Machine and deep Learning-Powered analysis of photovoltaic properties in 4-terminal FASnI3/CIGS tandem solar cells

A. Maoucha, T. Berghout, F. Djeffal,
Machine and deep Learning-Powered analysis of photovoltaic properties in 4-terminal FASnI3/CIGS tandem solar cells,
Materials Science and Engineering: B,
Volume 322,
2025,
118629,
ISSN 0921-5107,
https://doi.org/10.1016/j.mseb.2025.118629.
(https://www.sciencedirect.com/science/article/pii/S0921510725006531)
Abstract: This work presents a comprehensive numerical and machine learning-based analysis of lead-free four-terminal (4 T) FASnI3/CIGS tandem thin-film solar cells. Using SCAPS-1D, we evaluated the photovoltaic performance of the top and bottom sub-cells under various material and structural configurations. The FASnI3 top cell and CIGS bottom cell were optimized individually, achieving power conversion efficiencies (PCE) of 18.80 % and 15.47 %, respectively. Machine learning (ML) and deep learning (DL) approaches were employed to identify key performance-influencing parameters. Feature importance analysis revealed that the buffer layer donor density significantly impacts the Jsc and fill factor of the bottom sub-cell, while the top sub-cell’s performance is predominantly governed by the electron transport layer and perovskite properties. Despite higher complexity in the top cell’s behavior, attributed to environmental variability, the ML/DL framework effectively pinpointed the most critical design factors. These findings contribute to the accelerated development of high-efficiency and sustainable lead-free tandem solar cells.
Keywords: Lead-free; CIGS; 4T tandem; FASnI3; Machine Learning; Deep Learning