Assessing the sensitivity of electron momentum densities and Fermi surfaces to different exchange-correlation approximations.

  • Eddie I Harris-Lee

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

Discrepancies between experimental and theoretical electronic properties, including electron momentum densities and Fermi surfaces, have been investigated. Attention has been given to the different ways that the different discrepancies may originate. Although it predicts the properties of a wide variety of periodic solids remarkably accurately, the local density approximation for density-functional theory is not without shortcomings. Here, the primary focus is on testing alternative first principles exchange-correlation approximations which have recently come to wider prominence. In particular, GW approximations, the strongly constrained and appropriately normed approximation for density-functional theory, and the power functional for reduced-density-matrix-functional theory have been tested.

The magnetic Compton profiles (MCPs) of Ni are not predicted satisfactorily when the LDA is used. It is found to be possible to predict the MCPs of Ni almost flawlessly (almost entirely to within the experimental error) with a non-self-consistent GW approximation. The interpretation of this result is discussed.

The Group V and Group VI transition metals share a common Fermi surface feature: the hole-like ellipsoids, which are centred on the N point in the Brillouin zone. These N ellipsoids arise from a band that has a significant proportion of p-character, in clear contrast to the other Fermi surface features. DFT approximations were found to overpredict the sizes of these N hole ellipsoids, whereas ‘quasi-particle self-consistent’ GW calculations predict sizes that are too small and non-self-consistent GW calculations (starting from the LDA) predict sizes that are much too small. Improvements to the accuracy of the Fermi surface prediction were found to coincide with a worsening of the prediction of magnetic properties.
Date of Award27 Sept 2022
Original languageEnglish
Awarding Institution
  • University of Bristol
SupervisorStephen B Dugdale (Supervisor) & Jude Laverock (Supervisor)

Cite this

'