Machine Learning-Guided SCAPS-1D Optimization of Lead-Free Cs2TiI4Br2 Perovskite Solar Cells for High-Efficiency and Sustainability
摘要
Photovoltaic solar technology’s primary challenges include toxicity, long-term stability, and elevated production costs. To address these issues, our research investigates the potential of lead-free noble metal halide perovskites, particularly Cs2TiI4Br2, which, due to the narrow bandgap, shows potential as an absorber material, excellent light absorption capabilities, and cost efficiency. We utilize SCAPS-1D simulations to analyze solar cell (SC) architectures that integrate Cs2TiI4Br2 with PCBM as the electron transport layer (ETL). Our goal is to determine the optimal photovoltaic parameters by examining how absorber thickness, temperature, defect, and doping concentrations influence device performance. Additionally, we explore various back and front contact materials to find the best electrode for optimized solar cells. Considering the configurations that were investigated, the most effective design was identified as Au (metal contact)/Cs2TiI4Br2 (absorber)/PCBM (ETL)/ITO, obtaining a power conversion efficiency (PCE) of 28.77%, an open-circuit voltage (VOC) of 0.82 V, a short-circuit current density (JSC) of 41.02 mA/cm2, and a fill factor (FF) of 85.70% at a temperature of 300 K. The Random Forest (RF) model forecasts the optimal PCE by combining various semiconductor descriptors as inputs. The model’s interpretability is improved by SHAP (Shapley Additive Explanations) analysis, which measures the relative contribution of each factor to the prediction result. The RF model attains a high coefficient of determination (R2 = 0.8875), indicating its strength and dependability in predicting photovoltaic performance. These discoveries underscore the promise of Cs2TiI4Br2 as a highly promising and sustainable absorbent material for next-generation, environmentally friendly solar applications.