Multiscale computational and machine learning insights into rare-earth-doped front electrode and La2NiMnO6 double perovskite material for photovoltaic applications and proposed fabrication pathways
摘要
This study presents an integrated computational, data-driven investigation, and photovoltaic performance of La2NiMnO6 (LNMO) for next-generation solar cell applications. Using SCAPS-1D and wxAMPS simulations combined with machine learning (ML) modeling, the device architecture FTO/WS2/LNMO/CuSCN/Au was optimized for high efficiency and stability. The SCAPS-1D simulations revealed excellent photovoltaic performance, yielding a short-circuit current density (Jsc) of 32.92 mA cm⁻² and a power conversion efficiency (PCE) of 23.87%, accompanied by an open-circuit voltage (Voc) of 0.8650 V and a fill factor (FF) of 83.81%. Furthermore, Tb doping of the FTO layer significantly improved electrical conductivity and interfacial quality, leading to an enhanced device efficiency of 24.34%. This improvement was reflected in increased photovoltaic parameters, with Jsc, Voc, and FF reaching 33.52 mA cm⁻², 0.8657 V, and 83.89%, respectively. Parametric analysis demonstrated that optimal absorber thickness (600–800 nm), low defect density (1014-1015 cm⁻³), and reduced series resistance significantly improve carrier generation, reduce recombination, and maximize performance. Temperature and illumination studies confirmed high thermal stability with moderate efficiency losses at elevated conditions. Cross-validation with wxAMPS simulations and comparison with reported LNMO-based architectures confirmed the reliability of the proposed model. Machine learning algorithms yielded high predictive accuracy, with XGBoost achieving R² > 0.9998 and MSE < 0.005 for PCE prediction. This multidisciplinary approach demonstrates that LNMO, when combined with rare-earth doping and optimized interfacial engineering, is a strong candidate for high-efficiency, environmentally benign, and thermally stable lead-free perovskite solar cells.