Saturation of Ruderman-Kittel-Kasuya-Yosida interaction damping in high-resistivity spin glasses.

Abstract

Saturation of the Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction damping at large resistivity or short mean free path in metallic alloys, as predicted in the Kaneyoshi model, is shown to account for the recently observed spin-glass freezing temperature, ${T}_{0}$, in such systems as AuFe quench-condensed films and ternary ${\mathrm{XY}}_{c}$${Z}_{x}$ alloys, where X=Au,Ag,..., is a nonmagnetic metal host, Y=Fe,Mn,..., is the dilute magnetic species of concentration c, and Z=Cu,...,Ti,Sb,..., is a nonmagnetic impurity of concentration x. Some deeper aspects of the c dependence of the characteristic RKKY interaction energy scale are discussed, with emphasis on the necessary distinction between quenched and ergodic situations in the randomly dilute alloys. A consequent logarithmic correction to the c-scaling laws (at the marginal dimensionalities d=p=3, where d is the electronic dimension of RKKY interaction varying as ${R}^{\mathrm{\ensuremath{-}}d}$, and p is the space dimension of the magnetic structure), in the form of ${T}_{0}$\ensuremath{\sim}c(-0.577-lnc${)}^{1/2}$, is shown to be due to broken dilatation invariance, by finite atomic size. The finite mean free path in real systems also breaks this invariance by providing a length scale. However, at the damping saturation limit a pseudo-c-scaling ${T}_{0}$\ensuremath{\sim}c reappears, as was found in the amorphous spin-glass LaAuGd. This, and related predictions of the ``typical environment'' approach to the quenched-random-averaging problem agree remarkably well with the new data that have recently become available.