Achieving over 28 % efficiency in inorganic halide perovskite Ca3AsI3: Optimization of electron transport layers via DFT, SCAPS-1D, and machine learning

Abstract

Recent advancements in solar technology underscore the promise of Ca₃AsI₃, a novel cubic perovskite with remarkable physical properties, photovoltaic performance, and machine learning (ML)-assisted predictive potential. This study systematically explores the physical characteristics of Ca₃AsI₃ using density functional theory (DFT) calculations. Thermodynamic stability, phonon properties, and tolerance factor evaluations confirm the high stability of Ca₃AsI₃, which is further validated by mechanical analyses and elastic property calculations. The bandgap analysis reveals a direct bandgap of 1.41 eV at the Γ point, confirming the semiconducting nature of Ca₃AsI₃, further supported by Partial Density of States (PDOS) results. Optical investigations demonstrate strong absorption across the visible to ultraviolet spectrum, as shown by dielectric functions, absorption coefficients, and conductivity measurements. These findings collectively highlight the significant potential of Ca₃AsI₃ perovskite for advanced solar energy applications. The findings were integrated into the Solar Cell Capacitance Simulator in 1 Dimension (SCAPS-1D) to assess the photovoltaic (PV) performance of various electron transport layers (ETLs), including Zinc sulfide (ZnS), Indium (III) sulfide (In₂S₃), Titanium dioxide (TiO₂), and Tungsten disulfide (WS₂). Optimization analysis focused on key parameters such as absorber thickness, ETL thickness, defect density, acceptor density, and interface defect density at the ETL/Ca₃AsI₃ junction. Additionally, temperature effects, quantum efficiency (QE), and current density-voltage (J-V) characteristics were explored. Under ideal conditions, the Al/FTO/ETL (ZnS, In₂S₃, TiO₂, WS₂)/Ca₃AsI₃/Au structures demonstrated maximum efficiencies of 26.59 %, 19.70 %, 27.95 %, and 28.66 %, respectively, highlighting the promising potential of these configurations. Additionally, we applied a Ridge regressor-based ML model to predict the performance of Ca-perovskite solar cells (PSCs) with TiO₂ as the ETL. A dataset of 2187 data points from SCAPS-1D simulations was used, varying parameters such as absorber thickness, defect density, and TiO₂ properties. The model demonstrated high accuracy with an RMSE of 0.556 and an R-squared value of 0.6917, confirming its effectiveness in modeling Ca-PSC characteristics. These findings highlight the potential of Ca₃AsI₃ as a promising material for optoelectronic applications.