1:
2: % ------------------------------- Summary ------------------------------
3: %
4: %`` Equilibrium spin current through the tunnelling junctions''
5: %
6: %
7: %
8: % Author (s): J. wang, K.S. Chan
9: %
10: %
11: % Last Update : 17/11/05
12: %
13: %
14: %
15: % Remarks :
16: %
17: % Documentstyle is [preprint,aps]{revtex} (PRB, PRL)
18: % Galley proof style is obtained by taking out the option >preprint<
19: %
20: % ---------------------------------- End -------------------------------
21: %
22: %
23: %
24: \tolerance = 10000
25: \documentstyle[aps,twocolumn,prb]{revtex}
26: %\documentstyle[preprint,aps,prb,epsfig,graphicx]{revtex}
27: \begin{document}
28:
29: \draft
30:
31: \title{ Equilibrium spin current through the tunnelling junctions}
32:
33: \author{J. Wang and K.S. Chan}
34: \address{$^{1}$Department of Physics and Materials Science, City University of Hong Kong, Tat Chee
35: Avenue, Kowloon, Hong Kong, P.R. China}
36:
37: \date{\today}
38: \maketitle
39:
40: \begin{abstract}
41: We study equilibrium pure spin current through tunnelling
42: junctions at zero bias. The two leads of the junctions connected
43: via a thin insulator barrier, can be either a ferromagnetic metal
44: (FM) or a nonmagnetic high-mobility two-dimensional electron gas
45: (2DEG) with Rashba spin orbital interaction (RSOI) or Dresselhaus
46: spin orbital interaction (DSOI). As a lead of a tunnelling
47: junction, the isotropic RSOI or DSOI in 2DEG can give rise to an
48: average effective planar magnetic field orthogonal or parallel to
49: the current direction. It is found by the linear response theory
50: that equilibrium spin current $\vec{J}$ can flow in the following
51: three junctions, 2DEG/2DEG, 2DEG/FM, and FM/FM junctions, as a
52: result of the exchange coupling between the magnetic moments,
53: $\vec{h}_{l}$ and $\vec{h}_{r}$, in the two electrodes of the
54: junction, i.e., $\vec{J}\sim\vec{h}_{l}\times\vec{h}_{r}$. An
55: important distinction between the FM and 2DEG with RSOI (DSOI)
56: lead is that in a strict one-dimensional case RSOI (DSOI) cannot
57: lead to equilibrium spin current in the junction since the two
58: spin bands are not spin-polarized as in a FM lead where Zeeman
59: spin splitting occurs.
60:
61: \end{abstract}
62:
63: \pacs{Pacs numbers: 73.23.-b, 72.25.Dc, 71.70. Ej}
64: \section{introduction}
65: Spin related transport in magnetic microstructures has drawn
66: considerable interests in research community for the purpose of
67: spintronics.\cite{1,2,3} In order to manipulate the spin degree of
68: freedom of an electron, one needs first to build a source of
69: spin-polarized electrons and efficiently injects spins into a
70: non-polarized medium. The ferromagnetic metal (FM) is an ideal
71: spin source for spin injection into semiconductor devices;
72: however, the injection efficiency experimentally measured is
73: extremely low due to the much larger conductivity of FM than that
74: of semiconductor.\cite{4,5,6} This issue has not yet been resolved
75: although it is fundamentally important. As an alternative solution
76: to the spin injection problem, the pure spin current was proposed
77: and has attracted much attention in recent years. The pure spin
78: current is composed of equal spin down and spin up electron
79: currents flowing along opposite directions without any charge
80: current.\cite{7} Several different methods have been proposed to
81: create pure spin current. Based on the spin pumping
82: principle,\cite{8,9} an alternating or inhomogeneous magnetic
83: field as a spin-pumping force could result in a pure spin current.
84: Experimentally, Stevens \emph{et al.}\cite{10} and Hubner \emph{et
85: al.}\cite{11} have independently realized the pure spin current by
86: using the quantum interference of two-color laser fields with
87: cross-linear polarization in ZnSe and GaAs semiconductors. The
88: transport of spin current by magnons have been theoretically
89: studied by several groups,\cite{12,13,14} e.g., Meier \emph{et
90: al.}\cite{12} demonstrated that by using a finite length spin
91: chain between magnetic reservoirs, pure spin current can be
92: generated without the transport of electrical charge.
93:
94: The equilibrium spin current (ESC) in a magnetic or nonmagnetic
95: system is also very attractive. The newly unearthed spin Hall
96: current, discovered by Murakami \emph{et al.}\cite{15} in
97: $p$-doped semiconductors and Sinova \emph{et al.}\cite{16} in
98: two-dimensional electron systems (2DEG), is actually a
99: dissipationless spin current, which is a transverse response to a
100: longitudinal external electric field $E_{x}$. The spin Hall
101: current in a 2DEG is generated by the Rashba spin orbital
102: interaction (RSOI). In a single RSOI system the spin state of an
103: electron is dependent on its momentum direction, so that at
104: equilibrium the spin current could flow in the system without
105: bias.\cite{17} As long as a system has noncollinear magnetic
106: order, a ESC could exist in the system.\cite{18} For instance,
107: K{\"{o}}{nig} \emph{et al.}\cite{19} found a dissipationless spin
108: current in a thin film ferromagnet, in which a spiral magnetic
109: order exists and the spin phase coherence can affect the
110: electronic transport properties.
