Performance optimization of CIGS thin-film solar cells via charge transport layer engineering: SCAPS-1D simulations coupled with machine learning

This work presents a comprehensive numerical investigation of thin-film Cu(In,Ga)Se 2 (CIGS) solar cells using the SCAPS-1D simulation tool. The CIGS absorber layer was first optimized in terms of thickness (600–1200 nm) and acceptor density, resulting in an increase in the simulated power conversion efficiency from 19.94% to 24.98%. Further optimization of the electron transport layer (ETL) identified CdSe as the most suitable ETL, yielding the highest efficiency of 25.11% for the Mo/CIGS/CdSe/ZnO/ITO configuration. To further improve device performance, 108 different solar-cell architectures were systematically designed and evaluated by incorporating various electron transport layers (ETLs) and hole transport layers (HTLs) to enhance charge extraction and suppress recombination losses. Detailed analysis of current–voltage characteristics, quantum efficiency, capacitance–voltage behavior, Mott–Schottky plots, and impedance spectra identified Mo/CuSbS 2 /CIGS/CdSe/ZnO/ITO as the best-performing architecture among HTL-based configurations. Its favorable behavior is attributed to improved band alignment, including a +0.15 eV spike-like conduction band offset at the CIGS/CdSe interface, which reduces interfacial recombination, and a +0.30 eV spike together with a −0.40 eV valence band offset at the CuSbS 2 /CIGS back interface, which establishes an effective back-surface field. In addition, machine-learning models including artificial neural networks, random forest, and XGBoost regressors were trained on the simulated dataset to uncover structure–performance relationships across all device configurations. In the primary machine-learning analysis, tree-based models achieved the best predictive performance, particularly for fill factor. SHAP analysis further revealed that ETL and HTL material selection is the most influential factor governing device efficiency. This integrated SCAPS–machine-learning framework provides both fundamental physical insight and an efficient strategy for the rational design of high-performance CIGS solar cells. • SCAPS-1D modeling reveals absorber and interface optimization pathways for high-efficiency CIGS • Systematic screening of 108 ETL/HTL stacks identifies CdSe/CuSbS 2 as optimal interfaces • Integrated SCAPS–machine learning framework enables accelerated CIGS device architecture design • SHAP analysis demonstrates transport layer selection as the dominant factor controlling efficiency • Electrical, CV and impedance analyses correlate band alignment with recombination mechanisms