Machine Learning Assisted Development of Sn/Pb Free Perovskites Solar Cells

This study focuses on Cs₃Bi₂I₉, a zero-dimensional perovskite absorber material with great potential for photovoltaic applications. We employed Machine Learning (ML) techniques to understand and optimize the performance of Cs3Bi2I9 perovskite solar cells. A prepared sub-dataset comprising 26,457 experimentally developed perovskite solar cell device data was utilized for ML calculations. The cleaned dataset contained 29 features and four target variables, including open circuit voltage (Voc), short circuit current density (Jsc), fill factor (FF), and power conversion efficiency (PCE). Among the models, Random Forest Regressor showcased the best performance. Predicted values for Voc, Jsc, FF, and PCE were determined to be 0.562 V, 9.544 mA/cm², 0.508, and 5.356%, respectively.<br/><br/>We characterized the different layers of the device using techniques such as X-ray Photoelectron Spectroscopy, Ultraviolet-Visible Spectroscopy, Atomic Force Microscopy, and Ellipsometry to gain insights into the individual layers' properties. The absorber layer bandgap was found to be 1.9 eV, which indicates that Cs₃Bi₂I₉exhibits favorable characteristics for indoor photovoltaic applications. The solution-processed spin-coated complete solar cells have the structure glass/FTO/c-TiO₂/Cs₃Bi₂I₉/CuSCN/Ag. Preliminary tests on the complete Cs₃Bi₂I₉ perovskite solar cell have shown the values for Voc, Jsc, FF, and PCE as 0.447 V, 29.08 mA/cm², 0.291, and 3.63%, respectively. These initial results indicate the potential of Cs₃Bi₂I₉ as a viable material for solar cell applications.