111:
112:
113: For the FM/FM tunnelling junction, a spontaneous ESC has been
114: verified\cite{20,21,22,23} flowing across the interface when the
115: two magnetic moments $\vec{h}_{l}$ and $\vec{h}_{r}$ in the two
116: leads of the junction are not collinear, and their exchange
117: interaction determines the ESC $\vec{J}_s$,
118: $\vec{J}_{s}\sim\vec{h}_{l}\times\vec{h}_{r}$. This indicates
119: non-collinear magnetizations in the two leads can lead to spin
120: flip and consequently a ESC through the junction. Nogueira
121: \emph{et al.}\cite{22} and Lee \emph{et al.}\cite{23} referred to
122: the ESC through the FM/FM junction as the spin supercurrent as an
123: analogy to the Josephson effect in superconductor junctions. The
124: spin current in a FM/FM junction was also shown to result in
125: magnetic moment reversal in one of the FM leads of the
126: junction.\cite{24}
127:
128: In 2DEG with RSOI or Dresselhaus spin orbital interaction (DSOI),
129: there is a pseudomagnetic field which could lead to ESC in the
130: non-magnetic system. This pseudomagnetic field is different from a
131: real magnetic field as it is dependent on the direction of
132: electron momentum and keeps the system's time reversal symmetry. A
133: natural question thus arises whether a spin supercurrent (or ESC)
134: could flow through a \emph{nonmagnetic} tunnelling junction in
135: which the electrodes are 2DEGs with either RSOI or DSOI. A recent
136: paper\cite{25} by B{\o}rkje and Sudb{\o} has been dedicated to
137: this issue and authors obtained a ESC in a 2DEG junction with RSOI
138: by rotating one of the 2DEG electrodes along the current direction
139: to change the direction of the pseudomagnetic field, so that the
140: product of two pseudomagnetic fields in two leads is nonzero,
141: meanwhile, the absence of ESC through the 2DEG/FM tunnelling
142: junction was also predicted. In this paper, we will study
143: systematically the ESC in three different junctions: 2DEG/2DEG,
144: 2DEG/FM, and FM/FM junctions, using the linear response theory. It
145: is found that ESC is present in all three junctions resulting from
146: the exchange coupling between the two magnetic moments in the two
147: electrodes of the junctions. Non-magnetic 2DEG with RSOI or DSOI
148: as an electrode in a tunnelling junction can lead to a ESC,
149: because only half of the electrons (those with $k_{x}>0$, if the
150: 2DEG is on the left of the junction) in an electrode contribute to
151: the tunnelling current so that the isotropic RSOI or DSOI can give
152: rise to an average planar effective magnetic field orthogonal or
153: parallel to the current direction. We also found that in a strict
154: one-dimensional (1D) case RSOI or DSOI cannot result in a ESC in
155: the junction because the 1D density of states is not
156: spin-polarized, which is very different from that of a 1D FM lead.
157:
158:
159: This paper is organized in the following way. In the second part,
160: we give a general formula of the ESC in a tunnelling junction
161: derived using the linear response theory. In the third part, we
162: analyze in detail the ESC in three junctions: 2DEG/2DEG, 2DEG/FM,
163: and FM/FM junctions. A conclusion was drawn in the last section.
164:
165: \section{formula}
166: We start our discussion with the derivation of a general formula
167: of ESC through a junction, consisting of two leads of either 2DEG
168: with RSOI (DSOI) or FM metal and an insulator barrier between
169: them. The Hamiltonian of such a tunnelling system is given by
170: \begin{mathletters}
171: \begin{eqnarray}
172: {\cal H}={\cal H}_{L}(C^{\dagger}_{k\sigma};C_{k\sigma})+{\cal
173: H}_{R}(C^{\dagger}_{q\sigma};C_{q\sigma})+{\cal H}_{T}
174: \end{eqnarray}
175: \begin{eqnarray}
176: {\cal
177: H}_{T}=\sum_{kq\sigma}(t_{kq}C_{k\sigma}^{\dagger}C_{q\sigma}+\emph{h.c.}),
178: \end{eqnarray}
179: \end{mathletters}
180: where ${\cal H}_{L(R)}$ is the Hamiltonian of a non-interacting
181: free-electron gas in the left (right) lead, whose specific form
182: will be present in next section; ${\cal H}_{T}$ is the tunnelling
183: part connecting two leads. $C_{k\sigma}^{\dagger}(C_{k\sigma})$
184: and $C_{q\sigma}^{\dagger}(C_{q\sigma})$ are the fermion creation
185: (annihilation) operators of the left and right leads,
186: respectively. The hopping matrix $t_{kq}$ is independent of spin,
187: so that there is no spin-flip when electrons tunnel through the
188: junction. In the following discussion the quantum number $k$ and
189: $q$ can also denote implicitly the left and right lead.
190:
191: We focus on the tunnelling ESC and the bulk ESC in the RSOI (DSOI)
192: lead is not considered here because it does not contribute to the
193: tunnelling current. The total spin in the left lead is
194: ${\mathbf{S}}_{L}=(\hbar/2)\sum_{k\alpha\beta}C_{k\alpha}^{\dagger}{
195: {{\sigma}}_{\alpha\beta}}C_{k\beta}$ where $\alpha$ and $\beta$
196: are spin indices, and ${{\sigma}}$ is the Pauli spin operator. The
197: time evolution of ${\mathbf{S}}_{L}$ in the Heisenberg picture is
198: given by
199: $\dot{\mathbf{S}}_{L}=(1/i\hbar)[{\mathbf{S}}_{L},H_{T}]$, and
200: thus the operator of spin current
201: $\hat{{\mathbf{J}}}_{s}=\partial{{\mathbf{S}}_{L}}/\partial{t}$
202: reads
203: \begin{equation}
204: {\hat{\mathbf{J}}}_{s}=\frac{-i}{2}\sum_{kq\alpha\beta}{{\sigma}}_{\alpha\beta}
205: (t_{kq}C_{k\alpha}^{\dagger}C_{q\beta}-t_{kq}^{*}C_{q\alpha}^{\dagger}C_{k\beta}).
206: \end{equation}
207: Here the coupling ${\cal H}_{T}$ is assumed to be very weak and we
208: only need to calculate the spin current to the lowest order of the
209: coupling. Since the spin current operator $\hat{\mathbf{J}}_{s}$
210: itself is already linear in $t_{kq}$, and thus according to the
211: Kubo formula the spin current is to first order in ${\cal{H}}_{T}$
212: given by
213: \begin{mathletters}
214: \begin{eqnarray}
215: {{\mathbf{J}}}_{s}(t)=\int_{-\infty}^{t}dt^{\prime}e^{-0^{+}(t-t^{\prime})}\Pi_{R}(t,t^{\prime}),
216: \end{eqnarray}
217: \begin{eqnarray}
218: \Pi_{R}(t,t^{\prime})=-i\theta(t-t^{\prime})\langle[\hat{{\mathbf{J}}}_{s}(t),{\cal
219: H}_{T}(t^{\prime}) ]\rangle,
220: \end{eqnarray}
221: \end{mathletters}
222: where $\langle\cdots\rangle$ is the thermal statistical average on
223: two independent leads ${\cal H}_{L}+{\cal H}_{R}$ (the unperturbed
224: part of $\cal{H}$), which are assumed to be in local equilibrium.
225: $\Pi_{R}(t,t^{\prime})$ is the retarded Green's function. After
226: some straightforward algebra and keeping the average only in lead
227: ${\cal H}_{L}$ or ${\cal H}_{R}$, we could express Eq.~3 in the
228: steady state as
229: \begin{eqnarray}
230: {\mathbf{J}}_{s}=\frac{-1}{2}\int\frac{d\omega_{1}}{2\pi}\frac{d\omega_{2}}{2\pi}
231: \sum_{kq}|t_{kq}|^{2}Tr\{ \nonumber \\
232: \frac{G_{k}^{<}(\omega_{1}){{\sigma}}
233: G_{q}^{>}(\omega_{2})-G_{k}^{>}(\omega_{1}){{\sigma}}
234: G_{q}^{<}(\omega_{2})}{i(\omega_{2}-\omega_{1}-i0^{+})}
235: +\nonumber \\
236: \frac{G_{q}^{>}(\omega_{2}){{\sigma}}
237: G_{k}^{<}(\omega_{1})-G_{q}^{<}(\omega_{2}){{\sigma}}
238: G_{k}^{>}(\omega_{1})}{i(\omega_{1}-\omega_{2}-i0^{+})}
239: \},
240: \end{eqnarray}
241: where the trace $Tr$ is over the spin space, $G_{k(q)}^{<(>)}$ are
242: the Keldysh Green's functions in the left (right) lead and can be
243: readily solved for leads of free-electron model. Their
244: definitions are $G_{k(q),\alpha\beta}^{<}(t,t^{\prime})=i\langle
245: C_{k(q)\beta}^{\dagger}(t^{\prime})C_{k(q)\alpha}(t)\rangle$ and
246: $G_{k(q),\alpha\beta}^{>}(t,t^{\prime})=-i\langle
247: C_{k(q)\alpha}(t) C_{k(q)\beta}^{\dagger}(t^{\prime})\rangle$,
248: respectively. The Fourier transform of the Green's function is given
249: by $G_{k(q)}^{<(>)}(t-t^{\prime})=
250: \int\frac{d\omega}{2\pi}G_{k(q)}^{<(>)}(\omega)e^{-i\omega(t-t^{\prime})}$.
251:
252: It is noted that Eq.~4 above is a general formula of spin current
253: through a tunnelling junction and the bias could be reflected in
254: the local equilibrium Green's functions $G_{k(q)}^{<(>)}$ of the
255: two leads. For zero bias on the junction considered in this
256: article, i.e., zero charge current flow through the junction, we
257: can further simplify Eq.~4 by putting
258: $1/(x+i0^{+})={\sf{P}}(1/x)-i\pi\delta(x)$ with $\sf{P}$ denoting
259: the principle integral,
260: \begin{mathletters}
261: \begin{eqnarray}
262: {\mathbf{J}}_{s}=\frac{1}{2}\int\frac{d\omega_{1}}{2\pi}\frac{d\omega_{2}}{2\pi}
263: \sum_{kq}|t_{kq}|^{2}{\sf{P}}\frac{f(\omega_{2})-f(\omega_{1})}{i(\omega_{2}-\omega_{1})}
264: Tr\{ \nonumber \\
265: A_{k}(\omega_{1}){{\sigma}}
266: A_{q}(\omega_{2})-A_{q}(\omega_{2}){{\sigma}}A_{k}(\omega_{1}) \},
267: \end{eqnarray}
268: \begin{eqnarray}
269: A_{k(q)}(\omega)=i(G^{r}_{k(q)}(\omega)-G^{a}_{k(q)}(\omega)).
270: \end{eqnarray}
271: \end{mathletters}
272: Here $A_{k(q)}(\omega)$ is the spectral function of the left
273: (right) lead with $G^{r(a)}(\omega)$ denoting the retard
274: (advanced) Green's function and $f(\omega)$ denoting the Fermi
275: distribution function. The ESC in Eq.~5 is mainly determined by
276: the difference between two Hermitian-conjugate matrices in spin
277: space, so that the off-diagonal term in the spectral function
278: $A_{k(q)}$ is utmost important to the formation of a ESC in a
279: tunnelling junction. For FM or 2DEG with RSOI (DSOI), this
280: off-diagonal term in spin space exists when two magnetizations
281: (pseudomagnetic fields in 2DEG) are not collinear in the two
282: leads, as a result, a nonzero ESC can flow through the junction
283: according to Eq.~5. In the following section we will discuss the
284: ESC in three specific junctions: 2DEG/2DEG, 2DEG/FM, and FM/FM
285: junctions.
286:
287: \section{results and discussion}
288: \subsection{2DEG/2DEG tunnelling junction}
289: In this section we study the ESC through a 2DEG/2DEG junction with
290: spin orbital interaction. First we review briefly the
291: characteristics of a 2DEG with RSOI, which stems from the
292: structure inversion asymmetry of the confining potential of a
293: quantum well.\cite{26} The magnitude of the RSOI strength could be
294: modulated by a vertical electric field $E\vec{z}$ ($\vec{z}$ is
295: the unit vector in the $z$ direction in Fig.~1) applied on the
296: 2DEG. The Hamiltonian of the 2DEG with RSOI is
297: \begin{mathletters}
298: \begin{eqnarray}
299: {\cal{H}}=\frac{\mathbf{p}^{2}}{2m_s}+{\cal{H}}_{rsoi},
300: \end{eqnarray}
301: \begin{eqnarray}
302: {\cal{H}}_{rsoi}=\frac{\lambda}{\hbar}(\sigma_y p_x-\sigma_x p_y),
303: \end{eqnarray}
304: \end{mathletters}
305: where $m_s$ is the effective mass of electrons in the 2DEG, $p_x$
306: and $p_y$ are the two components of the momentum operator
307: $\textbf{p}$, and $\lambda$ is the coupling constant and was
308: estimated to be $\sim 3.9\times 10^{-12}$ eV m in an InGaAs/InAlAs
309: heterostructure.\cite{27} $\sigma_x$ and $\sigma_y$ are the Pauli
310: matrices, and the $z$-axis is taken as the spin quantum axis.
