797 publications from this institution
Langreth and Mehl (LM) and co-workers have developed a useful spin-density functional for the correlation energy of an electronic system. Here the LM functional is improved in two ways: (1) The natural separation between exchange and correlation is made, so that the density-gradient expansion of each is recovered in the slowly varying limit. (2) Uniform-gas and inhomogeneity effects beyond the randomphase approximation are built in. Numerical results for atoms, positive ions, and surfaces are close to the exact correlation energies, with major improvements over the original LM approximation for the ions and surfaces.
Talman and co-workers have presented a realization of the exact Kohn-Sham density-functional theory, neglecting correlation. From their numerical results we conclude that the occupied orbital energy eigenvalues of the local-spin-density (LSD) approximation for exchange and correlation are close to the exact Kohn-Sham orbital energies (apart from a constant shift), but that the latter do not accurately predict the removal energies of tightly-bound electrons in atoms, molecules, and solids. For the calculation of these removal energies, we propose an add-on, single-shot self-interaction correction (SIC) to the LSD orbital energies, based on a simplification and representation-invariant transformation of the original SIC method. This correction's relationship to the Dyson mass operator is briefly discussed.