine learning-interpreted numerical analysis of CsSnBr3 perovskite solar cells with diverse electron transport layers

Abstract

Among emerging photovoltaic technologies, perovskite solar cells (PSCs) are getting increasingly more research attention for being efficient and cheap. Using SCAPS-1D simulation software, this work uses CsSnBr₃ as the absorber material, which is placed between the electron transport layer (ETL) made up of PC₆₀BM, CdS, SnS₂, and CdZnS, and the hole transport layer (HTL). The thickness of the absorber and other factors like defect and doping densities are varied to indicate an optimized performance. The optimized configuration, FTO/SnS₂/CsSnBr₃/MoO₃/Au, demonstrates a power conversion efficiency (PCE) of 19.58% relative to alternative structures. Three additional structures have been optimized, and all four structures demonstrate ideal solar cell behavior, as confirmed by variations in series and shunt resistance, current–voltage (J–V), quantum efficiency (QE) characteristics, and generation and recombination rates. The structures exhibit optimal efficiency at room temperature. Additionally, a machine learning-assisted analysis has been employed to identify the device design parameters having the highest influence on the performance. Correlation mapping, regression modeling, and SHAP interpretability confirm that absorber doping, ETL donor density, and HTL acceptor density predominantly control the V_OC, J_SC, FF, and overall PCE. The integrated SCAPS-ML framework provides deeper physical insights and accelerates the optimization of CsSnBr₃-based perovskite solar cells.