Assessing Cu3BiS3 for Thin-Film Photovoltaics: A Systematic DFT Study Comparing LCAO and PAW Across Multiple Functionals

Cu₃BiS₃ (CBS) has emerged as a promising earth-abundant absorber for thin-film photovoltaics, offering a sustainable alternative to conventional technologies. However, ab initio studies on its optoelectronic properties remain scarce and often yield contradictory results. This study systematically examines the influence of two density functional theory (DFT) methodologies, linear combination of atomic orbitals (LCAO) and projector augmented wave (PAW), on the structural and electronic properties of CBS, aiming to establish a reliable computational framework for future research. With this in mind, we also assessed the impact of a wide range of exchange-correlation (XC) functionals within both methods, including 6 from the local density approximation (LDA) family (HL, PW, PZ, RPA, Wigner, XA), 10 from the generalized gradient approximation (GGA) family (BLYP, BP86, BPW91, GAM, KT2, PBE, PBEsol, PW91, RPBE, XLYP), 2 meta-GGA functionals (SCAN, R2SCAN), and the hybrid HSE06 functional. Both LCAO and PAW consistently predict an indirect bandgap for CBS across all XC functionals, aligning with most previous DFT studies but contradicting experimental reports of a direct transition. The LDA and meta-GGA functionals systematically underestimated the CBS bandgap (<1 eV), with further reductions upon structural relaxation. GGA functionals performed better, with BLYP and XLYP yielding the most experimentally consistent results. The hybrid HSE06 functional substantially overestimated the bandgap (1.9 eV), with minimal changes after relaxation. The calculated hole and electron effective masses reveal strong anisotropy along the X, Y, and Z crystallographic directions. Additionally, CBS exhibits an intrinsic p-type nature, as the Fermi level consistently lies closer to the valence band maximum across all methods and functionals. However, the PAW method generally predicted more accurate lattice parameters than LCAO; the best agreement with experimental values was achieved using the PW91 (1.2% deviation) and HSE06 (0.9% deviation) functionals within LCAO. Based on these findings, we recommend the PW91 functional with LCAO for structural optimizations in large supercell studies of CBS dopants and/or defects and BLYP/XLYP for electronic properties.