Quantum :Monte Carlo calculation of the Fe atom

Abstract A variety of earlier calculations for the third-row transition element iron using argon-core pseudopotentials had failed to give good agreement with experimental measurements of the ionization potential, the electron affinity, and excitation energies. This paper reports QMC calculations which give excellent agreement and which are likely to be the most accurate of any type of calculation for the Fe atom in the 1990s. For these the author used neon-core pseudopotentials (leaving 16 electrons, including 3s and 3p electrons, in the valence space) for both variational and fixednode diffusion QMC calculations. The pseudopotential used was an ab initio pseudopotential tested against all-electron results in limited configuration interaction calculations. Relativistic effects were included. The complete Hamiltonian was modified slightly to facilitate treatment of the nonlocal part of the pseudopotential expression. The trial functions for importance sampling and specifying nodes were linear combinations of Slater determinants along with Jastrow functions. Six low-lying neutral states were examined in addition to the ground states of the anion and cation. Recovery of valence correlation energies was 0.6 to 0.8 hartrees for all of these, typically 0.1 to 0.2 hartrees greater that of coupled cluster and configuration interaction calculations. Most impressive is the agreement with experimental measurements: ionization potential, DQMC 7.67(6) eV vs. expt. 7.87 eV; 5  D  3  F transition, DQMC 4.24(9) eV vs. expt. 4.07 eV; 5  D  5  F transition, DQMC 0.84(6) eV vs. expt.