1: \documentclass[aps,prb,twocolumn,epsf,floatfix]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{amsmath}
4: \begin{document}
5: \title{Spin Relaxation in a Two-Dimensional Electron Gas in a Perpendicular
6: Magnetic Field}
7: \author{A.A. Burkov and Leon Balents}
8: \affiliation{\small Department of Physics, University of California,
9: Santa Barbara, CA 93106}
10: \date{\today}
11:
12: \begin{abstract}
13: We consider the problem of spin relaxation in a two-dimensional electron gas
14: (2DEG) in a perpendicular magnetic field.
15: We assume that the spin relaxation is induced by the Rashba spin-orbit (SO)
16: interaction, which appears due to the inversion asymmetry of the confining
17: potential.
18: Our solution is based on a microscopic evaluation of the spin density response
19: function of the 2DEG with impurities and SO interaction.
20: We derive explicit expressions for the transverse and longitudinal
21: spin relaxation rates.
22: Our analysis shows, in particular, that the spin relaxation rates exhibit
23: {\it magnetoquantum oscillations}, which are analogous to the
24: Shubnikov-de Haas oscillations of the electrical resistivity.
25: These oscillations can be observed, for example, in time-resolved optical
26: experiments.
27:
28: \end{abstract}
29:
30: \maketitle
31: \section{Introduction}
32: \label{sec:1}
33: The study of spin relaxation in semiconductor
34: 2DEGs is an important area of the emerging field of {\it spintronics}.\cite{Awschalom02}
35: The dominant spin relaxation mechanism in these
36: systems can typically be associated \cite{Dyakonov86} with the Rashba SO
37: interaction,\cite{Rashba84} which exists due to the inversion asymmetry of
38: the confining potential in semiconductor heterostructure based 2DEG systems.
39: A simple semiclassical picture of the spin relaxation, due to Dyakonov and
40: Perel (DP),\cite{Dyakonov71,Dyakonov86} is that since the Rashba
41: interaction has the form of a momentum-dependent magnetic
42: field $\lambda \hat z
43: \cdot [\boldsymbol{\tau} \times {\bf p}]$, it induces precession of the spin
44: of a moving electron.
45: This precession leads to the randomization of the spin at long times.
46: The DP spin relaxation rate is
47: given, to leading order in $\lambda$, by the simple expression
48: $1/\tau_{sf} \sim (\lambda p_F/\hbar)^2 \tau$,
49: where $p_F$ is the Fermi momentum and $\tau$ is the elastic scattering time.
50: Quasiclassically, $\lambda p_F/\hbar$ is simply the spin-precession frequency,
51: associated with the Rashba field.
52: The distinguishing feature of the DP spin relaxation is that the relaxation
53: rate is smaller in more disordered systems,
54: since impurity scattering disrupts the Rashba spin precession by randomizing
55: the electron's momentum.
56:
57: It is well-known that transport processes in 2DEG systems are strongly
58: influenced by the application of a perpendicular magnetic field.
59: The perpendicular field quantizes electron's energy spectrum, which
60: manifests spectacularly in Shubnikov-de Haas (SdH) oscillations of the
61: resistivity and eventually leads to the quantum Hall effect.
62: In this paper we address the question of how a perpendicular
63: magnetic field influences the Rashba SO interaction-induced spin relaxation
64: in semiconductor 2DEG systems.
65: This problem has been addressed before by several
66: authors, both in the
67: semiclassical \cite{Ivchenko73,Margulis83,Martin02,Puller03} and in the
68: quantum limits,\cite{Frenkel91,Dikman96,Bychkov99,Nazarov00,Khaetskii01,Dikman03} but a complete and rigorous analysis is still lacking.
69:
70: In this paper we will provide such an analysis.
71: We will mainly be interested in the regime of moderate magnetic fields, such
72: that $\omega_c \tau \lesssim 1$ (here $\omega_c$ is the cyclotron frequency),
73: and $\epsilon_F \gg \hbar \omega_c$.
74: Our analysis can, however, be generalized to the quantum Hall limit, when
75: $\epsilon_F \sim \hbar \omega_c$ and $\omega_c \tau$ is large.
76: Our theory is based on a
77: microscopic evaluation of the spin density response function of a disordered
78: 2DEG system with the Rashba SO interaction in a perpendicular magnetic field.
79: We obtain explicit analytical expressions for the longitudinal and
80: transverse spin relaxation rates:
81: \begin{eqnarray}
82: \frac{1}{\tau_z} &=& \frac{8 \lambda^2 m \epsilon_F \tau /\hbar^2}
83: {1 + (\omega_c \tau)^2}, \nonumber \\
84: \frac{1}{\tau_{\perp}} &=& \frac{1}{2\tau_z},
85: \nonumber
86: \end{eqnarray}
87: which reduce to the well-known Dyakonov-Kachorovskii
88: \cite{Dyakonov86} expressions in the zero-field limit.
89: We show, in particular, that the application of a perpendicular magnetic
90: field leads to {\it quantum
91: oscillations} of the spin relaxation rate, which are analogous to the
92: well-known quantum oscillations of transport coefficients.
93: These oscillations arise from the magnetic field dependence of the
94: scattering time $\tau$, which will be calculated below.
95:
96: The paper is organized as follows.
97: In section \ref{sec:2} we calculate the disorder-averaged Green's
98: function of a 2DEG system with the Rashba SO interaction in a perpendicular
99: magnetic field using the self-consistent Born approximation (SCBA).
100: In section \ref{sec:3} the spin-density response function of our system
101: is calculated by summing SCBA self-energy and ladder vertex corrections to the
102: bare polarization diagram.
103: The transverse and longitudinal spin relaxation rates
104: are then extracted from the poles of this response function.
105: \section{Density of states and disorder self-energy}
106: \label{sec:2}
107: The simplest single-particle Hamiltonian describing the dynamics of
108: electrons in a semiconductor 2DEG in a perpendicular magnetic field
109: can be written as:
110: \begin{equation}
111: \label{eq:1}
112: H_0=\frac{\boldsymbol{\pi}^2}{2m} + \lambda \hat z \cdot [\boldsymbol{\tau}
113: \times \boldsymbol{\pi}] - \frac{\Delta_z}{2} \tau^z,
114: \end{equation}
115: where $\boldsymbol{\pi}={\bf p}+(e/c){\bf A}$ is the kinetic momentum,
116: $\boldsymbol{\tau}$ is the spin operator and $\Delta_z=g\mu_B B$ is the Zeeman
117: energy.
118: The second term in Eq.(\ref{eq:1}) describes the Rashba SO interaction.
119: Hamiltonian $H_0$ can be easily diagonalized \cite{Das90,Falko92,Schliemann03}
120: if one notices that the Rashba term mixes only neighboring Landau levels
121: with opposite spin directions.
122: The eigenstates are thus given by:
123: \begin{equation}
124: \label{eq:2}
125: |n,a\rangle = u_{na} |n,\downarrow \rangle + v_{na} |n-1,\uparrow \rangle,
126: \end{equation}
127: where $a=1,2$ and $|n,\sigma \rangle$ is the $n$-th Landau level eigenstate
128: with spin $\sigma=\uparrow,\downarrow$.
129: The corresponding eigen-energies are:
130: \begin{equation}
131: \label{eq:3}
132: \epsilon_{na} = \hbar \omega_c n + (-1)^a \sqrt{\left(\frac{\hbar \omega_c +
133: \Delta_z}{2}\right)^2 + 2 \lambda^2 m \hbar \omega_c n},
134: \end{equation}
135: where $\omega_c = eB/mc$.
136: Note that $n=1,2,...$ for $a=1$, but $n=0,1,...$ for $a=2$.
137: The amplitudes $u_{na}$ and $v_{na}$ are given by:
138: \begin{eqnarray}
139: \label{eq:4}
140: u_{na}&=&i \sqrt{\frac{1}{2} + \frac{(-1)^a(\hbar \omega_c + \Delta_z)/2}
141: {\sqrt{(\hbar \omega_c + \Delta_z)^2/4 + 2 \lambda^2 m \hbar \omega_c n}}},
142: \nonumber \\
143: v_{na}&=&(-1)^{a+1}\sqrt{\frac{1}{2} +
144: \frac{(-1)^{a+1}(\hbar \omega_c + \Delta_z)/2}
145: {\sqrt{(\hbar \omega_c + \Delta_z)^2/4 + 2 \lambda^2 m \hbar \omega_c n}}}.
146: \nonumber \\
147: \end{eqnarray}
148:
149: Using the above basis of single-particle eigenstates, the Hamiltonian
150: of the 2DEG system in the presence of impurity scattering potential can be
151: written as:
152: \begin{equation}
153: \label{eq:5}
154: H=\sum_{nak}\epsilon_{na} c^{\dag}_{nak} c^{\vphantom\dag}_{nak} +
155: \sum_{nak,n'a'k'} V_{na,n'a'}(k,k') c^{\dag}_{nak} c^{\vphantom\dag}_{n'a'k'},
156: \end{equation}
157: where index $k$ denotes orbit-center quantum numbers in the Landau gauge and
158: \begin{equation}
159: \label{eq:6}
160: V_{na,n'a'}(k,k')=\sum_{\sigma} \int d{\bf r} \Psi^*_{nak} ({\bf r}\sigma)
161: V({\bf r}) \Psi^{\vphantom *}_{n'a'k'}({\bf r} \sigma),
162: \end{equation}
163: are the matrix elements of the impurity potential $V({\bf r})$ in the basis
164: (\ref{eq:2}).
165: The wavefunctions $\Psi_{nak}({\bf r}\sigma)$ are given by:
166: \begin{eqnarray}
167: \label{eq:7}
168: \Psi_{nak}({\bf r} \uparrow)&=&v_{na} \phi_{n-1,k}({\bf r}), \nonumber \\
169: \Psi_{nak}({\bf r} \downarrow)&=&u_{na} \phi_{n,k}({\bf r}),
170: \end{eqnarray}
171: where $\phi_{n,k}({\bf r})$ are the Landau-level eigenfunctions in the Landau
172: gauge.
173:
174: Our analysis of the spin relaxation in a system, described by Eq.(\ref{eq:5}),
175: is based on a calculation of the spin density response function, which
176: is similar to Ando's calculation of the conductivity of a 2DEG in a
177: perpendicular field.\cite{Ando}
178: Our calculation is technically more complicated than Ando's due to the
179: presence of the SO interactions and also due to the fact that vertex
180: corrections to the polarization bubble,
181: which vanish in the case of the conductivity calculation, are crucial
182: in our case.
183:
184: Following Ando, we use the self-consistent Born approximation (SCBA)
185: to find the disorder-averaged Green's function.
186: This is a very good approximation for the high Landau level filling factors
187: that we are assuming.
188: SCBA equation for the retarded disorder self-energy in our case reads:
189: \begin{equation}
190: \label{eq:8}
191: \Sigma_{na}^R(\epsilon) = \sum_{n'a'} \Gamma_{na,n'a'}
192: G^R_{n'a'}(\epsilon),
193: \end{equation}
194: where
195: \begin{equation}
196: \label{eq:9}
197: \Gamma_{na,n'a'} = \sum_{k'} \langle V_{na,n'a'}(k,k') V_{n'a',na}(k',k)
198: \rangle,
199: \end{equation}
200: and
201: \begin{equation}
202: \label{eq:10}
203: G^R_{na}(\epsilon) = \frac{1}{\epsilon-\epsilon_{na}-
204: \Sigma_{na}^R(\epsilon)},
205: \end{equation}
206: is the retarded SCBA Green's function.
207: Angular brackets in Eq.(\ref{eq:9}) denote disorder average.
208:
209: Matrix elements of $\Gamma$ can be easily evaluated for $\delta$-function
210: impurity potential and are given by:
211: \begin{equation}
212: \label{eq:11}
213: \Gamma_{na,n'a'}=\frac{\hbar^2 \omega_c}{2\pi\tau_0}\left(|u_{na}|^2
214: |u_{n'a'}|^2 + |v_{na}|^2 |v_{n'a'}|^2\right),
215: \end{equation}
216: where $\tau_0$ is the elastic impurity scattering time in the absence of
217: the magnetic field.
218:
219: Eq.(\ref{eq:8}) is too complicated to be solved analytically without
220: approximations.
221: Fortunately, only very minor simplification is needed to make this
222: equation solvable.
223: The main complication is the dependence of the matrix $\Gamma$ on the
224: Landau level index $n$.
225: This dependence is, however, inessential, since it appears entirely
226: due to the term $2\lambda^2 m\hbar \omega_c n$ in Eq.(\ref{eq:4}).
227: The dependence on $n$ in this term can be safely ignored, since
228: the Rashba SO interaction can be assumed to be a weak perturbation, in the
229: sense that $\lambda p_F \ll \epsilon_F$.
230: In this case we can make a replacement
231: $n \rightarrow \epsilon_F/\hbar \omega_c$
232: in the Rashba term in Eqs.(\ref{eq:3}) and (\ref{eq:4}), which makes the
233: amplitudes $u_{na}$ and $v_{na}$ and, consequently, $\Gamma$, independent of
234: $n$.
235: This simple approximation allows us to solve Eq.(\ref{eq:8})
236: analytically.
237:
238: It is convenient to introduce the following new notation.
239: Let
240: \begin{equation}
241: \label{eq:12}
242: \Delta = 2 \sqrt{\left(\frac{\hbar \omega_c +\Delta_z}{2} \right)^2 +
243: 2\lambda^2 m \epsilon_F}.
244: \end{equation}
245: Then $\epsilon_{na} = \hbar \omega_c n + (-1)^a \Delta/2$.
246: Also define:
247: \begin{eqnarray}
248: \label{eq:13}
249: &&u_1 = i \cos(\vartheta), \,\, v_1 = \sin(\vartheta), \nonumber \\
250: &&u_2 = i \sin(\vartheta), \,\, v_2 = - \cos(\vartheta),
251: \end{eqnarray}
252: where
253: \begin{equation}
254: \label{eq:14}
255: \cos(\vartheta) = \sqrt{\frac{1}{2} - \frac{\hbar \omega_c + \Delta_z}
256: {2\Delta}}.
257: \end{equation}
258: Using this notation, the matrix elements of $\Gamma$ are given by:
259: \begin{eqnarray}
260: \label{eq:15}
261: \Gamma_{11} &=& \Gamma_{22} = \frac{\hbar^2 \omega_c}{2 \pi \tau_0}
262: \left(1 - \frac{1}{2} \sin^2(2\vartheta) \right), \nonumber \\
263: \Gamma_{12} &=& \Gamma_{21} = \frac{\hbar^2 \omega_c}{\pi \tau_0}
264: \sin^2(2\vartheta).
265: \end{eqnarray}
266: Let us also make the usual assumption that the real part of the disorder
267: self-energy can be absorbed into the chemical potential.
268: Then SCBA equation can be written as:
269: \begin{equation}
270: \label{eq:16}
271: \textnormal{Im} \Sigma^R_a(\epsilon) = \sum_{n,a'} \Gamma_{aa'}
272: \textnormal{Im} G^R_{na'} (\epsilon) =
273: -2 \pi^2 \ell^2 \sum_{a'} \Gamma_{aa'} \varrho_{a'}(\epsilon),
274: \end{equation}
275: where
276: \begin{equation}
277: \label{eq:17}
278: \varrho_a(\epsilon) = - \frac{1}{2\pi^2 \ell^2} \sum_n \textnormal{Im}
279: G^R_{na}(\epsilon),
280: \end{equation}
281: is the density of states and $\ell=\sqrt{\hbar c/eB}$ is the magnetic length.
282: Writing out Eq.(\ref{eq:17}) explicitly, we have:
283: \begin{eqnarray}
284: \label{eq:18}
285: &&\varrho_a(\epsilon) = - \frac{1}{2\pi^2 \ell^2} \nonumber \\
286: &\times& \sum_{n=0}^{\infty}
287: \frac{\textnormal{Im} \Sigma^R_a(\epsilon)}
288: {\left[\epsilon - \hbar \omega_c n - (-1)^a \Delta - \hbar \omega_c
289: \delta_{a,1}\right]^2 + \left[\textnormal{Im} \Sigma^R_{a}(\epsilon)\right]^2}.
290: \nonumber \\
291: \end{eqnarray}
292: The sum over Landau level indices in the above equation can be done
293: using Poisson summation formula: \cite{Golub01}
294: \begin{equation}
295: \label{eq:19}
296: \sum_{n=0}^{\infty} f(n) = \frac{f(0)}{2} + \sum_{p=-\infty}^{\infty}
297: \int_0^{\infty} dn f(n) e^{2\pi i p n}.
298: \end{equation}
299: Up to this point, our calculations were applicable to magnetic fields of
300: general strength.
301: To proceed further, we will restrict ourselves to the regime when
302: $\epsilon_F \tau /\hbar \gg 1$ and $\epsilon_F \gg \hbar \omega_c$.
303: (The calculation we present can be done in other regimes as well, but the
304: actual procedure will be a little different.)
305: This allows us to neglect the term $f(0)/2$ in Eq.(\ref{eq:19}).
306: Then, changing the integration variable to $x=\hbar \omega_c n - \epsilon
307: +(-1)^a \Delta-\hbar \omega_c \delta_{a,1}$ and extending the lower limit
308: of integration to $-\infty$, which is justified by the above assumptions, we
309: obtain:
310: \begin{eqnarray}
311: \label{eq:20}
312: &&\varrho_a(\epsilon) = -\frac{1}{2\pi^2\ell^2}
313: \sum_{p=-\infty}^{\infty}
314: \int_{-\infty}^{\infty}\frac{dx}{\hbar \omega_c} \nonumber \\
315: &\times& \exp\left(2\pi i p
316: \frac{x+\epsilon-(-1)^a\Delta-\hbar\omega_c\delta_{a,1}}{\hbar\omega_c}\right)
317: \frac{\textnormal{Im} \Sigma^R_a(\epsilon)}{x^2 +
318: \left[\Sigma^R_a(\epsilon)\right]^2}. \nonumber \\
319: \end{eqnarray}
320: Calculating the above integral, we obtain:
321: \begin{eqnarray}
322: \label{eq:21}
323: \varrho_a(\epsilon) &=& \frac{m}{2\pi \hbar^2}\left[1 +
324: 2 \sum_{p=1}^{\infty}e^{-\pi p/\omega_c \tau_0} \right. \nonumber \\
325: &\times&\left.\cos\left(2\pi p \frac{\epsilon - (-1)^a\Delta -
326: \hbar\omega_c\delta_{a,1}}
327: {\hbar \omega_c}\right) \right],\nonumber \\
328: \end{eqnarray}
329: where we have replaced $\textnormal{Im} \Sigma^R_a(\epsilon)$, which appears
330: in the exponent after integration, by its zero-field value $-\hbar/2\tau_0$.
331: The oscillatory term in the above expression is slightly different from
332: the usual expression, which is periodic in $\hbar \omega_c$, due to the
333: presence of the Rashba and Zeeman splittings.
334: In what follows we will ignore this modification of the density of states,
335: putting $\lambda=0$ and $\Delta_z=0$ in the oscillatory term.
336: This does not mean, of course, that we are ignoring SO interactions
337: altogether, since they still enter in the matrix elements of $\Gamma$.
338: Then the density of states becomes independent of the index $a$ and
339: is given by the following simple expression:
340: \begin{equation}
341: \label{eq:22}
342: \varrho(\epsilon)=\varrho_0\left[1 + 2 \sum_{p=1}^{\infty} e^{-\pi p/\omega_c
343: \tau_0} \cos\left(\frac{2\pi p \epsilon}{\hbar \omega_c} - \pi p
344: \right)\right],
345: \end{equation}
346: where $\varrho_1(\epsilon)=\varrho_2(\epsilon)=\varrho(\epsilon)/2$ and
347: $\varrho_0=m/\hbar^2 \pi$ is the zero-field density of states.
348: If the magnetic field is not too large, i.e. if $\omega_c\tau \lesssim 1$,
349: only the first term in the sum over $p$ in Eq.(\ref{eq:22}) needs to be
350: retained.
351: Then the oscillatory term in the density of states is purely sinusoidal.
352: Substituting Eq.(\ref{eq:22}) in Eq.(\ref{eq:16}) we find that SCBA
353: self-energy is also independent of the index $a$ and is given by:
354: \begin{equation}
355: \label{eq:23}
356: \textnormal{Im}\Sigma^R(\epsilon) \equiv - \frac{\hbar}{2\tau(\epsilon)} =
357: - \frac{\pi \hbar^3}{2 m \tau_0} \varrho(\epsilon).
358: \end{equation}
359: The above disorder self-energy reduces to the usual expression $-\hbar/2\tau_0$
360: in the zero-field limit. At finite fields, the oscillatory term in the
361: density of states in Eq.({\ref{eq:23}) is at the origin of a number of
362: magnetoquantum oscillation phenomena, including the SdH oscillations of the
363: electrical resistivity.
364: As will be shown later, it also leads to similar magnetoquantum oscillation
365: effects in the spin relaxation rate.
366:
367: \section{Spin density response function}
368: \label{sec:3}
369: We can now calculate spin density response function of a disordered 2DEG system
370: with the Rashba SO interactions in a perpendicular magnetic field.
371: Transverse and longitudinal spin relaxation rates can be found from the
372: poles of the corresponding spin density response functions.
373: The calculation is similar to the analogous calculation that can be done at
374: zero field. \cite{Burkov03}
375: We will use $\hbar=1$ units in the rest of this section, except for the final
376: results.
377:
378: We start from the general density matrix response function:
379: \begin{eqnarray}
380: \label{eq:24}
381: &&\chi_{\sigma_1\sigma_2,\sigma_3\sigma_4}({\bf r}-{\bf r}',t-t') =
382: \nonumber \\
383: &-&i\theta(t-t') \langle \left[\hat \varrho^{\dag}_{\sigma_1
384: \sigma_2}({\bf r},t),\hat \varrho^{\vphantom \dag}_{\sigma_3\sigma_4}({\bf r}'
385: ,t')\right]
386: \rangle,
387: \end{eqnarray}
388: where $\hat \varrho_{\sigma_1\sigma_2}({\bf r},t)=\Psi^{\dag}_{\sigma_2}
389: ({\bf r},t) \Psi_{\sigma_1}({\bf r},t)$ is the generalized density operator,
390: whose expectation value is the density matrix.
391: As usual, the calculation is most conveniently done by analytical
392: continuation of the imaginary time response function. \cite{Senna82}
393: Fourier transformed imaginary time response function can be written as:
394: \begin{equation}
395: \label{eq:25}
396: \chi_{\sigma_1\sigma_2,\sigma_3\sigma_4}({\bf q},i\Omega) =
397: \frac{1}{\beta} \sum_{i\omega} P_{\sigma_1\sigma_2,\sigma_3\sigma_4}
398: ({\bf q},i\omega,i\omega+i\Omega),
399: \end{equation}
400: where
401: \begin{eqnarray}
402: \label{eq:26}
403: &&P_{\sigma_1\sigma_2,\sigma_3\sigma_4}({\bf r}-{\bf r}',i\omega,i\omega+
404: i\Omega) = \nonumber \\
405: &&\langle {\cal G}_{\sigma_3\sigma_1}({\bf r}',{\bf r},i\omega)
406: {\cal G}_{\sigma_2\sigma_4}({\bf r},{\bf r}',i\omega+i\Omega)\rangle
407: \end{eqnarray}
408: is the polarization bubble diagram
409: and ${\cal G}$ are imaginary time Green's functions.
410: $P$ can be calculated by summing all the SCBA diagrams
411: for the disorder self-energy and all the ladder vertex corrections
412: to the polarization bubble, which constitutes a conserving approximation
413: for the density matrix response function.
414: The result can be written in matrix notation as $P=P^0 D$, where
415: $P^0$ is the ``bare'' polarization bubble with only the self-energy corrections
416: included:
417: \begin{eqnarray}
418: \label{eq:27}
419: &&P^0_{\sigma_1\sigma_2,\sigma_3\sigma_4}({\bf r}-{\bf r}',i\omega,i\omega+
420: i\Omega) = \nonumber \\
421: &&{\cal G}_{\sigma_3\sigma_1}({\bf r}',{\bf r},i\omega)
422: {\cal G}_{\sigma_2\sigma_4}({\bf r},{\bf r}',i\omega+i\Omega),
423: \end{eqnarray}
424: with ${\cal G}$ here being the disorder-averaged Green's function.
425: The vertex part, or, as it is often called, diffusion propagator, $D$
426: satisfies the following matrix equation:
427: \begin{equation}
428: \label{eq:28}
429: D = {\bf 1} + \gamma P^0 D,
430: \end{equation}
431: where $\gamma \equiv \langle V^2({\bf r})\rangle$.
432: The solution of this equation is
433: \begin{equation}
434: \label{eq:29}
435: D = \left[{\bf 1} - \gamma P^0\right]^{-1}.
436: \end{equation}
437:
438: \begin{figure}[t]
439: \includegraphics[width=8cm]{qsr_fig1.eps}
440: \caption{(a) SCBA equation for the disorder-averaged Green's function;
441: (b) Equation for the vertex function $D$. Dashed lines denote impurity
442: potential correlator $\langle V({\bf r}) V({\bf r}')\rangle$.}
443: \label{diagram}
444: \end{figure}
445:
446: We can now calculate the retarded real time response function by analytically
447: continuing Eq.(\ref{eq:25}) to real frequencies.
448: The result can be written as an integral along the branch cut of the
449: SCBA Green's function:
450: \begin{eqnarray}
451: \label{eq:30}
452: &&\chi({\bf q},\Omega) = \int_{-\infty}^{\infty} \frac{d \epsilon}{2 \pi i}
453: n_F(\epsilon) \left[P({\bf q},\epsilon+i\eta,\epsilon+\Omega+i\eta)\right.
454: \nonumber \\
455: &-&\left.P({\bf q},\epsilon-i\eta,\epsilon+\Omega+i\eta)
456: +P({\bf q},\epsilon-\Omega-i\eta,\epsilon+i\eta)\right. \nonumber \\
457: &-&\left.P({\bf q},\epsilon-\Omega-i\eta,\epsilon-i\eta)\right].
458: \end{eqnarray}
459: At low frequencies and low temperatures Eq.(\ref{eq:30}) can be simplified to:
460: \begin{equation}
461: \label{eq:31}
462: \chi({\bf q},\Omega) = \frac{i\Omega}{2\pi}P({\bf q},\epsilon_F-i\eta,
463: \epsilon_F+\Omega+i\eta)+\frac{\varrho(\epsilon_F)}{2}.
464: \end{equation}
465: Thus, the problem reduces to calculating the matrix
466: $P^0({\bf q},\epsilon_F-i\eta,\epsilon_F+\Omega+i\eta)$.
467: We will in fact be interested only in the relaxation of uniform spin
468: polarization and thus will put ${\bf q} = 0$ henceforth.
469: The calculation of $P^0$ in our case closely resembles
470: the analogous calculation in the zero-field problem. \cite{Burkov03}
471:
472: We start from the definition of $P^0$ Eq.(\ref{eq:27}), using
473: the disorder-averaged Green's functions calculated in section \ref{sec:2}:
474: \begin{eqnarray}
475: \label{eq:32}
476: &&P^0_{\sigma_1\sigma_2,\sigma_3\sigma_4}(\epsilon_F-i\eta,
477: \epsilon_F+\Omega++i\eta)
478: \nonumber \\
479: &=&\frac{1}{L^2} \int d{\bf r} d {\bf r}' G^A_{\sigma_3\sigma_1}({\bf r}',
480: {\bf r},\epsilon_F) G^R_{\sigma_2\sigma_4}({\bf r},{\bf r}',\epsilon_F+\Omega)
481: \nonumber \\
482: &=&\frac{1}{L^2} \int d{\bf r} d{\bf r}' \sum_{nak,n'a'k'}
483: \frac{\Psi^{\vphantom *}_{nak}({\bf r}',\sigma_3)
484: \Psi^*_{nak}({\bf r},\sigma_1)}
485: {\epsilon_F -\epsilon_{na} - i/2\tau} \nonumber \\
486: &\times&\frac{\Psi^{\vphantom *}_{n'a'k'}({\bf r},\sigma_2)
487: \Psi^*_{n'a'k'}({\bf r}',
488: \sigma_4)}{\epsilon_F +\Omega-\epsilon_{n'a'} + i/2\tau},
489: \end{eqnarray}
490: where $\tau \equiv \tau(\epsilon_F)$ and the wavefunctions are given by
491: Eq.(\ref{eq:7}).
492: The spatial integrals in the above equation can be easily evaluated
493: using the orthonormality of the Landau gauge eigenfuctions and
494: are given by:
495: \begin{eqnarray}
496: \label{eq:33}
497: \int d{\bf r} \Psi^*_{nak}({\bf r},\uparrow)
498: \Psi^{\vphantom *}_{n'a'k'}({\bf r},\uparrow)
499: &=& v^*_a v^{\vphantom *}_{a'} \delta_{n,n'} \delta_{k,k'},\nonumber \\
500: \int d{\bf r} \Psi^*_{nak}({\bf r},\downarrow)
501: \Psi^{\vphantom *}_{n'a'k'}({\bf r},
502: \downarrow)&=& u^*_a u^{\vphantom *}_{a'} \delta_{n,n'} \delta_{k,k'},
503: \nonumber \\
504: \int d{\bf r} \Psi^*_{nak}({\bf r},\uparrow)
505: \Psi^{\vphantom *}_{n'a'k'}({\bf r},
506: \downarrow)&=& v^*_a u^{\vphantom *}_{a'} \delta_{n,n'+1}
507: \delta_{k,k'},\nonumber \\
508: \int d{\bf r} \Psi^*_{nak}({\bf r},\downarrow)
509: \Psi^{\vphantom *}_{n'a'k'}({\bf r},
510: \uparrow)&=& u^*_a v^{\vphantom *}_{a'} \delta_{n,n'-1} \delta_{k,k'}.
511: \end{eqnarray}
512: The sums over Landau level indices in Eq.(\ref{eq:32}) can be done as follows.
513: We first decouple the product of two Green's functions appearing in
514: (\ref{eq:32}) as:
515: \begin{eqnarray}
516: \label{eq:34}
517: &&\frac{1}{(\epsilon_F-\epsilon_{na}-i/2\tau)(\epsilon_F+\Omega-
518: \epsilon_{n'a'}+i/2\tau)} \nonumber \\
519: &=&\frac{1}{\Omega + (-1)^a\Delta - (-1)^{a'}\Delta + i/\tau} \nonumber \\
520: &\times&\left(
521: \frac{1}{\epsilon_F-\epsilon_{na}-i/2\tau} - \frac{1}{\epsilon_F+\Omega-
522: \epsilon_{n'a'}+i/2\tau} \right). \nonumber \\
523: \end{eqnarray}
524: The real parts of the decoupled sums over Landau level indices can then be
525: neglected by our assumption that
526: $\epsilon_F \gg \hbar \omega_c, \lambda p_F, \hbar \Omega$.
527: The imaginary parts are simply proportional to the density of
528: states at Fermi energy $\varrho(\epsilon_F)$.
529: Thus we obtain the following expressions for the matrix elements
530: of $P^0$:
531: \begin{eqnarray}
532: \label{eq:35}
533: \gamma P^0_{\uparrow\uparrow,\uparrow\uparrow} &=& \gamma
534: P^0_{\downarrow\downarrow,\downarrow\downarrow} = \left(1-\frac{1}{2}
535: \sin^2(2\vartheta)\right)f_0(\Omega) \nonumber \\
536: &+&\frac{\sin^2(2 \vartheta)}{4}
537: \left[f_+(\Omega) + f_-(\Omega)\right],\nonumber \\
538: \gamma P^0_{\uparrow\uparrow,\downarrow\downarrow} &=&
539: \gamma P^0_{\downarrow\downarrow,\uparrow\uparrow} =
540: \frac{\sin^2(2\vartheta)}{4} \left[2 f_0(\Omega) - f_+(\Omega) -
541: f_-(\Omega)\right], \nonumber \\
542: \gamma P^0_{\uparrow\downarrow,\uparrow\downarrow} &=&
543: \frac{\sin^2(2\vartheta)}{2} f_0(\Omega+\omega_c) + \sin^4(\vartheta)
544: f_+(\Omega+\omega_c) \nonumber \\
545: &+& \cos^4(\vartheta) f_-(\Omega+\omega_c),\nonumber \\
546: \gamma P^0_{\downarrow\uparrow,\downarrow\uparrow} &=&
547: \frac{\sin^2(2\vartheta)}{2} f_0(\Omega-\omega_c) + \sin^4(\vartheta)
548: f_-(\Omega-\omega_c) \nonumber \\
549: &+& \cos^4(\vartheta) f_+(\Omega-\omega_c),
550: \end{eqnarray}
551: where functions $f_0$ and $f_{\pm}$ are given by:
552: \begin{eqnarray}
553: \label{eq:36}
554: f_0(\Omega) &=& \frac{1}{1-i\Omega \tau}, \nonumber \\
555: f_{\pm}(\Omega) &=& \frac{1}{1 - i\Omega \tau \pm i\Delta \tau}.
556: \end{eqnarray}
557: All other matrix elements of $P^0$ vanish.
558:
559: To calculate the transverse and longitudinal spin relaxation rates we need to
560: find the poles of the
561: spin density response function, or, equivalently, zeroth of the inverse
562: diffusion propagator $D^{-1}$.
563: It is most easily done by transforming $D^{-1}$ to the physical space
564: of charge and spin density components, which is accomplished by multiplying
565: it on both sides with Pauli matrices:
566: \begin{equation}
567: \label{eq:37}
568: D^{-1}_{\alpha\beta} = \frac{1}{2} \tau^{\alpha}_{\sigma_1\sigma_2}
569: D^{-1}_{\sigma_1\sigma_2,\sigma_3\sigma_4} \tau^{\beta}_{\sigma_4\sigma_3},
570: \end{equation}
571: where $\alpha,\beta=c,x,y,z$ and $\tau^c$ is the identity matrix.
572: Thus we obtain:
573: \begin{eqnarray}
574: \label{eq:38}
575: D^{-1}_{cc}(\Omega)&=&1 - f_0(\Omega), \nonumber \\
576: D^{-1}_{zz}(\Omega)&=&1 - f_0(\Omega) + \frac{\sin^2(2\vartheta)}{2}
577: \left[2 f_0(\Omega)\right. \nonumber \\
578: &-&\left.f_+(\Omega) - f_-(\Omega)\right], \nonumber \\
579: D^{-1}_{+-}(\Omega)&=&1 - \frac{\sin^2(2\vartheta)}{2} f_0(\Omega+\omega_c)-
580: \sin^4(\vartheta) f_+(\Omega+\omega_c) \nonumber \\
581: &-&\cos^4(\vartheta) f_-(\Omega+\omega_c).
582: \end{eqnarray}
583: Clearly $D^{-1}_{cc}(0) = 0$, which is a consequence of charge conservation.
584: For the spin part of the inverse diffusion propagator, to leading order in
585: $\lambda p_F \tau /\hbar$ and $\Omega \tau$, and assuming that
586: $\Delta_z \ll \hbar \omega_c$, we obtain:
587: \begin{eqnarray}
588: \label{eq:39}
589: \tau^{-1} D^{-1}_{zz}(\Omega) &=& -i\Omega + \frac{1}{\tau_z}, \nonumber \\
590: \tau^{-1} D^{-1}_{+-}(\Omega) &=& -i\Omega + i\hbar^{-1} \Delta_z +
591: \frac{1}{\tau_{\perp}},
592: \end{eqnarray}
593: where the longitudinal and transverse spin relaxation rates are given by:
594: \begin{eqnarray}
595: \label{eq:40}
596: \frac{1}{\tau_z} &=& \frac{8 \lambda^2 m \epsilon_F \tau /\hbar^2}
597: {1 + (\omega_c \tau)^2}, \nonumber \\
598: \frac{1}{\tau_{\perp}} &=& \frac{1}{2\tau_z}.
599: \end{eqnarray}
600: Eq.(\ref{eq:40}) is our main result.
601: The form of these expressions is rather similar to the well-known
602: Dyakonov-Kachorovskii expressions \cite{Dyakonov86} for the spin relaxation
603: rates in the zero field case, except for the factor $1/[1+(\omega_c \tau)^2]$,
604: which describes the suppression of spin relaxation by the perpendicular
605: field.
606: The mechanism of this suppression can be understood in quasiclassical
607: terms by imagining electrons moving along classical cyclotron orbits
608: and undergoing impurity scattering and the Rashba spin precession.
609: Orbital motion affects the spin precession, since the direction of
610: the Rashba field is always transverse to the direction of electron's velocity,
611: and thus changes as the electron moves along a circular cyclotron orbit.
612: However, if $\omega_c \tau \ll 1$, impurity scattering randomizes the
613: direction of electron's motion and the influence of the magnetic field
614: on the orbital motion is then negligible.
615: On the other hand, when $\omega_c\tau \gg 1$, electrons can move around
616: cyclotron orbits almost freely and thus the Rashba spin precession is
617: averaged out.
618: This picture of the suppression of the DP spin relaxation by a perpendicular
619: magnetic field has been discussed before by several authors. \cite{Ivchenko73,Margulis83,Martin02}
620: However, we are not aware of a rigorous derivation of a simple analytical
621: expression for the spin relaxation rate, such as Eq.(\ref{eq:40}).
622:
623: Probably the most noteworthy feature of our result is the oscillatory
624: dependence on the magnetic field, which appears due to the oscillatory
625: term in the scattering rate $1/\tau$ in Eqs.(\ref{eq:22}) and (\ref{eq:23}).
626: If $\omega_c \tau < 1$, we may neglect the $(\omega_c\tau)^2$ term in the
627: denominator in Eq.(\ref{eq:40}) and leave only the first term in the sum over
628: $p$ in Eq.(\ref{eq:22}).
629: In this case spin relaxation rates will exhibit purely sinusoidal SdH-like
630: oscillations.
631: Similar effects have been discussed earlier in the context
632: of nuclear spin-lattice relaxation in quantum Hall systems.\cite{Vagner88,Klitzing90,MacDonald91}
633: However, the possibility of such magnetoquantum oscillations in the SO-induced
634: spin relaxation rates has never been discussed before.
635: Such oscillations have in fact been observed in recent experiments
636: on spin relaxation in InGaAs quantum wells.\cite{Awschalom03}
637:
638: In conclusion, we have considered the problem of spin relaxation in
639: a 2DEG with the Rashba SO interactions in a perpendicular magnetic field.
640: We have given a rigorous microscopic derivation of the transverse and
641: longitudinal spin relaxation rates from the poles of the spin density
642: response function in the limit $\epsilon_F \gg \hbar \omega_c$.
643: It is possible to extend our calculation to other regimes,
644: including the quantum Hall regime $\epsilon_F \sim \hbar \omega_c$.
645: Our results demonstrate that magnetoquantum oscillation effects should
646: be observed in spin relaxation phenomena in 2DEGs.
647: One of the advantages of our approach is that it can be easily extended to
648: study another very interesting and still unexplored question of what is the
649: effect of Coulomb interactions on the spin relaxation in semiconductor 2DEGs.
650: This question will be addressed in a future publication.
651:
652: We would like to acknowledge helpful discussions with David Awschalom,
653: Yogesh Joglekar, Allan MacDonald and Vanessa Sih.
654: This work was supported by DARPA/ONR N00014-99-1-1096, by the NSF under
655: grant DMR-9985255 and by the Sloan and Packard foundations.
656:
657:
658: \begin{thebibliography}{99}
659: \bibitem{Awschalom02} {\it Semiconductor Spintronics and Quantum Computation},
660: edited by D.D. Awschalom, D. Loss and N. Samarth, (Springer Verlag, 2002).
661: \bibitem{Dyakonov86} M.I. Dyakonov and V.Yu. Kachorovskii, Sov. Phys. Semicond.
662: {\bf 20}, 110 (1986).
663: \bibitem{Rashba84} Yu.A. Bychkov and E.I. Rashba,
664: J. Phys. C {\bf 17}, 6093 (1984).
665: \bibitem{Dyakonov71} M.I. Dyakonov and V.I. Perel, Sov. Phys. JETP {\bf 33},
666: 1053 (1971); Sov. Phys. Solid State {\bf 13}, 3023 (1972).
667: \bibitem{Ivchenko73} E.L. Ivchenko, Fiz. Tverd. Tela {\bf 15},
668: 1566 (1973) [Sov. Phys. Solid State {\bf 15}, 1048 (1973)].
669: \bibitem{Margulis83} A.D. Margulis and V.A. Margulis, Fiz. Tverd. Tela
670: {\bf 25}, 1590 (1983) [Sov. Phys. Solid State {\bf 25}, 918
671: (1983)].
672: \bibitem{Martin02} F.X. Bronold, I. Martin, A. Saxena and D.L. Smith,
673: Phys. Rev. B {\bf 66}, 233206 (2002).
674: \bibitem{Puller03} V.I. Puller, L.G. Mourokh, N.J.M. Horing and A.Yu. Smirnov,
675: Phys. Rev. B {\bf 67}, 155309 (2003).
676: \bibitem{Frenkel91} D.M. Frenkel, Phys. Rev. B {\bf 43}, 14228 (1991).
677: \bibitem{Dikman96} S.M. Dikman and S.V. Iordanskii, Sov. Phys. JETP {\bf 83},
678: 128 (1996).
679: \bibitem{Bychkov99} W. Apel and Yu.A. Bychkov, Phys. Rev. Lett. {\bf 82},
680: 3324 (1999).
681: \bibitem{Nazarov00} A.V. Khaetskii and Yu.V. Nazarov, Phys. Rev. B {\bf 61},
682: 12639, (2000).
683: \bibitem{Khaetskii01} A.V. Khaetskii, Physica E {\bf 10}, 27 (2001).
684: \bibitem{Dikman03} S. Dickmann, cond-mat/0304515 (unpublished).
685: \bibitem{Das90} B. Das, S. Datta and R. Reifenberger, Phys. Rev. B {\bf 41},
686: 8278 (1990).
687: \bibitem{Falko92} V.I. Falko, Phys. Rev. B {\bf 46}, 4320 (1992);
688: Phys. Rev. Lett. {\bf 71}, 141 (1993).
689: \bibitem{Schliemann03} J. Schliemann, J.C. Egues and D. Loss, Phys. Rev. B
690: {\bf 67}, 085302 (2003).
691: \bibitem{Ando} T. Ando and Y. Uemura, J. Phys. Soc. Japan {\bf 36}, 959
692: (1974); T. Ando, {\it ibid.} {\bf 37}, 1233 (1974).
693: \bibitem{Golub01} A. Isihara and L. Smr$\check{c}$ka, J. Phys. C: Solid
694: State Phys. {\bf 19}, 6777 (1986);
695: N.S. Averkiev, L.E. Golub, S.A. Tarasenko and M. Willander,
696: J. Phys. Condens. Matter {\bf 13}, 2517 (2001); S.A. Tarasenko,
697: Fiz. Tverd. Tela {\bf 44}, 1690 (2002) [Phys. Solid State {\bf 44}, 1769
698: (2002)].
699: \bibitem{Burkov03} A.A. Burkov and A.H. MacDonald, cond-mat/0311328
700: (unpublished).
701: \bibitem{Senna82} A. Houghton, J.R. Senna and S.C. Ying, Phys. Rev. B
702: {\bf 25}, 2196 (1982); S.M. Girvin, M. Johnson and P.A. Lee,
703: {\it ibid.} {\bf 26}, 1651 (1982).
704: \bibitem{Vagner88} I.D. Vagner and T. Maniv, Phys. Rev. Lett. {\bf 61}, 1400
705: (1988).
706: \bibitem{Klitzing90} A. Berg, M. Dobers, R.R. Gerhardts and K.v. Klitzing,
707: Phys. Rev. Lett. {\bf 64}, 2563 (1990).
708: \bibitem{MacDonald91} D. Antoniou and A.H. MacDonald, Phys. Rev. B {\bf 43},
709: 11686 (1991).
710: \bibitem{Awschalom03} V. Sih and D.D. Awschalom, to be published.
711:
712: \end{thebibliography}
713:
714: \end{document}
715:
716:
717:
718:
719:
720:
721:
722:
723:
724:
725:
726: