Accurate descriptions of the electronic properties of molecules, solids and other materials are crucial in the development of many modern technologies, and are therefore the subject of much scientific research. In tandem with experimental research, computational simulations play a huge role. Density functional theory (DFT) is the most popular method for performing ground-state electronic structure calculations due to its accuracy in many situations along with its computational efficiency. Its time-dependent extension (TDDFT) allows the modelling of the dynamics of systems of interacting electrons. (TD)DFT is an exact theory in principle, but requires approximations in practice, specifically to the many-body effects of electron exchange and correlation (xc), which are described by some unknown universal functional. Despite much success, it often proves inadequate in many scenarios of increasing interest. Therefore, we aim to contribute to the development of improved approximate density functionals. \n \nIn this thesis, we take the approach of studying small model systems for which we can compute the exact functionals. These systems are designed to exhibit the key features present in realistic quantum systems, such as atoms and molecules, which the exact universal (TD)DFT functional is capable of describing, but are often ill-described by the commonly used approximations. By analysing the nature of these exact functionals, we gain valuable insight into their fundamental properties, together with identifying where the current approximations are lacking, thereby allowing us to suggest how these approximations may be improved.
Density-Functional Theory (DFT), although widely used and very successful in the calculation of several observables, fails to correctly describe strongly correlated materials. In the first part of this work we, therefore, introduce reduced-density-matrix-functional theory (RDMFT) which is one possible way to treat electron correlation beyond DFT. Within this theory the one-body reduced density matrix (1-RDM) is used as the basic variable. Our main interest is the calculation of the fundamental gap which proves very problematic within DFT. In order to calculate the fundamental gap we generalize RDMFT to fractional particle numbers M by describing the system as an ensemble of an N and an N+1 particle system (with N≤ M≤N+1). For each fixed particle number, M, the total energy is minimized with respect to the natural orbitals and their occupation numbers. This leads to the total energy as a function of M. The derivative of this function with respect to the particle number has a discontinuity at integer particle number which is identical to the gap. In addition, we investigate the necessary and sufficient conditions for the 1-RDM of a system with fractional particle number to be N-representable. Numerical results are presented for alkali atoms, small molecules, and periodic systems. Another problem within DFT is the description of non-relativistic many- electron systems in the presence of magnetic fields. It requires the paramagnetic current density and the spin magnetization to be used as basic variables besides the electron density. However, electron-gas-based functionals of current-spin-density-functional Theory (CSDFT) exhibit derivative discontinuities as a function of the magnetic field whenever a new Landau level is occupied, which makes them difficult to use in practice. Since the appearance of Landau levels is, intrinsically, an orbital effect it is appealing to use orbital-dependent functionals. We have developed a CSDFT version of the optimized effective potential (OEP) method which allows for the use of explicitly orbital-dependent functionals in the context of CSDFT. We present the derivation of corresponding equations and show results for a quantum dot in external magnetic fields in the second part of this thesis.
Abstract This article is a short introduction to the modern computational techniques used to tackle the many-body problem in materials. The aim is to present the basic ideas, using simple examples to illustrate strengths and weaknesses of each method. We will start from density-functional theory (DFT) and the Kohn–Sham construction—the standard computational tools for performing electronic structure calculations. Leaving the realm of rigorous density-functional theory, we will discuss the established practice of adopting the Kohn–Sham Hamiltonian as approximate model. After recalling the triumphs of the Kohn–Sham description, we will stress the fundamental reasons of its failure for strongly-correlated compounds, and discuss the strategies adopted to overcome the problem. The article will then focus on the most effective method so far, the DFT+DMFT technique and its extensions. Achievements, open issues and possible future developments will be reviewed. The key differences between dynamical (DFT+DMFT) and static (DFT+ U ) mean-field methods will be elucidated. In the conclusion, we will assess the apparent dichotomy between first-principles and model-based techniques, emphasizing the common ground that in fact they share.
We establish theoretical foundations for embedding a correlated wavefunction in an environment formed by Kohn-Sham orbitals. We show that introducing an approximation that equates two, in principle, distinct kinetic-energy functionals yields an embedding functional identical to the projection-based wavefunction-in-density functional theory (DFT) formulation of Miller and co-workers. We demonstrate that this functional is inherently nonvariational: its minimum is not guaranteed to coincide with the exact ground-state energy and remains bounded from above by it. Building on this formal framework, we analyze the dominant sources of error in projection-based density matrix renormalization group-in-DFT embedding with approximate exchange-correlation (xc) functionals. Using molecules with dissociating covalent bonds as a diagnostic example, we demonstrate that the primary source of error is the nonadditive exchange-correlation energy describing the nonclassical coupling between the active subsystem and its environment. Eliminating the fractional-spin error by employing a pair-density xc functional (pair-density functional theory) instead of a semilocal generalized gradient approximation (GGA) does not remedy this deficiency, because the inaccuracy stems from self-interaction effects at the subsystem-environment interface.
The use of Density Functional Theory (DFT) has become a standard procedure in the recent years for the simulation of nanostructures. However it is well known that DFT suffers from severe limitations when covering the description of strongly correlated systems for which those approximations are not good.\n\nStrongly correlated systems can be described by means of tight-binding models. Such models are much simpler than the more realistic real-space models as they depend on few parameters and in some cases can be solved algebraically.\n\nThe first three chapters of this thesis are devoted to the use of the DFT machinery to the description of such lattice models in order to assess the true limitations when DFT is applied to strongly correlated systems.\n\nThe fourth and fifth chapters are oriented to the direct application of DFT to the study of two different and interesting problems: the search of molecular magnets with large magnetic anisotropy energy (MAE) and the study of molecular junctions using graphene leads.\n\nIn the first chapter we derive the exact functional and the exchange-correlation energies that let us building the Kohn-Sham (KS) potentials of the single and double site Anderson-Hubbard models in order to find out if the exact DFT gives the correct description of such systems. We find that this description is in general incomplete.\n\nThe second chapter presents an analysis of the double-site asymmetric Hubbard model with 2 electrons. The properties of its ground and excited states are discussed. We then use the Mean-Field (MF) and the Bethe-ansatz LDA (BALDA) approximations to assess their performance as a function of the external potential-interactions ratio. We then apply the concepts of DFT to this model, establishing its KS description and the corresponding adiabatic connection. A parametrized functional is also devised using a method that allows improving the accuracy of any approximated functional whenever the relation between the kinetic and the potential correlation terms is known.\n\nThe third chapter covers the DFT estimation of the fundamental gap in strongly correlated systems. We study the asymmetric Hubbard dimer. We find the exact quasi-particle excitation spectrum, including the ionization potential and electron affinity, which let us derive the real gap. We set a method to calculate the KS potentials for non-integer particle numbers as a necessary step to find the KS Green's function. We prove that the Koopmans' and the Janak's theorems are obeyed. We find that the KS spectrum is in general incorrect, except for the HOMO by definition, when strong correlations are dominant. The KS gap is always smaller than the true gap with interactions and we find the role of the derivative discontinuity on this.\n\nIn the fourth chapter we focus on the practical use of DFT and we test several organometallic molecules based with the objective of finding stable molecular magnets at room temperatures. The current technology uses thin-sheets for manufacturing magnetic bits for data storage. Different alternatives such as the use of single molecular magnets or clusters of transition metals to reduce the physical size of a bit. Heavy metal dimers may show huge magnetic anisotropies provided by a strong spin-orbit coupling (SOC). We analyze the magnetic properties of several organometallic molecules based on transition metals to determine if such huge anisotropies survive when the metal is encapsulated into the organic molecule.\n\nThe fifth chapter analyzes the use of graphene in the development of nanoelectronic circuits based on molecular junctions. The graphene has unique properties excellently matching the requirements for ideal electrodes. However the use of graphene nanoelectronics is difficulty because the shape of the graphene edges can change arbitrarily the coupling and switching functionality of the molecules attached. The aim of this chapter is to investigate the electronics and transport properties of different edge morphologies of graphene-wedge single-molecule junctions. It is shown that some graphene-wedge BDT and BPD junctions show negative differential resistance (NDR), spin-filtering (SP) behavior and Fano resonances.