Systematic improvement and applications of density functional theory in complex chemical systems

Abstract

Density functional theory (DFT) has become the de facto method for quantum mechanical simulations of molecules and solids because of its high performance to cost ratio. In this thesis, we discuss two aspects of DFT simulations in complex chemical systems: systematic improvement of the accuracy of density functional approximations and proper utilization of DFT methods for efficient modeling of electronic properties. We first develop the many-pair expansion (MPE) method, which is a density functional hierarchy that systematically corrects any deficiencies of an approximate density functional to converge to the exact energy. We show that MPE gives accurate results for 1D/2D Hubbard and ID Peierls-Hubbard models, suggesting its ability to remove strong correlation errors. Applying MPE to unsaturated hydrocarbons in the Pariser-Parr-Pople lattice model, we find that it deals very well with dispersion interactions. Afterwards, we describe our efforts to implement MPE for molecular systems. A new density decomposition method, self-attractive Hartree (SAH), is developed to generate localized and smooth fragment densities. The SAH decomposition is shown to be useful for extracting chemical bonding information directly from the electron density and further applied to develop a simple and accurate hydrogen bonding strength indicator. Using SAH fragment densities, we demonstrate that MPE provides accurate description of reaction energies and bond breaking processes for a few small molecules, even with a low-level starting functional and low orders of expansion. To show how DFT methods can be properly utilized to obtain electronic properties of interest, we employ the theoretical investigation of organic light-emitting diodes (OLEDs) as an example. We adopt a hybrid quantum mechanics/molecular mechanics (QM/MM) approach to reveal the charge and energy mechanisms of a host-guest phosphorescent OLED in condensed phase, emphasizing the importance of incorporating environment effects. We then show successful computational design of new thermally activated delayed fluorescence (TADF) materials using conventional time-dependent DFT method, while point out the need for better excited-state DFT methods. Finally, we develop efficient computational screening protocols to study TADF materials based on a restricted open-shell Kohn-Sham approach.