1: \documentclass[prl,twocolumn,showpacs,floatfix,amsfonts,superscriptaddress]{revtex4}

2: 

3: 

4: \usepackage{graphicx,graphics,color,epsfig}% Include figure files

5: \usepackage{bm}

6: \usepackage{amsmath}

7: \usepackage{amssymb}

8: %\usepackage[showrefs]{refcheck}

9: 

10: 

11: \newcommand{\bS}{{\bf S}}

12: \newcommand{\bd}{{\bf d}}

13: \newcommand{\bB}{{\bf B}}

14: \newcommand{\bQ}{{\bf Q}}

15: \newcommand{\bk}{{\bf k}}

16: \newcommand{\bv}{{\bf v}}

17: \newcommand{\bl}{{\bf l}}

18: \newcommand{\br}{{\bf r}}

19: \newcommand{\bp}{{\bf p}}

20: \newcommand{\s}{{\sigma}}

21: \newcommand{\bA}{{\bf A}}

22: \newcommand{\ba}{{\bf a}}

23: \newcommand{\bJ}{{\bf J}}

24: \newcommand{\bE}{{\bf E}}

25: \newcommand{\bb}{{\bf b}}

26: \newcommand{\bsig}{\mbox{\boldmath{$\sigma$}}}

27: \newcommand{\Det}{\,{\rm{Det}\,}}

28: \newcommand{\Tr}{\,{\rm{Tr}\,}}

29: \newcommand{\tr}{\,{\rm{tr}\,}}

30: \newcommand{\bn}{{\bf \nabla}}

31: 

32: 

33: \newcommand{\sign}{\,{\rm{sign}\,}}

34: \newcommand{\bomega}{\bar{\Omega}}

35: \newcommand{\wL}{\omega_{L}}

36: \newcommand{\COMMENT}{\large\bf}

37: \newcommand{\FigWidth}{\columnwidth}

38: \newcommand{\beqa}{\begin{eqnarray}}

39: \newcommand{\eeqa}{\end{eqnarray}}

40: \newcommand{\bI}{\bf{I}_s}

41: 

42: 

43: \renewcommand{\Re}{{\rm Re}}

44: \renewcommand{\Im}{{\rm Im}}

45: 

46: 

47: \usepackage{mathptm}    %clears up some problems for generating PDF file

48: \usepackage{dcolumn}                    % Align table columns on decimal point

49: \usepackage{bm}                         % bold math

50: \usepackage{graphicx}

51: 

52: 

53: 

54: \usepackage{times}

55: 

56: 

57: \newcommand{\PRG}{PuRhGa$_5$}

58: \newcommand{\PCG}{PuCoGa$_5$}

59: 

60: 

61: \begin{document}

62: 

63: 

64: \title{Voltage dependence of Landau-Lifshitz-Gilbert damping of a spin in

65: a current driven tunnel junction}

66: 

67: 

68: 

69: 

70: 

71: 

72: \author{Hosho Katsura}

73: \email[Electronic address: ]{katsura@appi.t.u-tokyo.ac.jp}

74: \affiliation{Department of Applied Physics, University of Tokyo,

75: Hongo, Bunkyo-ku, Tokyo 113-8656, Japan}

76: 

77: 

78: \author{Alexander V. Balatsky}

79: \email{avb@lanl.gov}  \affiliation{Theoretical Division, Los Alamos

80: National Laboratory, Los Alamos, New Mexico 87545, USA}

81: 

82: 

83: \author{Zohar Nussinov}

84: \email{zohar@wuphys.wustl.edu}

85: \affiliation{Department of Physics, Washington University, St.

86: Louis, MO 63160, USA}

87: 

88: 

89: \author{Naoto Nagaosa}

90: \affiliation{Department of Applied Physics, University of Tokyo,

91: Hongo, Bunkyo-ku, Tokyo 113-8656, Japan} \affiliation{CERC, AIST

92: Tsukuba Central 4, Tsukuba 305-8562, Japan} \affiliation{CREST,

93: Japan Science and Technology Agency (JST)}

94: 

95: 

96: 

97: \begin{abstract}

98: We present a theory of Landau-Lifshitz-Gilbert damping $\alpha$ for

99: a localized  spin ${\vec S}$ in the junction coupled to the

100: conduction electrons in both leads under an applied volatege $V$. We

101: find the voltage dependence of the damping term reflecting the

102: energy dependence of the density of states. We find the  effect is

103: linear in the voltage and cotrolled by particle-hole asymmetry of

104: the leads.

105: \end{abstract}

106: 

107: 

108: \pacs{75.80.+q, 71.70.Ej, 77.80.-e}

109: 

110: 

111: \date{\today}

112: \maketitle

113: 

114: 

115: 

116: \section{Introduction}

117: 

118: 

119: Spintronics is an emerging subfield that holds the potential to

120: replace conventional electronic devices with spintronic analogues

121: where the manipulation, control, and readout of spins will

122: enable novel functionality with no or little electronic charge

123: dynamics \cite{spintronics}. In order to realize this

124: promise, the spin dynamics of the small scale devices needs to be well

125: controlled. One of the most pressing questions concerns

126: a set up which would preserve coherence and allow a manipulation of spins.

127: In most systems, the relevant spin degrees of freedom are coupled to

128: some bath, such as a fermionic bath of electrons. The detailed

129: dynamics of single spins when in contact with such a bath

130: plays a pivotal role in addressing decoherence in potential

131: spintronic systems.

132: 

133: 

134: The conventional way to treat this problem is via a Caldeira-Leggett

135: approach where the external bath is modeled by collective

136: excitations which are capable of destroying coherent spin dynamics.

