Quasiparticle calculations in atoms and many-body core-valence partitioning

Abstract

The central work of this thesis is many-body Green's Function Method calculations in atoms using Hedin's GW approximation with various vertex corrections, leading to polarizabilities and quasiparticle energies, which are electron addition (removal) energies corresponding to the lowest few unoccupied (highest few occupied) electron states. Issues of gauge symmetry, conservation laws, Fermion statistics (crossing symmetry), and shake-up effects are discussed. We find GW with generalized RPA vertex corrections treats accurately ion core dipole polarizabilities and corrections to binding energies for one electron to stripped cores for many types of elements. Ordinary GW improves significantly over single-body method eigenvalues in open-shell atoms, though is still quite inaccurate in predicting s $-$ d promotion energies in iron series elements. This suggests limitations in the applicability of GW in highly correlated solids. In addition to studying atomic many-body theory, we use our GW atom results to formulate an explicitly many-body approach to core-valence partitioning by fitting "core-polarization potentials" to GW's corrections beyond Hartree-Fock. This is a means of deriving an appropriate, effective valence Hamiltonian which is more rigorous than are usual approaches such as Hartree-Fock or local-density-functional theory. We test our valence Hamiltonians by carrying out virtually exact valence calculations in atoms and molecules, obtaining definitely improved agreement with experiment of predicted quantities. We also present local density-functional results in solids indicating that our method of core-valence partitioning should also affect solid-state many-body calculations.