From machine learning based on electronic structure design to strain engineering for improving optoelectronic performance

Lead-based perovskites have widespread applications in the photovoltaic field, with conversion efficiencies exceeding 26.41%. However, the toxicity and instability of lead pose significant barriers to large-scale commercialization. To overcome these limitations, double perovskite materials with s⁰+s² and s⁰+s⁰ electronic configurations were designed, and six machine learning models were employed to effectively predict the bandgap. Further optimize the optoelectronic properties of Cs₂AgSbX₆ (X=Cl, Br) by achieving a tunable bandgap through strain engineering. Within the range of -4% to 4%, the band gap responds almost linearly to external strain. Compression significantly reduces the band gap by 17% to 27%, enhances the interaction of anti-bonding orbitals, and increases the dispersion of band edge states, thereby lowering the effective mass of holes and improving carrier mobility. The device simulations indicate that tensile strain can effectively improve the theoretical efficiency of solar cells. This study elucidates the microscopic mechanisms of lattice structure and optoelectronic performance changes under strain, accelerating the exploration of high-performance perovskite photovoltaic devices through theoretical investigation. It emphasizes the potential of lattice strain engineering to enhance Photoelectric properties and expand applications in the optoelectronics field.