137: Often, spin dynamics is described by a Landau-Lifshitz-Gilbert equation

138: \cite{ll1,ll2}:

139: \begin{equation} {\frac {d{\vec S(t)}}{dt}} = - {\vec

140: S(t)}\times{\vec h} -\alpha {\vec S(t)}\times{\frac{d{\vec

141: S(t)}}{dt}}, \label{LLG2}

142: \end{equation}

143: where ${\vec h}$ is, up to constant prefactors, the external

144: magnetic field and the coefficient $\alpha$ captures the damping due

145: to the external bath. A caricature of the solution of this equation

146: \cite{Chika} is provided in Fig.\ref{FIG:FIG2}. There are standard

147: methods to calculate $\alpha$ in an equilibrium situation when, say,

148: one considers a spin in a Fermi liquid \cite{Heinrich,Halperin}.

149: 

150: In the current publication, we address a related novel question

151: concerning the effect of an applied voltage bias on the Gilbert

152: coefficient $\alpha$. Our work complements the recent results of

153: \cite{Onoda} wherein the effects of the ``retarded'' electronic

154: contributions in the equations of motion for a system of spins were

155: studied. Both such retarded correlations \cite{Zhu03} as well as

156: additional ``Keldysh'' correlations generally manifest themselves in

157: the single spin equations of motion, see e.g. \cite{nsabz} for

158: general spin equations of motion entailing the effects of both

159: correlations. In the current work, we examine the voltage dependence

160: of Gilbert damping. For the sake of clarity, we depart from the

161: Keldysh contour formalism of \cite{Onoda,Zhu03,nsabz}, and use

162: a Caldeira-Leggett approach.

163: 

164: 

165: In what follows, we consider the case of a junction between two

166: electrodes that contains one spin $\vec{S}$, see Fig.\ref{FIG:FIG1}.

167: This spin $\vec{S}$ may be the spin of a single magnetic impurity or

168: it may portray the spin of a cluster at low temperature when the

169: spins in the cluster are locked. Upon applying a finite bias between

170: the electrodes of Fig.\ref{FIG:FIG2}, a current flow is generated.

171: Thereafter, at long times, the system is at a steady but

172: non-equilibrium state so long as the voltage bias $V$ is applied. We

173: will focus on the voltage dependence of the damping term $\alpha(V)$

174: in Eq.(1). We find that the change in the density of states

175: associated with the chemical potential gradient across the junction

176: triggers a modification to the damping $\alpha$ that is {\em linear}

177: in voltage and is proportional to the particle-hole assymmetry of

178: the density of states. The scale of the correction is set by the

179: Fermi energy of the metal in the leads $E_F$ and by particle-hole

180: asymmetry in the density of states:

181:  \beqa \alpha(V) = \alpha_0+\alpha_1(V)

182: = \alpha_0 ( 1+ O(1)eV/E_F). \label{voltage+} \eeqa

183:  This result

184: vividly illustrates the presence of voltage induced damping in such

185: junctions. Spin unpolarized electrons tunneling across the junction

186: interact via exchange interaction with the spin $\vec{S}$ and

187: produce random magnetic fields that disorder the local spin. This

188: noise augments that already present equilibrium magnetic noise in a

189: Fermi liquid bath. Such a behavior of $\alpha(V)$ with the external

190: voltage is in line with the works of \cite{hooley}. An analysis of

191: a related single spin problem in a Josephson junction (instead of

192: the normal junction studied here) was advanced in

193: \cite{Zhu03,nsabz,bulaevskii}.

194: 

195: 

196: We will shortly derive the effective single spin action from which

197: the principle equation of motion of Eq.(\ref{LLG2}) follows. Several

198: technical details of our derivation are given in the appendices.

199: 

200: %Below we will present our results. We will start with the

201: %Hamiltonian and obtain effective action get equations of motion

202: %etc.

203: 

204: 

205: 

206: 

207: \begin{figure}

208: \centerline{\includegraphics[width=0.9\columnwidth]{fig2.eps}}

209: \caption{Sketch of the dissipative spin dynamics.

210: Panel (a) depicts a cartoon of the Larmor

211: precession of the spin about the direction of an applied magnetic

212: field (B). In panel (b), a caricature of the spin dynamics in the

213: presence of Landau-Lifshitz-Gilbert damping is shown.}

214: \label{FIG:FIG2}

215: \end{figure}

216: 

217: 

218: \begin{figure}

219: \centerline{\includegraphics[width=0.9\columnwidth]{fig1.eps}}

220: \caption{Magnetic impurity coupled to two electrodes. $\mu_{\rm L}$

221: and $\mu_{\rm R}$ denotes the chemical potentials of the left and

222: right leads respectively.  The voltage drop across the two

223: electrodes is $eV= \mu_{L} - \mu_{R}$.} \label{FIG:FIG1}

224: \end{figure}

225: 

226: 

227: 

228: 

229: 

230: 

231: 

232: \section{THE SYSTEM AND OUR PRINCIPLE RESULTS}

233: 

234: 

235: The physical system under consideration in this publication is

236: illustrated in Fig.\ref{FIG:FIG1}. It consists of two (left(L) and

237: right(R)) electrodes across which a voltage bias is applied; a

238: magnetic impurity ($\vec{S}$) is situated in between (or lies on one

239: of) the electrodes. An external magnetic field ${\vec{B}}$ is

240: present. In the absence of effects stemming from conduction

241: electrons in the tunneling barrier, the single spin would precess at

242: the Larmor precession frequency about the applied field direction

243: (see, e.g. panel (a) of Fig.\ref{FIG:FIG2}). With the external

244: circuit elements present, the spin motion becomes dissipative (as

245: schematically shown in panel (b) of Fig.\ref{FIG:FIG2}).

246: 

247: 

248: With the spin embedded in the tunneling barrier, the work function

249: is modified and the conduction electron tunneling matrix element is

250: supplanted by a Kondo like exchange term $ J(\vec{S} \cdot

251: \vec{\sigma}_{c})$, with $\vec{\sigma}_{c}$ denoting the conduction

252: electron spin. In what follows, we will dispense with the $c$

253: subscript. The Hamiltonian governing this system is given by

254: \begin{eqnarray}

255: {\cal H} &=& {\cal H}_e + {\cal H}_s + {\cal H}_T,

256: \\

257: {\cal H}_e &=& \sum_{\alpha,k,\sigma} \xi_{\alpha k} c_{\alpha k

258: \sigma}^{\dagger}c_{\alpha k \sigma}, \nonumber

259: \\

260: {\cal H}_s &=& - {\vec h} \cdot {\vec S(t)}, \nonumber

261: \\

262: {\cal H}_T &=& {\frac{1}{\Omega}}{\sum_{\alpha,k,\sigma}}{\sum_{\beta,p,\sigma'}}

263: c_{\alpha k \sigma}^{\dagger} (T_{\alpha\beta})_{\sigma \sigma'}

264: c_{\beta p \sigma'}, \label{Ham}

265: \end{eqnarray}

266: where $c_{\alpha k \sigma}^{\dagger}~(c_{\alpha k \sigma})$ creates

267: (annihilates) an electron with momentum $k$ and spin $\sigma \in \{

268: \uparrow,\downarrow \}$ in the lead $\alpha \in \{{\rm L}, {\rm

269: R}\}$. The abbreviation $\xi_{\alpha k} = \epsilon_{\alpha k}

270: -\mu_{\alpha}$, where $\epsilon_{\alpha k}$ is the energy of

271: electron with momentum $k$ in (the lead) $\alpha$ and $\mu_{\alpha}$

272: is the chemical potential in (the lead) $\alpha$. The second term in

273: Eq.(\ref{Ham}), ${\cal H}_s$, is Zeeman energy of the spin in an

274: external magnetic field $\vec B$. Here, $\vec h \equiv g \mu_B {\vec

275: B}$ with $g$ gyromagnetic ratio and $\mu_B$ the Bohr magneton. The

276: last term in Eq.(\ref{Ham}), ${\cal H}_T$, represents both Kondo

277: coupling and direct tunneling process, where the amplitudes

278: $\{T_{\alpha \beta}\}=\{ T_{{\rm L}{\rm L}},T_{{\rm L}{\rm

279: R}},T_{{\rm R}{\rm L}},T_{{\rm R}{\rm R}} \}$ are tunneling matrix

280: elements and their explicit forms are

281: \begin{eqnarray}

282: T_{LL} &=& J_{LL}({\vec \sigma} \cdot {\vec S(t)}), \nonumber

283: \\

284: T_{RR} &=& J_{RR}({\vec \sigma} \cdot {\vec S(t)}), \nonumber

285: \\

286: T_{LR} &=& T_{RL} = (T_0 \delta_{\sigma \sigma'} + J_{LR}({\vec

287: \sigma} \cdot {\vec S(t)})). \label{tunnel_amp}

288: \end{eqnarray}

289: Here, $T_0$ is the direct tunneling matrix element and $J_{\alpha

290: \beta}$ is the Kondo coupling, 'while $\Omega$ denotes the Volume of

291: each lead (assumed, for simplicity, to be the same)'. Typically, from

292: the expansion of the work function for tunneling,

293: ${J_{LR}}/{T_0} \sim J/U$, where $U$ is the height of a

294: spin-independent tunneling barrier and $J$ the magnitude of the spin

295: exchange interaction ~\cite{Zhu_Balatsky}. Typical values of the

296: ratio the spin dependent to spin independent tunneling amplitudes in

297: Eq.(\ref{tunnel_amp}), $J_{\alpha \beta}/T_{0}$, are

298: ${\cal{O}}(10^{-1})$, with a typical Fermi energy $E_{F}^{R}$ of the

299: order of several electron-volts. From the Hermiticity of the

300: Hamiltonian, we can  find that the matrix element $(T_{\alpha

301: \beta})_{\sigma \sigma'}$ satisfies $((T_{\alpha \beta})_{\sigma

302: \sigma'})* = (T_{\beta \alpha})_{\sigma' \sigma}.$

303: 

304: 

305: In the up and coming, we derive the effective action for the single

306: impurity spin via an imaginary time path integral formalism. The

307: full action is given by

308: \begin{eqnarray}

309: S = \int_0^{\beta}d{\tau} \sum_{\alpha k \sigma} c_{\alpha k

310: \sigma}^{\dagger} \partial_{\tau} c_{\alpha k \sigma} +

311: iS\omega(\vec S(\tau)) +\int_0^{\beta}d{\tau}{\cal H}(\tau),

312: \label{action}

313: \end{eqnarray}

314: where the second, Wess-Zumino-Novikov-Witten(WZNW), term in Eq.(\ref

315: {action}) depicts the Berry phase accumulated by the spin

316: \cite{Fradkin}. In our action, we have the following quadratic form

317: of fermions,

318: \begin{eqnarray}

319: && \int_0^{\beta} d{\tau} {\frac{1}{\Omega}} \sum_{\alpha k \sigma}

320: \sum_{\beta p \sigma'}  c_{\alpha k \sigma}^{\dagger}

321: (\delta_{\alpha \beta} \delta_{\sigma \sigma'}(\partial_{\tau} +

322: \xi_{\alpha k}) + (T_{\alpha \beta})_{\sigma \sigma'}) c_{\beta p

323: \sigma'} \nonumber

324: \\

325: && \equiv \int_0^{\beta} d{\tau} {\frac{1}{\Omega}} \sum_{\alpha k

326: \sigma} \sum_{\beta p \sigma'} c_{\alpha k \sigma}^{\dagger}

327: ((M_{\alpha \beta}(\tau))_{\sigma \sigma'})_{kp} ~ c_{\beta p \sigma'}.

328: \label{quad}

329: \end{eqnarray}

330: We may integrate over the lead electrons to obtain the effective

331: action for the spin

332: \begin{equation}

333: S_{\rm {eff}}(\vec S(\tau)) \sim i S \omega({\vec S(\tau)}) +

334: \int_0^{\beta} d{\tau} {\vec h} \cdot {\vec S(\tau)} -{\rm{ln}}{\rm {det}} M , \label{eff}

335: \end{equation}

336: where ${\rm {det}} M$ means \textit{functional determinant} of $M.$

337: From the third term in Eq.(\ref {eff}), we obtain a quadratic

338: non-local in time interaction of the spin with itself, $\vec

339: S(\tau)$ as

340: \begin{equation}

341: \Delta S(\vec S(\tau)) = -2 \int d{\tau} \int d{\tau'} {\vec S(\tau)}

342: \cdot {\vec S(\tau')} K(\tau-\tau'), \label{K+}

343: \end{equation}

344: where

345: \begin{eqnarray}

346: K(\tau) &=& \sum_{\alpha, \beta \in \{ \rm{L,R} \}} K_{\alpha

347: \beta}(\tau), \nonumber

348: \\

349: K_{\alpha \beta} &=& \sum_{k \in \alpha} \sum_{p \in \beta}

350: {\frac{J_{\alpha \beta} J_{\beta \alpha}}{2}} {\frac{1}{\beta}}

351: \sum_{\omega_m} {\frac {f(\xi_k)-f(\xi_p)}{i\omega_m +\xi_k -\xi_p}}

352: e^{-i \omega_m \tau}, \label{Kt}

353: \end{eqnarray}

354: with $f(\xi)$ denotes the Fermi distribution function (see APPENDIX

355: A). The effective action $\Delta S$ of Eq.(\ref{eff}) can be

356: decomposed into two (trivial and non-trivial) components as $\Delta

357: S = \Delta S_{\rm loc} + \Delta S_{\rm{dis}}$, with

358: \begin{eqnarray}

359: \Delta S_{\rm loc} &=& -2K(\omega=0) \int d{\tau} ({\vec S(\tau)})^2,

360: \nonumber

361: \\

362: \Delta S_{\rm{dis}} &=& \int d{\tau} \int d{\tau}' ({\vec

363: S(\tau)}-{\vec S({\tau})})^2 K(\tau-{\tau}'). \label{locdis}

364: \end{eqnarray}

365: Here, $K(\omega=0)$ is the zero-frequency Fourier component of

366: $K(\tau)$. The first term ($\Delta S_{\rm loc}$) is a trivial

367: constant as $S(\tau)^2 = S^2$. The nonlocal part ($\Delta

368: S_{\rm{dis}}$) represents the dissipative effect due to the coupling

369: between $S(\tau)$ and electrons bath. The integral kernel $K(\tau)$

370: is calculated in the same way as the Caldeira-Leggett theory

371: \cite{Leggett,Leggett2} leading to

372: \begin{equation}

373: K(\tau) = \int_0^{\infty} d{\omega} J(\omega) \frac{{\rm

374: cosh}({{\beta}/2}-|\tau|){\omega}}{{\rm sinh}({\beta \omega}/2)},

375: \label{K}

376: \end{equation}

377: where $J(\omega)$ is the \textit{spectral density} and its explicit

378: form is

379: \begin{equation}

380: J(\omega) = \sum_{\alpha \beta} \sum_{k \in \alpha} \sum_{p \in

381: \beta} {\frac{J_{\alpha \beta} J_{\beta \alpha}}{2}} [f(\xi_k)

382: -f(\xi_p)] \delta (\omega+ \xi_k -\xi_p). \label{spec}

383: \end{equation}

384: The details of the derivation of Eq.(\ref{spec}) are provided in

385: APPENDIX B.

386: The spectral density of Eq.(\ref{spec}), $J(\omega)$, is estimated

387: as

388: \begin{eqnarray}

389: J(\omega) &=& \sum_{\alpha \beta} {\frac{J_{\alpha \beta} J_{\beta

390: \alpha}}{2}} \int_{E_F^{\alpha}}^{\infty} d{\xi_{\alpha}}

391: N(\xi_{\alpha})\int_{E_F^{\beta}}^{\infty} d{\xi_{\beta}} N(\xi_{\beta})

392: \nonumber

393: \\

394: && \times [f(\xi_{\alpha})-f(\xi_{\beta})]

395: \delta(\omega-\xi_{\alpha}-\xi_{\beta}) \nonumber

396: \\

397: &\sim& \sum_{\alpha \beta}{\frac{J_{\alpha \beta} J_{\beta

398: \alpha}}{2}} N(\xi_{\alpha}=0) N(\xi_{\beta}=0) \omega,

399: \label{spec2}

400: \end{eqnarray}

401: where $E_F^{\alpha}$ denotes the Fermi Energy of the lead $\alpha$.

402: It is obvious that $J(\omega)$ in Eq.(\ref{spec2}) is proportional

403: to $\omega$, i.e., $J(\omega)$ is Ohmic. If spectral density is

404: expressed as $J(\omega)={\eta \omega}/{2 \pi}$, then

405: the Gilbert coefficient $\alpha$ in Eq.(\ref{LLG2}) is exactly equal

406: to $\eta$. By varying the total action with respect to the spin

407: ${\delta S}/{\delta {\vec S(\tau)}}=0$, we immediately obtain the

408: Landau-Lifshitz-Gilbert equation with $\alpha={\frac{2

409: \pi}{\omega}}J(\omega)$(see APPENDIX C). In other words, the voltage

410: dependence of $\alpha$ in Eq.(\ref{LLG2}) is identically the same as

411: that of $J(\omega)$. We next examine the voltage dependence of

412: $J(\omega)$.

413: 

414: If we apply a voltage leading to a chemical drop of

415: $\Delta\mu_L-\Delta\mu_R=eV$. Assuming, for example,that

416:  the

417: net charge on both right and left leads is unchanged, we also have

418:  $D^L(E_F)\Delta\mu_L+D^R(E_F)\Delta\mu_R=0$.

419: With these constraints we get

420: \begin{eqnarray}

421: \Delta\mu_L&=&\frac{D^R(E_F)}{D^L(E_F)+D^R(E_F)}eV,

422: \nonumber \\

423: \Delta\mu_R&=&-\frac{D^L(E_F)}{D^L(E_F)+D^R(E_F)}eV,

424: \end{eqnarray}

425: the Gilbert coefficient $\alpha$ may be approximated as\\

426: \begin{eqnarray}

427: \alpha(V) &\sim& 2\pi\Bigl(\frac{J_{LL}^2}{2} D^L(E_F+\Delta\mu_L)^2 + \frac{J_{RR}^2}{2}

428: D^R(E_F+\Delta\mu_R)^2 \nonumber\\

429: &&+J_{LR}^2 D^L(E_F+\Delta\mu_L) D^R(E_F+\Delta\mu_R)\Bigr)

430: \nonumber

431: \end{eqnarray}

432: \begin{eqnarray}

433: &\sim& 2\pi \Bigl(

434: \frac{[J_{LL}D^L(E_F)]^2}{2} + \frac{[J_{RR} D^R(E_F)]^{2}}{2} + J^2_{LR}D^L(E_F)D^R(E_F)\nonumber \\

435: &&+(J^2_{LL}D^L(E_F)\frac{\partial{D^L(E_F)}}{\partial{E_F}}

436: +   J^2_{LR}D^R(E_F)\frac{\partial{D^L(E_F)}}{\partial{E_F})})\Delta\mu_L

437: \nonumber \\

438: &&+(J^2_{RR}D^R(E_F)\frac{\partial{D^R(E_F)}}{\partial{E_F}}

439: +   J^2_{LR}D^L(E_F)\frac{\partial{D^R(E_F)}}{\partial{E_F})})\Delta\mu_R)

440: \Bigr) \nonumber

441: \end{eqnarray}

442: \begin{eqnarray}

443: &=&2\pi \Bigl(

444: \frac{[J_{LL}D^L(E_F)]^2}{2} + \frac{[J_{RR}D^R(E_F)]^{2}}{2}

445: + J^2_{LR}D^L(E_F)D^R(E_F)\nonumber \\

446: &+& \frac{eV}{D^L(E_F)+D^R(E_F)}

447: (J^2_{LL}D^L(E_F)D^R(E_F)\frac{\partial{D^L(E_F)}}{\partial{E_F}}

448: \nonumber \\

449: &-& J^2_{RR}D^L(E_F)D^R(E_F)\frac{\partial{D^R(E_F)}}{\partial{E_F}}

450: \nonumber \\

451: &+& J^2_{LR}D^R(E_F)D^R(E_F)\frac{\partial{D^L(E_F)}}{\partial{E_F}}

452: \nonumber \\

453: &-& J^2_{LR}D^L(E_F)D^L(E_F)\frac{\partial{D^R(E_F)}}{\partial{E_F}})\Bigr).

454: \label{Gil}

455: \end{eqnarray}

456: The change in the density of states associated with the chemical

457: potential gradient across the junction triggers a modification of

458: the damping $\alpha$ that is linear in voltage. For typical Fermi

459: energy $E_{F}^{L/R}$ of the order of several electron-volts, the

460: voltage dependence of $\alpha$ may become very notable. This voltage

461: driven effect may be expressed in terms of $\alpha_0$ and

462: $\alpha_1(V)$ with $ \alpha(V)=\alpha_0 + \alpha_1$

463: (Eq.(\ref{voltage+})). Here \begin{eqnarray}

464: \alpha_0= \pi \bigl(

465: J_{LL}^2[D(E_F^L)]^2 + 2J_{LR}^2 D(E_F^L)

466: D(E_F^R)+J_{RR}^2[D(E_F^R)]^2 \bigl), \nonumber

467: \\

468: \alpha_1= \frac{1}{4 e|T_{0}|^{2}[D(E_F^L) + D(E_F^R)]} \Bigl(

469: \frac{I_{o}^{L}}{[D(E_F^L)]^2} + \frac{I_{o}^{R}}{[D(E_F^R)]^2}

470: \Bigl) \nonumber

471: \\ \times \bigl[ D(E_F^L)

472: D(E_F^R) \bigl( J_{LL}^{2} \frac{\partial D^{L}(E_F)}{\partial

473: E_{F}} - J_{RR}^{2} \frac{\partial D^{R}(E_F)}{\partial E_{F}}

474: \bigl) \nonumber

475: \\+ J_{LR}^{2} \bigl( [D(E_F^R)]^2 \frac{\partial D^{L}(E_F)}{\partial E_{F}}

476: - [D(E_F^L)]^2 \frac{\partial D^{R}(E_F)}{\partial E_{F}} \bigl)

477: \bigl].

478: %\alpha_0

479: %{\frac{2(J_{LL}^2+J_{LR}^2)}{J_{LL}^2+2J_{LR}^2+J_{RR}^2}} =I_o

480: %{\frac{1}{eE_F^R}} {\frac{J_{LL}^2+J_{LR}^2}{2|T_0|^2}}.

481: \label{central++}

482: \end{eqnarray}

483: 

484: 

485: 

486: %It points out that the

487: %steady non-equilibrium current produces a correction to the

488: %Landau-Lifshiz-Gilbert damping  of the spin.

489: 

490: 

491: \section{Conclusions}

492: In conclusion, we present a theoretical study of

493: Landau-Lifshitz-Gilbert damping (Eq.(\ref{LLG2})) for a localized

494: spin ${\vec S}$ in a junction. The exchange interactions between the

495: localized spin and tunneling electrons leads to additional

496: dissipation of the spin motion, see Fig.(\ref{FIG:FIG2}). In the

497: presence of an applied voltage bias $V$, the damping coefficient,

498: i.e., Gilbert damping, is modified in linear order in $V$ for the

499: leads with particle-hole asymmetry in the Density of States. 

500: 

501: 

502: \section{Acknowledgements}

503: 

504: Work at LANL was supported by the US DOE under LDRD X1WX.

505: 

506: 

507: 

508: \section{Appendix A: DERIVATION OF THE EFFECTIVE ACTION}

509: Here we will give a detailed derivation of the effective action for

510: a spin. From Eq.(\ref {eff}), we can extract a quadratic form of

511: spins with the aid of the well known identity ${\rm {ln}}~{\rm

512: {det}}M ={\rm Tr}~{\rm {ln}}M.$  In order to tabulate the expansion

513: of ${\rm Tr}~{\rm {ln}}M$ perturbatively, we define matrices $M_0$

514: and $M_1$,

515: 

516: \begin{eqnarray}

517: M_{KP}&&=(M_0)_{KP}+(M_1)_{KP}

518: \nonumber

519: \\

520: (M_0)_{KP} &&\equiv ((-i\omega + \xi_{\alpha k})\delta_{\alpha \beta} \delta_{\sigma \sigma'})\delta_{KP}

521: \nonumber

522: \\

523: (M_1)_{KP} &&\equiv {\frac{1}{\sqrt{\beta}}}(T_{\alpha \beta}(\omega-\omega'))_{\sigma \sigma'},

524: \nonumber

525: \end{eqnarray}

526: where $K \equiv (k, \omega)$ and $P \equiv (p, \omega')$ with fermionic Matsubara frequencies $\omega$ and $\omega'$.

527: Employing the expansion ${\rm ln}(1+x) = - \sum_{n=1}^{\infty}

528: {\frac {(-x)^n}{n}}$, we can  write the effective action as

529: \begin{equation}

530: S_{\rm {eff}}(\vec S(\tau)) \sim S_0 + {\rm {Tr}}~{\rm {ln}}M_0 +

531: {\rm {Tr}}(M_0^{-1}M_1) -{\frac{1}{2}}{\rm {Tr}} (M_0^{-1}M_1)^2 +

532: \cdots, \label{Seff}

533: \end{equation}

534: where $S_0$ is the sum of the first and the second term in

535: Eq.(\ref{eff}). The third term in Eq.(\ref{eff}) (and consequent

536: last term shown in Eq.(\ref{Seff})) is the first non-trivial

537: contribution to the spin equation of motion.

538: Its evaluation is straightforward,

539: \begin{eqnarray}

540: &&{\rm Tr}(M_0^{-1}M_1)^2

541: \nonumber

542: \\

543: &&= \sum (M_0^{-1})_{K_1 K_1}(M_1)_{K_1 K_2}(M_0^{-1})_{K_2 K_2}(M_1)_{K_2 K_1}

544: \nonumber

545: \\

546: &&={\frac{1}{2 \beta}}\sum (-i\omega + \xi_{\alpha k})^{-1} (T_{\alpha \beta}(\omega-\omega'))_{\sigma \sigma'}

547: \nonumber

548: \\

549: &&(-i\omega' + \xi_{\beta p})^{-1} (T_{\beta \alpha}(\omega'-\omega))_{\sigma' \sigma},

550: \nonumber

551: \end{eqnarray}

552: where, repeated indices are implicitly summed over.

553: Then, we find

554: \begin{eqnarray}

555: &&\Delta S = {\frac{1}{2}} {\rm Tr}(M_0^{-1}M_1)^2

556: \nonumber

557: \\

558: &&={\frac{1}{2 \beta}} \sum_{\omega\omega'} \sum_{\alpha \beta} \sum_{\sigma \sigma'}

559: \sum_{k \in \alpha} \sum_{p \in \beta}(-i\omega + \xi_{\alpha k})^{-1} (T_{\alpha \beta}(\omega-\omega'))_{\sigma \sigma'}

560: \nonumber \\

561: & &(-i\omega' + \xi_{\beta p})^{-1} (T_{\beta \alpha}(\omega'-\omega))_{\sigma' \sigma}

562: \nonumber

563: \\

564: &&=-\sum_{\alpha \beta} \sum_{k \in \alpha} \sum_{p \in \beta} \sum_{\omega_m}

565: J_{\alpha \beta} J_{\beta \alpha} {\frac{f(\xi_k)-f(\xi_p)}{i\omega_m +\xi_k -\xi_p}}

566: {\vec S}(-\omega_m) \cdot {\vec S}(\omega_m)

567: \nonumber

568: \\

569: &&\equiv -\int d\tau \int d\tau' {\vec S}(\tau) \cdot {\vec S}(\tau') 2K(\tau-\tau').

570: \nonumber \end{eqnarray}

571: Here, $K(\tau-\tau')$ denotes the integral kernel defined in

572: Eq.(\ref{Kt}). Upon invoking the identity $2{\vec S(\tau)}\cdot{\vec

573: S({\tau}')} = (({\vec S(\tau)})^2+({\vec S({\tau}')})^2) -({\vec

574: S(\tau)}-{\vec S({\tau}')})^2$, the effective action becomes that of

575: Eq.(\ref{locdis}).

576: 

577: \section{APPENDIX B: THE DERIVATION OF THE SPECTRAL DENSITY}

578: We return to Eq.(\ref{Kt}) derived in Appendix A, and express the

579: sum as a contour integral following standard procedures,

580: e.g.\cite{Mahan}, to obtain

581: \begin{widetext}

582: \begin{eqnarray}

583: K_{\alpha \beta} &=& {\frac{J_{\alpha \beta} J_{\beta \alpha}}{2}}

584: \sum_{k \in \alpha}  \sum_{p \in \beta} \oint {\frac{dz}{2 \pi i}}

585: \Bigl( {\frac {e^{-z \tau}}{e^{\beta z}-1}} \theta(-\tau) + {\frac

586: {e^{-z \tau}}{1-e^{-\beta z}}} \theta(\tau) \Bigr) {\frac

587: {f(\xi_k)-f(\xi_p)}{z +\xi_k -\xi_p}} \nonumber

588: \\

589: &=& P \int_{- \infty}^{\infty} d{\omega} \Bigl( {\frac {J_{\alpha

590: \beta} J_{\beta \alpha}}{2}} \Bigr)

591:               \sum_{k \in \alpha}  \sum_{p \in \beta} [f(\xi_k)-f(\xi_p)]

592: \delta(\omega + \xi_k - \xi_p)

593:               \Bigl({\frac {e^{-\omega \tau}}{e^{\beta \omega}-1}}

594: \theta(-\tau) + {\frac {e^{-\omega \tau}}{1-e^{-\beta \omega}}}

595: \theta(\tau)  \Bigr) \nonumber

596: \\

597: &=& P \int_0^{\infty} d{\omega} \Bigl( -{\frac {J_{\alpha \beta}

598: J_{\beta \alpha}}{2}} \sum_{k \in \alpha}  \sum_{p \in

599: \beta}[f(\xi_k)-f(\xi_p)] \delta(\omega + \xi_k - \xi_p)  \Bigr)

600: \frac{{\rm cosh}({{\beta}/2}-|\tau|){\omega}}{{\rm sinh}({\beta

601: \omega}/2)} \nonumber

602: \\

603: &\equiv&  \int_0^{\infty} d{\omega} J(\omega) \frac{{\rm

604: cosh}({{\beta}/2}-|\tau|){\omega}}{{\rm sinh}({\beta \omega}/2)}

605: \label{cont}

606: \end{eqnarray}

607: \end{widetext}

608: where $J(\omega)$ denotes the \textit{spectral density} in

609: Caldeira-Leggett theory. The standard contour employed here is shown

610: in Fig.(\ref{FIG:FIG3}). The symbol $P$ in Eq.(\ref{cont}) denotes

611: the principal part of the integral.

612: \begin{figure}

613: \centerline{\includegraphics[width=0.9\columnwidth]{fig3.eps}}

614: \caption{The standard contour employed in Eq.(\ref{cont}) in order

615: to evaluate the Matsubara sum of Eq.(\ref{Kt}). The crosses along

616: the imaginary axis denote the Matsubara frequencies.}

617: \label{FIG:FIG3}

618: \end{figure}

619: 

620: 

621: 

622: 

623: %\section{Appendix C: A DERIVATION OF THE SPECTRAL DENSITY}

624: The spectral density $J(\omega)$ in Eq.(\ref{spec2})

625: \begin{widetext}

626: \begin{eqnarray}

627: J(\omega)=\frac{J_{LL}^2}{2} \int_{-E_F^L}^{\infty} d{\xi_L}

628: N(\xi_L) \int_{-E_F^L}^{\infty} d{\xi_L'} N(\xi_L')

629: [f(\xi_L)-f(\xi_L')]\delta(\omega+{\xi_L}-{\xi_L'}) \nonumber

630: \\

631: +\frac{J_{RR}^2}{2} \int_{-E_F^R}^{\infty} d{\xi_R} N(\xi_R)

632: \int_{-E_F^R}^{\infty} d{\xi_R'} N(\xi_R')

633: [f(\xi_R)-f(\xi_R')]\delta(\omega+{\xi_R}-{\xi_R'}) \nonumber

634: \\

635: +J_{LR}^2 \int_{-E_F^L}^{\infty} d{\xi_L} N(\xi_L)

636: \int_{-E_F^R}^{\infty} d{\xi_R} N(\xi_R)

637: [f(\xi_L)-f(\xi_R)]\delta(\omega+{\xi_L}-{\xi_R}) \nonumber

638: \\

639: \sim  \Bigl( \frac{J_{LL}^2}{2} N(\xi_L =0) N(\xi_L' =0) +

640: \frac{J_{RR}^2}{2} N(\xi_R =0) N(\xi_R' =0)

641: + J_{LR}^2 N(\xi_L =0) N(\xi_R =0) \Bigr) \omega

642: \nonumber

643: \\

644: =\Bigl( \frac{J_{LL}^2}{2} D^L(E_F^L) D^L(E_F^L) + \frac{J_{RR}^2}{2}

645: D^R(E_F^R) D^R(E_F^R)

646:             + J_{LR}^2 D^L(E_F^L) D^R(E_F^R) \Bigr) \omega,

647: \label{aspec}

648: \end{eqnarray}

649: \end{widetext}

650: 

651: where $D^{L/R}(E^{L/R}_F)$ denotes the \textit{density of states} at the Fermi

652: energy level of the left/right lead.

653: \\

654: If we apply a voltage leading to a chemical potential drop of

655: $E^L_F-E^R_F=(E_F +\Delta\mu_L)-(E_F+\Delta\mu_R)=eV$, then

656: Eq.(\ref{spec}) follows. This, in turn, leads to Eq.(\ref{Gil}).

657: then Eq.(\ref {spec}) follows.

658: 

659: 

660: 

661: 

662: \section{APPENDIX C: THE SPIN EQUATION OF MOTION}

663: 

664: 

665: If $J(\omega) = \frac{\alpha \omega}{2 \pi}$ then, from

666: Eq.(\ref{K}), the non-local in time kernel of the action

667: (Eq.(\ref{K+})) is $K(\tau) \sim {\frac{\alpha}{2 \pi}}

668: {\frac{1}{{\tau}^2}}$. We  thus obtain from Eq.(\ref{locdis}),

669: \begin{equation}

670: \Delta S_{dis} =  {\frac{\alpha}{2 \pi}} \int d{\tau} \int d{\tau}'

671: {\frac {({\vec S(\tau)}-{\vec S({\tau}')})^2} {(\tau-{\tau}')^2}}.

672: \end{equation}

673: The functional derivative of $\Delta S_{\rm {dis}}$ with respect to

674: $\vec S(\tau)$ is

675: \begin{eqnarray}

676: {\frac{\delta \Delta S_{dis}}{\delta S(\tau)}} =

677: {\frac{\alpha}{\pi}} \int d{\tau}'{\frac {({\vec S(\tau)}-{\vec

678: S({\tau}')})} {(\tau-{\tau}')^2}} \nonumber

679: \\ = {\frac{\alpha}{\pi}} \int d{\tau}' {\frac{1}{(\tau-{\tau}')}}

680: {\frac{d}{d{\tau}'}}({\vec S(\tau)}-{\vec S({\tau}')}) = i \alpha

681: {\frac {d}{d{\tau}}}{\vec S(\tau)} \label{corr}

682: \end{eqnarray}

683: From the free portion of the action (the first two terms of

684: Eq.(\ref{eff})), we have

685: \begin{equation}

686: {\frac{\delta S_0}{\delta {\vec S(\tau)}}} = i {\frac{1}{S^2}}

687: {\frac{d{\vec S(\tau)}}{d{\tau}}} \times {\vec S(\tau)} + \vec h.

688: \label{free}

689: \end{equation}

690: Adding Eqs.(\ref{corr},\ref{free}), equating the sum to zero, cross

691: multiplying with $\vec{S}(\tau)$, and changing $\tau \to it$, we

692: obtain Eq.(\ref{LLG2}).

693: 

694: 

695: 

696: \begin{thebibliography}{20}

697: 

698: 

699: \bibitem{spintronics} D. D. Awschalom, M. E. Flatte, and N.Samarth,

700: {\em Spintronics}, Scientific American, pp. 67-73, June 2002.

701: 

702: 

703: \bibitem{ll1} T. L. Gilbert, Phys. Rev. {\bf 100}, 1243 (1955)

704: \bibitem{ll2} L. Landau and E. M. Lifshitz, Physik Zeits. Soviets. {\bf 8},

705: 

706: 

707: 153 (1935)

708: 

709: 

710: \bibitem{Chika} S. Chikazumi, {\em Physics of Ferromagnetism}, (Oxford

711: university press, Oxford, 1997)

712: 

713: 

714: 

715: 

716: \bibitem{Heinrich} B. Heinrich, D. Fraitov\'a, and V. Kambersk\'y, Phys.

717: Status. Solidi, {\bf 23}, 501 (1967).

718: 

719: 

720: \bibitem{Halperin}  Y. Tserkovnyak, G. A. Fiete and B. I. Halperin, Appl.

721: Phys. Lett. {\bf 84}, 5234 (2004).

722: 

723: 

724: \bibitem{Onoda}  M. Onoda and N. Nagaosa,

725: cond-mat/0509058

726: 

727: 

728: \bibitem{Zhu03} Jian-Xin Zhu, Z. Nussinov, A. Shnirman and A. V.

729: Balatsky, Phys. Rev. Lett., {\bf 92}, 107001 (2004).

730: 

731: 

732: \bibitem{nsabz} Z. Nussinov, A. Shnirman, D. P. Arovas,

733: A. V. Balatsky, and J-X. Zhu, Phys. Rev. B {\bf 71}, 214520 (2005)

734: 

735: 

736: 

737: \bibitem{hooley} O. Parcollet and C. Hooley, Phys.

738: Rev. B {\bf 66}, 085315 (2002); L. N. Bulaevskii, M. Hruska, and G.

739: Ortiz, Phys. Rev. B {\bf 68}, 125415 (2003)

740: 

741: 

742: %\bibitem{SHE} S. Murakami, N. Nagaosa, and S.-C. Zhang, Science {\bf 301},

743: % 1248 (2003); J. Sinova et al., Phys. Rev. Lett. {\bf 92}, 126603 (2004)

744: 

745: 

746: 

747: \bibitem{bulaevskii}

748: L. Bulaevskii, M. Hruska, A. Shnirman, D. Smith, and Yu. Makhlin, Phys.

749: Rev. Lett. {\bf 92}, 177001 (2004)

750: 

751: 

752: \bibitem{Zhu_Balatsky}

753: J-X. Zhu, and A. V. Balatsky, Phys. Rev. B {\bf 67}, 174505 (2003)

754: 

755: 

756: \bibitem{Fradkin} E. Fradkin, {\em Field Theories of Condensed Matter

757: Systems} (Addison-Wesley, Redwood City, 1991).

758: 

759: 

760: \bibitem{Leggett} A. O. Caldeira and A. J. Leggett, Ann. Phys, {\bf 149},

761: 374(1983).

762: 

763: 

764: \bibitem{Leggett2} A. O. Caldeira and A. J. Leggett, Phys. Rev. Lett, {\bf

765: 46}, 211(1981).

766: 

767: 

768: 

769: 

770: %\bibitem{Ohm} The correction term in Eq.(\ref{Gil})  formally can be expressed as

771: 

772: 

773: 

774: %A single Ohmic current form for $\alpha_{1}$ is readily achieved

775: %($\frac{I_{o}^{L}}{[D(E_F^L)]^2} = \frac{I_{o}^{R}}{[D(E_F^R)]^2}$).

776: %a variety of other equivalent forms for $\alpha^{1}$ are

777: % a single current expression).

778: 

779: \bibitem{Mahan} G. D. Mahan, {\em Many-Particle Physics} (Kluwer, New York,

780: 2000).

781: 

782: 

783: \end{thebibliography}

784: 

785: 

786: \end{document}

787: