Resonant band-electron–f-electron scattering theory for highly correlated actinide systems

In earlier studies we recognized that the highly correlated behavior of the f electrons within moderately delocalized light actinide (uranium, neptunium, plutonium) systems is linked to the non-f-band behavior via the hybridization process. By transforming the hybridization into a band-electron--f-electron resonant scattering from the correlated multiplet states of the actinide ions, and considering only the scattering processes that involve f electrons in the ${m}_{l}$=0, ${m}_{s}$=\ifmmode\pm\else\textpm\fi{}1/2 states (for quantization along the interionic axis) which dominate the two-ion interactions, our earlier work explained the main features of the anisotropic magnetic equilibrium behavior for the PuSb system but failed to reproduce the correct polarization (longitudinal) for the long-period antiferromagnetic structure observed in the temperature range below the N\'eel temperature. In this paper we include the next-to-dominant scattering channels (single-site scattering processes involving f electrons with ${m}_{l}$=\ifmmode\pm\else\textpm\fi{}1, ${m}_{s}$=\ensuremath{\mp}1/2). This refinement changes the angular dependence of the anisotropic interaction, and successfully yields the ferromagnetic to longitudinally polarized long-period antiferromagnetic phase transition as is experimentally observed. Excellent agreement with experiment for the correlation length anisotropy is also obtained. For the magnetic excitation behavior in the ferromagnetic phase pertinent to PuSb at T=0, the theory gives a spectrum with two polarized branches at the zone boundary for q along the [100] (transverse-to-moment) direction. In fact, the predicted excitation behavior is rather remarkable. The appearance of two polarized branches rather than a single branch at the zone boundary occurs only over an extremely narrow range of crystal-field splitting. We choose the crystal-field splitting to give two branches, and this unadjustably yields excitation energies that are very close to the experimental values. An only slightly different crystal-field value would give neither two branches nor correct excitation energies.