First-Principles DFT Investigation of CsSn0.5Ge0.5I3 and Machine Learning-Assisted Numerical Simulation of Lead-Free Solar Cells
TL;DR
This integrated approach offers a practical pathway for designing high-performance, stable, and environmentally sustainable perovskite solar cells (PSCs) by identifying absorber layer thickness as the dominant factor influencing efficiency.
Abstract
The optoelectronic properties of the lead-free perovskite CsSn0.5Ge0.5I3 are investigated by first-principles calculations and numerical simulations using SCAPS-1D. The energy-level alignment between transport layers and the perovskite layer is evaluated, resulting in the establishment of the PCBM/CsSn0.5Ge0.5I3/PEDOT:PSS structure. Key parameters, including bulk defect density, layer thickness, and electrode materials, are optimised, and the effects of resistance, illumination intensity, thermal stability, and carrier generation-recombination rates on device performance are analysed. The optimal device structure FTO/PCBM/CsSn0.5Ge0.5I3/PEDOT:PSS/C achieves a power conversion efficiency (PCE) of 24.50% and a fill factor (FF) of 80.01%. Machine learning (ML) algorithms are applied to predict photovoltaic parameters, with Random Forest (RF) exhibiting the highest accuracy. SHAP analysis identifies absorber layer thickness as the dominant factor influencing efficiency, providing guidance for experimental optimisation. This integrated approach offers a practical pathway for designing high-performance, stable, and environmentally sustainable perovskite solar cells (PSCs).