Spin transport at finite temperatures: A first-principles study for ferromagnetic$|$nonmagnetic interfaces

TL;DR

This study uses first-principles calculations to determine how spin-memory loss and related parameters at ferromagnetic|nonmagnetic interfaces vary with temperature, providing insights into spin transport phenomena relevant for spintronic devices.

Contribution

It presents the first-principles calculation of temperature-dependent spin-memory loss and interface resistance for Co|Pt and Py|Pt interfaces, incorporating thermal disorder effects.

Findings

Calculated temperature-dependent spin-memory loss parameter δ.

Determined interface resistance AR_I as a function of temperature.

Provided insights into spin transport at ferromagnetic|nonmagnetic interfaces.

Abstract

Symmetry lowering at an interface leads to an enhancement of the effect of spin-orbit coupling and to a discontinuity of spin currents passing through the interface. This discontinuity is characterized by a "spin-memory loss" (SML) parameter $\delta$ that has only been determined directly at low temperatures. Although $\delta$ is believed to be significant in experiments involving interfaces between ferromagnetic and nonmagnetic metals, especially heavy metals like Pt, it is more often than not neglected to avoid introducing too many unknown interface parameters in addition to often poorly known bulk parameters like the spin-flip diffusion length $l_{\rm sf}$. In this work, we calculate $\delta$ along with the interface resistance $AR_{\rm I}$ and the spin-asymmetry parameter $\gamma$ as a function of temperature for Co$|$Pt and Py$|$Pt interfaces where Py is the ferromagnetic Ni$_{80}$Fe$_{20}$ alloy, permalloy. We use first-principles scattering theory to calculate the conductance as well as local charge and spin currents, modeling temperature-induced disorder with frozen thermal lattice and, for ferromagnetic materials, spin disorder within the adiabatic approximation. The bulk and interface parameters are extracted from the spin currents using a Valet-Fert model generalized to include SML.