1: \documentclass[

2: %preprint,

3: showpacs,

4: floatfix,

5: aps,prl,amsmath,

6: %nofootinbib,

7: twocolumn,

8: superscriptaddress,

9: %groupaddress,

10: %eqsecnum

11: ]{revtex4}

12: 

13: 

14: %\documentclass[twocolumn,aps,prb,epsf]{revtex4}

15: 

16: \usepackage{epsfig}

17: \usepackage{bm}

18: %

19: \usepackage{epsf}

20: \usepackage{array}

21: %

22: 

23: 

24: \usepackage[varg]{txfonts}

25: 

26: 

27: \newcommand{\etal}{ {\em et al.}}

28: 

29: \newcommand{\be}{\begin{equation}}

30: \newcommand{\ee}{\end{equation}}

31: 

32: \newcommand{\bea}{\begin{eqnarray}}

33: \newcommand{\eea}{\end{eqnarray}}

34: 

35: \newcommand{\bes}{\begin{split}}

36: \newcommand{\ees}{\end{split}}

37: 

38: 

39: 

40: 

41: 

42: \newcommand{\req}[1]{Eq.~(\ref{#1})}

43: \newcommand{\reqs}[1]{Eqs.~(\ref{#1})}

44: \newcommand{\rref}[1]{(\ref{#1})}

45: \newcommand{\mylabel}[1]{\label{#1}

46: %\tag{#1}

47: }

48: 

49: 

50: \newcommand{\wc}{\omega_{\rm c}}

51: \newcommand{\pF}{p_{\rm F}}

52: \newcommand{\Tc}{T_{\rm c}}

53: \newcommand{\Rc}{R_{\rm c}}

54: \newcommand{\vF}{v_{\rm F}}

55: \newcommand{\tautr}{\tau_{\rm tr}}

56: \newcommand{\tauin}{\tau_{\rm in}}

57: \newcommand{\tauint}{\tilde \tau_{\rm in}}

58: \newcommand{\tauq}{\tau_{0}}

59: \newcommand{\taul}{\tau_{\rm sm}}

60: \newcommand{\taus}{\tau_{\rm sh}}

61: \newcommand{\tautrl}{\tilde \tau_{\rm sm}}

62: 

63: 

64: 

65: \newcommand{\w}{\omega}

66: 

67: \newcommand{\W}{\Omega}

68: \newcommand{\e}{\epsilon}

69: \newcommand{\nn}{\mbox{\boldmath $i$}}

70: \newcommand{\mm}{\mbox{\boldmath $e$}}

71: \renewcommand{\r}{\mbox{\boldmath $r$}}

72: \renewcommand{\P}{\mbox{\boldmath $P$}}

73: \newcommand{\p}{\mbox{\boldmath $p$}}

74: \newcommand{\R}{\mbox{\boldmath $R$}}

75: \renewcommand{\k}{\mbox{\boldmath $k$}}

76: \renewcommand{\v}{\mbox{\boldmath $v$}}

77: \newcommand{\E}{\mbox{\boldmath $E$}}

78: \renewcommand{\j}{\mbox{\boldmath $j$}}

79: \newcommand{\nnabla}{\mbox{\boldmath $\nabla$}}

80: \newcommand{\F}{\mbox{\boldmath $F$}}

81: \newcommand{\s}{{\hat{\sigma}}}

82: \renewcommand{\t}{{\hat{\tau}}}

83: \newcommand{\q}{\mbox{\boldmath $q$}}

84: \newcommand{\z}{\mbox{\boldmath $z$}}

85: \newcommand{\Q}{\mbox{\boldmath $Q$}}

86: \newcommand{\sgn}{\mbox{sgn}}

87: \renewcommand{\>}{\rangle}

88: \newcommand{\<}{\langle}

89: \newcommand{\ptau}{\partial_\tau}

90: \newcommand{\pt}{\partial_t}

91: \newcommand{\px}{\partial_x}

92: \newcommand{\bzeta}{\mbox{\boldmath $\zeta$}}

93: \newcommand{\bbeta}{\mbox{\boldmath $\eta$}}

94: \newcommand{\bchi}{\mbox{\boldmath $\chi$}}

95: \newcommand{\St}{\mbox{St}}

96: \newcommand{\Arg}{\mbox{$\mathrm Arg$}}

97: \newcommand{\Arctan}{\mbox{$\mathrm Arctan$}}

98: \renewcommand{\Re}{\mbox{$\mathrm Re$} }

99: \renewcommand{\Im}{\mbox{$\mathrm Im$} }

100: \newcommand{\tr}{\mbox{$\mathrm tr$} }

101: 

102: \newcommand{\vare}{\varepsilon}

103: \newcommand{\lH}{\lambda_H}

104: 

105: 

106: 

107: \begin{document}

108: 

