Contribution of Spin Polarization to Transferred Hyperfine Effects in Iron-Series Fluorides

Abstract

Theoretical interpretations of the transferred hyperfine spectra in iron-series fluorides have followed several distinct models for explaining the large effective magnetic fields which are observed at the F19 nucleus. The unpairing of the electron spins in the fluoride orbitals, responsible for these fields, is described as the result of either (1) an admixture of covalent bonding into the purely ionic configuration1 (i.e., the metal ion 3d wavefunctions have been augmented by small amounts of fluorine functions of the proper symmetry), or (2) the unpairing action of the Pauli principle on the wholly ionic configuration which affects those fluoride orbitals, which have the same spin as the cation 3d orbital, differently from those orbitals which have opposite spin.2 In either method the unpaired fluorine s electrons produce an isotropic hfs (from the contact part of the Fermi interaction) whereas the 2p electrons are responsible for the anisotropic interaction. Although existing analyses of transferred hyperfine spectra in iron-series fluorides have generally been considered successful, they have, of necessity, omitted a number of terms which contribute to the hyperfine field.3 One of these is the spin (or exchange) polarization of the F− electrons by the net spin density of the unfilled 3d shells on the cation. This polarization contribution exists in addition to the usual polarization terms arising from the nonorthogonality of anion and cation wave function. A major reason why accurate computations of transferred hyperfine effects including spin polarization cannot be done is our current inability to describe accurately many-center exchange interactions in solids. We have carried out several spin-polarized Hartree-Fock calculations for an Mn–F–Mn system, such as exists in KMnF3, using different crude exchange ``potentials'' in order to give some indication of the magnitude of the effect. To make them consistent with methods (1) and (2) above, the exchange potentials were derived from either purely ionic or partially covalent 3d spin densities. Results for both the isotropic and anisotropic parts of the interaction have been obtained. The ionic and covalent ``potentials'' each give contributions to the isotropic term which are several times larger in magnitude but differ in sign with experiment.4 A surprising feature of the calculations (in view of the large calculated effect) is the relative insensitivity of the computational results to the assumed form of the cation spin density, i.e., whereas the effect on the 3d spin density is considerable when going from the ionic to the strongly covalent configuration, the anisotropic interaction is changed by only 5% while the isotropic interaction is reduced by 40% (which is still twice the experimental value). Experiments on4 Cr3+, where the isotropic covalent term does not occur, show the isotropic spin polarization contribution (plus any other small terms) to be appreciably smaller than these calculations suggest. Thus, difference appears to be primarily due to the poor description of interatomic exchange used in these calculations. However, no such experimental data are available for the anisotropic spin polarization terms which our calculations suggest are relatively more important than the isotropic part. A detailed report of this work will be submitted to The Physical Review.