Table of Contents
Fetching ...

Reconciling Kubo and Keldysh Approaches to Fermi-Sea-Dependent Nonequilibrium Observables: Application to Spin Hall Current and Spin-Orbit Torque in Spintronics

Simao M. Joao, Marko D. Petrovic, J. M. Viana Parente Lopes, Aires Ferreira, Branislav K. Nikolic

Abstract

Quantum transport studies of spin-dependent phenomena in solids commonly employ the Kubo or Keldysh formulas for the nonequilibrium density operator in the steady-state linear-response regime. Its trace with operators of interest, such as the spin density, spin current density, etc., gives expectation values of experimentally accessible observables. For local quantities, these formulas require summing over the manifolds of {\em both} Fermi-surface and Fermi-sea states. However, debates have been raging in the literature about the vastly different physics the two formulations can apparently produce, even when applied to the same system. Here, we revisit this problem using infinite-size graphene with proximity-induced spin-orbit and magnetic exchange effects as a testbed. By splitting this system into semi-infinite leads and central active region, in the spirit of Landauer formulation of quantum transport, we prove the {\em numerically exact equivalence} of the Kubo and Keldysh approaches via the computation of spin Hall current density and spin-orbit torque in both clean and disordered limits. The key to reconciling the two approaches are the numerical frameworks we develop for: ({\em i}) evaluation of Kubo(-Bastin) formula for a system attached to semi-infinite leads, which ensures continuous energy spectrum and evades the need for commonly used phenomenological broadening introducing ambiguity; and ({\em ii}) proper evaluation of Fermi-sea term in the Keldysh approach, which {\em must} include the voltage drop across the central active region even if it is disorder free.

Reconciling Kubo and Keldysh Approaches to Fermi-Sea-Dependent Nonequilibrium Observables: Application to Spin Hall Current and Spin-Orbit Torque in Spintronics

Abstract

Quantum transport studies of spin-dependent phenomena in solids commonly employ the Kubo or Keldysh formulas for the nonequilibrium density operator in the steady-state linear-response regime. Its trace with operators of interest, such as the spin density, spin current density, etc., gives expectation values of experimentally accessible observables. For local quantities, these formulas require summing over the manifolds of {\em both} Fermi-surface and Fermi-sea states. However, debates have been raging in the literature about the vastly different physics the two formulations can apparently produce, even when applied to the same system. Here, we revisit this problem using infinite-size graphene with proximity-induced spin-orbit and magnetic exchange effects as a testbed. By splitting this system into semi-infinite leads and central active region, in the spirit of Landauer formulation of quantum transport, we prove the {\em numerically exact equivalence} of the Kubo and Keldysh approaches via the computation of spin Hall current density and spin-orbit torque in both clean and disordered limits. The key to reconciling the two approaches are the numerical frameworks we develop for: ({\em i}) evaluation of Kubo(-Bastin) formula for a system attached to semi-infinite leads, which ensures continuous energy spectrum and evades the need for commonly used phenomenological broadening introducing ambiguity; and ({\em ii}) proper evaluation of Fermi-sea term in the Keldysh approach, which {\em must} include the voltage drop across the central active region even if it is disorder free.
Paper Structure (5 equations, 3 figures)

This paper contains 5 equations, 3 figures.

Figures (3)

  • Figure 1: Doubly-proximitized Zollner2020 infinite graphene sheet viewed as a two-terminal Landauer setup Imry1999Nazarov2009Baranger1989Caroli1971Fisher1981 for computational quantum transport Waintal2024 with its CA region being an armchair nanoribbon (of finite length $L = 40$ Å and width $W = 15$ Å) attached to semi-infinite nanoribbons of the same kind. The nanoribbon is periodically repeated along the $y$-axis to reproduce bulk behavior of an infinite sheet Liu2012d. We employ this setup in the calculation of SH current density [Fig. \ref{['fig:fig2']}] and SO torque [Fig. \ref{['fig:fig3']}], within the shaded-in-yellow middle strip, via both Kubo Eq. \ref{['eq:rhokubo']} and Keldysh Eq. \ref{['eq:rhokeldyshlr']}.
  • Figure 2: Spin Hall current density---obtained by tracing its operator Wang2016a with (a) Fermi-surface, (b) Fermi-sea, and (c) total density matrices---in Kubo [Eq. \ref{['eq:rhokubo']}] vs. Keldysh [Eq. \ref{['eq:rhokeldyshlr']}] approaches employing the same retarded GF [Eq. \ref{['eq:retardedgfleads']}] of doubly-proximitized graphene. The parameters in Eq. \ref{['eq:hamiltonian']} are set as $\lambda_\mathrm{RSO}=J_{\text{sd}}=0.1$ eV and the disorder strength is $D=0.3$ eV. The spin current density is averaged over $200$ disorder configurations. Convergence with respect to $k_y$-point sampling was also established.
  • Figure 3: Similar to Fig. \ref{['fig:fig2']}, but for even $\mathbf{T}^e$ (or damping-like Ralph2008Belashchenko2019) and odd $\mathbf{T}^o$ (or field-like Ralph2008Belashchenko2019) SO torques. Each row was obtained by tracing the torque operator Belashchenko2019Nikolic2018 with (a),(b),(g),(h) Fermi-surface; or (c),(d),(i),(j) Fermi-sea density matrix; as well as (e),(f),(k),(l) total density matrix within Kubo Eq. \ref{['eq:rhokubo']} vs. Keldysh Eq. \ref{['eq:rhokeldyshlr']}. In panels (a)--(f), the CA region in Fig. \ref{['fig:fig1']} is clean, while in panels (g)--(l) it contains Anderson disorder of strength $D=0.5$ eV. Panels (e) and (f) show additional (black and green) curves obtained from conventional~Freimuth2014Mahfouzi2018Mahfouzi2020Ghosh2018MedinaDuenas2024 Kubo calculations on periodic lattices [i.e., by using Eq. \ref{['eq:retardedgf']} plugged into Eq. \ref{['eq:rhokubo']}]. The parameters in Eq. \ref{['eq:hamiltonian']} read as $\lambda_\mathrm{RSO} = J_{\text{sd}} = 0.3$ eV. Calculations with disorder employ $200$ configurations.