Exploring Ruddlesden–Popper Perovskite Solar Cells Performance Using First‐Principles Calculations, Processing Parameters Simulation, and Machine Learning

Nida Amin, Zia Ur Rehman, Javid Ullah, Ibrar Ahmad, Haris Haider, Khizar Hayat, Abdullah Shah, Aseel Smerat, Wejdan Deebani, Said Karim Shah

This study presents computational insights into Ruddlesden–Popper (RP), Cs 2 SnI 2 Cl 2 ‐based lead‐free perovskite solar cells (PSCs). The optical and electronic characteristics of Cs 2 SnI 2 Cl 2 were examined using first‐principles calculations, followed by simulation of the device performance parameters. Due to high absorption at 15.4 eV, optical conductivity in the range of 5–20 eV, and the direct bandgap of 1.4 eV, it can be used as an absorber layer (AL) for PSCs. Using SCAPS‐1D, certain essential parameters were optimized, including AL thickness, acceptor doping concentration ( N A ), defect density ( N t ), series ( R s )/shunt ( R sh ) resistances, sun intensity, and temperature. The optimized efficiency of 31.42% was achieved after tuning of these parameters at 360 K. XGBoost, among the various machine learning (ML) models, demonstrates the best predictive accuracy with the highest R 2 (0.98) and lowest error metrics, as compared to random Forest (RF) R 2 (0.97) and Ridge regression R 2 (0.88). The SHAP analysis further confirms that AL thickness, N A , N t , R sh , R s , and temperature are the most influential parameters controlling PCE. The conventional simulation, combined with the ML techniques, guided by the XGBoost‐optimized inputs, demonstrates the predictive accuracy of device processing performance of 31.3%, which is in good agreement with the actual PCE of the PSC.