Unravelling Hybrid AI–SCAPS‐1D Framework for 29% Efficient CsSnGeI 3 /DNMO Dual‐Absorber Perovskite Solar Cells

ABSTRACT This study introduces a lead‐free perovskite solar cell (PSC) integrating CsSnGeI 3 and Dy 2 NiMnO 6 as dual absorbers in the FTO/C60/CsSnGeI 3 /Dy 2 NiMnO 6 /Au configuration, designed to boost both efficiency and environmental safety. SCAPS‐1D simulations reveal that the optimized device delivers a power conversion efficiency (PCE) of 28.43%, with a short‐circuit current density (Jsc) of 42.61 mA/cm 2 , an open‐circuit voltage (V OC ) of 0.81 V, and a fill factor (FF) of 82.10%. The dual‐absorber design enhances spectral utilization and charge separation, surpassing single‐absorber devices based on CsSnGeI 3 (23.68%) and Dy 2 NiMnO 6 (23.31%). To accelerate optimization, machine learning (ML) techniques were employed for performance prediction and parameter evaluation. A Random Forest model, trained on 80% of the dataset with 100 estimators, accurately predicted device efficiency and identified critical features influencing PCE. Additional models, including KNeighbors Regressor, CatBoost (yielding an R 2 of 0.9997, MAE of 0.0132, and RMSE of 0.0210) , and Ridge Regressor, further validated performance trends. ML predictions for the dual‐absorber cell aligned with simulations, projecting a PCE of 29.19%, Voc of 0.82 V, Jsc of 42.69 mA/cm 2 , and FF of 83.69%. The integration of SCAPS‐1D simulations with data‐driven modelling establishes a powerful framework for advancing next‐generation, stable, and eco‐friendly PSCs with enhanced performance.