The many-body perturbation theory in the Hedin's $GW$ approximation is nowadays a common tool to carry out calculations in solids and molecular systems. Nevertheless a lot of theoretical issues concerning the self-interaction problem remain unresolved in the basic theory of this approximation. In this paper we carry out calculations in some very simple exactly solvable model systems and demonstrate that at this level of approach there are self-interaction problems in the correlation part that seriously affect the final results. This is especially important in the low-density regime. We also have carried out the calculations using different input data coming from the local-density approximation (LDA), generalized gradient approximation (GGA), and exact solutions of the Schr\"odinger equation to check how much affected are the final ${G}_{0}{W}_{0}$ results by the election of the input data. We have found that the final results inherit the errors from the input data and that only good results can be achieved if the input data are very close to (or are) the exact ones.
The many-body perturbation theory in the Hedin's GW approximation is nowadays a common tool to carry out calculations in solids and molecular systems. Nevertheless a lot of theoretical issues concerning the self-interaction problem remain unresolved in the basic theory of this approximation. In this paper we carry out calculations in some very simple exactly solvable model systems and demonstrate that at this level of approach there are self-interaction problems in the correlation part that seriously affect the final results. This is especially important in the low-density regime. We also have carried out the calculations using different input data coming from the local-density approximation (LDA), generalized gradient approximation (GGA), and exact solutions of the Schr"odinger equation to check how much affected are the final G0W0 results by the election of the input data. We have found that the final results inherit the errors from the input data and that only good results can be achieved if the input data are very close to (or are) the exact ones.
In many-body perturbation theory (MBPT) the self-energy $\ensuremath{\Sigma}=iGW\ensuremath{\Gamma}$ plays a key role since it contains all the many-body effects of the system. The exact self-energy is not known; as a first approximation one can set the vertex function $\ensuremath{\Gamma}$ to unity which leads to the GW approximation. The latter properly describes the high-density regime, where screening is important; in the low-density regime, instead, other approximations are proposed, such as the T matrix, which describes multiple scattering between two particles. Here we combine the two approaches. Starting from the fundamental equations of MBPT, we show how one can derive the T-matrix approximation to the self-energy in a common framework with GW. This allows us to elucidate several aspects of this formulation, including the origin of, and link between, the electron-hole and the particle-particle T matrix, the derivation of a screened T matrix, and the conversion of the T matrix into a vertex correction. The exactly solvable Hubbard molecule is used for illustration.
A systematic investigation of two approximate self-interaction corrections (SICs), Perdew-Zunger SIC and Lundin-Eriksson SIC, and the local-density approximation (LDA) is performed for a model Hamiltonian whose exact many-body solution and exact LDA are known. Both SICs as well as LDA are applied in the calculation of ground-state energies, ground-state densities, energy gaps, and impurity densities of one-dimensional Hubbard chains differing in size, particle number, and interaction strength. The orbital-dependent potentials arising from either SIC are treated within the optimized-effective potential method, which we reformulate for the Hubbard model. The delocalization tendency of LDA is confronted with the localization tendency of SIC. A statistical analysis of the resulting data set sheds light on the role of SIC for weakly and strongly interacting particles and allows one to assess the performance of each methodology.
A systematic comparison of three approximate self-interaction corrections (SICs), Perdew-Zunger SIC, Lundin-Eriksson SIC and extended Fermi-Amaldi SIC, is performed for a model Hamiltonian whose exact many-body solution and exact local-density approximation (LDA) are known. For each of the three proposals we compare its implementation only for the potential, only for the energy, i.e., a post-LDA evaluation of the SIC energy), to none of them, i.e., a standard LDA calculation) and to both. Each of the resulting 10 permutations of methodologies is applied to 420 Hubbard chains differing in size, particle number and interaction strength. A statistical analysis of the resulting data set reveals trends and permits to assess the performance of each methodology. Overall, but not in each individual case, a post-LDA application of Perdew-Zunger SIC emerges as the recommended methodology.
There is increasing interest in many-body perturbation theory as a practical tool for the calculation of ground-state properties. As a consequence, unambiguous sum rules such as the conservation of particle number under the influence of the Coulomb interaction have acquired an importance that did not exist for calculations of excited-state properties. In this paper we obtain a rigorous, simple relation whose fulfilment guarantees particle-number conservation in a given diagrammatic self-energy approximation. Hedin's G0W0 approximation does not satisfy this relation and hence violates the particle-number sum rule. Very precise calculations for the homogeneous electron gas and a model inhomogeneous electron system allow the extent of the nonconservation to be estimated.
We introduce a scheme through which the ground-state electronic density n and total energy Ev [n] of inhomogeneous interacting systems can be determined in a self-consistent (SC) way. It is based on a SC perturbation expansion for the single-particle Green function G. Our employed expansion is argued to be valid under all conditions provided that the ground-state charge density of the interacting system is pure-state non-interacting v representable. The formalism discussed in this paper has been applied to a quasi-one-dimensional periodic system. It has even been possible to carry out SC calculations with the dynamically screened exchange diagram (the so-called GW diagram, where W is the dynamically screened electron-electron interaction function) for the self-energy operator Σ. A coherent mode of interpretation of a number of our results entails that the overwhelming success of the GW framework in relatively accurately reproducing the experimental energy gaps of a number of semiconductors and insulators may be a direct consequence of non-self-consistency effects; whereas a non-self-consistent calculation in our tests increases the value of the local-density approximation (LDA) gap in the range of 40–65% of the ‘experimental’ gaps, self-consistency effects reduce this range to 14–17%. The same line of argument leads us to conclude that, in our model, 26–35% of the LDA error in the gap energies is due to the absence of a well known discontinuity in the LDA exchange-correlation potential, and the remaining part is due to the Kohn-Sham eigenvalues not being identifiable with one-particle excitation energies. A natural conclusion that may be drawn from these observations is that a correct description of excitation energies (e.g. of the gap energy) cannot be achieved without including, at least approximately, the vertex part of Σ, which is entirely absent in the GW scheme.
The total energy and electron addition and removal spectra can, in principle, be obtained exactly from the one-body Green's function (GF). In practice, the GF is obtained from an approximate self-energy. In the framework of many-body perturbation theory, we derive different expressions that are based on an approximate self-energy, but that yield nevertheless, in principle, the exact exchange-correlation contribution to the total energy for any interaction strength. Response functions play a crucial role, which explains why, for example, ingredients from time-dependent density functional theory can be used to build these approximate self-energies. We show that the key requirement for obtaining exact results is the consistent combination of ingredients. Also when further approximations are made, as it is necessary in practice, this consistency remains the key to obtain good results. All findings are illustrated using the exactly solvable symmetric Hubbard dimer.