311:
312: The electron eigenfunctions of Eq.~(6) are
313: $\psi_{\pm}({\mathbf{r}})=(1/\sqrt{2})e^{i{\mathbf{k}\cdot
314: r}}\left( \begin{array}{c}
315: \mp ie^{-i\theta_{r}} \\
316: 1 \\
317: \end{array}\right)$
318: with $\tan\theta_{r}=k_{y}/k_{x}$, and the corresponding
319: eigenvalues are $\varepsilon_{\pm}=\hbar^{2}k^2/2m_{s}\pm\lambda
320: k$. This wavefunction of electron reminds us that the spin orbital
321: interaction can give rise to a magnetic field parallel to the
322: plane, which is dependent on the direction of the electron
323: momentum. From Eq.~6b, the effective (pseudo) magnetic field is
324: $\vec{B}\sim \vec{z}\times \mathbf{p}$. Different from a real
325: magnetic field, the RSOI keeps the system's isotropy and
326: time-reversal symmetry $\mathbf{B(k)}=\mathbf{-B(-k)}$, and the
327: 2DEG does not exhibit a real spin splitting like in the Zeeman
328: effect. This will result in a distinction between the ESC in a
329: 2DEG junction and the ESC in a FM junction.
330:
331: Although the pseudomagnetic field $B$ in 2DEG is isotropic, only
332: half of the electrons (electrons with $k_{x}>0$ on the left and
333: $x$ is the current direction in Fig.~1) in the 2DEG contribute to
334: the tunnelling current, as a result, the average pseudomagnetic
335: field of the tunnelling electrons is nonzero and along the $y$
336: direction ($B\vec{y}$). If we could change the direction of the
337: pseudomagentic field in one of the 2DEG, a ESC would flow through
338: a 2DEG/2DEG junction. A celebrated example is the RSOI/DSOI
339: junction, where one 2DEG lead possesses RSOI while the other has
340: DSOI. By exchanging the spin axis
341: $\sigma_{x}\longleftrightarrow\sigma_{y}$, RSOI can be transformed
342: into DSOI as
343: \begin{equation}
344: {\cal{H}}_{dsoi}=\frac{\lambda}{\hbar}(\sigma_x p_x-\sigma_y p_y).
345: \end{equation}
346: This coupling is due to the lack of the bulk inversion symmetry in
347: the material.\cite{28} Thus in the RSOI/DSOI tunnelling junction,
348: the average pseudomagnetic field in the DSOI lead is along the $x$
349: direction, which is different from that in the RSOI lead, so that
350: a nonzero ESC $\sim B_{rsoi}\vec{y}\times B_{dsoi}\vec{x}$ exists
351: as shown below in Eq.~11. In Ref.~25, B{\o}rkje and Sudb{\o}
352: attempted to rotate one 2DEG lead along the current direction
353: ($x$-direction in Fig.~1) so as to rotate the orientation of the
354: pseudomagnetic field. However this method may not create a ESC
355: through the RSOI/RSOI junction as discussed below. The rotation of
356: the average psuedomagnetic field could be also obtained by
357: rotating physically a quasi-one-dimensional 2DEG along the
358: $z$-axis by an azimuthal angle $\alpha$, which is schematically
359: shown in Fig.~1. Since the pseudomagnetic field $B\sim
360: \vec{z}\times \mathbf{p}$ is dependent on the direction of the
361: electron momentum, an oblique incident electron will feel
362: pseudomagentic fields with different orientations when it tunnels
363: through a RSOI/RSOI junction where the energy band edges of the
364: two 2DEG leads are different. This change in psuedomagnetic field
365: direction could lead to a ESC in the junction. The energy band
366: shift in one 2DEG lead may be obtained by applied a gate voltage
367: or mechanical strain to the lead.
368:
369: To understand the relation of a rotated pseudomagnetic field and
370: ESC, we consider a rotation of the spin $x-y$ axes by an angle
371: $\alpha$ with respect to the spatial $x-y$ axes. The rotated angle
372: $\alpha$ of the pseudomagnetic field could be
373: incorporated into the Pauli matrices $\sigma_x(\alpha)=\left(%
374: \begin{array}{cc}
375: 0 & e^{i\alpha} \\
376: e^{-i\alpha} & 0 \\
377: \end{array}%
378: \right)$ and $\sigma_y(\alpha)=\left(%
379: \begin{array}{cc}
380: 0 & -ie^{i\alpha} \\
381: ie^{-i\alpha} & 0 \\
382: \end{array}%
383: \right)$. Here $\alpha$ is macroscopic in that it applies to all
384: the electrons and may denote the rotation angle of the average
385: pseudomagnetic field.
386:
387: In second quantization formalism, Eq.~6 can be rewritten as
388: \begin{equation}
389: {\cal{H}}=\sum_{k}\left(%
390: \begin{array}{cc}
391: C_{k\uparrow}^{\dagger} & C_{k\downarrow}^{\dagger} \\
392: \end{array}%
393: \right)
394: \left(%
395: \begin{array}{cc}
396: \varepsilon_{k} & -i\lambda ke^{i\theta} \\
397: i\lambda ke^{-i\theta} & \varepsilon_{k} \\
398: \end{array}%
399: \right)
400: \left(%
401: \begin{array}{c}
402: C_{k\uparrow} \\
403: C_{k\downarrow} \\
404: \end{array}%
405: \right),
406: \end{equation}
407: where $\varepsilon_{k}=\hbar^2k^2/2m_{s}-\mu$ with $\mu$ the
408: chemical potential, and $e^{i\theta}=e^{i(\alpha-\theta_{r})}$.
409: With $\theta_{r}$ replaced by $\theta_{d}$
410: ($\tan\theta_{d}=k_{x}/k_{y}$) as well as $\lambda$ by $-\lambda$,
411: this Hamiltonian can represent a free 2DEG with DSOI in Eq.~7. The
412: retard (advanced) Green's function of this free electron model
413: with RSOI is easily given by
414: \begin{equation}
415: G^{r(a)}(\omega)=\left(%
416: \begin{array}{cc}
417: \omega\pm i0^{+}-\varepsilon_{k}& i\lambda ke^{i\theta} \\
418: -i\lambda ke^{-i\theta} & \omega\pm i0^{+}-\varepsilon_{k} \\
419: \end{array}%
420: \right)^{-1}.
421: \end{equation}
422: Before we present the final expression of ESC of 2DEG/2DEG junction by
423: substituting $G^{r(a)}$ above into Eq.~5, we first focus on the
424: spectral function $A_{k}(\omega)$ of 2DEG with RSOI
425: \begin{equation}
426: A_{k}=\left(%
427: \begin{array}{cc}
428: \pi [\delta_{+}+\delta_{-}] & -i\pi [\delta_{-}-\delta_{+}]e^{i\theta} \\
429: i\pi [\delta_{-}-\delta_{+}]e^{-i\theta} & \pi [\delta_{+}+\delta_{-}] \\
430: \end{array}%
431: \right),
432: \end{equation}
433: where $\delta_{+}=\delta(\omega-\varepsilon_k+\lambda k)$ and
434: $\delta_{-}=\delta(\omega-\varepsilon_k-\lambda k)$ with $\delta$
435: denoting the delta function. In the formula of ESC (Eq.~5), the
436: summation term $\sum_{k}A_{k}$ exists in the case of weak
437: connection, i.e., $t_{kq}=t$ independent of energy. In a
438: \emph{strict one-dimensional} case, the off-diagonal terms of this
439: summation vanish $\sum_{k}A_{k}(12)=\sum_{k}A_{k}(21)=0$; thus,
440: there is no ESC through the junction as discussed earlier (Eq.~5).
441: This result stems from the fact that the 1D density of states of
442: the two spin bands of an electron gas with RSOI or DSOI is not
443: spin-polarized, which is very different from the FM case. In the
444: 1D case the density of states is inversely proportional to the
445: electron velocity, and the velocities of electrons in the two
446: spin bands with the same energy under RSOI or DSOI are identical
447: so that the summation of the off-diagonal terms in Eq.~10 over $k$
448: vanishes. In other words, the perpendicular incidence of electrons
449: from one RSOI (DSOI) lead into another RSOI (DSOI) lead will
450: conserve their spins whereas this is not the case for the
451: obliquely incident electrons.\cite{29} Therefore, the rotation of
452: a 2DEG along the current direction (the $x$-axis in Fig.~1) in the
453: 2DEG/2DEG tunnelling junction will give rise to two 2DEG leads not
454: in the same plane, and if the middle insulator barrier is very
455: thin, the electron tunnelling in such a junction may be a 1D
456: transport, i.e., only electrons with velocity along the
457: $x$-direction could tunnel through the junction due to the
458: momentum mismatch. Thus the ESC may vanish in this scheme.
459:
460:
461: After straightforward algebra, the ESC for the two-dimensional
462: RSOI/DSOI junction (two 2DEGs in the same plane) is given by
463: \begin{mathletters}
464: \begin{eqnarray}
465: J_{s}^x=0,
466: \end{eqnarray}
467: \begin{eqnarray}
468: J_{s}^y=0,
469: \end{eqnarray}
470: and
471: \begin{eqnarray}
472: J_{s}^{z}=\frac{\hbar}{2\pi
473: e^2}G_{N}E_{r}f(-E_{r})\sin(\alpha_L-\alpha_R-\pi/2).
474: \end{eqnarray}
475: \end{mathletters}
476: Here $\alpha_{L(R)}$ is the rotation angle of the spin axes of the
477: left (right) lead that changes the direction of the pseudomagnetic
478: field. $E_{r}=\hbar^{2}k_{r}^{2}/2m_{s}$ is the Rashba energy with
479: $k_{r}=\lambda m_{s}/\hbar^{2}$, $f(-E_{r})$ is the Fermi
480: distribution function.
481: $G_N=\frac{e^{2}}{\hbar}2\pi|t|^{2}\rho_{L}\rho_{R}$ is the
482: conductance of the normal 2DEG junction without spin-orbital
483: interaction, $\rho_{L(R)}$ is a constant density of state of 2DEG.
484: $\pi/2$ phase difference in Eq.~11c comes from the exchange of
485: spin axes when DSOI is compared with RSOI. Here we have assumed
486: for simplicity the spin-orbital coupling constants of two leads
487: are same $\lambda_{L}=\lambda_{R}=\lambda$ and $t_{kq}=t$ is
488: independent of energy.
489:
490: From Eq.~11, only ESC polarized along the $z$ direction is nonzero
491: since the two pseudomagnetic fields in RSOI/DSOI junction lie in
492: the $xy$ plane, their exchange coupling will lead to spin current
493: polarized along the $z$ direction. The phase (rotation angle)
494: $\alpha$ of the pseudomagnetic field is cannonically conjugate to
495: the spin $s_z$ so that they satisfy the commutation relation
496: $[s_z,\alpha]=-i$, which means the spin component $s_z$ is not a
497: conserved quantity as the RSOI (DSOI) can change $s_z$ by rotating
498: the spin, and ESC could hence flow in these junctions. The physics
499: is similar to the Josephson current in a superconductor
500: junction,\cite{22,23} in which the macroscopic phase $\phi$
501: resulted from the condensation of Cooper pairs is conjugate to the
502: particle number $N$ and $[\hat{N},\phi]=-i$.\cite{30} The
503: continuity equation of the spin current in 2DEG with RSOI(DSOI) is
504: $\partial s/\partial t+\Delta \cdot J_{s}=
505: S$, where $J_{s}={\bf Re}\left[ \Psi^{\dagger} \hat{\vec{v}}s\Psi\right]$
506: is the spin current density with $\hat{\vec{v}}$ being the
507: velocity operator, and $S={\bf Re}\left[
508: \Psi^{\dagger}(\vec{B}\times \sigma)\Psi\right]$ is a source term
509: from the pseudo-magnetic field by rotating the spin component
510: $s_z$.\cite{31} A similar source term appears also in the
511: continuity equation of the current of superconductor.\cite{32}
512:
513: \subsection{2DEG/FM junction}
514: We turn to discuss ESC in an 2DEG/FM junction. In the last
515: subsection it was stated that RSOI (DSOI) in 2DEG can be
516: equivalent to a pseudomagnetic field along the $y$ ($x$) direction
517: with $\alpha=0$ when electrons tunnel through the junction. When
518: one of the 2DEG leads is substituted by a FM, a ESC could also
519: flow through the 2DEG/FM junction from the point of view of the
520: magnetic field. The Hamiltonian of the FM in the right side of the
521: junction is given in a simple framework of the Stoner model by
522: \begin{equation}
523: {\cal{H}}_{fm}=\sum_{q}\left(%
524: \begin{array}{cc}
525: C_{q\uparrow}^{\dagger} & C_{q\downarrow}^{\dagger} \\
526: \end{array}%
527: \right)
528: \left(%
529: \begin{array}{cc}
530: \varepsilon_{q}+h & 0 \\
531: 0 & \varepsilon_{q}-h \\
532: \end{array}%
533: \right)
534: \left(%
535: \begin{array}{c}
536: C_{q\uparrow} \\
537: C_{q\downarrow} \\
538: \end{array}%
539: \right).
540: \end{equation}
541: Here the spin quantum axis is set to be along the local magnetic
542: moment of FM, $\varepsilon_q=\hbar^{2}q^2/2m-\mu$ and $h$ (use
543: energy as the unit here) is a molecular field in the FM. Then the
544: local retarded Green's function is
545: \begin{equation}
546: g^{r}_{\uparrow(\downarrow)}=\frac{1}{\omega+
547: i0^{+}-\varepsilon_q\mp h}.
548: \end{equation}
549: In the common spin quantum axis set along the $z$-direction for
550: the 2DEG/FM junction, i.e. the normal of the 2DEG plane, the
551: Green's functions of the FM lead above should be transformed
552: according to
553: \begin{mathletters}
554: \begin{eqnarray}
555: G^{r(a)}=U\left(%
556: \begin{array}{cc}
557: g^{r(a)}_{\uparrow} & 0 \\
558: 0 & g^{r(a)}_{\downarrow} \\
559: \end{array}%
560: \right)U^{\dagger},
561: \end{eqnarray}
562: \begin{eqnarray}
563: U=\left(%
564: \begin{array}{cc}
565: \cos\frac{\theta}{2} & -\sin\frac{\theta}{2}e^{-i\phi} \\
566: \sin\frac{\theta}{2}e^{i\phi} & \cos\frac{\theta}{2} \\
567: \end{array}%
568: \right),
569: \end{eqnarray}
570: \end{mathletters}
571: where $U$ is the unitary transformation matrix, ($\theta$,$\phi$)
572: denotes the direction of the magnetic moment of the FM lead, which
573: in terms of the unit vectors of the cartesian coordinates should
574: be written as ($\sin\theta\cos\phi
575: \vec{x}$,$\sin\theta\sin\phi\vec{y}$,$\cos\theta\vec{z}$). For the
576: specific RSOI/FM junction we consider, the rotation angle in the
577: left lead is set as $\alpha_{L}=0$. Substituting Eq.~14 and Eq.~9
578: into Eq.~5, we obtain the final result as
579: \begin{mathletters}
580: \begin{eqnarray}
581: J_{s}^{x}=\int
582: d\omega_{1}d\omega_{2}|t|^{2}{\sf{P}}\frac{f(\omega_{2})-f(\omega_{1})}
583: {\omega_{2}-\omega_{1}}\times\nonumber \\
584: \chi_{fm}(\omega_{2})\chi_{2deg}(\omega_{1})\cos\theta
585: \end{eqnarray}
586: \begin{eqnarray}
587: J_{s}^{y}=0
588: \end{eqnarray}
589: \begin{eqnarray}
590: J_{s}^{z}=\int
591: d\omega_{1}d\omega_{2}|t|^{2}{\sf{P}}\frac{f(\omega_{2})-f(\omega_{1})}
592: {\omega_{2}-\omega_{1}}\times\nonumber \\
593: \chi_{fm}(\omega_{2})\chi_{2deg}(\omega_{1})\sin\theta\cos\phi
594: \end{eqnarray}
595: \end{mathletters}
596: with
597: \begin{mathletters}
598: \begin{eqnarray}
599: \chi_{fm}(\omega)=\sum_{q}\{\delta(\omega-\varepsilon_{q}-h)-\delta(\omega-\varepsilon_{q}+h)\}
600: \end{eqnarray}
601: \begin{eqnarray}
602: \chi_{2deg}(\omega)=\sum_{k}
603: {^\prime}\{\delta(\omega-\varepsilon_k-\lambda
604: k)-\delta(\omega-\varepsilon_k+\lambda k)\}.
605: \end{eqnarray}
606: \end{mathletters}
607: Here the prime $\prime$ on the summation in Eq.~16b means that the
608: incident angle $\theta_{r}$ of electron from 2DEG into FM has
609: been integrated out from $-\pi/2$ to $\pi/2$, which denotes only
610: $k_x>0$ electrons contribute to the tunnelling current in our
611: calculation. $\chi(\omega)$ is the difference of the density of
612: states of two spin bands in FM and 2DEG. For the 1D case
613: $\chi_{2deg}=0$ as discussed earlier; whereas for the FM cases
614: with any dimensions , $\chi_{fm}\neq 0$ if the molecular field $h$
615: is nonzero. Thus in 1D RSOI(DSOI)/FM junction, ESC does not exist.
616:
617: Eq.~15 indicates that in a RSOI/FM junction at zero bias a ESC can
618: also flow through the junction and RSOI in the left lead acts as a
619: magnetic field with direction along the $y$-direction. According
620: to the exchange coupling between the two magnetic moments in the
621: two sides of the junction, $J_{s}^{y}$ is zero while
622: $J_{s}^{x}\sim\cos\theta$ and $J_{s}^{z}\sim\sin\theta\cos\phi$
623: are nonzero. If RSOI in the left lead is replaced by DSOI to form
624: a DSOI/FM junction, the ESC of Eq.~15 is different,
625: $J_{s}^{x}=0$, $J_{s}^{y}\sim\cos\theta$, and
626: $J_{s}^{z}\sim\sin\theta\sin\phi$ because in this case DSOI gives
627: rise to an average pseudomagnetic field along the $x$-direction.
628: The results we obtained here indicate again the RSOI (DSOI) in the
629: tunnelling junction can play the same role as in a FM metal.
630:
631: \subsection{FM/FM}
632: For completeness we present in this subsection the results of the
633: ESC through a FM/FM junction, which are same as those found in
634: literatures.\cite{21,22,23} The spin quantum axis is taken as the
635: magnetic moment direction of the left FM lead
636: $\vec{{\mathbf{h}}}_{L}$($0$,$0$,$1$) and the magnetic moment in
637: the right FM lead is same as that in the last subsection
638: $\vec{{\mathbf{h}}}_{R}$($\sin\theta\cos\phi $,
639: $\sin\theta\sin\phi$, $\cos\theta$). Using the Green's function
640: (Eq.~14) of the FM lead, we can obtain the formula of ESC as
641: \begin{mathletters}
642: \begin{eqnarray}
643: J_{s}^{x}=-\frac{1}{2}\int
644: d\omega_{1}d\omega_{2}|t|^{2}{\sf{P}}\frac{f(\omega_{2})-f(\omega_{1})}
645: {\omega_{2}-\omega_{1}}\times\nonumber \\
646: \chi_{fm}^{R}(\omega_{2})\chi_{fm}^{L}(\omega_{1})\sin\theta\sin\phi,
647: \end{eqnarray}
648: \begin{eqnarray}
649: J_{s}^{y}=\frac{1}{2}\int
650: d\omega_{1}d\omega_{2}|t|^{2}{\sf{P}}\frac{f(\omega_{2})-f(\omega_{1})}
651: {\omega_{2}-\omega_{1}}\times\nonumber \\
652: \chi_{fm}^{R}(\omega_{2})\chi_{fm}^{L}(\omega_{1})\sin\theta\cos\phi,
653: \end{eqnarray}
654: \begin{eqnarray}
655: J_{s}^{z}=0.
656: \end{eqnarray}
657: \end{mathletters}
658: Here $\chi_{fm}^{L(R)}$ is the difference between the density of
659: states of the two spin bands of the left (right) FM lead as shown
660: in Eq.~16a. In a simpler form, Eq.~17 can be written as
661: ${\mathbf{J}}_{s}\sim
662: \vec{{\mathbf{h}}}_{L}\times\vec{{\mathbf{h}}}_{R}$, which means
663: the ESC comes from the exchange coupling between the two
664: magnetizations in the left and right FM leads. Only the spin
665: component along
666: $\vec{{\mathbf{h}}}_{L}\times\vec{{\mathbf{h}}}_{R}$ is not a
667: conserved quantitiy since both magnetizations in the two sides of
668: the junction can flip the spin. The ESC Eq.~11 and Eq.~15 for the
669: 2DEG/2DEG and 2DEG/FM junction can also be expressed as the form
670: of vector product of (pseudo)magnetic moments.
671:
672: In earlier works,\cite{24} this spin current in the FM/FM junction
673: was found to transport spin torque and result in magnetization
674: reversal in an unfixed FM lead. The spin current can also induce
675: an electric field, which may be used to experimentally detect the
676: ESC through magnetic junctions.\cite{12,13} We wish to point out
677: here the ESC stemming from the exchange coupling of two magnetic
678: moments in two leads is achieved based on the approximation of the
679: linear response, beyond which it is unclear whether the above
680: results are still valid, thus it is worthwhile to study the
681: behavior of ESC at the limit of the strong coupling.
682:
683: \section{summary}
684: We have presented a detailed analysis of the dissipationless ESC
685: through three tunnelling junctions, 2DEG/2DEG, 2DEG/FM, and FM/FM.
686: For these three junctions, the ESC comes from the exchange
687: coupling between two magnetic (or pseudomagnetic) moments in the
688: two leads in the linear approximation. Although in 2DEG with RSOI
689: or DSOI, the pseudomagnetic field is isotropic and keeps the time
690: reversal symmetry, only electrons with $k_{x}>0$ contribute to the
691: tunnelling current when 2DEG with RSOI or DSOI is used as the left
692: lead of the junction as in the 2DEG/2DEG and 2DEG/FM junctions, so
693: that the tunnelling electrons would feel an average nonzero
694: pseudomagnetic field along the $y$-direction for RSOI or the
695: $x$-direction for DSOI. The average pseudomagnetic fields have the
696: same effect as a real magnetic moment and the ESC could flow in
697: both 2DEG/2DEG and 2DEG/FM junctions. For the RSOI/DSOI junction,
698: a spontaneous ESC could form since the pseudomagnetic fields
699: associated with RSOI or DSOI are inherently different. While for
700: RSOI/RSOI or DSOI/DSOI junctions, one possibility is to rotate one
701: of the leads to change the orientation of the pseudomagnetic
702: field, or we could shift the energy band of one of the 2DEG
703: because the pseudomagnetic field is dependent on the direction of
704: the electron momentum $\mathbf{P}$, which in turn is determined by
705: the band edge. This is different from the magnetic moment in a FM
706: electrode. Another distinction between the \emph{2DEG} with RSOI
707: (DSOI) and FM lead is that in strict 1D case RSOI (DSOI) cannot
708: result in a ESC through the junction, while this is not the case
709: for 1D FM leads.
710:
711: ACKNOWLEDGEMENT This work is supported by HongKong Research Grant
712: Council, Project No: CityU 100303.
713:
714:
715: \begin{references}
716: \bibitem{1}
717: I. Z\"{u}tic, J. Fabian, and S.D. Sarma, Rev. Mod. Phys. {\bf 76},
718: 323 (2004).
719:
720: \bibitem{2}
721: S.A. Wolf, D.D. Awschalom, R.A. Buhrman, J.M. Daughton, S. von
722: Moln${\acute{\text{a}}}$r, M.L. Roukes, A.Y. Chtchelkanova, and
723: D.M. Treger, Science {\bf294}, 1488 (2001).
724:
725: \bibitem{3}
726: G.A. Prinz, Phys. Today {\bf 48}({\bf{4}}), 58 (1995).
