The accuracy of the pseudopotential approximation. III. A comparison between pseudopotential and all-electron methods for Au and AuH

Abstract

The quality of the pseudopotential approximation has been tested thoroughly by calculating spectroscopic properties of the gold atom and ground state AuH for eight different effective core potentials using Hartree–Fock, second-order Mo/ller–Plesset and coupled cluster methods. The pseudopotential valence basis set {φ}v for Au was chosen to be identical for all pseudopotentials, a subset of the all-electron basis set {φ}v⊂{φ}AE, and the condition was applied that all sets are of near basis set limit quality. The pseudopotential results are compared with data obtained from nonrelativistic, scalar relativistic Douglas–Kroll and fully relativistic four-component all-electron calculations. The variation between the results obtained for all valence electron small-core pseudopotentials and all electron Douglas–Kroll calculations is found to be small (for the Stuttgart pseudopotential Δre=0.001 Å, ΔDe=0.03 eV, Δωe=9 cm−1, Δμe=0.04 D). Sizable differences to all electron results are only found for the 11 valence electron large-core pseudopotentials. The effects of the basis set superposition error on spectroscopic constants were investigated. Calculated coupled cluster electron affinities and ionization potentials for gold and spectroscopic properties for AuH were found to be in excellent agreement with available experimental data. The variation between the different small-core pseudopotentials for one particular spectroscopic property is shown to be less than the error due to the incompleteness of electron correlation procedure or the basis set and approximately of the same size as the basis set superposition error. The results show that scalar relativistic effects for valence properties are perfectly described by the pseudopotential approximation.