cond-mat0112466/T1.tex
1: 
2: 
3: % ----------------------------------------------------------------------------
4: % Article T1 2001,
5: % Revtex format
6: % ----------------------------------------------------------------------------
7: \documentclass[twocolumn,a4paper,aps,amssymb,dvipdfm]{revtex4}
8: % ----------------------------------------------------------------------------
9: \usepackage[dvips]{graphicx}
10: %\usepackage{dcolumn}
11: % ----------------------------------------------------------------------------
12: %
13: \begin{document}
14: 
15: \title{ The longitudinal spin relaxation of 2d electrons in Si/SiGe
16: quantum wells in a magnetic field}
17: 
18: \author{Z. Wilamowski$^{1,2}$, W. Jantsch$^1$ }
19: \address{$^1$Institut f\"{u}r Halbleiterphysik, Johannes Kepler Universit\"{a}t, A-4040 Linz,
20: Austria.}
21: \address{$^2$Institute of Physics, Polish Academy of Sciences, Al. Lotnikow 32/46, PL 0668 Warsaw, Poland}
22: 
23: \date{\today}
24: \begin{abstract}
25:  The longitudinal spin relaxation time, $T_1$, in a Si/SiGe quantum well is determined
26: from the saturation of the ESR signal.  We find values of a few
27: microseconds.  Investigations of $T_1$ as a function of Fermi
28: energy, concentration of scattering centers and of the momentum
29: scattering time, $\tau_k$, lead to the conclusion that for high
30: electron mobility the spin relaxation is ruled by the
31: Dyakonov-Perel (DP) mechanism while for low mobility the
32: Elliott-Yaffet mechanism dominates. The DP relaxation is caused by
33: Bychkov-Rashba coupling. Evaluation of the DP mechanism shows that
34: $T_1^{-1}$ for high electron mobility can be effectively reduced
35: by an external magnetic field. The effect of the degenerate
36: Fermi-Dirac statistics on the DP process is discussed.
37: 
38: 
39: 
40: \end{abstract}
41: \pacs{PACS Numbers:71.18.+y, 71.20.Mq, 71.70.-d, 72.15.Lh,
42: 72.25.-b}
43: 
44:  \maketitle
45: 
46: 
47: \section{Introduction}
48: 
49:  Looking for spin systems suitable for spintronics or
50: quantum computing devices, the longitudinal spin relaxation time,
51: $T_1$, is of basic importance.  $T_1$ is ruled by spin-flip
52: processes and it corresponds to the characteristic spin memory
53: time. In this paper we investigate $T_1$ in the high mobility 2d
54: electron gas in a Si/SiGe quantum well, where electrons can be
55: easily manipulated by illumination with light and by an electric
56: field \cite{R1,R2}.  We show that for this material system, which
57: magnetically is one of the cleanest, $T_1$ is of the order of
58: microseconds whereas the time needed for a spin manipulation by a
59: microwave magnetic field is by more than two orders of magnitude
60: shorter.
61: 
62: We also investigate the mechanism for spin relaxation.  Analyzing
63: the spin relaxation rate as a function of the momentum scattering
64: rate, $\tau_k^{-1}$, allows to distinguish the Elliott-Yaffet (EY)
65: mechanism \cite{R3} and the D'yakonov-Perel (DP) mechanism
66: \cite{R4,R5}.  The EY mechanism describes the probability of a
67: spin-flip in a momentum scattering event.  This probability is
68: ruled by spin-orbit coupling and the resulting admixture of a
69: state with opposite spin projection \cite{R3,R10}:
70: 
71: \begin{equation}\label{1}
72: (T_1^{-1})_{EY}=\alpha_{EY}\tau_k^{-1}
73: \end{equation}
74: 
75: The DP relaxation, in contrast, originates from a zero field spin
76: splitting of the conduction band states \cite{R4,R5,R10}.  For a
77: Si quantum well the zero field splitting is described by the
78: Bychkov-Rashba (BR) term \cite{R3,R10}:
79: \begin{equation}\label{2}
80: {\cal H}_{BR}=\alpha_{BR}(\bf k\times \bf{\sigma})\cdot
81: \hat{\bf{e}}_z
82: \end{equation}
83: which was shown to exist also for single sided modulation doped Si
84: quantum wells \cite {R7}.
85: 
86: 
87: Here $\sigma$ stands for the vector of a Pauli spin-matrix of a
88: conduction electron \cite{R8}, $\bf k$ is the k-vector
89: proportional to the electron momentum, $\hat{\bf{e}}_z$   is a
90: unit vector perpendicular to the 2d layer and $\alpha _{BR}$ is
91: the Rashba parameter that depends on the spin-orbit coupling and
92: details of the interface \cite{R9}.  Momentum scattering causes
93: also a time dependent modulation of the BR interaction.  As a
94: consequence, the probability for spin-flips becomes finite.  For
95: non-quantizing magnetic field, the DP mechanism \cite{R5,R10} is
96: expected to be proportional to the momentum scattering time:
97: \begin{equation}\label{3}
98: (T_1^{-1})_{DP}=\Omega_{BR}^2\tau_k
99: \end{equation}
100: where the frequency $\Omega _{BR}$  is proportional to the
101: k-vector and the BR parameter $\alpha_{BR}$:
102: \begin{equation}\label{4}
103: {\bf\Omega_{BR}}^2=\alpha _{BR}{\bf k}/2\hbar
104: \end{equation}
105: 
106: 
107: In this paper we present results obtained from conduction electron
108: spin resonance (CESR) spectroscopy.  Simultaneous measurements of
109: CESR, which allows to evaluate $T_1$, and cyclotron resonance
110: (CR), which allows \cite{R1,R2} to estimate $\tau_k$, permit the
111: evaluation of $T_1$ as a function of $\tau_k$.  Such data are
112: obtained from samples with different donor and electron
113: concentrations.
114: 
115: 
116: \section{Samples and experimental results}
117: 
118: Samples were grown by molecular beam epitaxy on 1000 $\Omega$cm
119: Si(001) substrates, which show complete carrier freeze-out below
120: 30 K.  A 20 nm thick Si channel with tensile in- plane strain was
121: deposited on a strain-relaxed Si$_{0.75}$Ge$_{0.25}$ buffer layer,
122: which consists of a 0.5 $\mu$m thick Si$_{0.75}$Ge$_{0.25}$ layer
123: on top of a 2 $\mu$m thick Si$_{1-x}$Ge$_x$ layer with
124: compositional grading.  The upper Si$_{0.75}$Ge$_{0.25}$  barrier
125: was modulation doped with a 12.5 nm thick, nominally undoped
126: spacer layer, and capped with 5 nm of Si.  Three modulation doped
127: Si/SiGe structures with different donor concentrations were
128: examined.  The electron concentration was changed by the light
129: illumination.
130: 
131: 
132:     All measurements were performed with a the standard X-band ESR spectrometer, at a
133: microwave frequency 9.4 GHz.  The sample were situated in the
134: center of the rectangular TM$_{201}$ cavity, at the maximum of the
135: magnetic component of microwave field (which is perpendicular to
136: the applied magnetic field).  The sample layer was oriented to be
137: perpendicular to the applied magnetic field (and to electric
138: component of microwave field).
139: 
140: 
141:  \begin{figure}[t]
142:  \begin{center}
143:  \includegraphics[width=7cm]{Fig1.eps}
144:  \end{center}
145:  \caption{ (a) Longitudinal spin relaxation time of the
146:  2d electron gas in a Si/SiGe quantum well
147: as a function of the sheet electron concentration, $n_s$.  (b)
148: Concentration dependence of the momentum scattering rate,
149: $\tau_k^{-1}$, as evaluated from the cyclotron resonance
150: linewidth. Different symbols stand for samples with different
151: donor concentrations in the doping layer.}
152: 
153: \label{fig1}
154: \end{figure}
155: 
156: 
157: 
158: The spin resonance has an exceedingly narrow linewidth in the
159: range $3 \div 10 \mu$T.  In spite of the fact that the sample was
160: situated in the minimum of the electric microwave field and
161: perpendicular to it, the strong absorption due to CR was well
162: observed allowing for to monitor the carrier density from the
163: integral absorption and the momentum scattering rate from the CR
164: linewidth \cite{R2}.
165: 
166: 
167: In Fig. 1 the longitudinal spin relaxation time, $T_1$, and the
168: momentum relaxation rate, $\tau_k^{-1}$, are plotted as a function
169: of the electron concentration, $n_s$.  The parameter $T_1$ has
170: been evaluated from the saturation of the ESR signal amplitude and
171: the ESR line broadening at high microwave power \cite{R12,R13}.
172: Estimating the quality factor of the loaded cavity from the
173: resonance dip width, we obtain the amplitude of the magnetic
174: component of the microwave field of 1.1 G at a microwave power of
175: 200 mW.  The data for different samples are marked by different
176: symbols.  The results for the spin relaxation time, $T_1$, vary in
177: the range of 1 to 5 $\mu$s.  For different samples $T_1$ is
178: different and it depends on the electron concentration.  The
179: momentum scattering rate varies with $n_s$ by an order of
180: magnitude.  The increase of the momentum scattering is related to
181: the screening breakdown and an increase of the potential
182: fluctuations at low Fermi energy \cite {R1}.  Samples with a
183: higher doping level show also a higher $\tau_k^{-1}$.
184: 
185: 
186: 
187: The dependence of $T_1$  on $n_s$ is governed by the complex
188: dependence of the relaxation rate on the Fermi k-vector and of the
189: dependence of $\tau_k^{-1}$ on the electron concentration.  In
190: order to follow the dependence of the spin relaxation rate,
191: $T_1^{-1}$, on the momentum relaxation, $\tau_k^{-1}$, our data
192: are plotted in Figs. 2a and 2b in two different ways.  In Fig. 2a
193: the spin relaxation rate, $T_1^{-1}$, is given as a function of
194: the momentum scattering rate, $\tau_k^{-1}$.  In Fig. 2b, the spin
195: relaxation rate, $T_1^{-1}$, is normalized by the electron
196: concentration, $n_s$. This normalization allows to account for the
197: dependence of $T_1^{-1}$ on the BR parameter, $\alpha_{BR}$, and
198: to study the dependence of the DP rate $T_1^{-1}$ on
199: $\tau_k^{-1}$.
200: 
201: 
202: 
203: \begin{figure}[t]
204: \begin{center}
205: \includegraphics[width=7cm]{Fig2.eps}
206: \end{center}
207:  \caption{ (a) Spin relaxation rate, $T_1^{-1}$, as a function of
208: momentum scattering rate,$\tau_k^{-1}$.  It shows that for the
209: strong scattering the EY process dominates the spin relaxation.
210: The spin relaxation normalized by the electron concentration,
211: $n_s$, is shown in Fig (b).  It shows that the spin relaxation for
212: the high electron mobility, $\tau_k^{-1}<3 \cdot 10^{11}$
213: s$^{-1}$, is dominated by the DP mechanism.  Solid line: DP
214: relaxation rate for a BR parameter of $\alpha_{BR}= 1.1 \cdot
215: 10^{11}$ eV·cm \cite{R7} and degenerate statistics (Eqs.(7-8)).
216: Dash-dotted line: non-degenerate statistics (Eqs. (6-7)).  Dashed
217: straight line: from Eq. (2) which describes the DP relaxation for
218: non-degenerate statistics in the absence of an external magnetic
219: field.} \label{fig2}
220: \end{figure}
221: 
222: 
223: 
224: \section{The spin relaxation caused by the Elliott-Yafet mechanism}
225: 
226: Comparison of the data in the two figures demonstrates the
227: existence of two different ranges with different spin relaxation
228: behavior.  For high scattering rate, $\tau_k^{-1}>3\cdot10^{11}$
229: s$^{-1}$, the spin relaxation is simply proportional to the
230: momentum scattering rate indicating the EY process as the
231: dominating one.  The EY coefficient is independent both of the
232: electron concentration and the doping level within the
233: experimental error.  The solid line in Fig. 2a corresponds to
234: $\alpha_{EY} = 1.0\cdot 10^{-6}$.
235: 
236: 
237: 
238: For low momentum scattering rate, $\tau_k^{-1} < 3\cdot10^{11}$
239: s$^{-1}$, the spin relaxation rate is bigger than expected from
240: the EY mechanism.  Moreover, $T_1^{-1}$  depends on the electron
241: concentration in the high electron mobility range.  On the other
242: hand, the normalized spin relaxation (see Fig. 2b) is
243: characterized by a systematic dependence, common for all
244: investigated samples.  As we argue below the observed dependence
245: in the high mobility range is well described by the DP mechanism.
246: 
247: 
248: 
249: \section{The D'yakonov-Perel mechanism of the relaxation}
250: 
251: 
252: The prediction of the DP scattering rate, as described by Eqs. (3)
253: and (4) is marked by the dashed line in Fig. 2.  For the BR
254: parameter we took the value $\alpha_{BR} = 0.55\cdot 10^{-12}$ eV
255: cm evaluated earlier from the analysis of the linewidth and
256: g-factor anisotropy in the same samples \cite{R7}.   No
257: correlation between the dashed line and the experimental data is
258: recognizable. Eq. (3) stands, however, for the case of a weak
259: external magnetic field, when the momentum scattering rate is much
260: smaller as compared to the Zeeman frequency, and it does not
261: consider cyclotron motion and Landau quantization. In an external
262: magnetic field, because of the cyclotron motion, the electron
263: velocity changes its direction all the time.  The time correlation
264: function of the k-vector, and consequently the correlation
265: function of the effective BR field seen by an electron, is
266: describes by:
267: \begin{equation}\label{5}
268: \langle {\bf k} \cdot {\bf k}(t)\rangle \propto \langle {\bf
269: \Omega}_{BR} \cdot {\bf \Omega}_{BR}(t)\rangle=\Omega_{BR}^2
270: e^{i\omega t-\tau_k^{-1}t}
271: \end{equation}
272: 
273: 
274: The corresponding probabilities of spin-up and down flips are
275: obtained by the Fourier component of the correlation function
276: (Eq.(5)) at the Zeeman frequency $\omega_o$:
277: \begin{equation}\label{6}
278: W_{\pm}=\frac{\Omega_{BR}^2}{2} \frac{\tau_k}{1+(\omega_c \pm
279: \omega_o)^2 \tau_k^2}
280: \end{equation}
281: 
282: 
283: The longitudinal spin relaxation time for a single electron is
284: equal to:
285: \begin{equation}\label{7}
286: (T_1^{-1})_{DP}=W_{+}+W_{-}
287: \end{equation}
288: 
289: 
290: Eqs. (6) and (7) describe the DP relaxation in an external
291: magnetic field.  For a short momentum relaxation time, Eq.(7)
292: becomes equivalent to Eq.(3).  But for quantizing magnetic field,
293: where $\omega_c \tau_k>1$ , the DP relaxation rate is expected to
294: be reduced by the denominator in Eq. (6).  The dependence
295: corresponding to Eqs. (6-7) is shown in Fig. 2 by the dash-dotted
296: line.  The reduction of the spin relaxation caused by the external
297: magnetic field is well visible.  Moreover, for low scattering
298: rate, $\omega_c \tau_k \gg 1$ , the DP relaxation rate is expected
299: to be proportional to the momentum scattering.
300: 
301: 
302: For an electron gas, in which the final states to which electron
303: can be scattered, are partially occupied the evaluation of the
304: mean spin relaxation rate of the whole electron system requires
305: thermodynamic averaging.  The scattering probability and the
306: momentum relaxation rate depend on energy.  These quantities are
307: proportional to the population of empty states:
308: $\tau_k^{-1}(\varepsilon)=\tau_{ko}^{-1} [1-f_{FD}(\varepsilon)]$
309: where $\tau_{ko}^{-1}$ is the momentum relaxation rate (as used in
310: the Boltzmann equation approach) and $f_{FD}(\varepsilon)$ is the
311: Fermi-Dirac distribution function. For moderate magnetic field,
312: where $\hbar \omega_c<k_BT<E_F$ the dependence of the BR frequency
313: on energy can be neglected and the mean value of the transition
314: probability, $\langle W_{\pm}\rangle$, weighted by the derivative
315: of the Fermi-Dirac distribution function, is described by:
316: \begin{equation}\label{8}
317: \langle W_{\pm}\rangle=\frac{\Omega_{BR}^2}{2}\int
318: f'_{FD}(\varepsilon)\frac{\tau_k(\varepsilon)}{1+(\omega_c \pm
319: \omega_o)^2\tau_k^2(\varepsilon)} d\varepsilon
320: \end{equation}
321: 
322: 
323: The solid line in Fig. 2b corresponds to the DP scattering rate as
324: described by Eqs. (7) and (8).  For the BR parameter again a value
325: of $0.55\cdot 10^{-12}$ eV cm has been taken.  The theoretical
326: curve, without any other fitting parameter, fits well to the
327: experimental points for high electron mobility.  For the highest
328: mobility the effect of a moderate magnetic field (B=0.34 T) is the
329: reduction of the DP mechanism by about two orders of magnitude.
330: In the limit of very high electron mobility, the DP relaxation
331: rate for degenerate statistics (Eq. (7-8)) is by a factor 2
332: smaller as compared to the non-degenerate case described by Eqs.
333: (6-7)
334: 
335: 
336:  For the high momentum scattering rate the DP relaxation rate is
337: expected to tend to the solution for weak magnetic field.  But for
338: degenerate statistics (solid line), where the final states are
339: partially occupied, the spin relaxation rate is by a factor 2
340: bigger as compared to the non-degenerate case (dashed and
341: dash-dotted lines) for a given momentum relaxation rate.
342: 
343: \section{Conclusions}
344: 
345: In conclusion, we have shown that:
346: \begin{itemize}
347: \item the DP mechanism dominates for high mobility structures but the
348: quantization due to the applied magnetic field leads to a
349: considerable reduction of the DP relaxation rate.  In Eq. (3) a
350: reduction factor of about $1+\tau_k^2\omega_c^2$ must be
351: introduced, where $\omega_c$ is the cyclotron frequency (compare
352: Eq. (3) and Eqs. (6-7)).  As a consequence, for weak momentum
353: scattering the reduced DP spin relaxation rate is proportional to
354: $\tau_k^{-1}$, in contrast to Eq. (3).
355: 
356: \item the value of the BR parameter, $\alpha_{BR}$,
357: as determined from $T_1$ turns out to be the same within the
358: experimental accuracy as previously evaluated from the anisotropy
359: of the CESR linewidth (dephasing time, $T_2$) and the g-factor
360: \cite{R7}.
361: 
362: \item for low mobility samples the spin relaxation is dominated by the EY mechanism.  We find
363: an EY coefficient of: $\alpha_{EY}= 1.0 \cdot 10^{-6}$, which is
364: common for all samples and, for the investigated range of
365: parameters (the Fermi energy does not exceed 2.5 meV),
366: $\alpha_{EY}$ does not depend on the electron concentration.
367: 
368:  \end{itemize}
369: 
370: 
371: \acknowledgments
372: 
373: We thank F. Sch\"{a}ffler (JKU)
374:  for generously providing samples and for
375: helpful discussions.  Work supported within the KBN grant 2 P03B
376: 007 16 in Poland and in Austria by the Fonds zur F\"{o}rderung der
377: Wissenschaftlichen Forschung, and \"{O}AD, both Vienna.
378: 
379: 
380: 
381: 
382:  \begin{thebibliography}{}
383: 
384: 
385:  \bibitem{R1}
386:  Z. Wilamowski, N. Sandersfeld, W. Jantsch, D.T\"{o}bben, F. Sch\"{a}ffler,
387:  Phys. Rev. Lett. {\bf 87},026401 (2001)
388: 
389: 
390:   \bibitem{R2}
391:   W. Jantsch, Z. Wilamowski, N. Sandersfeld and F. Sch\"{a}ffler,
392:   Phys. stat. sol. (b) {\bf 210}, 643 (1998), and:
393:   Z. Wilamowski and W. Jantsch, Physica
394:   E 10, 17-21(2001)
395: 
396: 
397:   \bibitem{R3}
398:   R.J. Elliott,
399:   Phys. Rev. 96, 266 (1954) and
400:   Y. Yafet,
401:   Sol. St. Phys. {\bf 14}, 1 (1963)
402: 
403: 
404:   \bibitem{R4}
405:   F.G. Pikus and G.E. Pikus, Phys. Rev. B51, 16928 (1995)
406: 
407: 
408:   \bibitem{R5}
409:   W. Knap, C. Skierbiszewski, A. Zduniak, E. Litwin-Staszewska, D. Bertho, F. Kobbi, J.L.
410: Robert, G.E. Pikus, F.G. Pikus, S.V. Iordanskii, V. Mosser, K.
411: Zekentes, Yu. B. Lyanda-Geller,
412: Phys. Rev. B{\bf 53}, 3912 (2001)
413: 
414: 
415:    \bibitem{R10}
416: F.G. Pikus and G.E. Pikus, Phys. Rev. B{\bf 51}, 16928 (1995)
417: 
418: 
419:    \bibitem{R6}
420:    E. I. Rashba, Fiz. Tverd. Tela (Leningrad) {\bf 2}, 1224 (1969) [Sov. Phys. Solid State {\bf 2}, 1109
421: (1960)]
422: 
423: 
424:   \bibitem{R7}
425:   Z. Wilamowski, W. Jantsch and U. R\"{o}ssler, unpublished and: cond-mat/0110342
426: 
427: 
428:   \bibitem{R8}
429:   The coefficient $\alpha_{BR}$, as defined by Equ. (3), is half as big as compared to the case when
430:   $\sigma$ stands for the spin operator.
431: 
432: 
433:   \bibitem{R9}
434:   J.A. Majewski, P. Vogl and P. Lugli,
435:   Proc. 25th Int conf Phys. Semic. Osaka 2000, (Eds. N. Miura and T. Ando), p791 (Springer, Berlin2001)
436: 
437: 
438:   \bibitem{R12}
439:   C.P. Poole: Electron Spin Resonance (J. Wiley, N.Y. 1983 (2nd. Ed.))
440: 
441: 
442:   \bibitem{R13}
443:   Z. Wilamowski and W. Jantsch,
444:   Physica E  {\bf 10}, 17-21 (2001)
445: 
446: 
447:  \end{thebibliography}
448: 
449: \end{document}
450: