Abstract
Quantum chemical calculations of metal clusters in proteins for redox studies require both computational feasibility as well as accuracies of at least ∼50 mV for redox energies but only ∼0.05 Å for bond lengths. Thus, optimization of spin-unrestricted density functional theory (DFT) methods, especially the hybrid generalized gradient approximation functionals, for energies while maintaining good geometries is essential. Here, different DFT functionals with effective core potential (ECP) and full core basis sets for [Fe(SCH(3))(4)](2-/1-) and [Fe(SCH(3))(3)](1-/0), which are analogs of the iron-sulfur protein rubredoxin, are investigated in comparison to experiment as well as other more computationally intensive electron correlation methods. In particular, redox energies are calibrated against gas-phase photoelectron spectroscopy data so no approximations for the environment are needed. B3LYP gives the best balance of accuracy in energy and geometry compared B97gga1 and BHandH and is better for energies than Møller-Plesset perturbation theory series (MP2, MP3, MP4SDQ) and comparable to coupled cluster [CCSD, CCSD(T)] methods. Of the full core basis sets tested, the 6-31G** basis sets give good geometries, and addition of diffuse functions to only the sulfur significantly improves the energies. Moreover, a basis set with an ECP on only the iron gives a less accurate but still reasonable geometries and energies.