109: \title{Non-linear Resistivity of a Two-Dimensional Electron Gas in a

110:   Magnetic Field}

111: \author{M. G. Vavilov}

112: \affiliation{Department of Physics, University of Wisconsin, Madison, WI 53706 }

113: \author{I. L.  Aleiner}

114: \affiliation{ Physics Department, Columbia University, New

115:   York, NY 10027 }

116: \author{L. I. Glazman}

117: \affiliation{Theoretical Physics Institute, University of Minnesota, Minneapolis, MN 55455 }

118: 

119: \date{November 5, 2006}

120: 

121: \begin{abstract}

122: We develop a theory of nonlinear response to an electric field of a

123: two-dimensional electron gas (2DEG) placed in a classically strong

124: magnetic field. The latter leads to a non-linear current-voltage

125: characteristic at a relatively weak electric field. The origin of

126: the non-linearity is two-fold: the formation of a non-equilibrium

127: electron distribution function, and the geometrical resonance in the

128: inter-Landau-levels transitions rates. We find the dependence of the

129: current-voltage characteristics on the electron relaxation rates

130: in the 2DEG.

131: \end{abstract}

132: \pacs{73.40.-c, 73.50.Pz, 73.43.Qt, 73.50.Fq }

133: \maketitle

134: 

135: A magnetic field applied to a two-dimensional electron gas (2DEG)

136: changes the energy spectrum and

137: dynamics of electrons. It leads to a modification of the transport

138: characteristics of the 2DEG even at relatively weak magnetic fields, at

139: which the Landau levels~\cite{Ando} are not resolved yet and

140: the Quantum Hall effect~\cite{prange} is not developed.

141: The most well-studied modification of that

142: kind is the Shubnikov -- de Haas (SdH) oscillations in the resistivity

143: of 2DEG. Its observation, however, is restricted to fairly low

144: temperatures, $T\lesssim\hbar\wc$, so that the thermal broadening

145: $T$ of the electron distribution is small compared to the Landau

146: quantization energy $\hbar\wc$. In the low-temperature limit, the

147: strength of the oscillations is controlled by the Dingle factor,

148: $\lambda= [-\pi/(\wc\tauq)]$; it yields information about the

149: ``quantum'' lifetime $\tauq$ of the electron~\cite{Ando}

150: due to scattering off disorder.

151: 

152: Recently it was realized~\cite{du02} that the effect of a magnetic

153: field on the dc non-linear transport, unlike the SdH oscillations of

154: the linear resistivity, is not confined to low temperatures.

155: Oscillations of the differential resistivity with the magnetic field

156: at a finite level of current observed in Ref.~\cite{du02} persisted to

157: quite high temperatures (about $4$K), while the conventional SdH

158: effect was fully smeared out by temperature. This finding was

159: confirmed in later experiments~\cite{vitkalov05,vitkalov06,zudov06}

160: where the differential resistivity was measured both as a function of

161: the applied magnetic field and transport current. The oscillations of

162: the nonlinear resistance were associated with the geometrical

163: resonance in the electron transitions between the Landau

164: levels~\cite{du02} that arises from the commensurability of

165: the period in the spatial oscillations of the density of states (DOS) and

166: the diameter $2\Rc$ of an electron cyclotron trajectory.

167: Although this was a plausible explanation of the

168: effect, it remained unclear, why the oscillations are so weakly

169: sensitive to the temperature and which parameters of the 2DEG control

170: the amplitude of the oscillations.

171: 

172: Another notable effect of magnetic fields on the

173: nonlinear transport in 2DEG was reported in Ref.~\cite{vitkalov06} and

174: deals with the region of relatively small current densities. In that

175: regime, a sharp drop in the differential resistivity was observed.

176: The effect was attributed to the modification of the electron energy

177: distribution caused by the current~\cite{DVAMP}. Clearly, this modification

178: depends on the energy relaxation rate, and the mechanisms behind the

179: observations reported in Refs.~\cite{du02} and \cite{vitkalov06} seem

180: quite different from each other.

181: 

182: The goal of our work is to show that the two seemingly different

183: phenomena are essentially two manifestations of the electron kinetics

184: described by a standard Boltzmann equation for a weakly disordered

185: 2DEG in the presence of electric and magnetic fields. We demonstrate

186: that the low-current nonlinearity~\cite{vitkalov06} is the consequence

187: of the variation of the occupation factors of electron states: the

188: non-equilibrium population of states renders the transitions normally

189: contributing to the dissipative current ineffective. At high currents,

190: the effect of the electric and magnetic fields on the electron motion

191: becomes important. The oscillations in the $I$-$V$ characteristic are

192: associated with the geometric resonance in the electron transitions.

193: 

194: We evaluate the dissipative component of the electric current density

195: in a 2DEG placed in a perpendicular magnetic field $B$ as a function

196: of electric field characterized by the dimensionless parameter

197: $\zeta$,

198: \be \zeta=\pi \frac{2 eE\Rc}{\hbar\wc},\ \ \ \Rc=\frac{v_{\rm

199:     F}}{\wc},\ \ \ \wc=\frac{eB}{m_{\rm e}c},

200: \label{zeta}

201: \ee

202: proportional to the ratio of the work of electric field

203: associated with the displacement of the guiding center of

204: a cyclotron trajectory by $2\Rc$ to the Landau quantization energy

205: $\hbar\wc$. The displacement occurs due to the electron scattering off

206: an impurity, and does not exceed $2\Rc$ in a single scattering act.

207: This geometrical constraint leads to the oscillations of the current

208: with $\zeta$; each oscillation corresponds to an increase by

209: $\hbar\wc$ of the maximal energy acquired by an electron from the

210: electric field in a single scattering event. The maximal

211: displacement of the guiding center is reached for the scattering angle $\pi$

212: (backscattering), thus the amplitude of oscillations is proportional to

213: the corresponding scattering rate $1/\tau(\pi)$.

214: 

215: The ``preferred'' values of the energy absorbed by an electron from

216: the electric field are multiples of $\hbar\wc$ because of the

217: oscillations of the electron DOS associated with the

218: Landau quantization. It is interesting to note however, that even a

219: strong electric field does not result in developing a substantial

220: modulation in the electron energy distribution with the period

221: $\hbar\wc$. The reason for that is the dual role electric field

222: plays. On one hand, it promotes the build-up of electron distribution

223: at the energies corresponding to the maxima of the DOS.

224: On the other hand, it increases the electron diffusion in energy

225: space, the corresponding coefficient of the spectral diffusion being

226: proportional to the Joule losses. The latter effect wins over the

227: former one, and the electron distribution in energy gets smoother at

228: higher fields. As the result, the oscillatory part of the $I$-$V$

229: characteristic reflects the modulation of the electron transition

230: rates with the field, rather the modifications in the electron

231: distribution function; the amplitude of oscillations provides

232: information about the backscattering rate $1/\tau(\pi)$,

233: hardly accessible in other experiments.

234: 

235: 

236: In this work we express the dissipative current in terms  of the

237: inelastic relaxation rate $1/\tauin$ and harmonics $1/\tau_n$

238: of the elastic electron scattering rate $1/\tau(\theta)$ on angle $\theta$:

239: \be

240: \frac{1}{\tau(\theta)}=

241: \sum_{n=-\infty}^{+\infty}

242: \frac{e^{in\theta}}{\tau_{n}},\quad

243: \tau_n=\tau_{-n}.

244: \label{tau}

245: \ee

246: %with $\tau_{n}=\tau_{-n}$.

247: Typically, the ``quantum scattering time'' $\tauq$ is short,

248: $\tau_0\ll \tau_{\rm in}$. However, the transport relaxation time,

249: defined as $1/\tautr=1/\tauq-1/\tau_1$, may be in an arbitrary

250: relation with $\tau_{\rm in}$. We show that the measurements of the

251: dissipative current as a function of $\zeta$ at small ($\zeta \lesssim

252: \sqrt{\tauq/\tauin}$) and large ($\gtrsim 1$) values of $\zeta$

253: reveal the rates of inelastic relaxation and of the back-scattering

254: off disorder, respectively.

255: 

256: In the following, we consider the limit of high temperatures $T\gtrsim

257: \hbar\wc/2\pi^2$, when the non-linear resistivity is observed, but the

258: SdH oscillations are already suppressed~\cite{du02}. We also limit our

259: analysis to the case of ``classically strong'' magnetic fields, {\it

260:   i.e.}, we assume that $\wc\tauq\lesssim 1$ while $\wc\tautr\gg

261: 1$. The former condition allows us to keep only the first harmonic in

262: the oscillations of the DOS

263: \be

264: \begin{split}

265: &\nu(\vare)= \nu_0\left(1-2\lambda\cos\frac{2\pi\vare}{\hbar\wc}

266: \right),\quad \lambda=  e^{-\pi/\wc\tauq}. \label{dos}

267: \end{split}

268: \ee

269: The condition $\tautr/\tauq\gg 1$ is routinely met in semiconductor

270: heterostructures, and the domain of magnetic fields $1/\tautr\ll \wc\ll

271: 1/\tauq$ is quite wide.

272: 

273: The dissipative part of the electric current

274: \be

275: j_{\rm d}=2e\vF\int d\vare \nu(\vare)

276: \int \cos\varphi

277: f(\vare,\varphi)\frac{d\varphi}{2\pi}

278: \label{jdc}

279: \ee

280: is determined by the stationary electron distribution function

281: $f(\vare,\varphi)$, which is the solution of the following

282: kinetic equation

283: \be

284: -\wc\partial_\varphi f(\vare,\varphi)=

285: {\rm St}_{\varphi}\{f(\vare,\varphi)\}+

286: {\rm St}_{\rm in}\{f(\vare,\varphi)\}.

287: \label{kineq}

288: \ee

289: Here $\varphi$ is the angle the electron momentum makes with the

290: direction of the electric field. The first term in the right hand side

291: of \req{kineq} is the collision integral for electron scattering off

292: disorder:

293: \be

294: \begin{split}

295: {\rm St}_{\varphi}  \{f\}=

296: \int\frac{

297: \nu(\vare+ W_{\varphi\varphi'})}{\nu_0}

298: \frac{f(\vare+ W_{\varphi\varphi'},\varphi')

299: -

300: f(\vare,{\varphi})}{\tau(\varphi-\varphi')}\frac{d\varphi'}{2\pi}.

301: \label{Stel}

302: \end{split}

303: \ee

304: Here $W_{\varphi\varphi'}=eE \Rc [\sin{\varphi'}-\sin\varphi]$

305: is the work of the electric field in the course of the shift

306: $\Rc \bm{z}\times[\bm{n}_{\varphi'}-\bm{n}_\varphi]$ of the

307: guiding center of the cyclotron trajectory, see Fig.~\ref{fig:explanation};

308: unit vector $\bm{z}$ is perpendicular to the 2DEG plane, and

309: $\bm{n}_\varphi=\{\cos\varphi,\ \sin\varphi\}$ is directed along the electron momentum.

310: The rate of such scattering events is given by $1/\tau(\varphi'-\varphi)$ and

311: characterized by its harmonics $1/\tau_n$, see \req{tau}.

312: The imbalance between scattering ``in'' and ``out'' terms is determined

313: by the difference of the corresponding distribution functions

314: $f(\vare+W_{\varphi\varphi'},\varphi')$ and

315: $f(\vare,\varphi)$.

316: 

317: \begin{figure}

318: \epsfxsize=0.25\textwidth

319: %\vspace*{0.3\textwidth}

320: \centerline{\epsfbox{figure1.eps}}

321: \caption{Electron scattering off impurity changes the momentum direction from

322: $\bm{n}_\varphi$ to $\bm{n}_{\varphi'}$ and the position of the guiding center

323: shifts by $\Rc\bm{z}\times[\bm{n}_{\varphi'}-\bm{n}_\varphi]$.}

324: \label{fig:explanation}

325: \end{figure}

326: 

327: The inelastic relaxation at sufficiently low temperatures is

328: dominated by the electron-electron interaction, represented by

329: the inelastic collision integral

330: \begin{eqnarray}

331: &&{\rm St}_{\rm in}\left\{ f (\vare) \right\}  =

332: \int d\vare'\int dE   M(E, \vare, \vare')

333: \label{Stin}

334: \\

335: &&\times \left[

336: \tilde  f (\vare)f(\vare_+)\tilde f (\vare')

337: f(\vare'_-)

338: -f(\vare)\tilde f (\vare_+)f(\vare') \tilde f (\vare'_-)

339: \right].

340: \nonumber

341: \end{eqnarray}

342: Here $\tilde f(\vare) \equiv 1-f(\vare)$, $\vare_{+}= \vare + E$, $\vare_{-}'=

343: \vare' - E$ and $M(E,\,\vare,\,\vare')$ describes the

344: dependence of the matrix element of the screened Coulomb interaction on the

345: transferred energy $E$ and the electron energies $\vare$ and

346: $\vare'$.

347: 

348: To the first order in  $1/\wc\tautr\ll 1$ we look for a solution

349: of the kinetic equation~\req{kineq} in the form:

350: \begin{eqnarray}

351: &&f(\vare,\varphi)=

352: \left[1-\frac{eE\cos\varphi \Rc^2}{\vF\tautr}\partial_\vare\right]

353: f_{T}(\vare)

354: \label{fgen}

355: \\

356: &&

357: +\lambda \left\{I_1\sin\frac{2\pi \vare}{\hbar\wc}

358: +\left[A_1

359: \cos\frac{2\pi \vare}{\hbar\wc}+\lambda A_2\right]\cos\varphi

360: \right\} \partial_\vare f_{T}(\vare).

361: \nonumber

362: \end{eqnarray}

363: The first term in \req{fgen} is the result for $\lambda=0$, when the

364: effect of quantization by magnetic field is neglected. The second term

365: in \req{fgen} is proportional to $\lambda$ and oscillates with

366: $\vare/\wc$. Here we assume  that $T\gg \wc$, so that we can

367: separate fast oscillatory dependence on energy of $f(\vare,\varphi)$

368: with period $\hbar\wc$ and smooth energy dependence of

369: $f_T(\vare)$ on the scale

370: of temperature $T$ of the 2DEG. We assume that $\hbar\wc \zeta\ll

371: T$ and that $f_{T}(\vare)$ is close to the Fermi distribution

372: function.

373: 

374: The oscillations amplitude $I_1(\vare)$

375: of isotropic in momentum space component of the distribution function

376: \req{fgen} is determined by

377: \begin{subequations}

378: \label{kineq:amp}

379: \be

380: \frac{I_1(\vare)}{ \tau_{\rm in}(\vare)}\sin\frac{2\pi\vare}{\hbar\wc}=

381: \left\langle{\cal K}^{(0)}_{\varphi\varphi'}(\vare)  I_1

382: +

383: {\cal K}^{(1)}_{\varphi\varphi'}(\vare)

384: \right\rangle

385: .

386: \label{kineq:iso}

387: \ee

388: For the amplitudes $A_{1,2}$ of the anisotropic component in \req{fgen}

389: we have

390: \be

391: \frac{\wc}{2}A_1\cos\frac{2\pi\vare}{\hbar\wc}=

392: \left\langle \sin\varphi\left[{\cal K}^{(0)}_{\varphi\varphi'}(\vare)

393: I_1 +

394: {\cal K}^{(1)}_{\varphi\varphi'}(\vare)  \right]

395: \right\rangle,

396: \label{kineq:j}

397: \ee

398: \be

399: \frac{\wc}{4}A_2=

400: -\left\langle \sin\varphi\cos\frac{2\pi(\vare+W_{\varphi\varphi'})}

401: {\hbar\wc}{\cal K}^{(0)}_{\varphi\varphi'}(\vare)

402: \right\rangle I_1.

403: \label{kineq:j2}

404: \ee

405: \end{subequations}

406: Here $\langle\dots \rangle$ stands for averaging over

407: angle variables $\varphi$ and $\varphi'$.

408: The kernels ${\cal K}^{(0,1)}_{\varphi\varphi'}(\vare)$ are given by

409: \be

410: \begin{split}

411: {\cal K}^{(0)}_{\varphi\varphi'}(\vare) & =

412:  \frac{\sin[2\pi(\vare+W_{\varphi\varphi'})/\hbar\wc]

413: -\sin[2\pi\vare/\hbar\wc]

414: }{\tau(\varphi-\varphi')},

415: \\

416: {\cal K}^{(1)}_{\varphi\varphi'}(\vare) & =-2

417:  \cos\frac{2\pi(\vare+W_{\varphi\varphi'})}{\hbar\wc}

418: \frac{W_{\varphi\varphi'}} {\tau(\varphi-\varphi')}.

419: \end{split}

420: \ee

421: In the left hand side of \req{kineq:iso} we used the linearized

422: form of the inelastic collision integral \req{Stin}. The

423: inelastic relaxation rate $1/\tauin(\vare)$ of the oscillating

424: component of the distribution function \req{fgen} was

425: calculated in~\cite{DVAMP}:

426: \be

427: \frac{1}{\tauin(\vare)}=\frac{\pi^2T^2+\vare^2}{4\pi E_{\rm F}}

428: \ln\frac{\kappa v_{\rm F}}{{\rm max}\{T,\sqrt{\wc^{3}\tautr}\}},

429: \ee

430: where $E_{\rm F}$ is the Fermi energy and $\kappa=4\pi e^2\nu_0$.

431: Solving \req{kineq:iso} with respect to $I_1$, we obtain

432: \begin{subequations}

433: \label{fresult}

434: \be

435: I_1(\vare)  =   \frac{-2 eE\Rc[d\gamma(\zeta)/d\zeta]}

436: {\tauin^{-1}(\vare)+\tauq^{-1}-\gamma(\zeta)},\ \

437: \gamma(\zeta)=\sum_n\frac{J_n^2(\zeta)}{\tau_n}.

438: \label{fis}

439: \ee

440: Next we substitute $I_1(\vare)$ into \reqs{kineq:j} and

441: \rref{kineq:j2} and find

442: \be

443: \begin{split}

444: A_1(\vare) & = -(2eE\Rc/\wc\tautr)

445: \left[\Gamma_{1}(\zeta,\tauin(\vare))+\Gamma_{2}(\zeta)

446: \right],\\

447: A_2(\vare) & =(2eE\Rc/\wc\tautr) \Gamma_{1}(\zeta,\tauin(\vare))

448: \label{fan}.

449: \end{split}

450: \ee

451: \end{subequations}

452: Here we introduced the following notations:

453: \be

454: \frac{\Gamma_{1}(\zeta,\tauin)}{\tautr} =-\frac{[d\gamma(\zeta)/d\zeta]^2}

455: {\tauin^{-1}+\tauq^{-1}-\gamma(\zeta)},\ \

456: \frac{\Gamma_{2}(\zeta)}{\tautr}  = - \frac{d^2\gamma(\zeta)}{d\zeta^2},

457: \label{G12}

458: \ee

459: and $J_n(\zeta)$ are the Bessel functions.

460: 

461: The isotropic in momentum and oscillatory in energy

462: component of the distribution function $f(\vare,\varphi)$

463: results in the $\Gamma_1(\zeta,\tauin)$ contribution

464: to the amplitudes $A_{1,2}$ of anisotropic part

465: of  $f(\vare,\varphi)$~\cite{DVAMP}. The second contribution,

466: containing $\Gamma_2(\zeta)$, is coming from the second term in

467: the r.h.s. of \req{kineq:j}. This contribution was studied in

468: Refs.~\cite{ryzhii,durst03,VA03} and originates directly from the

469: effects of electric fields on the

470: collision integral for scattering off disorder.

471: 

472: 

473: 

474: Substituting the distribution function \req{fgen} with oscillating components

475: \req{fresult} into \req{jdc} and integrating over energy,

476: we obtain the dissipative current:

477: \be

478: j_{\rm d}=\sigma_{\rm D}%\frac{e^2 \Rc^2\nu_0}{2\tautr}

479: E[1+2\lambda^2 F(\zeta)],\ \

480: F(\zeta)=2\Gamma_{1}(\zeta,\tauint)+

481: \Gamma_{2}(\zeta),% \sigma_{\rm D}=\frac{e^2 \Rc^2\nu_0}{2\tautr}

482: \label{current}

483: \ee

484: where $\sigma_{\rm D}=e^2\Rc^2\nu_0/2\tautr$ is the Drude

485: conductivity at large Hall angle, $\wc\tautr\gg 1$.

486: To perform the integration and

487: simplify further analysis we replaced the inelastic relaxation rate

488: $1/\tauin(\vare)$ by a parameter $1/\tauint$, which can be chosen as

489: $1/\tauint\sim 1/\tauin(\vare=T)$.

490: 

491: Equation~\rref{current} for the dissipative current together with

492: \reqs{fis} and \rref{G12} for functions $\gamma(\zeta)$, $\Gamma_1(\zeta,\tauin)$

493: and $\Gamma_2(\zeta)$ constitute

494: the central result of the paper. Below we discuss the properties

495: of functions $\Gamma_1(\zeta,\tauint)$ and $\Gamma_2(\zeta)$, which

496: determine the non-linear response of the dissipative current $j_{\rm

497: d}$, \req{current}. Then we consider a specific model for

498: $1/\tau_n$ in \req{tau} and analyze the

499: non-linear behavior of the current within the model.

500: 

501: 

502: 

503: For weak electric fields, $\zeta\ll 1$, we obtain the following

504: expressions for functions $\Gamma_{1}(\zeta,\tauint)$ and

505: $\Gamma_{2}(\zeta)$ expanding the Bessel functions to lowest order in

506: $\zeta^2$:

507: \begin{subequations}

508: \label{smallz}

509: \begin{eqnarray}

510: \Gamma_{1}(\zeta,\tauint)

511: & = & -\frac{(\tauint/\tautr)\zeta^2}{1+(\tauint/2\tautr)\zeta^2},

512: \label{smallz:k}

513: \\

514: \Gamma_{2}(\zeta) & = & 1-

515: \frac{3}{8}\tautr\zeta^2\left[\frac{3}{\tau_0}-\frac{4}{\tau_1}+\frac{1}{\tau_2}

516: \right].

517: \label{smallz:d}

518: \end{eqnarray}

519: \end{subequations}

520: The $\zeta^2$ term in the denominator of \req{smallz:k} is

521: legitimate in the limit $\tauin\gg\tautr$, which may take place at

522: sufficiently low electron temperatures. We also note that

523: \req{smallz:k} coincides with the result of Ref.~\cite{DVAMP}

524: in the absence of microwave fields.

525: 

526: In the strong-field limit, $\zeta\gg 1$, we find

527: \be

528: \frac{\Gamma_{1}(\zeta,\tauint)}{\tautr}

529: \propto -\frac{\tau_0}{\tau^2(\pi)}

530: \frac{\cos^2\zeta}{\zeta^2}, \ \

531: \frac{\Gamma_{2}(\zeta)}{\tautr}

532: \propto \frac{1}{\tau(\pi)}\frac{\sin

533: 2\zeta}{\zeta},

534: \label{asymptotes}

535: \ee

536: where $1/\tau(\pi)=\sum_n e^{i\pi n}/\tau_n$ is the

537: back-scattering rate off disorder. The $\Gamma_2$-contribution,

538: arising from the effect of electric field on the collision integral,

539: is larger than the $\Gamma_1$-contribution, which arises from the

540: stationary out-of-equilibrium component of the distribution function.

541: The latter contribution not only decays faster with the increase of

542: $\zeta$ than the former one, but also contains an additional

543: parameter $\tauq/\tau(\pi)$ which is small for smooth disorder. The

544: amplitude of oscillations of current \req{current} decays

545: proportionally to $1/\zeta$ at $\zeta\gg 1$, but the oscillations in

546: differential conductivity

547: $$

548: \sigma=\partial j_{\rm d}/\partial E \propto \partial [\zeta\Gamma_2(\zeta)]/\partial

549: \zeta\propto \cos 2\zeta

550: $$

551: do not vanish; its maxima and minima are situated at $\zeta=\pi k/2$ with integer

552: $k$. At smaller values of $\zeta$, term $\Gamma_1(\zeta)$ also

553: contributes to $\sigma$ and results in dependence of the oscillations amplitude

554: on $\zeta$ as well as in some shift of maxima and minima from $\zeta=\pi

555: k/2$.

556: 

557: To discuss the properties of the non-linear current \req{current} in a

558: broad range of electric fields, we consider a specific model for the

559: harmonics $1/\tau_n$ of the scattering rate off disordered potential

560: due to charged impurities inside or in the proximity of 2DEG:

561: \be

562: \frac{1}{\tau_{n}}=

563: \frac{1}{\taul}\frac{1}{1+\chi n^2}+

564: \frac{\delta_{n,0}}{\taus},\quad \chi\ll 1.

565: \label{taun}

566: \ee

567: Here $1/\taus$ is the scattering rate off impurities inside the 2DEG,

568: which produce sharp ($\delta$-correlated) potential for

569: electrons. Charged impurities in the proximity of

570: 2DEG produce a smooth potential  resulting in

571: electron scattering on small angle $\theta \sim \sqrt{\chi}\ll 1$,

572: where $\chi$ can be estimated as $\chi\sim (\lambda_{\rm F}/\xi)^2$,

573: with $\lambda_{\rm F}$ and $\xi$ being the Fermi wavelength

574: and the correlation length of the disorder potential, respectively.

575: The two restrictions for validity of \req{current}

576: on the strength of magnetic field

577: ($\wc\tautr\gg 1$ and $\wc\tauq\lesssim 1$)

578: can be satisfied simultaneously for $\chi\ll 1$.

579: 

580: 

581: \begin{figure}

582: \epsfxsize=0.35\textwidth

583: %\vspace*{0.3\textwidth}

584: \centerline{\epsfbox{figure2.eps}}

585: \caption{Function $F(\zeta)$ for disorder

586:   described by \req{taun} with $\taul=\taus/30$ for several values of

587:   $\tauint$ and $\chi$: $\tauint=10\taus$, $\chi=0.02$ (solid);

588:   $\tauint=10\taus$, $\chi=0.01$ (dashed); $\tauint=0.5\taus$, $\chi=0.01$

589:   (dotted).  Note that $F(\zeta)$ has a peak at small $\zeta$

590:   determined by the inelastic relaxation time $\tauint$ and oscillates

591:   at large $\zeta$.  At intermediate

592:   $\zeta\sim 1$ the positions of maxima and minima of $F(\zeta)$ depend

593:   on the relation between various parameters of the model.}

594: \label{fig3}

595: \end{figure}

596: 

597: For the disorder characterized by harmonics of the scattering rate

598: \req{taun}, we obtain $ \gamma(\zeta)=J_0^2(\zeta)/\taus+

599: 1/(\taul\sqrt{1+\chi\zeta^2})$. Here we omitted the term, which

600: arises from the back-scattering off smooth disorder and yields

601: exponentially small ($\exp(-\pi/\sqrt{\chi})$) contribution to

602: $1/\tau(\pi)$, cf.  \req{asymptotes}. Substituting this expression for

603: $\gamma(\zeta)$ in~\req{G12}, we can evaluate the

604: current~\req{current} at arbitrary $\zeta$. At $\zeta\ll

605: 1/\sqrt{\chi}$ we have

606: \be \frac{F(\zeta)}{\tautr}

607: =\frac{1}{\tautrl}

608: -\frac{2[2J_0(\zeta)J_1(\zeta)/\taus+\zeta/\tautrl]^2}

609: {\tauint^{-1}+\taus^{-1}\![1-J_0^2(\zeta)]+\tautrl^{-1}\zeta^2/2}-

610: \frac{[J_0^2(\zeta)]''}{\taus},

611: \label{broad}

612: \ee

613: where $1/\tautrl=\chi/\taul$ is the smooth disorder contribution to

614: the transport scattering rate.

615: Equation~\rref{broad} covers both the regime of relatively weak fields,

616: where inelastic scattering is important, and the regime of strong

617: fields, exhibiting prominent oscillations. Function $F(\zeta)$ has a

618: sharp feature at $\zeta\sim \sqrt{\taus/\tauin}$. At these fields, the

619: spectral diffusion of electrons caused by electric field becomes

620: comparable with the inelastic relaxation. At stronger fields, $\zeta\gtrsim

621: \sqrt{\tautrl/\tauin}$, the two kinds of disorder, smooth and

622: ``sharp'', yield two separate contributions to $F(\zeta)$,

623: \be

624: %\begin{split}

625:  \frac{F(\zeta)}{\tautr} =

626:  \frac{1-3\chi\zeta^2-2(1+\sqrt{1+\chi\zeta^2}) }{\tautrl(1+\chi\zeta^2)^{5/2}}

627:  -\frac{[J_0^2(\zeta)]''}{\taus}.

628: %\end{split}

629: \ee

630: Here only the sharp component of disorder

631: contributes to the oscillatory behavior of $F(\zeta)$.

632: 

633: We calculated the dissipative

634: component of electric current in response to the applied dc electric

635: field of arbitrary strength within self-consistent Born approximation

636: \cite{Ando}. We show that the non-linear component of the current

637: consists of two contributions. One contribution arises due to the

638: formation of the out-of-equilibrium component of the distribution

639: function, oscillating as a function of energy. The second contribution

640: is the result of modification by electric field of electron scattering

641: amplitudes off the disorder potential.  We showed that the former

642: contribution is important at relatively weak fields, while the

643: latter one dominates in the high-field domain.  There, the non-linear

644: contribution to the current oscillates as a function of the applied

645: electric field.  The amplitude of oscillations of the differential

646: conductivity does not decrease with the increase of electric field

647: (and at fixed magnetic field).  It may be necessary to take into

648: account the effect of heating on the quantum scattering time and thus

649: on the Dingle factor $\lambda$ in order to explain the suppression of

650: oscillations observed in~\cite{zudov06}.

651: 

652: Finally, we considered the limit $\wc\tauq\lesssim 1$, and therefore

653: assumed that the oscillations in the DOS are described by one harmonic

654: with period in energy $\hbar\wc$, \req{dos}. In stronger fields, the

655: DOS remains periodic in energy with the same period $\hbar\wc$, but

656: contains higher harmonics~\cite{Ando,VA03}. These higher harmonics in

657: the DOS result in a more complicated form of the oscillatory part of

658: the non-linear resistivity.

659: 

660: The authors are thankful to S. Vitkalov

661: and M. Zudov for discussions of experiments.  Part of this work

662: was performed during the visit to the Aspen Center for Physics. The

663: work was supported by  the W. M. Keck Foundation and NSF grants

664: DMR 02-37296, DMR 04-39026 and DMR-0408638.

665: 

666: %%%%%%

667: %%%%%%\bibliographystyle{prsty}

668: %%%%%%\bibliographystyle{prsty}

669: %%%\bibliography{NL2DEG}

670: 

671: 

672: \begin{thebibliography}{10}

673: 

674: \bibitem{Ando}

675: T. Ando, A. Fowler, and F. Stern, Rev. Mod. Phys. {\bf 54},  437  (1982).

676: 

677: \bibitem{prange}

678: R.~E. Prange and S.~M. Girvin, {\em The Quantum Hall Effect}, 2nd ed.

679:   (Springer, ADDRESS, 1990).

680: 

681: \bibitem{du02}

682: C. L. Yang,  J. Zhang, R. R. Du, J. A. Simmons and

683: J. L. Reno, Phys. Rev. Lett. {\bf 89},  076801  (2002).

684: 

685: \bibitem{vitkalov05}

686: A. A. Bykov, J. Zhang, Sergey Vitkalov, A. K. Kalagin

687: and A. K. Bakarov, Phys. Rev. B {\bf 72},  245307  (2005).

688: 

689: \bibitem{vitkalov06}

690: J. Zhang,  S. Vitkalov,  A. A. Bykov, A. K. Kalagin and

691: A. K. Bakarov, cond-mat/0607741.

692: 

693: \bibitem{zudov06}

694: W. Zhang, H. -S. Chiang,  M. A. Zudov,

695: L. N. Pfeiffer and K. W. West, cond-mat/0608727.

696: 

697: \bibitem{DVAMP}

698: I. A. Dmitriev, M. G. Vavilov, I. L. Aleiner, A. D. Mirlin and

699: D. G. Polyakov, Phys. Rev. B {\bf 71},  115316  (2005).

700: 

701: \bibitem{VA03}

702: M.~G. Vavilov and I.~L. Aleiner, Phys. Rev. B {\bf 69},  035303  (2004).

703: 

704: \bibitem{durst03}

705: A.~C. Durst, S. Sachdev, N. Read, and S.~M. Girvin, Phys. Rev. Lett. {\bf 91},

706:   086803  (2003).

707: 

708: \bibitem{ryzhii}

709: V.~I. Ryzhii, R. A. Suris, and B. Shchamkhalova, Sov. Phys. Semicond. {\bf 20},

710:   1299  (1986).

711: 

712: \end{thebibliography}

713: 

714: 

715: 

716: \end{document}

717: