Electronic Excitations in Transition Metal Dioxide and Open-Shell Molecular Systems

Many-body perturbation within the GW approximation is a powerful method to calculate the electron removal and addition energies in a wide variety of bulk and molecular systems. Over the last decade, there have been an increasing number of studies that have benchmarked the performance of this method in molecular systems. Most of these studies have focused on closed-shell and sp-bonded molecules. In this thesis, I investigate the performance of different flavors of the GW approximation in two systems: (i) negatively charged 3d-transition metal dioxide molecules, and (ii) systems that exhibit multiplet splitting in their photoelectron spectra. These are systems of both technological and scientific interest, and they present significant computational challenges for straightforward application of GW methods due to enhanced electron correlations and their open-shell character. For the first system, I present results and analyses of photoelectron spectra of early 3d-transition metal dioxide molecular anions TMO2- (TM = Sc, Ti, V, Cr, Mn) computed using semi-local and hybrid density functional theory (DFT) and the GW approximation with perturbative and eigenvalue self-consistent formalisms. The results are compared with each other and with experimental photoelectron spectroscopy data. I show that perturbative GW with a particular fraction of exact exchange (25%) on top of Perdew-Burke-Ernzerhof exchange-correlation functional provides excellent agreement with experiment by mitigating self-interaction error. I also show that an eigenvalue self-consistent formalism with updates in the Green’s function is a reasonably accurate choice for computing electron removal energies. In the second project, I investigate the performance of two different perturbative GW approaches with hybrid functional DFT starting points for five sp-bonded open-shell molecules, NO2, NF2, ClO2, O2, and S2, containing multiplet-split orbitals in their photoelectron spectra. My results show that (i) typical semi-local exchange-correlation functionals do not provide good agreement with experimental data for multiplet-split orbitals, and (ii) it is possible to accurately predict multiplet splitting in open-shell molecules by adding a fraction of exact exchange in the DFT starting point.