Interface Defects and Recombination in HTL-Free and Bilayer Cs2SnI6 Perovskite Solar Cells: Numerical Modeling and Machine Learning Analysis.

Perovskite solar cells (PSCs) are a promising technology for sustainable energy generation. Among lead-free perovskites, Cs2SnI6 offers high structural stability and environmental compatibility, making it ideal for next-generation PSCs. This work systematically investigates Cs2SnI6-based HTL-free and bilayer PSC architectures using SCAPS-1D simulations, focusing on interface defects and defect-induced recombination. The bilayer design incorporates CsSnBr3 as a secondary absorber, while suitable transport layers and back contacts are applied for each structure. Key parameters─including absorber thickness, defect density, interface defect density, interface defect energy levels, and electron and hole interface capture cross sections─are optimized, yielding maximum power conversion efficiencies of 27.92% for the HTL-free device and 24.08% for the bilayer device. Machine learning models (ANN, RF, MLR) are also employed to predict device performances, with SHAP analysis revealing defect density as the dominant factor for HTL-free devices and absorber thickness for bilayer structures. The ANN (6-15-4) model achieves the highest accuracy (R2 = 0.9968). This integrated simulation-machine learning framework clarifies defect-driven performance behavior and provides practical guidance for interface optimization, highlighting the potential of stable, lead-free PSCs for green energy applications.