727:
728: \bibitem{4}
729: G. Schmidt, D. Ferrand, L.W. Molenkamp, A.T. Filip, and B.J. van
730: Wees, Phys. Rev. B {\bf 62}, R4790 (2000).
731: \bibitem{5}
732: P.R. Hammar, B.R. Bennet, M.J. Yang, and M. Johnson, Phys. Rev.
733: Lett. {\bf83}, 203 (1999).
734: \bibitem{6}
735: H.J. Zhu, M. Ramsteiner, H. Kostial, M. Wassermeier, H.P.
736: Sch$\ddot{\text{o}}$nherr, and K.H. Ploog, Phys. Rev. Lett.
737: {\bf87}, 016601 (2001).
738:
739: \bibitem{7}
740: A. Brataas, Y. V. Nazarov, and G.E.W. Bauer, Phys. Rev. Lett. {\bf
741: 84}, 2481 (2000); A. Brataas, Y. Tserkovnyak, G.E.W. Baure, and
742: B.I. Haplerin, Phys. Rev. B {\bf 66}, 060404 (2002).
743:
744: \bibitem{8}
745: P. Sharma and P.W. Brouwer, Phys. Rev. Lett. {\bf 91}, 166801
746: (2003).
747:
748: \bibitem{9}
749: B.G. Wang, J. Wang, and H. Guo, Phys. Rev. B {\bf 67}, 092408
750: (2003); Q.F. Sun, H. Guo, and J. Wang, Phys. Rev. Lett. {\bf 90},
751: 258301 (2003).
752:
753: \bibitem{10}
754: M.J. Stevens, A.L. Smirl, R.D.R. Bhat, A.Najmaie, J.E. Sipe, and
755: H.M. van Driel, Phys. Rev. Lett. {\bf 90}, 136603 (2003).
756:
757: \bibitem{11}
758: J. Hubner, W.W. Ruhle, M. Klude, D. Hommel, R.D.R. Bhat, J.E. Sipe,
759: H.M. van Driel, Phys. Rev. Lett. {\bf 90}, 216601 (2003).
760:
761: \bibitem{12}
762: F. Meier and D. Loss, Phys. Rev. Lett. {\bf 90}, 167204 (2003).
763:
764: \bibitem{13}
765: F. Sch{\"{u}}tz, M. Kollar, and P. Kopietz, Phys. Rev. Lett. {\bf
766: 91}, 017205 (2003).
767:
768: \bibitem{14}
769: B.G. Wang, J. Wang, J. Wang, and D.Y. Xing, Phys. Rev. B {\bf 69},
770: 174403 (2004).
771:
772: \bibitem{15}
773: S. Murakami, N. Nagaosa, and S.-C. Zhang, Science {\bf 301}, 1368
774: (2003).
775:
776: \bibitem{16}
777: J. Sinova, D. Culcer, Q. Niu, N.A. Sinitsyn, T. Jungwirth, and
778: A.H. MacDonald, Phys. Rev. Lett. {\bf 92}, 126603 (2004).
779:
780: \bibitem{17}
781: E.I. Rashba, Phys. Rev. B {\bf 68}, 241315 (2003).
782:
783: \bibitem{18}
784: P.Bruno and V.K. Dugaev, cond-mat/0508640.
785:
786: \bibitem{19}
787: J. K\"{o}nig, M.C. B\o nsager, and A.H. MacDonald, Phys. Rev.
788: Lett. {\bf 87}, 187202 (2001).
789:
790: \bibitem{20}
791: H.Katsura, N. Nagaosa, A.V. Balatsky, Phys. Rev. Lett. {\bf 95},
792: 057205 (2005); P. Chandr, P. Coleman, adn A.I. Larkin, J. Phys.
793: Cond. Mat. {\bf 2}, 7933 (1990).
794:
795: \bibitem{21}
796: M.Braun, J. Konig, and J. Martinek, Superlattices and
797: Microstrutures {\bf 37}, 333 (2005).
798:
799: \bibitem{22}
800: Nogueira and K.H. Bennemann, Europhys. Lett. {\bf 67}, 620 (2004).
801:
802: \bibitem{23}
803: Y.L. Lee and Y.W. Lee, Phys. Rev. B {\bf 68}, 184413 (2003).
804:
805: \bibitem{24}
806: L. Berger, Phys. Rev. B {\bf 54}, 9353 (1996); J.C. Slonczewski,
807: J. Magn. Magn. Mater. {\bf 159}, L1 (1996).
808:
809: \bibitem{25}
810: K. B{\o}rkje and A. Sudb{\o}, cond-mat/0506024.
811: \bibitem{26}
812: Y.A. Bychkov and E.I. Rashba, J. Phys. C {\bf 17}, 6039 (1984).
813:
814: \bibitem{27}
815: J. Nitta, T. Akazaki, H. Takayanagi, T. Enoki, Phys. Rev. Lett.
816: {\bf 78}, 1335 (1997).
817:
818: \bibitem{28}
819: G. Dresselhaus, Phys. Rev. {\bf 100}, 580 (1955).
820:
821: \bibitem{29}
822: V.M. Ramaglia, D. Bercioux, V. Cataudella, G.De Filippis, C.A.
823: Perroni, and F. Ventriglia, Eur. Phys. J. B {\bf 36}, 365 (2003).
824:
825: \bibitem{30}
826: P.W. Anderson, Lectures on the many-body problem, edited by E.R.
827: Cainello (Academic press, New York) 1964.
828:
829: \bibitem{31}
830: Q.F. Sun and X.C. Xie, Cond-Mat/0502317 (2005).
831:
832: \bibitem{32}
833: G.E. Blonder, M. Tinkham, and T.M. Klapwijk, Phys. Rev. B {\bf
834: 25}, 4515 (1982).
835:
836: \end{references}
837:
838:
839: \begin{figure}
840: \caption{Schematic of a tunnelling junction with two quasi-1D
841: electrodes with RSOI (DSOI) connected by an insulator barrier
842: (I.B.). The left electrode could be physically rotated along
843: $z$-direction by angle $\alpha$, which is equivalent to rotating
844: the direction of the average pseudo-magnetic moment in RSOI
845: (DSOI). The current flows along $x$-axis. }
846: \end{figure}
847:
848:
849:
850:
851:
852: \end{document}
853: