Abstract
Understanding the structure of metal-ligand complexes is essential for catalyst design, materials development, and biochemical modeling. Metal carbonyls are especially relevant due to their diverse structures and electronic features. Here, we benchmarked seventeen density functionals (B3LYP, BP86, CAM-B3LYP, M06, M06L, PBE, PBE0, r(2)SCAN, r(2)SCAN-3c, revPBE, revTPSS, RPBE, TPSS, TPSS0, TPSSh, ω B97, and ω B97X) combined with three dispersion schemes (D3zero, D3BJ, D4) and also tested calculations without dispersion correction, totaling fifty-four approaches. Their ability to reproduce geometries, structural parameters, and CO stretching frequencies was assessed for thirty-four Mn(I) and Re(I) carbonyls obtained from the CCDC. Relative electronic energies were further compared using DLPNO-CCSD(T) calculations, alongside evaluation of computational cost. Our results highlight that hybrid meta-GGA and meta-GGA functionals, particularly TPSSh(D3zero) and r(2)SCAN(D3BJ, D4), offer the best balance between accuracy and efficiency, providing reliable structures, vibration properties, and energetics consistent with high-level DLPNO-CCSD(T) references.