Elimination of local-spin-density-approximation errors from electronic-structure calculations of Gd

All previous electronic-structure calculations of Gd have obtained a density of states N(EF) much larger than the experimental value. By comparing Hartree-Fock with local-spin-density-approximation (LSDA) atomic calculations we previously concluded that the minority-spin 4f resonance just above EF in the metal, which is responsible for the enhanced N(EF), is an artifact of the LSDA. We present here a calculation for the Gd crystal in which the conduction electrons experience a Hartree-Fock potential due to the core electrons and a LSDA potential due to themselves. The resonance is gone, N(EF) is less than the experimental value (many-body effects will increase it), and the lattice constants are in near-perfect agreement with experiment.