LOCAL-DENSITY APPROXIMATION - COHESION IN THE TRANSITION-METALS AND S-]D PROMOTION IN THE TRANSITION-METAL ATOMS

Systematics in cohesive energies (H(coh)) and d-non-d atomic promotion energies have been examined for the transition elements treated within the local-density approximation (LDA). Cohesive energies involve the energy of the solid as compared with that of a reference state in the free atom; going from one atomic configuration to another (e.g., d(n-2)S2-->d(n-1)S) for that reference state involves a promotion energy. Errors in the LDA's ability to calculate promotion energies are then translated into changes in calculated H(coh). Employing the local-spin-density approximation (LSDA), scalar-relativistic values of the promotion energies have been obtained for atomic states of maximum spin multiplicity of the neutral atoms in the 3d, 4d, and 5d rows for those configurations for which experimental spectral data are available for comparison. The intent is that by scanning all three rows and those cases for which there are experimental data that those factors contributing to the LDA's shortcomings in describing electron-electron interactions in the transition elements may become better defined. Previously, other workers have obtained d(n-2)S2-->d(n-1)S promotion energies for the 3d row, indicating that the LDa (or LSDA) significantly overestimates the stability of d valence electrons as compared with the non-d. The more extensive results obtained here indicate that, while often significant, s-->d promotion energy errors are sometimes essentially zero valued. This variation in s-->d promotion energy behavior has implications for what might be presumed to be the shortcomings of the LDA as applied to atoms and, in turn, to solids. Given the promotion energies, the consequences of choosing different references states in estimates of H(coh) for the 4d and 5d rows are explored and an envelope of H(coh) values defined. No matter what the choice of reference state, the LDA significantly overestimates H(coh) in the middle of the transition-element rows, a result consistent with previous estimates. This error becomes small upon going to the noble metals and, as has not been generally recongnized, is essentially zero valued for the beginning members of each row. These matters are discussed.