Integrated first-principles, SCAPS-1D, and a ML framework for numerical modelling and performance optimization of BaXS3 (X = Zr, Hf)-based perovskite solar cells.

Recently, chalcogenide perovskites (CPs) have come to the forefront as a promising lead-free alternative for solar cell devices due to their enhanced stability, eco-friendly elemental composition, and potential for more economical, scalable manufacturing. However, there are few studies on the structural, electronic, optical, mechanical, and phonon characteristics of BaXS3 (X = Zr, Hf) for photovoltaic applications. In this regard, our detailed DFT-based study reveals that BaXS3 (X = Zr, Hf) exhibits a direct band gap and possesses anisotropic charge-carrier transport with lower effective mass. Moreover, the optical characteristics predict a significant coefficient of absorption (∼105 cm-1) and a low refractive index, suggesting their suitability as an absorber layer in photovoltaic solar cells (PSCs). Additionally, the mechanical properties indicate that the compounds are ductile and resistant to deformation, further supporting that our title materials are promising for solar energy conversion absorber layers. Finally, we simulated BaXS3 (X = Zr, Hf) as an absorber layer in SCAPS-1D, and yielded a maximum PCE of 25.2% through feature optimisation. In addition, we applied some ML algorithms to forecast efficiency in order to reduce simulation complexity and computational time. Exhibiting an exceptional value of R2 of 99.99% and a lower MSE of 0.002, the XGB algorithm outperforms alternative methods, enabling a more thorough investigation and understanding of these absorbers' potential in thin-film devices. Our research clarifies their functions in photovoltaics, helps screen for desired features, and yields key properties essential for device fabrication.