Composition-dependent electronic, optical, and thermoelectric properties of CsIn1-xGaxBr3 from DFT and machine learning

A combined density functional theory (DFT) and machine-learning (ML) study is conducted to examine the composition-dependent characteristics of lead-free perovskite materials CsIn1-xGaxBr3 (0 ≤ x ≤ 1) for absorption-based optoelectronic and thermoelectric applications. First-principles calculations employing PBE and PBEsol exchange–correlation functionals demonstrate a continuous lattice expansion from 11.20 Å (CsGaBr3) to 11.68 Å (CsInBr3) in accordance with Vegard’s law. In addition, the negative formation energies between −0.40 and −0.44 eV/atom affirming thermodynamic stability throughout the alloy series. Analysis of the electronic structure indicates that all compositions are indirect-gap semiconductors characterized by Γ-L transitions, with a pronounced nonlinear bandgap bowing that decreases from 0.856 eV (CsGaBr3) to a minimum of 0.505 eV at x = 0.5, followed by a reopening toward Ga-rich compositions. Optical spectra reveal a systematic red shift of the absorption edge at intermediate Ga concentrations and an augmented dielectric response in Ga-rich alloys, resulting from heightened Ga/Br orbital hybridization and conduction-band dispersion. Boltzmann transport model predicts n-type thermoelectric performance, with Seebeck coefficients approaching 1.2 × 10−3 V K−1 and a figure of merit ZT reaching 0.8 for CsGaBr3 and CsInBr3 at 300 K. A physics-informed machine-learning surrogate model, trained on DFT-calculated bandgaps (using PBE and PBEsol), precisely reproduces the compositional trend (R² = 0.972), facilitating rapid interpolation across the complete alloy spectrum. These findings establish CsIn1-xGaxBr3 as a versatile, eco-friendly perovskite framework and illustrate the efficacy of combining first-principles simulations with machine learning to expedite materials design for infrared absorption-based optoelectronic and thermoelectric applications.