Machine learning-driven optimization of lead-free La0.6Ce0.4Mn0.9Cd0.1O3 perovskites for sustainable photovoltaic applications

Abstract

Perovskite solar cells (PSCs) have rapidly emerged as a leading solar energy technology due to their high-power conversion efficiencies, low-temperature processing, and flexible design. However, the toxicity and instability of conventional lead-based halide perovskites have prompted a shift toward sustainable alternatives. In this study, we investigate La₀.₆Ce₀.₄Mn₀.₉Cd₀.₁O₃ (LCMCO), a novel lead-free oxide-based perovskite, as a potential absorber for single-junction PSCs. The incorporation of Mn and Cd enables favourable optoelectronic properties, with an estimated bandgap of ∼1.65  eV, suitable for visible light absorption. LCMCO exhibits structural robustness and temperature-dependent electrical performance, reinforcing its promise for green energy applications. A detailed numerical analysis using SCAPS-1D was performed to evaluate device performance, employing TiO₂ and NiO as non-toxic, energy-aligned ETL and HTL, respectively. Device parameters such as absorber thickness, doping level, interface defect density, and series/shunt resistance were systematically optimized. Simulations indicate a power conversion efficiency of up to ∼28 %, validating LCMCO’s potential as a viable, eco-friendly alternative to toxic perovskites. To enhance the analysis, a machine learning (ML) model was developed to predict the performance metrics of the LCMCO-based device, achieving a high accuracy of ∼97.75 %. The integration of ML underscores its effectiveness in accelerating material optimization and performance forecasting in solar cell research. This work highlights LCMCO as a promising candidate for sustainable, high-efficiency photovoltaic technologies.