1-370 GHz EPR linewidths for K3CrO8: a comprehensive test for the Anderson-Weiss model.

Electron paramagnetic resonance (EPR) measurements have been carried out over the frequency range of 1-370 GHz on single crystals of potassium peroxychromate (K3CrO8) with the view of examining the current models of exchange narrowing of EPR signals in solids. K3CrO8 has a simple (tetragonal) lattice structure, can be grown as single crystals pure or diluted with an isostructural diamagnetic host K3NbO8, and its paramagnetism can be described by a very simple (S = 12, I = 0) spin Hamiltonian. The measurements were made at various orientations of single crystals in the Zeeman field, with emphasis on the principal directions of the g-tensor. For essentially all orientations, the linewidth decreases monotonically for measurements at resonance frequencies, omega0, from 1 to about 100 GHz, and then starts to increase at higher omega0. In order to delineate the spin exchange effects from other sources of line broadening, the measurements were repeated with a diluted spin system, K3NbO8 containing approximately/= 0.5 mole % of K3CrO8, representing the broadening effect of all the magnetic field dependent terms, such as the broadening due to the g-strain and sample holder/waveguide magnetization at the high field utilized, up to 14 T. Using these data, the K3CrO8 linewidths were analyzed in terms of the current models of spin exchange narrowing in three-dimensional systems. A reasonably good agreement was found with the Anderson-Weiss model, when modified for various line broadening effects. The accuracy of the analysis procedure was confirmed by the comparison of the presently determined values of the exchange constant, J, and the dipolar field, Hp, with their values obtained by dc magnetic susceptibility measurements and theoretical analysis, respectively; the agreement was within 5% for J (=1.35 K) and about 25% for Hp (160 G). However, some deviations and unusual splittings were noted in measurements at 370 GHz, whose origin remains unclear.