1: \documentclass[prl,aps,showpacs,twocolumn,unsortedaddress]{revtex4}

2: \usepackage{graphics,bm}

3: \usepackage{amssymb}

4: \usepackage{epsfig}

5: \usepackage{epsf}

6: 

7: \begin{document}

8: 

9: \title{Thermally-Assisted Current-Driven Domain Wall Motion}

10: 

11: \author{R.A. Duine}

12: \email{duine@physics.utexas.edu}

13: \homepage{http://www.ph.utexas.edu/~duine}

14: 

15: \author{A. S. N\'u\~nez}

16: \email{alnunez@physics.utexas.edu}

17: \homepage{http://www.ph.utexas.edu/~alnunez}

18: 

19: \author{A.H. MacDonald}

20: \email{macd@physics.utexas.edu}

21: \homepage{http://www.ph.utexas.edu/~macdgrp}

22: 

23: \affiliation{The University of Texas at Austin, Department of

24: Physics, 1 University Station C1600, Austin, TX 78712-0264}

25: \date{\today}

26: 

27: \begin{abstract}

28: Starting from the stochastic Landau-Lifschitz-Gilbert equation, we

29: derive Langevin equations that describe the nonzero-temperature

30: dynamics of a rigid domain wall. We derive an expression for the

31: average drift velocity of the domain wall $\langle \dot{r}_{\rm

32: dw} \rangle$ as a function of the applied current, and find

33: qualitative agreement with recent magnetic semiconductor

34: experiments. Our model implies that at any nonzero temperature

35: $\langle \dot{r}_{\rm dw} \rangle$ initially varies linearly with

36: current, even in the absence of non-adiabatic spin torques.

37: \end{abstract}

38: 

39: \pacs{72.25.Pn, 72.15.Gd}

40: 

41: \maketitle

42: 

43: % definitions

44: \def\bx{{\bf x}}

45: \def\bk{{\bf k}}

46: \def\half{\frac{1}{2}}

47: \def\args{(\bx,t)}

48: 

49: \noindent {\it Introduction} --- Possibilities opened up by modern

50: nanofabrication capabilities have motivated renewed interest in

51: current-induced domain wall motion, a phenomenon first predicted

52: and subsequently observed in seminal work by Berger

53: \cite{berger1984,freitas1985}. The theoretical

54: \cite{tatara2004,zhang2004,thiaville2004,barnes2005,tserkovnyak2005,kohno2006,

55: nguyen2006} and experimental

56: \cite{grollier2003,tsoi2003,yamaguchi2004,klaui2005,yamanouchi2004,yamanouchi2006}

57: study of domain wall motion induced by spin transfer torques

58: \cite{slonczewski1996,berger1996,tsoi1998,myers1999} is currently

59: one of the most active subfields of spintronics \cite{wolf2001}.

60: Recent research has highlighted a number of fundamentally

61: interesting issues that are currently under spirited debate. One

62: controversy concerns intrinsic

63: pinning\cite{tatara2004,barnes2005}, {\it i.e.}, domain walls that

64: are stationary up to a critical current in the absence of spatial

65: inhomogeneity.  An intellectually distinct but phenomenologically

66: related debate surrounds theories of nonadiabatic spin torques,

67: which differ widely in their predictions

68: \cite{zhang2004,thiaville2004,barnes2005,tserkovnyak2005,kohno2006}.

69: It turns out that in the presence of these torques, a domain wall

70: is never intrinsically pinned

71: \cite{zhang2004,thiaville2004,barnes2005}.

72: 

73: This Letter is motivated primarily by the recent experiments of

74: Yamanouchi {\it et al.} \cite{yamanouchi2006}, in which

75: current-induced domain wall motion was studied over five orders of

76: magnitude of average wall velocity.  An important conclusion of

77: these authors is that at low temperatures the domain wall

78: undergoes creep motion, {\it i.e.}, that the domain wall does not

79: move rigidly. Yamanouchi {\it et al.} arrive at this conclusion

80: because the effects of nonzero temperature, when treated

81: \cite{tatara2005} in a rigid domain wall approximation, seem to

82: lead to results that are irreconcilable with experiment. In this

83: Letter we demonstrate that a systematic theory of the influence of

84: a thermal bath on the current-driven motion of a rigid domain wall

85: leads to results that are in qualitative agreement with

86: experiment.  In the following sections we first explain our theory

87: of nonzero-temperature domain wall motion and then discuss its

88: implications for recent experiments.

89: 

90: 

91: \noindent {\it Drift velocity of a rigid domain wall} --- Our

92: starting point is the stochastic Landau-Lifschitz-Gilbert (LLG) equation

93: \cite{brown1963,kubo1970,ettelaie1984,garciapalacios1998,heinonen2004,safonov2005,rossi2005}

94: for the direction of magnetization $\hat \Omega$

95: \begin{equation}

96: \label{eq:stochasticLLG}

97:   \frac{\partial \hat \Omega}{\partial t} =\hat \Omega \bm{\times} \left (

98:   {\bf H} + \bm {\eta} \right) - \alpha \, \hat \Omega

99:   \bm{\times}  \frac{\partial \hat \Omega}{\partial t}~,

100: \end{equation}

101: where ${\bf H}$ is the effective field defined by ${\bf H}(\bx) =

102: - \delta E_{\rm MM}[\hat{\Omega}]/(\hbar \delta

103: \hat{\Omega}(\bx))$ and $E_{\rm MM}[\hat{\Omega}]$ is the

104: micromagnetic energy functional. In Eq.~(\ref{eq:stochasticLLG})

105: ${\bm \eta}$ is a gaussian stochastic magnetic field with zero

106: mean and correlations

107: \begin{equation}

108: \label{eq:noisecorrsfull}

109:   \langle \eta_\sigma (\bx,t) \eta_{\sigma'} (\bx',t') \rangle

110:   = \sigma  \delta (t-t') a^3 \delta

111:   (\bx-\bx') \delta_{\sigma\sigma'}~,

112: \end{equation}

113: where $a^3$ is the (local) volume of the finite element grid. The

114: strength of the noise is given by $\sigma = 2 \alpha k_{\rm B}

115: T/\hbar$, proportional to the Gilbert damping parameter $\alpha$

116: and to the thermal energy $k_{\rm B} T$. ($\hbar$ is Planck's

117: constant which relates energy and frequency.) In using this

118: expression for the strength of the fluctuations we neglected the

119: influence of current on thermal magnetization fluctuations

120: \cite{foros2005} which is higher order \cite{duine2006} than the

121: spin-torque effects studied here.

122: 

123: The effective field can be separated into magnetic energy and spin transfer

124: torque contributions ${\bf H}=\left.{\bf H}\right|_{0}+\left.{\bf

125: H}\right|_j$. For a ferromagnet with an easy $z$ axis and a hard $y$ axis we have

126: \begin{equation}

127: \label{eq:efffieldstatic}

128:  \left. {\bf H} \right|_0 =  J \nabla^2 \hat

129:  \Omega + 2 \omega_{\rm i}\Omega_z - 2 \omega_{\rm o} \Omega_y~ + H_{\rm ext} \hat{z},

130: \end{equation}

131: where $J$ is the spin stiffness, $H_{\rm ext}$ is an external

132: field in the easy-axis direction and $\omega_{\rm i}$ and

133: $\omega_{\rm o}$ are respectively the easy axis and hard-plane

134: anisotropy constants. Assuming only locality implied by smooth

135: magnetization textures, the spin-transfer torque can be separated

136: quite generally into contributions parallel and perpendicular to

137: the spatial derivative of the magnetization:

138: \begin{equation}

139: \label{eq:efffieldcurrent}

140:   \left. {\bf H} \right|_j   \times \hat{\Omega} = v_{\rm s} \frac{\partial \hat \Omega}{\partial r}

141:    + \beta v_{\rm s} \hat \Omega \bm{\times} \frac{\partial \hat \Omega}{\partial

142:    r}~,

143: \end{equation}

144: where the gradient is taken in the direction of current flow. In

145: the absence of spin-orbit coupling, it follows from total spin

146: conservation that $\beta = 0$ and that the {\em spin velocity}

147: $v_{\rm s} \equiv (j_{\uparrow} - j_{\downarrow})/(-e

148: (n_{\uparrow} - n_{\downarrow}))$, where $j_{\sigma}$ and

149: $n_{\sigma}$ are majority and minority spin contributions to the

150: currents and spin-densities in the collinear limit. For realistic

151: ferromagnets spin and orbital degrees of freedom are coupled, and

152: the microscopic theory of $v_s$ and $\beta$ is more challenging

153: and still controversial

154: \cite{zhang2004,thiaville2004,barnes2005,tserkovnyak2005,kohno2006}.

155: The term proportional to $\beta$, the nonadiabatic spin transfer

156: torque, plays a central role in the theory of current-driven

157: domain wall motion.

158: 

159: In the absence of current and noise, Eq.~(\ref{eq:stochasticLLG})

160: admits time-independent solutions corresponding to domain walls.

161: In terms of the angles $\theta$ and $\phi$ defined by $\hat \Omega

162: = (\sin \theta \cos \phi, \sin \theta \sin \phi, \cos \theta)$,

163: the solution corresponding to an isolated domain wall centered at

164: $r_{\rm dw}$ is $\phi=0$ and $\cos \theta (x-r_{\rm dw}) =\tanh

165: [(x-r_{\rm dw})/\lambda]$, where the domain wall width

166: $\lambda=\sqrt{J/(2\omega_{\rm i}})$\,\cite{tatara2004}. Rigid

167: domain wall motion is described by elevating $r_{\rm dw} \to

168: r_{\rm dw}(t)$ and $\phi(x,t) \to \phi_0(t)$ to the role of

169: collective dynamical variables\cite{tatara2004}. The Langevin

170: equations which describe their stochastic dynamics are

171: \begin{eqnarray}

172: \label{eq:langevinvarpars}

173:     \dot \phi_0 + \alpha \frac{\dot  r_{\rm dw}}{\lambda} &=& \frac{\beta v_{\rm

174:     s}}{\lambda} -H_{\rm ext} + \eta_{\phi} ~; \\

175: \label{eq:langevinvarpars2} \frac{\dot r_{\rm dw}}{\lambda} -

176: \alpha \dot \phi_0 &=&\omega_{\rm o} \sin (2\phi_0) + \frac{v_{\rm

177: s}}{\lambda} + \eta_r ~.

178: \end{eqnarray}

179: These equations can be derived heuristically by enforcing

180: consistency with the stochastic LLG equations at the center of the

181: domain wall. Alternately these equations can be derived by noting

182: that the probability distribution $P[\hat \Omega,t]$, generated by

183: Eqs.~(\ref{eq:stochasticLLG})~and~(\ref{eq:noisecorrsfull}) can be

184: written as a path integral $P[\hat \Omega,t] = \int d [\hat

185: \Omega] \delta [\hat{\Omega} \cdot \hat{\Omega} - 1]

186: e^{-S[\hat\Omega]}$, with effective action

187: \cite{duine2006,zinnjustinbook,duine2002}

188: \begin{equation}

189: \label{eq:effactionfull}

190:   S[\hat \Omega] = \int^{t}dt'\int\!\frac{d\bx}{a^3} \frac{1}{2 \sigma}

191:  \left(

192:    \frac{\partial \hat \Omega}{\partial t'} \times \hat \Omega - {\bf H} + \alpha

193:    \frac{\partial \hat \Omega}{\partial t'}

194:   \right)^2~.

195: \end{equation}

196: Inserting the domain-wall solution with time-dependent $r_{\rm

197: dw}$ and $\phi_0$ into this action gives rigid domain wall

198: probabilities specified by the effective action

199: \begin{eqnarray}

200: \label{eq:effactionsrdwandphi}

201:  && S[r_{\rm dw},\phi_0] = \int^t dt' \frac{N}{2\sigma}\left[

202:   \left( \dot \phi_0 + \alpha \frac{\dot  r_{\rm dw}}{\lambda} - \frac{\beta v_{\rm s}}{\lambda}+H_{\rm ext}\right)^2

203:   \right. \nonumber \\

204:   && +

205:   \left( \frac{\dot r_{\rm dw}}{\lambda} - \alpha \dot \phi_0

206:    -\omega_{\rm o } \sin (2\phi_0) - \frac{v_{\rm s}}{\lambda}\right)^2

207:   \nonumber \\

208:    && \left.

209:    -\frac{4}{3} \omega_{\rm o} ^2 \sin^4 \phi_0

210:    \right]~,

211: \end{eqnarray}

212: where $N=2\lambda A/a^3$ is the number of spins in a domain wall

213: with cross-sectional area $A$. If we ignore the last term in this

214: effective action, which vanishes in any case in the gaussian

215: fluctuation limit,

216: Eqs.~(\ref{eq:langevinvarpars})~and~(\ref{eq:langevinvarpars2})

217: are recovered. This minor inconsistency in the theoretical

218: treatment can be traced to the constant azimuthal angle across the

219: domain wall in the variational {\em ansatz}. The final term must

220: also be dropped if the rigid domain wall approximation action is

221: to reproduce the correct equilibrium probability distribution

222: function. (Note that even when $\beta=0$ the Boltzmann equilibrium

223: distribution is approached in the steady state.) The advantage of

224: this approach is that we can read off the strengths of the

225: gaussian noise terms $\eta_\phi$ and $\eta_r$ which are given by

226: \begin{equation}

227: \label{eq:noiseonvarpars}

228:  \langle \eta_\phi (t) \eta_\phi (t') \rangle

229:  = \langle \eta_r (t) \eta_r (t') \rangle

230:  = \frac{\sigma}{N} \delta (t-t')~.

231: \end{equation}

232: Eqs.~(\ref{eq:langevinvarpars}),~(\ref{eq:langevinvarpars2}),~and~(\ref{eq:noiseonvarpars})

233: generalize the variational approach to current-induced domain-wall

234: motion of Ref.~\cite{tatara2004} to finite temperature. Tatara and

235: Kohno \cite{tatara2004} derive their equations from an energy

236: functional, an approach that has been criticized recently by

237: Barnes and Maekawa \cite{barnes2005}. Our derivation does not

238: appeal to an energy functional. Therefore, in addition to deriving

239: the correct form of the noise terms required for a consistent

240: treatment of thermal fluctuations, it also provides an alternative

241: justification of the zero-temperature approach of Tatara and

242: Kohno.

243: 

244: These equations can be explored numerically without difficulty.

245: However, it turns out to be possible to make some analytic

246: progress. Solving for $\dot{\phi_0}$ alone  specializing to purely

247: current-driven motion ($H_{\rm ext}=0$) we find that

248: \begin{equation}

249: \label{eq:langevinphionly}

250: (1 + \alpha^2) \dot \phi_0 = -  \alpha \omega_{\rm o} \sin (2\phi_0)

251:  + \frac{(\beta - \alpha) v_{\rm s}}{\lambda}+ \eta_{\phi} - \alpha \eta_{r}~.

252: \end{equation}

253: This is the equation of motion for an overdamped Brownian particle

254: which has been studied extensively with a large number of

255: different physical motivations.  Our theory for the equation of

256: motion of the magnetization tilt should be contrasted with the

257: treatment in Ref.~\cite{tatara2005} which ultimately arrives at an

258: underdamped Brownian particle equation to describe nonzero

259: temperature domain wall motion.

260: 

261: The average $\langle \dot \phi_0 \rangle$ can be calculated

262: exactly for overdamped Brownian motion in a tilted periodic

263: potential \cite{riskenbook}.  Inserting this well known result in

264: Eq.~(\ref{eq:langevinvarpars}) we find that

265: \begin{widetext}

266: \begin{equation}

267: \label{eq:fulleqndwvelocity}

268:    \langle \dot r_{\rm dw} \rangle

269: \!=\!\frac{\beta v_{\rm s}}{\alpha}-

270:   \frac{ 2\pi \lambda k_{\rm B} T \left[1\!-\!e^{

271:        2 \pi \hbar N  (\alpha-\beta) v_{\rm s}

272:        /(\alpha \lambda k_{\rm B} T) }\right]}

273:        {\hbar N \left\{\int_0^{2\pi}e^{ V(\phi)/(k_{\rm B} T)}

274:        d\phi \int_0^{2\pi}e^{- V(\phi')/(k_{\rm B} T)} d\phi'

275:    \!-\!\left[1\!-\!e^{

276:        2 \pi \hbar N  (\alpha-\beta) v_{\rm s}

277:        /(\alpha \lambda k_{\rm B} T) }\right] \int_0^{2\pi}e^{- V(\phi)/(k_{\rm B} T)}

278:        d\phi \int_0^{\phi}e^{V(\phi')/(k_{\rm B} T)}

279:        d\phi'\right\}}~,

280: \end{equation}

281: \end{widetext}

282: where

283: \begin{equation}

284: \label{eq:potentialphi} V(\phi_0)=- \hbar N \omega_{\rm o} \cos

285: (2\phi_0)/2+\hbar N(\alpha-\beta)v_{\rm s}\phi_0/(\alpha

286: \lambda)~,

287: \end{equation}

288: is the tilted-washboard potential experienced by the Brownian

289: ``$\phi_0$-particle'' \cite{tatara2004}. In the above equations we

290: have assumed that $\alpha^2 \ll 1$, for notational convenience.

291: 

292: It is well-known \cite{riskenbook} that at zero temperature the

293: equation for an overdamped particle in the tilted periodic

294: potential of Eq.~(\ref{eq:langevinphionly}) has solutions with

295: $\langle \dot \phi \rangle \neq 0$ only if $|(\alpha - \beta )

296: v_{\rm s} |> \alpha \lambda \omega_{\rm o}$. For $\beta=0$ the

297: zero temperature result for the average velocity of the domain

298: wall is $\langle \dot r_{\rm dw} \rangle \propto \sqrt{(v_{\rm

299: s}/v_{\rm sc})^2-1}$ \cite{tatara2004,riskenbook}, where the

300: critical current for depinning the domain wall is $v_{\rm sc} =

301: \lambda \omega_{\rm o}$ \cite{tatara2004}.  For $\beta \ne 0$, the

302: average domain wall velocity is nonzero at any finite value of

303: $v_s$ even at zero temperature.

304: 

305: 

306: \begin{figure}

307: \vspace{-0.5cm} \centerline{\epsfig{figure=v_j.eps}}

308:  \caption{Domain wall velocity as a function of current,

309:  for various temperatures and values of $\beta$. We take $\alpha=0.02$.

310:  The horizontal axis is normalized to $v_{\rm sc}=\lambda \omega_{\rm o}$.}

311:  \label{fig:v_j}

312: \end{figure}

313: 

314: In Fig.~\ref{fig:v_j} we plot the average domain-wall velocity

315: calculated from the full expression

316: [Eq.~(\ref{eq:fulleqndwvelocity})] for various temperatures and

317: values of $\beta$. These curves were calculated for Gilbert

318: damping parameter $\alpha=0.02$. We consider only $\beta<\alpha$

319: since for $\beta>\alpha$ the domain wall velocity would decrease

320: with increasing temperature in clear disagreement with experiment

321: \cite{yamanouchi2006}. One of the main conclusions of

322: Ref.~\cite{tatara2005} is that for $\beta = 0$ and small currents

323: the average domain-wall velocity $\langle \dot r_{\rm dw} \rangle

324: \propto \exp(C v_{\rm s})$ where $C$ is a constant. In the

325: low-temperature limit we find that for $v_{\rm s}/v_{\rm sc} < 1$

326: the average domain wall velocity is given by

327: \begin{widetext}

328: \begin{equation}

329:   \langle

330:   \dot r_{\rm dw} \rangle

331:    = \frac{\beta v_{\rm s}}{\alpha}

332:    + 2 \sqrt{\left(\lambda \omega_{\rm o}\right)^2 - \left( v_{\rm s}\right)^2}

333:    \exp\left\{ \left(\frac{\beta}{\alpha} -1 \right)

334:    \frac{N \hbar \omega_{\rm o}}{k_{\rm B} T}\left[

335:     \sqrt{1-\left( \frac{v_{\rm s}}{v_{\rm sc}}\right)^2} + \left(\frac{v_{\rm s}}{v_{\rm

336:     sc}}\right)

337:     \sin^{-1} \left(\frac{v_{\rm s}}{v_{\rm sc}} \right)\right]\right\}

338:   \sinh \left( \frac{\pi \hbar N \left( \alpha - \beta \right) v_{\rm s}}{\alpha \lambda k_{\rm B} T}

339: \right)~.

340: \label{lowT}

341: \end{equation}

342: \end{widetext}

343: This difference in estimated domain wall velocities is closely

344: analogous to the difference between the Langer-Ambegaokar

345: \cite{LA} and Halperin-McCumber \cite{HMcC} estimates of phase

346: slip resistance in thin superconducting wires. It follows from

347: Eq.~(\ref{lowT}) that the observation of linear dependence of

348: domain wall velocity on current does not necessarily imply that

349: $\beta \neq 0$. A careful analysis of the temperature dependence

350: of $\langle \dot r_{\rm dw} \rangle$ will be necessary to

351: determine the value of $\beta$ from low current experiments,

352: especially so if $N \hbar \omega_{\rm o}$ is not very large

353: compared to $k_{\rm B} T$. Because of thermal noise, even when

354: $\beta=0$ the domain is not intrinsically pinned at any nonzero

355: temperature.

356: 

357: \noindent {\it Comparison with experiment} --- The results

358: presented in Fig.~\ref{fig:v_j} look qualitatively  similar to the

359: experimental results of Yamanouchi {\it et al.}

360: \cite{yamanouchi2006}, in particular to the inset in Fig.~3 of

361: Ref.~\cite{yamanouchi2006}. A direct comparison with these

362: experimental results is complicated by the fact that they are

363: performed close to the critical temperature, with the result that

364: the magnetic anisotropy energy-density and the polarization of the

365: current depend on temperature. Moreover, the Gilbert-damping

366: parameter $\alpha$ may also depend on temperature. The fact that

367: curves at different temperatures in Fig.~\ref{fig:v_j} are not as

368: strongly offset vertically  from each other at $v_{\rm s}/v_{\rm

369: sc}>1$ as the experimental results is most likely due to these

370: additional temperature-dependent effects.

371: 

372: \begin{figure}

373: \vspace{-0.5cm} \centerline{\epsfig{figure=v_sqrtj.eps}}

374:  \caption{Test of the scaling $\ln \langle \dot r_{\rm dw} \rangle \propto \sqrt{v_{\rm s}}$,

375:  for various temperatures and values of $\beta$. We take

376:  $\alpha=0.02$.}

377:  \label{fig:v_sqrtj}

378: \end{figure}

379: 

380: One of the main results of Yamanouchi {\it et al.} is the

381: empirical finding that for small currents $\ln \langle \dot r_{\rm

382: dw} \rangle \propto \sqrt{ v_{\rm s}}$. From this the authors

383: conclude that the domain wall undergoes current-induced creep

384: motion at small currents. In Fig.~\ref{fig:v_sqrtj} we plot the

385: average domain wall velocity as a function of $\sqrt{v_{\rm s}}$

386: for various temperatures and values of $\beta$. From this we

387: observe that $\ln \langle \dot r_{\rm dw} \rangle$ is proportional

388: to the square root of the current over several decades of domain

389: wall velocity. It turns out that this approximate relationship is

390: valid for a larger range of parameter space with decreasing

391: temperature. In the experiments of Yamanouchi {\it et al.} a

392: typical velocity is $\langle \dot r_{\rm dw} \rangle \sim 1$ m

393: s$^{-1}$. Putting this equal to $\lambda \omega_{\rm o}$ we have

394: for $\lambda = 17$ nm and $T \sim 100$ K that $k_{\rm B} T/(N

395: \hbar \omega_{\rm o}) \sim 0.01$ for $N=3.4\times10^6$ Mn moments

396: (we take the density of moments $N_{\rm Mn}\sim 1$ nm$^{-3}$) in a

397: wall of cross-sectional area $A=2$ nm $\times$ $5$ $\mu$m. For

398: this value of $k_{\rm B} T/(N \hbar \omega_{\rm o})$ we conclude

399: from Fig.~\ref{fig:v_sqrtj} that $\ln \langle \dot r_{\rm dw}

400: \rangle \propto \sqrt{v_{\rm s}}$ over a large regime, and that

401: our theory is therefore in agreement with the experimental results

402: of Yamanouchi {\it et al.} \cite{yamanouchi2006}. Moreover, from

403: Fig.~\ref{fig:v_sqrtj} we also conclude that for $\beta=0.01$ this

404: scaling does not hold. Hence, an additional conclusion from the

405: experimental data is that $\beta$ is much smaller than  $\alpha$

406: in these ferromagnetic semiconductors.

407: 

408: A similar estimate for typical parameters of the metallic

409: nanowires used in studies of current-driven domain wall motion

410: \cite{grollier2003,tsoi2003,yamaguchi2004,klaui2005} leads to the

411: conclusion that $k_{\rm B} T/(N\hbar\omega_{\rm o})\sim0.0002$ and

412: therefore that temperature effects on rigid domain wall motion are

413: likely much less important in these systems. This difference

414: arises mainly because the density of moments is roughly $40$ times

415: higher in metallic systems.

416: 

417: In conclusion, we have presented a theory of the influence of

418: nonzero temperatures on current-driven motion of a rigid domain

419: wall, and found qualitative agreement with ferromagnetic

420: semiconductor experiments. An estimate of the bending energy of

421: the domain wall shows that these degrees of freedom are also

422: thermally accessible. It is, however, difficult to assess how they

423: influence translation of the domain wall. The qualitative

424: agreement of our results with the experiments of Yamanouchi {\it

425: et al.} \cite{yamanouchi2006} indicates that thermal activation of

426: rigid domain wall motion could play an important role in these

427: experiments. We expect that accounting for thermal fluctuations

428: will also be important in assessing the impact of intended and

429: unintended extrinsic domain wall pinning.   This work was

430: supported by the National Science Foundation under grants

431: DMR-0115947 and DMR-0210383, by a grant from Seagate Corporation,

432: and by the Welch Foundation.

433: 

434: 

435: 

436: 

437: \begin{thebibliography}{99}

438: %%% First by Berger:

439: \bibitem{berger1984} L. Berger, J. Appl. Phys. {\bf 55}, 1954

440: (1984).

441: \bibitem{freitas1985} P. P. Freitas and L. Berger, J. Appl. Phys. {\bf 57}, 1266

442: (1985).

443: 

444: 

445: %%% CI-DW theory

446: \bibitem{tatara2004} G. Tatara and H. Kohno, Phys. Rev. Lett. {\bf 92}, 086601

447: (2004); Phys. Rev. Lett. {\bf 96}, 189702 (2006).

448: \bibitem{zhang2004} S. Zhang and Z. Li, Phys. Rev. Lett. {\bf 93}, 127204 (2004).

449: \bibitem{thiaville2004}  A. Thiaville, Y. Nakatani, J. Miltat, and Y

450: Suzuki, Europhys. Lett. {\bf 69}, 990 (2005).

451: \bibitem{barnes2005} S. E. Barnes and S. Maekawa, Phys. Rev. Lett. {\bf 95},

452: 107204 (2005); S. E. Barnes, Phys. Rev. Lett. {\bf 96}, 189701

453: (2006).

454: \bibitem{tserkovnyak2005} Y. Tserkovnyak, A. Brataas, and G. E. W.

455: Bauer, cond-mat/0512715.

456: \bibitem{kohno2006} H. Kohno, G. Tatara, and J. Shibata,

457: cond-mat/0605186.

458: \bibitem{nguyen2006}  A. K. Nguyen, H. J. Skadsem, and A. Brataas,

459: cond-mat/0606498.

460: 

461: 

462: %%% CI-DW expt metals

463: \bibitem{grollier2003} J. Grollier, P. Boulenc, V. Cros, A. Hamzi, A. Vaurès,

464: A. Fert, and G. Faini, Appl. Phys. Lett. {\bf 83}, 509 (2003).

465: \bibitem{tsoi2003} M. Tsoi, R.E. Fontana, and S.S.P. Parkin, Appl. Phys. Lett. {\bf 83}, 2617

466: (2003).

467: \bibitem{yamaguchi2004} A. Yamaguchi, T. Ono, S. Nasu, K. Miyake, K. Mibu, and T.

468: Shinjo, Phys. Rev. Lett. {\bf 92}, 077205 (2004).

469: \bibitem{klaui2005} M. Kl\"aui, C. A. F. Vaz, J. A. C. Bland, W. Wernsdorfer,

470: G. Faini, E. Cambril, L. J. Heyderman, F. Nolting, and U.

471: R\"udiger, Phys. Rev. Lett. {\bf 94}, 106601 (2005).

472: 

473: %%% CD-DW expt FMSCs

474: \bibitem{yamanouchi2004} M. Yamanouchi, D. Chiba, F. Matsukura, and H.

475: Ohno, Nature {\bf 428}, 539 (2004).

476: \bibitem{yamanouchi2006} M. Yamanouchi, D. Chiba, F. Matsukura, T. Dietl, and H.

477: Ohno, Phys. Rev. Lett. {\bf 96}, 096601 (2006).

478: 

479: %%% STT Theory

480: \bibitem{slonczewski1996} J.C. Slonczewski, J. Mag. Mat. Mag. {\bf 159}, L1

481: (1996).

482: \bibitem{berger1996} L. Berger, Phys. Rev. B {\bf 54}, 9353 (1996).

483: 

484: %%% STT Experiment

485: \bibitem{tsoi1998} M. Tsoi, A. G. M. Jansen, J. Bass, W.-C. Chiang, M. Seck, V. Tsoi, and P.

486: Wyder, \prl {\bf 80}, 4281 (1998).

487: \bibitem{myers1999} E. B. Myers, D. C. Ralph, J. A. Katine, R. N. Louie, R.

488: A. Buhrman, Science {\bf 285}, 867 (1999).

489: 

490: %%% Spintronics general

491: \bibitem{wolf2001} S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. M. Daughton, S.

492: von Moln\'ar, M. L. Roukes, A. Y. Chtchelkanova, D. M. Treger,

493: Science {\bf 294}, 1488 (2001).

494: 

495: %%% Tatara finite T

496: \bibitem{tatara2005} G. Tatara, N. Vernier, and J. Ferr\'e, Appl. Phys. Lett. {\bf 86},

497: 252509 (2005).

498: 

499: %%% General References Stochastic LLG

500: \bibitem{brown1963} W.F. Brown, Jr., Phys. Rev. {\bf 130}, 1677

501: (1963).

502: \bibitem{kubo1970} R. Kubo and N. Hashitsume, Prog. Theor. Phys.

503: Suppl. {\bf 46}, 210 (1970).

504: \bibitem{ettelaie1984} R. Ettelaie and M.A. Moore, J. Phys. A {\bf

505: 17}, 3505 (1984).

506: \bibitem{garciapalacios1998} J. L. Garc\'ia-Palacios and

507: F. J. L\'azaro, \prb {\bf 58}, 14937 (1998).

508: \bibitem{heinonen2004} O.G. Heinonen and H.S. Cho,

509: IEEE Transactions on Magnetics {\bf 40}, 2227 (2004).

510: \bibitem{safonov2005} V. L. Safonov and H. N. Bertram, \prb {\bf

511: 71}, 224402 (2005).

512: \bibitem{rossi2005} E. Rossi, O. G. Heinonen, and A.H. MacDonald,

513: \prb {\bf 72}, 174412 (2005).

514: 

515: %%% effective T refs

516: \bibitem{foros2005} J\o rn Foros, A. Brataas, Y. Tserkovnyak, and G. E. W.

517: Bauer, \prl {\bf 95}, 016601 (2005).

518: \bibitem{duine2006} R.A. Duine, A.S. N\'u\~nez, and A.H. MacDonald, to

519: be submitted.

520: 

521: %%% ZJ book

522: \bibitem{zinnjustinbook} J. Zinn-Justin, {\it Quantum Field Theory

523: and Critical Phenomena} (Oxford, New York, 1989).

524: 

525: %%% stochastic business with PI's

526: \bibitem{duine2002} R.A. Duine and H.T.C. Stoof, Phys. Rev. A {\bf 65}, 013603

527: (2002).

528: 

529: %%% Risken book

530: \bibitem{riskenbook} H. Risken, {\it The Fokker-Planck Equation}

531: (Springer-Verlag, Berlin, 1984).

532: 

533: 

534: %%% relation to super wires

535: \bibitem{LA} J.S. Langer and V. Ambegaokar, Phys. Rev. {\bf 164}, 498 (1967).

536: 

537: \bibitem{HMcC} D.E. McCumber and B.I. Halperin, Phys. Rev. B {\bf 1}, 1054 (1970).

538: 

539: 

540: 

541: 

542: 

543: 

544: 

545: 

546: \end{thebibliography}

547: 

548: 

549: 

550: 

551: \end{document}

552: