1: %\input{tcilatex}
2: %%\input{tcilatex}
3:
4:
5: \documentclass[prb, twocolumn]{revtex4}
6: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
7: \usepackage{amsmath}
8: \usepackage{graphicx}
9:
10: \setcounter{MaxMatrixCols}{10}
11: %TCIDATA{OutputFilter=LATEX.DLL}
12: %TCIDATA{Version=4.00.0.2312}
13: %TCIDATA{LastRevised=Monday, October 17, 2005 15:36:40}
14: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
15: %TCIDATA{Language=American English}
16:
17: \input{tcilatex}
18:
19: \begin{document}
20:
21: \title{Numerical Study of Spin Hall Transport in a Two Dimensional Hole Gas
22: System}
23: \author{W.Q. Chen$^{1}$, Z.Y. Weng$^{1}$, and D.N. Sheng$^{2}$}
24: \affiliation{$^{1}$Center for Advanced Study, Tsinghua University, Beijing 100084\\
25: $^{2}$Department of Physics and Astronomy, California State University,
26: Northridge, CA 91330}
27:
28: \begin{abstract}
29: We present a numerical study of the spin Hall effect in a two-dimensional
30: hole gas (2DHG) system in the presence of disorder. We find that the spin
31: Hall conductance (SHC), extrapolated to the thermodynamic limit, remains
32: finite in a wide range of disorder strengths for a closed system on torus.
33: But there is no intrinsic spin Hall accumulation as induced by an external
34: electric field once the disorder is turned on. The latter is examined by
35: performing a Laughlin's Gedanken gauge experiment numerically with the
36: adiabatic insertion of a flux quantum in a belt-shaped sample, in which the
37: absence of level crossing is found under the disorder effect. Without
38: disorder, on the other hand, energy levels do cross each other, which
39: results in an oscillating spin-density-modulation at the sample boundary
40: after the insertion of one flux quantum in the belt-shaped system. But the
41: corresponding net spin transfer is only about one order of magnitude smaller
42: than what is expected from the bulk SHC. These apparently contradictory
43: results can be attributed to the violation of the spin conservation law in
44: such a system. We also briefly address the dissipative Fermi surface
45: contribution to spin polarization, which may be relevant to experimental
46: measurements.
47: \end{abstract}
48:
49: \pacs{72.10.-d,72.25.DC,73.43.-f}
50: \maketitle
51:
52: \section{Introduction}
53:
54: \label{sec:intro} Recently it has been proposed \cite{MNZ,Sinova1} that in
55: spin-orbit coupling (SOC)\ systems one may use an electric field to generate
56: transverse spin currents in the absence of external magnetic fields. It has
57: been argued \cite{MNZ,Sinova1} that such a spin Hall effect (SHE) is \emph{%
58: intrinsic}, contributed by \emph{all} the electrons below the Fermi energy,
59: and the corresponding spin currents are ``dissipationless''\ as in contrast
60: to the dissipative longitudinal charge currents which are only contributed
61: by the electrons close to the Fermi energy and are strongly subjected to
62: scattering effects.
63:
64: The original proposals for the SHE are for the disorder-free cases.\cite%
65: {MNZ,Sinova1} In the two-dimensional (2D) electron gas described by the
66: Rashba model,\cite{Sinova1} it was shown based on a perturbative approach
67: \cite{vertex1,vertex2} that the SHC is precisely cancelled by the vertex
68: correction once the disorder is turned on. On the other hand, the vertex
69: correction is found to vanish \cite{vertex_luttinger} for the
70: three-dimensional (3D) p-doped semiconductors described by the Luttinger
71: model \cite{MNZ} so that the SHC is still finite in the presence of weak
72: disorder. Numerical calculations \cite{kubo2,luttinger,kubo1} of the SHC
73: seem to support such perturbative results of the distinct behavior for two
74: models in the thermodynamic limit.
75:
76: Experimentally the signatures of spin polarization have been observed
77: recently in 2DHG system \cite{exp1} and 3D n-doped semiconductors,\cite{exp2}
78: which have generated a lot of excitement concerning whether they are due to
79: the intrinsic SHE mentioned above or some extrinsic effect.\cite{eshe}
80: Bernevig and Zhang have shown that the vortex correction does vanish in the
81: 2DHG \cite{2dhg} and 3D n-doped semiconductor \cite{ngaas} systems.
82: Furthermore, the mesoscopic SHE in the 2DHG is also found to be present
83: based on the nonequilibrium Green function method \cite{ne2dhg} and
84: Landauer-B\"{u}ttiker formula with attached leads,\cite{lb2dhg} similar to
85: (with larger magnitude than) the mesoscopic SHE found in the 2D Rashba model.%
86: \cite{LB1, LB3} But it is still unclear whether the bulk SHE in the 2DHG can
87: survive in the thermodynamic limit beyond the perturbative approach.
88:
89: In this paper, we perform numerical calculations for the 2DHG in the
90: presence of disorder. We first show that the bulk SHC calculated from the
91: Kubo formula is indeed robust against the disorder in extrapolation to the
92: thermodynamic limit, which is consistent with the vertex correction
93: calculation.\cite{2dhg} It is also similar to the behavior for the Luttinger
94: model,\cite{luttinger} but is in opposite to that\ of the 2D Rashba model.%
95: \cite{kubo2} But when we perform a Laughlin's Gedanken ``gauge experiment''\
96: on a belt-shaped sample to probe the spin transfer/accumulation due to the
97: SHE, we obtain a null result due to the anticrossing between energy levels,
98: which is quite similar to what has been previously seen in the 2D Rashba
99: model.\cite{kubo2} We point out that the absence of edge states in the 2DHG
100: system causes the general level repulsion with the turn on of disorder,
101: which leads to the disappearance of the net spin Hall accumulation in an
102: open system. Furthermore, at zero disorder case, energy levels do cross each
103: other and we find an oscillating spin-density-modulation at the sample
104: boundary after an adiabatic insertion of a flux quantum in the
105: above-mentioned Laughlin's gauge experiment. However, the corresponding net
106: spin transfer is only about one order of magnitude smaller than what is
107: expected from the calculated bulk SHC. We discuss the ``conflicting''\
108: results of a finite SHC but without an intrinsic spin accumulation in the
109: 2DHG, and point out that the underlying reason can be attributed to the
110: violation of the spin conservation law in such a SOC system, where the SHC
111: is no longer an unambiguous quantity for describing the spin transport. We
112: also address the addition dissipative Fermi surface contribution to spin
113: polarization, which may be relevant to experimental measurement.\cite%
114: {exp1,exp2}
115:
116: The remainder of the paper is organized as follows. In Sec. \ref{sec:shc},
117: we numerically compute the SHC through the Kubo formula in a tight-binding
118: model of the 2DHG system at different sample sizes and disorder strengths,
119: and perform finite-size scaling analysis. In Secs. \ref{sec:gelc}, \ref%
120: {sec:st} and \ref{sec:nr}, we perform a Laughlin's gauge experiment
121: numerically to determine the intrinsic spin transfer/accumulation due to the
122: adiabatic insertion of a magnetic flux quantum. And in Sec. \ref{sec:dis},
123: we compare the results with a system in a perpendicular external magnetic
124: field and demonstrate the relationship between the intrinsic spin transfer
125: (accumulation) and the existence of edge states. Finally, a summary is given
126: in Sec. V.
127:
128: \section{Spin Hall Conductance}
129:
130: \label{sec:shc}
131:
132: \subsection{Hamiltonian}
133:
134: We start with the 2DHG Hamiltonian proposed by Bernevig and Zhang\cite{2dhg}
135: \begin{equation}
136: H=(\gamma _{1}+\frac{5}{2}\gamma _{2})\frac{k^{2}}{2m}-\frac{\gamma _{2}}{m}(%
137: \vec{k}\cdot \vec{S})^{2}+\alpha (\vec{S}\times \vec{k})\cdot \hat{z}
138: \label{eq:con_ham}
139: \end{equation}%
140: For the convenience of the following numerical study, we further convert
141: this continuum Hamiltonian into a tight-binding version on the square
142: lattice. This may be realized by making the replacement $k_{\nu }\rightarrow
143: \sin k_{\nu }$ and $k_{\nu }^{2}\rightarrow 2(1-\cos k_{\nu })$. Two models
144: are apparently equivalent near the band bottom where $k_{\nu }\rightarrow 0$%
145: . The resulting Hamiltonian reads
146: \begin{eqnarray}
147: H &=&-t\sum_{\langle ij\rangle }(c_{i}^{\dagger }c_{j}+c_{j}^{\dagger
148: }c_{i})+V_{L}\sum_{i\nu }(c_{i}^{\dagger }S_{\nu }^{2}c_{i+\hat{\nu}}+H.c.)
149: \notag \\
150: &+&\frac{V_{L}}{4}\left( \sum_{i}c_{i}^{\dagger }\left\{ S_{x},S_{y}\right\}
151: (c_{i+\hat{x}+\hat{y}}-c_{i+\hat{x}-\hat{y}})+H.c.\right) \notag \\
152: &+&V_{R}\left( i\sum_{i}c_{i}^{\dagger }S_{y}c_{i+\hat{x}}-i\sum_{i}c_{i}^{%
153: \dagger }S_{x}c_{i+\hat{y}}+H.c.\right) \notag \\
154: &-&2V_{L}\sum_{i}c_{i}^{\dagger }\left[ (S_{x}^{2}+S_{y}^{2})-(\frac{9}{4}%
155: -S_{z}^{2})\langle k_{z}^{2}\rangle \right] c_{i} \notag \\
156: &&+4t\sum_{i}c_{i}^{\dagger }c_{i}\text{ }+\epsilon _{i}
157: \sum_{i}c_{i}^{\dagger }c_{i}\text{ } \label{H}
158: \end{eqnarray}%
159: where the electron annihilation operator $c_{i}$ has four components
160: characterized by the ``spin'' index $S_{z}=\frac{3}{2},\frac{1}{2},-\frac{1}{%
161: 2},-\frac{3}{2},$ respectively, and $i+\hat{\nu}$ ($\hat{\nu}=\hat{x},\hat{y}
162: $) denote the nearest-neighbors of site $i$. Here $t$ is the
163: nearest-neighbor hopping integral defined as
164: \begin{equation}
165: t=\frac{\hbar ^{2}}{2ma_{0}^{2}}(\gamma _{1}+\frac{5}{2}\gamma _{2}),
166: \end{equation}%
167: where $m$ is the mass of the electron and $a_{0}$ is the lattice constant. $%
168: V_{L}$ and $V_{R}$ represents the strengths of the Luttinger-type and
169: Rashba-type spin-orbital couplings, respectively:
170: \begin{eqnarray}
171: V_{L} &\equiv &\frac{2\gamma _{2}}{\gamma _{1}+\frac{5}{2}\gamma _{2}}t
172: \notag \\
173: V_{R} &\equiv &\frac{\alpha }{\gamma _{1}+\frac{5}{2}\gamma _{2}}t. \notag
174: \end{eqnarray}%
175: The parameters $\gamma _{1}$ and $\gamma _{2}$ in GaAs are given by $\gamma
176: _{1}=6.92$ and $\gamma _{2}=2.1$,\cite{haug} so we have $t\sim 1.45eV$ and $%
177: V_{L}\sim 0.345t$. We shall choose $V_{R}=0.02t$ for the Rashba-type
178: coupling. By assuming that the 2DHG is confined in a well of the thickness $%
179: a\sim 8.7$ nm, one approximately has $\langle k_{z}\rangle =0$ and $\langle
180: k_{z}^{2}\rangle \sim \left( \frac{\pi }{a/a_{0}}\right) ^{2}$, where $%
181: a_{0}\sim $ $0.565$ nm for GaAs. The corresponding energy spectrum for a
182: clean sample ($\epsilon_{i}=0$) is shown in Fig. \ref{fig:band}. Due to the
183: confinement in the $\hat{z}$ direction, a gap with $0.029t\sim 0.042eV$
184: opens up between the heavy hole (HH) band and the light hole (LH) band at $%
185: \Gamma $ point (see Fig. \ref{fig:band}). For simplicity, we shall set $t=1$
186: in the rest of paper. Finally, $\epsilon _{i}$ in Eq.(\ref{H}) accounts for
187: the on-site and spin-independent disorder strength, which is randomly
188: distributed within $[-W/2,W/2]$.
189:
190: \begin{figure}[t]
191: \resizebox{80mm}{!}{ \includegraphics{band.eps} } \centering
192: \caption{Band structure for a clean 2DHG system determined by Eq.( \protect
193: \ref{H}) (with $t=1)$. The HH and LH denote the heavy hole band and light
194: hole band, respectively.}
195: \label{fig:band}
196: \end{figure}
197:
198: \subsection{\protect\bigskip Spin Hall conductance}
199:
200: The SHC is defined by the Kubo formula\cite{kubo0}
201: \begin{equation}
202: \sigma _{SH}=\frac{2}{N}\left\langle \mathrm{Im}\sum_{E_{n}<E_{f}<E_{m}}%
203: \frac{\langle n|j_{x}^{z\mathrm{spin}}|m\rangle \langle m|j_{y}|n\rangle }{%
204: (E_{m}-E_{n})^{2}}\right\rangle , \label{eq:kubo}
205: \end{equation}%
206: where $N$ is the total number of lattice sites, $E_{f}$ denotes the Fermi
207: energy, $E_{m}$ ($E_{n}$) is the eigen-energy, and $\langle ...\rangle $ is
208: averaged over all disorder configurations. The charge current operator $%
209: j_{\mu }$ and spin current operator $j_{\mu }^{\nu \mathrm{spin}}$ are
210: defined as
211: \begin{eqnarray}
212: j_{\mu } &\equiv &ev_{\mu }, \notag \\
213: j_{\mu }^{\nu \mathrm{spin}} &\equiv &\frac{1}{2}\{v_{\mu },S_{\nu }\},
214: \notag
215: \end{eqnarray}%
216: where the velocity operator $\mathbf{v}$ is the conjugate operator of the
217: position operator $\mathbf{R}\equiv \sum_{i\sigma }\mathbf{r}_{i}n_{i\sigma
218: } $ ($n_{i\sigma }$ is the electron number operator at site $i$ with spin
219: index $\sigma $), is defined by the standard relation
220: \begin{equation}
221: v_{\mu }=\frac{i}{\hbar }[H,R_{\mu }]\text{ \ }.
222: \end{equation}
223:
224: \begin{figure}[t]
225: \resizebox{90mm}{!}{ \includegraphics{pure.eps} } \centering
226: \caption{$\protect\sigma_{SH}$ vs $E_f$ at $W = 0$. The inset shows the
227: whole Fermi energy region and the main panel shows the details around the
228: band bottom. The solid line is at $1000 \times 1000$ lattice, the dashed
229: line is at $24 \times 24$ lattice, and the open circles are at $24 \times 24$
230: lattice with averaged over 1000 BCs.}
231: \label{fig:pure}
232: \end{figure}
233:
234: For a pure sample ($\epsilon _{i}=0$), the Hamiltonian can be easily
235: diagonalized in momentum space, and we can determine\ $\sigma _{SH}$ vs $%
236: E_{f}$ for a very large lattice ($N=L^{2}$ with $L=1000$) as shown in the
237: inset of Fig. \ref{fig:pure}. Since the tight-binding model is a good
238: approximation of the original model only when $E_{f}$ is near the band
239: bottom, we shall focus on the regime of $0<E_{f}<0.6$ as represented in the
240: main panel of Fig. \ref{fig:pure} by the solid curve. One can see that $%
241: \sigma _{SH}$ in this region is a little bit larger than $10$ $\mathrm{%
242: e/8\pi }$ and very flat as a function of $E_{f}$ except for that within the
243: gap ($\sim 0.029)$ between the HH and LH bands.
244:
245: Because the momentum is no longer a good quantum number in the presence of
246: disorder, Eq.(\ref{H}) can not be diagonalized in the momentum space when $%
247: \epsilon _{i}\neq 0$. In the latter case, the maximal lattice size reachable
248: by the numerical calculation is usually much smaller than $L=1000$ (around $%
249: L=20-30$). To check what happens at smaller lattices, we compute $\sigma
250: _{SH}$ on an $L=24$ lattice in the pure system with a periodic boundary
251: condition (BC) and the result is illustrated by the dotted curve in Fig. \ref%
252: {fig:pure}. The corresponding $\sigma _{SH}$ exhibits very large
253: fluctuations whose magnitude can be several times larger than the converged
254: value at $L=1000$. To remove such finite-size fluctuations, we introduce an
255: average over different BCs. A general (twisted) BC is given by
256: \begin{eqnarray}
257: \psi (x+L,y) &=&e^{i\theta _{x}}\psi (x,y) \notag \label{eq:tbc} \\
258: \psi (x,y+L) &=&e^{i\theta _{y}}\psi (x,y)
259: \end{eqnarray}%
260: where $\theta _{\mu }$'s are defined within $[0,2\pi ]$ with $\theta _{\mu
261: }=0$ or $2\pi $ ($\mu =x,y)$ corresponding to the periodic BC. Since the SHC
262: should not be sensitive to the BCs in the thermodynamic limit, in principle,
263: $\sigma _{SH}$ in Eq. \eqref{eq:kubo} can be always defined as averaged over
264: both disorder and twisted BCs in such a limit. For a clean system, the
265: averaged $\sigma _{SH}$ at $L=24$ over $1000$ randomly generated BCs is
266: shown by the open circles in Fig. \ref{fig:pure}. One can clearly see that
267: the large fluctuations in $\sigma _{SH}$ at $L=24$ (dotted curve) have been
268: significantly reduced after the BC averaging, and the result (open circles)
269: coincides with $\sigma _{SH}$ obtained at $L=1000$ (solid line) very well.
270: It indicates that the finite-size effect at smaller size ($L=24$) can be
271: properly removed by the BC average. As a consequence, the BC averaged SHC
272: behaves much smoother than at a fixed BC, and is thus more suitable for a
273: finite-size scaling analysis. We note that the BC average is also required
274: by the gauge-invariant condition for SHC in a finite size system. When we
275: apply an electric field through a time-dependent vector potential, it acts
276: as a generalized boundary phase evolving with time (see below). Thus the SHC
277: averaged over time is gauge-invariant and equivalent to the BC averaged SHC.%
278: \cite{kubo2} %In the large system size limit, SHC becomes independed of the
279: %BC and thus the BC average is no longer needed.
280: Such a numerical method has been previously used in the study of the 2D
281: Rashba model and 3D Luttinger model.\cite{kubo2,luttinger} Similar boundary
282: phase averaged charge Hall conductance can be also related to a topological
283: invariant Chern number.\cite{kubo0}
284:
285: \begin{figure}[t]
286: \resizebox{60mm}{!}{ \includegraphics{impurity.eps} } \centering
287: \caption{$\protect\sigma _{SH}$ vs. $E_{f}$ at $W=0.2,$ $1,$ and $3$. The
288: solid curves are for $L=16$, and the dashed ones for $L=24$. (a) is for the
289: 2DHG system and (b) is for the spin-3/2 Rashba model by setting $V_L=0$.}
290: \label{fig:impurity}
291: \end{figure}
292:
293: Now we use the above method to study the sample-size dependence of $\sigma
294: _{SH}$ at $W=0.2$ for weak disorder strength; $W=1$ for intermediate
295: disorder; and $W=3$ for strong disorder. The calculated $\sigma _{SH}$ is
296: presented in Fig. \ref{fig:impurity} (a) for two sample sizes: $L=16$
297: [averaged over 20000 configurations (solid line)] and $L=24$ [averaged over
298: 5000 configurations (dashed line)]. The size-dependence of $\sigma _{SH}$ is
299: very weak for all three cases at different $E_{f}$'s. For comparison, we
300: have also done a similar calculation for a ``spin-3/2 (linear) Rashba
301: model'', which corresponds to Eq.~(\ref{H}) at $V_{L}=0$ and $V_{R}=0.5$.
302: SHC at $W=0.2,$ $1,$ and $3,$ with $L=16$ and $24,$ respectively, are shown
303: in Fig. \ref{fig:impurity}(b), which clearly decrease with the increase of
304: the sample length, similar to the result in the spin-1/2 Rashba model.\cite%
305: {kubo2}
306: %%where a scaling analysis shows $\sigma _{SH}\rightarrow 0$ when the sample
307: %%size goes to infinity.
308:
309: \begin{figure}[t]
310: \resizebox{80mm}{!}{ \includegraphics{dep.eps} } \centering
311: \caption{$W$ and sample-size dependence of $\protect\sigma _{SH}$ with $%
312: E_{f} $ averaged within the interval $[0.3,0.4]$. (a) $\protect\sigma _{SH}$
313: vs. $W $. The open squares are for $L=12,$ close squares for $L=16$, open
314: circles for $L=20$, and close circles for $L=24$. (b) $\protect\sigma _{SH}$
315: vs $L$. The lines from top to bottom are for $W=0.05,$ $0.2,$ $0.6,$ $1,$ $%
316: 1.5,$ $2,$ $2.5,$ $3,$ and $5$, respectively.}
317: \label{fig:dep}
318: \end{figure}
319:
320: To further investigate the effects of the lattice size and disorder, we
321: calculate the averaged $\sigma _{SH}$ within a chosen energy interval $%
322: [0.3,0.4]$ at various disorder strengths and sample sizes as shown in Fig. %
323: \ref{fig:dep}. $\sigma _{SH}$ is almost independent of sample sizes and
324: reduces monotonically with increasing $W$, which is different from the
325: spin-1/2 Rashba model but similar to the behavior in the 3D Luttinger model.%
326: \cite{luttinger} In the weak disorder limit, $\sigma _{SH}$ approaches to
327: the value slightly larger than $10$ $\mathrm{e/8\pi }$, which is quite close
328: to the value ($\sim 9e/8\pi $) obtained \cite{loss} in an analytic
329: calculation in a different model for 2DHG system. Thus we establish that SHC
330: is finite extrapolating to thermodynamic limit until the disorder strength
331: reaches a critical value $W_{c}=5.0$.
332:
333: \section{Intrinsic Spin Transfer and Accumulation}
334:
335: \label{sec:sa}
336:
337: In the last section, we have shown the numerical evidence that a finite $%
338: \sigma _{SH}$ can survive in the presence of disorder with the sample size
339: being extrapolated to the thermodynamic limit. This is in sharp contrast to
340: the 2D Rashba model in which $\sigma _{SH}$ vanishes in the thermodynamic
341: limit with the turn on of very weak disorder strength.\cite{kubo2} Such a
342: distinct behavior in $\sigma _{SH}$ is consistent with the perturbation
343: theories in which the vertex correction also results in opposite conclusions
344: on the fate of $\sigma _{SH}$ in two models, as mentioned in the
345: Introduction.
346:
347: In general, a finite conductance should lead to an accumulation of current
348: carriers at the edges (open boundaries) of the sample according to the
349: current conservation law. However, in the SOC system the spin current
350: conservation law does not exist as the spin is not conserved. Thus the
351: relation between the SHC and the spin accumulation is not straightforwardly
352: present.
353:
354: In this section, we will try to probe the spin transfer or accumulation
355: effect directly by performing numerically a Laughlin's gauge experiment,
356: which was first used in the IQHE system \cite{laughlin, halperin} to explain
357: why an integer number of charges can be transported across the sample in the
358: transverse direction when a longitudinal electric field is applied via an
359: adiabatic flux insertion process. This method has been recently generalized
360: to the SOC system by Sheng \textit{et al.}\cite{kubo2} in the study of the
361: SHE in the 2D Rashba model.
362:
363: \subsection{Gauge experiment and level crossing}
364:
365: \label{sec:gelc}
366:
367: The numerical gauge experiment will be performed on a belt-shaped sample
368: with a magnetic flux $\Phi $ adiabatically inserted at the center of the
369: ribbon [see Fig. \ref{fig:geometry}(a)]. Such a belt-shaped geometry with
370: the magnetic flux can be implemented in a square lattice sample shown in
371: Fig. \ref{fig:geometry}(b) by imposing a twisted BC along the $\hat{x}$
372: direction and an open BC along the $\hat{y}$ direction. Namely, with a gauge
373: transformation, one may impose the flux $\Phi $ by changing the BC along $%
374: \hat{x}$ direction from the periodic BC to a twist BC with the phase twist $%
375: \theta _{x}=\Phi $ indicated by the dashed line and the phase factor $%
376: e^{i\Phi }$ in Fig. \ref{fig:geometry}(b).
377:
378: \begin{figure}[t]
379: \resizebox{90mm}{!}{ \includegraphics{cyl.eps} } \centering
380: \caption{(a) The geometry of the sample in the gauge experiment. It is like
381: a belt enclosing a flux in the center. (b) The implementation of (a) in the
382: numerical calculation. The dashed line with a phase factor $e^{i\Phi }$
383: along the $\hat{x}$ axis means a twist BC, which represents the flux in (a).}
384: \label{fig:geometry}
385: \end{figure}
386:
387: Numerically the energy spectrum of single electron states at a fix $\Phi $
388: can be calculated in the geometry of Fig. \ref{fig:geometry}(b). Fig. \ref%
389: {fig:level_pure} shows such calculated energy spectra as a function of $\Phi
390: $ in the pure system. Note that the system is equivalent at $\Phi =0$ and $%
391: \Phi =2\pi $ as the Hamiltonian is periodic $H(\Phi =0)=H(\Phi =2\pi ),$ and
392: the energy spectrum is symmetric between $0\leq \Phi \leq \pi $ and $\pi
393: \leq \Phi \leq 2\pi $ due to a time-reversal symmetry, such that only the
394: first half is plotted in Fig. \ref{fig:level_pure}.
395:
396: \begin{figure}[t]
397: \resizebox{80mm}{!}{ \includegraphics{level_pure.eps} } \centering
398: \caption{The evolution of the single electron state energy with$\ \Phi $ for
399: the 2DHG on a $32\times 32$ clean sample ($W=0$).}
400: \label{fig:level_pure}
401: \end{figure}
402:
403: One may imagine changing the flux $\Phi $ adiabatically from $0$ to $2\pi $
404: to follow the corresponding change of the system. Generally there are two
405: possibilities for the evolution of the ground state with the electrons
406: filled below the Fermi energy $E_{f}$. If each single electron state always
407: anti-crosses with other states (level repulsion), it will eventually return
408: to the original state after the insertion of one flux quantum of $\Phi =2\pi
409: $ sufficiently slow (i.e., adiabatically such that there is no quantum
410: transition to other levels during this operation); On the other hand, if a
411: level crossing occurs for single electron states during the adiabatical
412: increase of $\Phi ,$ then the original ground state generally will evolve
413: into a different state through the change of the single state occupation
414: number, even though the single electron energy spectrum must remain the same
415: after the insertion of one flux quantum.
416:
417: A slow increase of the flux $\Phi $ is equivalent to applying a weak
418: electric field along the $\hat{x}$ direction:%
419: \begin{equation}
420: \epsilon _{x}=-\frac{1}{L_{x}}\frac{\partial \Phi }{\partial t}.
421: \label{electric}
422: \end{equation}%
423: Then the \emph{intrinsic} charge/spin Hall effect due to an electric field
424: in belt-shaped system will lead to the transfer of charge/spin along the $%
425: \hat{y}$-axis during the flux insertion. If there is to be charge/spin
426: accumulation at the open edges [Fig. \ref{fig:geometry}(a)], it means that
427: the ground state should not return to the original one after the adiabatic
428: flux insertion, which is the essence of Laughlin's gauge experiment for the
429: IQHE system. Therefore, the level crossing of single electron states serves
430: as a necessary condition for the existence of \emph{intrinsic} spin
431: transfer/accumulation for a SHE system. As clearly shown in Fig. \ref%
432: {fig:level_pure}, which is determined on an $L=32$ lattice, the single
433: electron energy levels do simply cross in the pure 2DHG system. The
434: resulting net spin transfer will be analyzed in the following section.
435: %%question if thus This property is thus consistent with the intrinsic
436: %%SHE in the pure sample discussed above.
437:
438: \subsection{Spin Transfer at $W = 0$}
439:
440: \label{sec:st}
441:
442: \begin{figure}[t]
443: \resizebox{80mm}{!}{ \includegraphics{band_o.eps} } \centering
444: \caption{The miniband structure of a $32 \times 32$ clean sample. Two arrows
445: represent the evolution of two states by tuning the parameter $\Phi $, which
446: start from the beginning at $\Phi =0$ and arrive at the end of the arrows at
447: $\Phi =2\protect\pi $ (see the text for details). The dashed line indicates
448: the Fermi energy.}
449: \label{fig:band_o}
450: \end{figure}
451:
452: %As a necessary condition of the spin transport, we have shown the level
453: %crossing in the pure sample in last section.
454: In this section, we calculate the spin transfer induced by tuning the flux $%
455: \Phi $ from $0$ to $2\pi $ in the pure case ($\epsilon_{i}=0$). Starting
456: from the periodic BC ($\Phi =0$), the momentum $k_x$ along $\hat{x}$
457: direction is a good quantum number. And due to the open boundary along the $%
458: \hat{y}$ direction, the electron states form a mini-band structure which is
459: shown in Fig. \ref{fig:band_o} and can be denoted as $|k_{x},n\rangle $,
460: where $n$ is the index of the mini-bands. If we choose an arbitrary Fermi
461: energy $E_{f}$, the ground state corresponds to that all the single particle
462: states below $E_{f}$ are occupied, whereas unoccupied above $E_{f}$. So the
463: spin density along $\hat{y}$ direction (labelled by $y\equiv j_{y}$) at the
464: initial state can be defined by
465: \begin{equation}
466: S_{z}^{i}(y)\equiv \sum_{E_{k_{x},n}<E_{f}}\langle
467: k_{x},n|\sum_{j_{x}}c_{j}^{+}S_{z}c_{j}|k_{x},n\rangle .
468: \end{equation}%
469: When one inserts the flux adiabatically from zero to $\Phi $, the state $%
470: |k_{x},n\rangle $ will evolve to $|P_{x}=k_{x}+\Phi /L_{x},n\rangle $, For $%
471: k_{x}>0$ and $<0$, the corresponding energies of the states will increase
472: and decrease, respectively, which is illustrated by arrows in Fig. \ref%
473: {fig:band_o}. Thus one finds a simple level crossing for two levels close to
474: each other with opposite signs of $k_{x}$, which represents the level
475: crossing shown in Fig. \ref{fig:level_pure} for pure system due to no mixing
476: term between different states in the Hamiltonian. After the insertion of one
477: flux quantum, the state $|k_{x},n\rangle $ evolves into a new state $%
478: |k_{x}+2\pi /L_{x},n\rangle $ at $\Phi =2\pi $. Then the spin density at the
479: final state is given by
480: \begin{equation}
481: S_{z}^{f}(y)=\sum_{E_{k_{x},n}<E_{f}}\langle k_{x}+\frac{2\pi }{L_{x}}%
482: ,n|\sum_{j_{x}}c_{j}^{+}S_{z}^{j}c_{j}|k_{x}+\frac{2\pi }{L_{x}},n\rangle .
483: \end{equation}%
484: The resulting change in the spin-density due to threading a flux quantum is $%
485: \Delta S_{z}(y)=S_{z}^{f}(y)-S_{z}^{i}(y)$.
486:
487: \begin{figure}[t]
488: \resizebox{90mm}{!}{ \includegraphics{spin_dis.eps} } \centering
489: \caption{The spin density changes between the state at the beginning and the
490: end of the procedure of increasing the flux from $0$ to $2 \protect\pi$ in
491: the pure samples. The main panel shows the data around one edge for various
492: lattice sizes, while the inset shows the whole result on a $500 \times 100$
493: lattice. The Fermi energy is fixed at $E_f = 0.4$. (a) is for the 2DHG; The
494: solid lines with open squares are for $1000 \times 100$ lattice, the dashed
495: lines with close squares are for $500 \times 500$ lattice, and the dotted
496: lines with open circles are for $1000 \times 1000$ lattice. (b) is for the
497: 2D Rashba model; The solid lines with open squares are for $100 \times 100$
498: lattice, the dashed lines with close squares are for $500 \times 100$
499: lattice, and the dotted lines with open circles are for $1000 \times 1000$
500: lattice.}
501: \label{fig:spin_dis}
502: \end{figure}
503:
504: $\Delta S_{z}(y)$ in a pure 2DHG with $500\times 100$ lattice is shown in
505: the inset of Fig. \ref{fig:spin_dis}(a), where the Fermi energy is $%
506: E_{f}=0.4 $. We recall that due to the time-reversal invariance, $%
507: S_{z}^{i}(y)=0$ everywhere. After adding the $2\pi $ flux quantum, the spin
508: density shows some strong peaks at the boundaries whereas with a much weaker
509: magnitude in the bulk.
510: %To see more clearly and consider the lattice size effect,
511: We show the spin-wave like modulation in $\Delta S_{z}(y)$ near one edge of
512: the belt-shaped sample at various lattice sizes in the main panel of Fig. %
513: \ref{fig:spin_dis}(a).
514: %%It is clearly that the size dependence of $\Delta S_z (y)$ is different
515: %%at the edge and at the edge and in the bulk is different.
516: Here $\Delta S_{z}(y)$ is almost sample-size independent at the edge,
517: whereas it reduces monotonically with the sample size in the bulk. So in the
518: thermodynamic limit, there should be only the spin density modulation
519: present near the edge. Similar results with smaller magnitude are also
520: obtained for an electron spin-$1/2$ Rashba model as shown in Fig. \ref%
521: {fig:spin_dis}(b) for comparison. Such a spin density modulation does
522: support a net spin transfer along the $\hat{y}$ direction, accompanying an
523: electric field generated along the $\hat{x}$ direction due to the inserting
524: flux according to Eq. (\ref{electric}), but the magnitude of the total
525: (integrated) $\Delta S_{z}\sim 0.2\hbar$ at one edge is much less than one
526: spin quantum, which is in sharp contrast to the large bulk SHC (the
527: calculated $\sigma _{SH}\sim 11 \mathrm{e/8\pi }$, that suggests a total
528: spin transfer of $5.5(\hbar/2)$ spin quanta). This large discrepancy between
529: two approaches indicates that the SHC is no longer an unambiguous quantity
530: for measuring the spin transport even in the pure case. In the following we
531: further explore the disorder case.
532:
533: \subsection{Disorder effect}
534:
535: \label{sec:nr}
536:
537: Now we consider the disorders effect.
538: %%The on-site disorder strength is denoted by $\epsilon _{i}$ in the
539: %%Hamiltonian (\ref{H}), which is randomly distributed within $[-W/2,W/2]$.
540: Fig. \ref{fig:level} shows the results for a chosen disorder configuration
541: with $W=0.05$ in a $32\times 32$ lattice. A careful examination of the
542: energy levels reveals that each eigenstate goes up and down, making several
543: large angle turns due to backward scattering. These energy levels never
544: cross with each other, except for at $\Phi=0$ and $\pi$, where two levels
545: become exactly degenerate (Kramers degeneracy). Then if one follows any
546: Kramers degenerated pair of states starting from $\Phi=0$ to $\Phi=2\pi$,
547: one will always go back to exactly the same pair of states. Namely, after
548: the adiabatic insertion of one flux quantum, all states evolve exactly back
549: to the starting states.
550:
551: \begin{figure}[t]
552: \resizebox{80mm}{!}{ \includegraphics{level.eps} } \centering
553: \caption{The evolution of the single electron states with$\ \Phi $ for the
554: 2DHG on $32\times 32$ lattice for a disordered sample with $W=0.05$. }
555: \label{fig:level}
556: \end{figure}
557:
558: \begin{figure}[t]
559: \resizebox{80mm}{!}{ \includegraphics{scatter.eps} } \centering
560: \caption{(a) The comparison of the evolutions of two levels for clean system
561: (dashed line) and disordered system with $W = 0.05$ (dashed line). The
562: levels are crossing at clean system and anti-crossing at disorder system.
563: The $|i_1 \rangle$ and $|i_2 \rangle$ are the initial states while the $|f_1
564: \rangle$ and $|f_2 \rangle$ are the final states, see the text for details.
565: (b) The process of a state $|i_1 \rangle$ scattered to $|f_2 \rangle$ by the
566: disorder. Such a process is leading to the anti-crossing as shown in (a). }
567: \label{fig:scattering}
568: \end{figure}
569:
570: For comparison with the $W=0$ case, in Fig. \ref{fig:scattering}(a) we
571: combine an enlarged plot of two adjacent energy levels in Fig.\ref%
572: {fig:scattering} together with the pure case, where the solid lines are for
573: the disordered case and the dashed lines are for the clean case. In the pure
574: case, an electron at the state $|i_{1}\rangle $ ($|i_{2}\rangle $) will
575: finally evolve to $|f_{1}\rangle $ ($|f_{2}\rangle $) with increasing flux
576: due to the level crossing. However, in contrast to the level crossing in the
577: pure system, two energy levels, $|i_{1}\rangle ,|i_{2}\rangle ,$ in the
578: disordered system generally show a level repulsion as one increases the flux
579: $\Phi ,$ to evolve into the final states of $|f_{2}\rangle $ and $%
580: |f_{1}\rangle ,$ respectively [see Fig. \ref{fig:scattering}(a)]. The level
581: repulsion in the disordered case represents a scattering process shown in
582: Fig. \ref{fig:scattering}(b).
583:
584: % which can be labelled by three quantum numbers:
585: % $n,\lambda ,P_{x}$ with the wavefunction given by $e^{iP_{x}}u_{n\lambda
586: % k_{x}}(y)$ and the energy $E_{n\lambda }(P_{x})$ forming 1D bands with
587: % $n=1,2,\cdots ,2L$ (for the open BC along the $y$-axis) and the
588: % \textquotedblleft helicity\textquotedblright\ $\lambda =\pm 1$ (in replacement
589: % of the spin index) as shown by the solid curves in the inset of Fig.
590: % \ref{fig:energy_spectrum}. $P_{x} = 2 k \pi/L + \theta / L$ is the momentum
591: % along $x$ direction which is the lattice momentum at $\theta = 0$ and
592: % increases for $k > 0$ and decreases for $k < 0$ with increasing the flux,
593: % shown as the arrows in fig. \ref{fig:energy_spectrum}. The helicity $\lambda$
594: % is unchanged under time reversal(T) transformation and tracks the sign of the
595: % contribution to the spin current.
596:
597: % \begin{figure}[t]
598: % \resizebox{80mm}{!}{ \includegraphics{band_o.eps} } \centering
599: % \caption{The inset is the Energy spectrum of a clean sample at $\protect%
600: % \theta =0$. The main panel shows two of these bands. The two arrows
601: % illustrate the evolution of the two states with tuning the parameter $%
602: % \protect\theta $, which start from the beginning of the arrows at
603: % $\protect%
604: % \theta =0$ and arrive at the end of the arrows at $\protect\theta
605: % =2\protect%
606: % \pi $(see the text for details). The dashed line indicate that the Fermi
607: % energy.}
608: % \label{fig:energy_spectrum}
609: % \end{figure}
610:
611: We have done similar calculations at different lattice sizes, such as $%
612: 8\times 8$, $16\times 16,$ and $24\times 24$. The results are all similar
613: except that the ``gaps'' characterizing the level repulsion decrease with
614: the lattice size, which are roughly proportional to $1/N$. According to the
615: discussion in Sec. \ref{sec:gelc}, the adiabatic insertion of the flux $\Phi
616: $ will not generate true spin transfer or accumulation as the ground state
617: simply evolves back after a flux quantum insertion. In fact, we have checked
618: various weak disorder strengths and always found the same generic result.
619:
620: Therefore, we conclude that there is no intrinsic spin Hall accumulation
621: (spin transfer) for the 2DHG system in the disorder case.
622: %%%similar to the case of the 2D Rashba model.\cite{kubo2}
623: This conclusion is in sharp contrast to the calculated SHC at small $W$ in
624: Sec. \ref{sec:shc}, which remains finite and close to the value of the pure
625: case when the disorder is weak. According to the discussion at the beginning
626: of Sec. \ref{sec:sa}, these two approaches, for a bulk system on a torus or
627: as an open system with two edges, are not necessarily in accordance with
628: each other, because the spin current continuity equation is broken. For
629: example, a backward scattering of an electron by an impurity shown by Fig. %
630: \ref{fig:scattering}(b) can simultaneously change the sign of its momentum
631: \emph{as well as} its spin direction due to the SOC. Then one can find that
632: the spin current is still additive (same sign) in this scattering process,
633: which contributes to the SHC accordingly, but there should be no true spin
634: transfer taking place because of the backward scattering of the spin carrier
635: -- electron which moves back in direction after the scattering.
636:
637: \subsection{Edge state}
638:
639: \label{sec:dis}
640:
641: \begin{figure}[t]
642: \resizebox{70mm}{!}{ \includegraphics{mag_level.eps} } \centering
643: \caption{The evolution of the single electron states with $\Phi$ in the
644: system with an external magnetic field on a $32 \times 32$ lattice. The
645: strength of the magnetic field is $\protect\pi/8$ per plaquette.}
646: \label{fig:mag_level}
647: \end{figure}
648:
649: To further understand the condition for the presence of level crossing in
650: the disordered case, we apply an external magnetic field perpendicular to
651: the 2D sample (with a flux strength $\pi /8$ per plaquette) and repeat the
652: above calculation. We find that most of the ``gaps'' still behave as $1/N$ ,
653: similar to those in the non-magnetic-field case. These are bulk states. But
654: we can also identify a few levels whose level-repulsion ``gaps'' reduce
655: exponentially with $N$ with the increase of the sample size. For example,
656: such kind of ``gap'' at $32\times 32$ lattice size is about $1/30$ of the
657: one at $16\times 16$ lattice size, an exponential dependence on system
658: length.
659: %instead of reducing only about $1/4$ for a level-repulsion ``gap''.
660: We show some of these ``level crossings'' in Fig. \ref{fig:mag_level} with
661: the same parameters as in Fig. \ref{fig:level}. The ``gaps'' in Fig. \ref%
662: {fig:mag_level} are much smaller than the ones in Fig. \ref{fig:level} and
663: hard to distinguish with the naked eye. %A careful examination clearly
664: The exponential decaying form of the gaps indicate the existence of level
665: crossings in the thermodynamic limit in the presence of a perpendicular
666: magnetic field, which corresponds to an IQHE. Then a net charge (spin) is
667: transported from one edge to the other correspondingly.\cite{kubo2}
668:
669: We have also systematically compared the spatial distribution of electron
670: density $\rho _{i}=\langle c_{i}^{\dagger }c_{i}\rangle $ of the states in
671: the 2DHG system with and without the presence of a perpendicular magnetic
672: field. %%%Fig. \ref{fig:dis} shows three
673: %%representative examples, with the top one for a state in the non magnetic
674: %%case, the middle one for a state which anticrosses with others in the
675: %%magnetic case, and the bottom one as a state which crosses with some other
676: %%states in the magnetic case. The top one shows an extended state for the SOC
677: %system. The middle one is a localized state due to the Anderson
678: %%localization. And the bottom one is an edge state which is localized at the
679: %%edge along the $\hat{y}$-direction but remains extended along the $\hat{x}$%
680: %%-direction.
681: It is found that the level crossing only occurs for the edge states.
682: Physically this is because in the thermodynamic limit the distance between
683: two edge states in the opposite sides of the sample goes to infinity, which
684: makes the scattering between them vanishes exponentially. Thus the absence
685: of level crossing (and the intrinsic spin Hall transfer) in the 2DHG system
686: can be attributed to the fact that there do not exist any edge states in
687: such a system in the absence of perpendicular external magnetic fields.
688:
689: \section{Conclusions and Discussions}
690:
691: \label{sec:dac}
692:
693: In the present work, we have numerically calculated the SHC through the Kubo
694: formula at different lattice sizes and impurity strengths. A reasonable
695: finite-size scaling analysis shows that the bulk SHC remains finite in the
696: thermodynamic limit in the disordered 2DHG up to a critical disorder
697: strength $W_{c}=5$. This result is consistent with the perturbative
698: calculation based on the vertex correction.\cite{2dhg}
699:
700: However, we have also found the strong evidence that the net spin Hall
701: accumulation or spin transfer disappears in the disordered case based on a
702: numerically performed Laughlin's gauge experiment.
703: % to probe the spin accumulation at the open boundaries of a
704: %belt-shaped sample upon an adiabatical insertion of a flux quantum.
705: We have interpreted the conflicting results of the SHC and direct spin
706: accumulation calculations as due to the fact that the spin is not conserved
707: in an SOC system and there is no longer a conservation law to govern the
708: spin current. Therefore the SHC is no more a unique quantity to characterize
709: the spin transport, where electron spin can relax and vanish in the bulk
710: without being transported across the sample by a transverse electric field.
711:
712: By comparing with the case in the presence of an applied perpendicular
713: magnetic field, which is in the IQHE regime, we have shown that the absence
714: of an intrinsic spin transfer (accumulation) can be related to that there
715: are no edge states in the 2DHG system, in contrast to the IQHE case.
716:
717: We further address the validity of the above adiabatic argument in the
718: thermodynamic limit. It is noted that since an intrinsic spin transfer
719: (accumulation) is always \emph{absent} in a finite size system according to
720: the above argument, it is very hard to imagine that such an effect could be
721: restored as an \emph{intrinsic} one by the Landau-Zener tunnelling \cite%
722: {kubo1} between two anticrossing levels in the thermodynamic limit. Of
723: course, the Landau-Zener tunnelling between the level repulsion gaps in the
724: large sample size limit (the level repulsion gap vanishes in a $1/N$
725: fashion, with the \emph{same }rate as the average level spacing vanishing in
726: this limit) can still contribute to \emph{dissipative} transport currents,
727: and similar to the conventional charge transport, one may expect a
728: dissipative spin transport term appearing, besides the null \emph{intrinsic}
729: spin transfer. Here the dissipative spin transport is qualitatively
730: different from the true level crossing with edge states. At a finite $W$, in
731: the longitudinal channel, the so-called \textquotedblleft Thouless
732: conductance\textquotedblright\ term can be obtained based on the sensitivity
733: of the energy level to the threading flux, giving rise to a finite $\sigma
734: _{xx}$,\cite{ando} which is subject to the detailed scattering mechanism as
735: contributed by the levels near the Fermi energy. In the present SOC system,
736: a dissipative spin Hall transfer related to this longitudinal conductance is
737: indeed possible, which similarly involves only the states near the Fermi
738: energy. We note that such a dissipative term may also be important for the
739: spin-polarization measured experimentally\cite{exp1,exp2}.
740:
741: The vanishing spin transfer revealed through Laughlin's Gedanken gauge
742: experiment for the present 2DHG and 2D electron Rashba model\cite{kubo2}
743: under an adiabatic procedure may be a generic behavior for a metallic system.%
744: \cite{haldane} It would be interesting to directly probe this property for
745: more SOC systems, including the 3D Luttinger model, although numerically it
746: is more challenging for the latter as it is much harder to do the
747: finite-size scaling for the level repulsion gaps in a 3D system. Finally, it
748: is interesting to note that several recent papers have studied topological
749: SHE in insulating electron systems with SOC.\cite{tshe} In these models, it
750: becomes clear that bulk intrinsic SHC can result in spin transfer and
751: accumulation, with the presence of spin-polarized edge states, which is
752: associated with a topological invariant. This is in agreement with our
753: general conclusion that the intrinsic spin transfer relies on the existence
754: of current-carrying edge state.
755:
756: \begin{acknowledgments}
757: \textbf{Acknowledgment:} We would like to thank F.D.M. Haldane, S.C. Zhang,
758: and M.W. Wu for insightful discussions. This work is supported by NSFC
759: grants 10374058 and 90403016 (ZYW), ACS-PRF 41752-AC10, the NSF
760: grant/DMR-0307170 (DNS). The computation of this project was partially
761: performed on the HP-SC45 Sigma-X parallel computer of ITP and ICTS, CAS.
762: \end{acknowledgments}
763:
764: \begin{thebibliography}{99}
765: \bibitem{MNZ} S. Murakami, N. Nagaosa, and S.C. Zhang, Science \textbf{301},
766: 1348 (2003).
767:
768: \bibitem{Sinova1} J. Sinova \emph{et al.}, Phys. Rev. Lett. \textbf{92}
769: 126603 (2004).
770:
771: \bibitem{vertex1} J. Inoue, G. E. W. Bauer, and L. W. Molenkamp, Phys. Rev.
772: B, \textbf{70}, 041303(R) (2004).
773:
774: \bibitem{vertex2} E.G. Mishchenko, A.V. Shytov, and B.I. Halperin, Phys.
775: Rev. Lett. \textbf{93, }226602 (2004).
776:
777: \bibitem{vertex_luttinger} S. Murakami, Phys. Rev. B \textbf{69}, 241202(R)
778: (2004).
779:
780: \bibitem{kubo2} D.N. Sheng, L.Sheng, Z.Y. Weng, and F.D.M. Haldane,
781: cond-mat/0504218.
782:
783: \bibitem{luttinger} W.Q. Chen, Z.Y. Weng, and D.N. Sheng, to appear in Phys.
784: Rev. Lett., 2005; cond-mat/0502570.
785:
786: \bibitem{kubo1} K. Nomura, Jairo Sinova, N. A. Sinitsyn, and A. H.
787: MacDonald, cond-mat/0506189; K. Nomura, J. Sinova, T. Jungwirth, Q. Niu and
788: A. H. MacDonald, Phys. Rev. B \textbf{71}, 041304(R) (2005).
789:
790: \bibitem{exp1} J. Wunderlich, B. K\"{a}stner, J. Sinova, and T. Jungwirth,
791: Phys. Rev. Lett. \textbf{94}, 047204 (2005)
792:
793: \bibitem{exp2} Y. K. Kato, R. C. Myers, A. C. Gossard, and D. D. Awschalom,
794: Science 306, 1910 (2004).
795:
796: \bibitem{eshe} Hans-Andreas Engel, Bertrand I. Halperin, and Emmanuel I.
797: Rashba, cond-mat/0505535
798:
799: \bibitem{2dhg} B. A. Bernevig, and S.C. Zhang, cond-mat/0411457
800:
801: \bibitem{ngaas} B. A. Bernevig, and S.C. Zhang, cond-mat/0412550.
802:
803: \bibitem{ne2dhg} S. Y. Liu, and X. L. Lei, cond-mat/0503352
804:
805: \bibitem{lb2dhg} M. W. Wu, and J. Zhou, cond-mat/0503616
806:
807: \bibitem{LB1} L. Sheng, D.N. Sheng, and C.S. Ting, Phys. Rev. Lett. \textbf{%
808: \ 94}, 016602 (2005); B. K. Nikoli\'{e}, L. P. Z\^{a}rbo and S. Souma,
809: cond-mat/0408693.
810:
811: \bibitem{LB3} E. M. Hankiewicz, L.W. Molenkamp, T. Jungwirth, and J. Sinova,
812: Phys. Rev. B \textbf{70}, 241301 (2004).
813:
814: \bibitem{kubo0} Q. Niu, D.~J. Thouless and Y.~S. Wu, Phys. Rev. B \textbf{31}%
815: , 3372 (1985).
816:
817: \bibitem{haug} see, \emph{Quantum Theory of the Optical and Electronic
818: Properties of Semiconductors}, H. Haug and S. W. Koch, (World Scientific,
819: 1990).
820:
821: \bibitem{loss} J. Schliemann and D. Loss, Phys. Rev. B \textbf{51}, 085308
822: (2005).
823:
824: \bibitem{laughlin} R. B. Laughlin, Phys. Rev. B \textbf{23}, 5632 (1981).
825:
826: \bibitem{halperin} B. I. Halperin, Phys. Rev. B \textbf{25}, 2185 (1982).
827:
828: \bibitem{ando} T. Ando, Phys. Rev. B \textbf{40}, 5325 (1989).
829:
830: \bibitem{haldane} F. D. M. Haldane, Phys. Rev. Lett. \textbf{93}, 206602
831: (2004).
832:
833: \bibitem{tshe} L. Kane and E. J. Mele, cond-mat/0411737; 0506581; L. Sheng,
834: D. N. Sheng, C. S. Ting, F. D. M. Haldane, cond-mat/0506589; B. A. Bernevig
835: and S. C. Zhang, cond-mat/0504147; X.-L. Qi, Y.-S. Wu and S. C. Zhang,
836: cond-mat/0505308.
837: \end{thebibliography}
838:
839: \end{document}
840: