Potential energy curves and electronic structure of 3d transition metal hydrides and their cations

We investigate gas-phase neutral and cationic hydrides formed by 3d transition metals from Sc to Cu with density functional theory (DFT) methods. The performance of two exchange-correlation functionals, Boese-Martin for kinetics (BMK) and Tao-Perdew-Staroverov-Scuseria (TPSS), in predicting bond lengths and energetics, electronic structures, dipole moments, and ionization potentials is evaluated in comparison with available experimental data. To ensure a unique self-consistent field (SCF) solution, we use stability analysis, Fermi smearing, and continuity analysis of the potential energy curves. Broken-symmetry approach was adapted in order to get the qualitatively correct description of the bond dissociation. We found that on average BMK predicted values of dissociation energies and ionization potentials are closer to experiment than those obtained with high level wave function theory methods. This agreement deteriorates quickly when the fraction of the Hartree-Fock exchange in DFT functional is decreased. Natural bond orbital (NBO) population analysis was used to describe the details of chemical bonding in the systems studied. The multireference character in the wave function description of the hydrides is reproduced in broken-symmetry DFT description, as evidenced by NBO analysis. We also propose a new scheme to correct for spin contamination arising in broken-symmetry DFT approach. Unlike conventional schemes, our spin correction is introduced for each spin-polarized electron pair individually and therefore is expected to yield more accurate energy values. We derive an expression to extract the energy of the pure singlet state from the energy of the broken-symmetry DFT description of the low spin state and the energies of the high spin states (pentuplet and two spin-contaminated triplets in the case of two spin-polarized electron pairs). The high spin states are build with canonical natural orbitals and do not require SCF convergence