One of the great challenges of modern condensed matter theory is to develop reliable and practical methods for describing the electronic structure of strongly correlated materials fully from first principles. It has been known for a long time that the widely used Kohn-Sham density functional theory within the local density approximation (LDA) often fails for these systems. This paper describes the theoretical development of including electron correlations beyond the LDA with emphasis on ab initio methods, starting from the highly successful LDA+U and the many-body Green's function-based GW method to the state-of-the-art combination of dynamical mean-field theory and GW. Apart from reviewing the development so far we also present some previously unpublished results.
We present the current-density functional theory for the superconductor immersed in the magnetic field. The order parameter of the superconducting state, transverse component of the paramagnetic current-density, and electron density are chosen as basic variables that uniquely determine the equilibrium properties of the system. In order to construct this theory, the development of the approximate form of the exchange-correlation (xc) energy functional is indispensable as well as the derivation of the effective single-particle equation which makes it possible to reproduce the equilibrium densities mentioned above. The rigorous expression of the xc-energy functional is derived using the technique of the coupling-constant integration. Furthermore, the approximate form of the xc energy functional is proposed such that the energy gap resulting from the effective single-particle equation is consistent with the attractive interaction energy of the system.
A density-functional theory is developed for fermions in one dimension, interacting via a $\ensuremath{\delta}$ function. Such systems provide a natural testing ground for questions of principle, as the local-density approximation should be highly accurate since for this interaction type the exchange contribution to the local-density approximation is intrinsically self-interaction-free. The exact-exchange contribution to the total energy is a local functional of the density. A local-density approximation for correlation is obtained using perturbation theory and Bethe ansatz results for the one-dimensional contact-interacting uniform Fermi gas. The ground-state energies are calculated for two finite systems, the analogs of helium and of Hooke's atom. The local-density approximation is shown to be excellent as expected.
A framework for developing new approximate electronic structure methods is presented, in which the correlation energy of a many-electron system in the ground state is computed as in the single-reference second-order many-body perturbation theory but with the reference one-body Hamiltonian modified by additional (generally non-local) potentials which are universal functionals of the Hartree-Fock density. The existence of such functionals that reproduce the exact correlation energy justifies the search for approximate models which may overcome some important deficiencies of the traditional density-functional methods and also perform better than the usual low-order wavefunction methods.
A perturbation theory is developed for the correlation energy
Ec[n], of a finite-density system, with respect to the coupling constant \ensuremath{\alpha} which multiplies the electron-electron repulsion operator in ${\mathit{H}}^{\mathrm{\ensuremath{\alpha}}}$=T^+\ensuremath{\alpha}V${\mathrm{^}}_{\mathit{e}\mathit{e}}$+
tsumi${\mathit{v}}_{\mathrm{\ensuremath{\alpha}}}$(
ri). The external potential ${\mathit{v}}_{\mathrm{\ensuremath{\alpha}}}$ is constrained to keep the gound-state density n fixed for all \ensuremath{\alpha}\ensuremath{\ge}0. ${\mathit{v}}_{\mathrm{\ensuremath{\alpha}}}$ is given completely in terms of functional derivatives at full charge (\ensuremath{\alpha}=1), from which
Ec[${\mathit{n}}_{\ensuremath{\lambda}}$]=
ec,2[n]+ ${\ensuremath{\lambda}}^{\mathrm{\ensuremath{-}}1}$
ec,3[n]+${\ensuremath{\lambda}}^{\mathrm{\ensuremath{-}}2}$
ec,4[n]+..., where each
ec,j[n] is expressed in terms of integrals involving Kohn-Sham determinants. Here, ${\mathit{n}}_{\ensuremath{\lambda}}$(x,y,x)=${\ensuremath{\lambda}}^{3}$n(\ensuremath{\lambda}x,\ensuremath{\lambda}y,\ensuremath{\lambda}z) and \ensuremath{\lambda}=${\mathrm{\ensuremath{\alpha}}}^{\mathrm{\ensuremath{-}}1}$. The identification of ${\mathrm{lim}}_{\ensuremath{\lambda}\ensuremath{\rightarrow}\mathrm{\ensuremath{\infty}}}$
Ec[${\mathit{n}}_{\ensuremath{\lambda}}$], which is a high-density limit, as the second-order energy
ec,2[n] allows one to compute bounds upon ${\mathrm{lim}}_{\ensuremath{\lambda}\ensuremath{\rightarrow}\mathrm{\ensuremath{\infty}}}$
Ec[${\mathit{n}}_{\ensuremath{\lambda}}$]; the bounds imply that ${\mathrm{lim}}_{\ensuremath{\lambda}\ensuremath{\rightarrow}\mathrm{\ensuremath{\infty}}}$
Ec[${\mathit{n}}_{\ensuremath{\lambda}}$]\ensuremath{\simeq}
Ec[n] for a large class of small atoms and molecules, and suggest that ${\mathrm{lim}}_{\ensuremath{\lambda}\ensuremath{\rightarrow}\mathrm{\ensuremath{\infty}}}$
Ec[${\mathit{n}}_{\ensuremath{\lambda}}$] should be of the same order of magnitude as
Ec[n] in finite insulators and semiconductors.Approximations to
Ec[n] should reflect all this. In contrast, perhaps the well-known overbinding of the local-density approximation (LDA) in molecules and solids is due, in part, to the fact that the LDA correlation energy is too sensitive to a coordinate scaling of n. Indeed, the LDA for
Ec[${\mathit{n}}_{\ensuremath{\lambda}}$] diverges when \ensuremath{\lambda}\ensuremath{\rightarrow}\ensuremath{\infty} because of the presence of the -ln(\ensuremath{\lambda}) term in the Gell-Mann and Brueckner high-density expression for the correlation energy, per particle, of a homogeneous density, which is infinite. In a sense, the derived perturbation expansion transforms the Gell-Mann and Brueckner expression into one that applies specifically to an inhomogeneous density which integrates to a finite number of electrons.
In an attempt to obtain a better description than the local-density approximation (LDA) for the electronic structure of realistic narrow-band systems, we have studied a perturbation-theory method for the self-energy and have tested it on the electronic properties of the periodic one-dimensional Anderson model. For a number of models we have calculated a wide range of properties such as quasiparticle energies, spectral densities, reduced hybridizations, and various magnetic correlation functions. For two different models we were able to compare with Monte Carlo results and found good agreement. In general, we have obtained a description of narrow-band systems that is more reasonable than the one given by the LDA.