High-efficiency InPbI3 perovskite solar cells: A multiscale approach from first-principles to machine learning

Abstract

This study investigates the structural, electronic, optical, mechanical, and photovoltaic properties of InPbI3 perovskite using a multiscale modeling approach combining density functional theory (DFT), numerical simulations via SCAPS-1D, and machine learning techniques. DFT calculations confirm that InPbI3 exhibits a stable cubic perovskite structure with a direct bandgap of 1.23 eV, strong visible light absorption, and favorable mechanical, dynamic, and thermal stability. The material's optical spectra reveal high absorption, suitable dielectric response, and low reflectivity-key attributes for solar harvesting. SCAPS-1D simulations of the FTO/ZnO/InPbI3/ Cu structure yield a peak PCE of 24.45 %, with JSC of 35.81 mA/cm2, VOC of 0.843 V, and FF of 81.01 %, optimized by absorber thickness, doping level, and low defect density. Performance sensitivity to series and shunt resistances, as well as thermal stability up to 380 K, further supports the device's practical viability. Additionally, machine learning models, particularly Random Forest, were employed to predict PCE with high accuracy (R2 = 0.9992), highlighting bandgap and electron affinity as critical features. These results validate InPbI3 as a promising candidate for next-generation perovskite solar cells, bridging theoretical predictions with device-level optimization.