Functional Keldysh Theory of Spin Torques

TL;DR

This paper develops a microscopic Keldysh formalism to analyze current-induced spin torques and thermal fluctuations in ferromagnets, providing new insights into the parameters governing magnetization dynamics.

Contribution

It introduces a functional Keldysh approach to derive a stochastic Landau-Lifschitz-Gilbert equation with calculated spin transfer torque parameters.

Findings

The non-adiabatic torque parameter β differs from Gilbert damping α.

β/α approaches one only in a simple isotropic-Stoner model with small exchange splitting.

Thermal fluctuations are incorporated naturally and validated at first order in electric field.

Abstract

We present a microscopic treatment of current-induced torques and thermal fluctuations in itinerant ferromagnets based on a functional formulation of the Keldysh formalism. We find that the nonequilibrium magnetization dynamics is governed by a stochastic Landau-Lifschitz-Gilbert equation with spin transfer torques. We calculate the Gilbert damping parameter $\alpha$ and the non-adiabatic spin transfer torque parameter $\beta$ for a model ferromagnet. We find that $\beta \neq \alpha$, in agreement with the results obtained using imaginary-time methods of Kohno, Tatara and Shibata [J. Phys. Soc. Japan 75, 113706 (2006)]. We comment on the relationship between $s-d$ and isotropic-Stoner toy models of ferromagnetism and more realistic density-functional-theory models, and on the implications of these relationships for predictions of the $\beta/\alpha$ ratio which plays a central role in domain wall motion. Only for a single-parabolic-band isotropic-Stoner model with an exchange splitting that is small compared to the Fermi energy does $\beta/\alpha$ approach one. In addition, our microscopic formalism incorporates naturally the fluctuations needed in a nonzero-temperature description of the magnetization. We find that to first order in the applied electric field, the usual form of thermal fluctuations via a phenomenological stochastic magnetic field holds.