Integrated SCAPS-1D and Machine Learning Approaches for Design Optimization of Cs 2 SnI 6 -Based Perovskite Solar Cells
Year: 2026
Type: conference-paper
Abstract: A lead-free, all-inorganic Cs2SnI6heterojunction perovskite solar cell (PSC) is numerically engineered in SCAPS-1D to identify transport-layer combinations and absorber-quality windows that can unlock high photovoltaic output while avoiding toxic materials. The modeled architecture adopts Al/FTO/WO3/Cs2SnI6/MoO3/Ni, and is compared with alternative ETLs (In3Se₄, CdS, WO3) and HTLs (P3HT, C6TBTAPH2, Zn3P2, MoO3). The ETL screening shows that WO3delivers the strongest photovoltaic response (Voc = 0.82 V, Jsc = 30.65 mA/cm2, FF = 85.26%, PCE= 21.35%), outperforming In3Se4and CdS, which exhibit less favorable band alignment. The spike-type interface of WO3/Cs2SnI6 (CBO = +0.12 eV) suppresses interfacial recombination, whereas the cliff-type alignments of In3Se4/Cs2SnI6(−0.48 eV) and CdS/Cs2SnI6(−0.38 eV) accelerate recombination losses, explaining the performance hierarchy. Subsequent HTL optimization identifies MoO3as the most selective back contact, enabling PCE= 29.56% with Voc = 1.11 V, Jsc = 30.66 mA/cm2, and FF = 86.50%. Physical parameters establish an optimal Cs2SnI6thickness of 1.0 µm and an acceptor density of 1015cm⁻3. Analyzing absorber defects, diffusion lengths, carrier lifetimes, and temperature primarily reduces Voc and efficiency by enhancing recombination. Furthermore, machine learning techniques were employed to optimize the performance prediction algorithm, yielding low MAE and RMSE values and a high R2of 0.9989, indicating excellent predictive accuracy. The study also examined material-level variations to support the design of optimized and cost-effective solar cell prototypes.
Institutions: Pabna University of Science and Technology, Tokai University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204489596