cond-mat0204570/wu.tex
1: %\documentstyle[preprint,aps]{revtex}
2: \documentstyle[aps,prbbib,twocolumn,epsf]{revtex}
3: 
4: \begin{document}
5: \draft
6: \title{Spin polarized parametric pumping}
7: 
8: \author{Wu Junling, Baigeng Wang, and Jian Wang$^{a)}$}
9: \address{Department of Physics, The University of Hong Kong, 
10: Pokfulam Road, Hong Kong, China\\
11: %2. Institute of Solid State Physics, Chinese Academy of Sciences,
12: %Hefei, Anhui, China 
13: }
14: \maketitle
15: 
16: \begin{abstract}
17: We have developed a general theory for a parametric pump consisting of 
18: a nonmagnetic system with two ferromagnetic leads whose magnetic moments 
19: orient at an angle $\theta$ with respect to each other. In this
20: theory, the leads can maintain at different chemical potential. As a
21: result, the current is driven due to both the external bias and the
22: pumping potentials. When both $\theta$ and external bias are zero, our
23: theory recovers the known theory. In particular, two cases are 
24: considered: (a). in the adiabatic regime, we have derived the
25: pumped current for arbitrary pumping amplitude and external bias. (b).
26: at finite frequency, the system is away from equilibrium, we have
27: derived the pumped current up to quadratic order in pumping amplitude.  
28: Numerical results show interesting spin valve effects for pumped
29: current. 
30: \end{abstract}
31: 
32: \pacs{73.23.Ad, 73.40.Gk, 73.40.-c, 72.10.Bg}
33: 
34: %\newpage
35: \section{Introduction}
36: Recently, there is considerable interest in the parametric 
37: pumping\cite{brouwer,aleiner,switkes,zhou,shutenko,wei1,aleiner1,avron1,levinson,brouwer2,buttiker1,avron2,wang1,vavilov,wei2,renzoni,makhlin,chu,wbg1,wang2,levinson2,levinson1,buttiker2,vavilov1,entin1,wbg3,wbg2,blaauboer}. 
38: The parametric pump is facilitated by cyclic variations of pumping 
39: potentials inside the scattering system and has been realized
40: experimentally by Switkes et al\cite{switkes}. On theoretical side,
41: much progress has been made towards understanding of various features
42: related to the parametric pump. This includes quantization of pumped 
43: charge\cite{aleiner,levinson,wang2,entin1}, the influence of discrete 
44: spatial symmetries and magnetic field\cite{shutenko,aleiner1}, the 
45: rectification of displacement current\cite{brouwer2}, as well as the 
46: inelastic scattering\cite{buttiker1} to the pumped current. 
47: The concept of optimal pump has been proposed with the lower bound for
48: the dissipation derived\cite{avron2}. Within the formalism of
49: time-dependent scattering matrix theory, the heat current and shot 
50: noise in the pumping process\cite{makhlin,buttiker2,vavilov1,wbg2} 
51: has also been discussed. Recently, the original adiabatic pumping theory 
52: has been extended to account for the effect due to finite 
53: frequency\cite{vavilov,wbg1}, Andreev reflection in the presence of 
54: superconducting lead\cite{wang1,blaauboer}, 
55: and strong electron interaction in the Kondo regime\cite{wbg3}. This
56: gives us more physical insight of parametric pumping. For instance, the
57: experimental observed anomaly of pumped current at $\phi=0$ and $\phi=\pi$
58: can be explained using finite frequency theory\cite{wbg1} as due to the 
59: quantum interference of different photon assisted processes. When
60: superconducting lead is present, the interference between the direct 
61: reflection and multiple Andreev reflection gives rise to an enhancement
62: of pumped current which is four times of that of normal system\cite{wang1}. 
63: It will be interesting to further extend the parametric theory to the
64: case where the ferromagnetic leads are present. With the theory
65: extended, many new physics are foreseen\cite{wu} which may lead to new 
66: operational paradigms for future spintronic devices\cite{review}. In 
67: this paper, we
68: have developed a parametric pumping theory for a nonmagnetic system with
69: two ferromagnetic leads whose magnetic moments orient at an angle 
70: $\theta$ with respect to each other. Our theory is based on nonequilibrium 
71: Green's function approach and focused on current perpendicular to plane 
72: geometry. Parametric pump generates current at zero external bias. It
73: would be interesting to see the interplay of the role played by pumping
74: potential and external bias if the leads are maintain at different
75: chemical potential\cite{levinson1}. Hence in our theory, the external
76: bias is also included. In the adiabatic regime, the pumped current is 
77: proportional to pumping frequency. In this regime, we have derived the 
78: parametric pumping theory for finite pumping amplitude. At the finite
79: pumping frequency, the system is away from equilibrium, we have performed
80: perturbation up to the second order in pumping amplitude and obtained 
81: the pumped current at finite frequencies. The newly developed theory
82: allows us to study the pumped current for a variety of parameters, such
83: as the pumping amplitude, pumping frequency, phase difference between
84: two pumping potentials, the angular dependence between the magnetization of
85: two leads, as well as the external bias. We have applied our theory 
86: to a tunneling magnetoresistance (TMR) junction\cite{slon}. Due to the 
87: reported room temperature operation of TMR, the fundmental principle and 
88: transport properties of TMR devices has attracted increasingly
89: attention\cite{moodera}. Our numerical results show interesting spin valve
90: effect for pumped current. The paper is organized as
91: follows. In section II, we derive the general theory of a parametric
92: pump in the presence of ferromagnetic leads. The numerical results and 
93: summary are presented in section III.
94: 
95: \section{General theory}
96: %GENERAL
97: 
98: The system we examine consists of a nonmagnetic system connected by two 
99: ferromagnetic 
100: electrodes to the reservoir. The magnetic moment ${\bf M}$ of the 
101: left electrode is pointing to the $z$-direction, the electric current 
102: is flowing in the $y$-direction, while the moment of the right 
103: electrode is at an angle $\theta$ to the $z$-axis in the $x-z$ plane. 
104: The Hamiltonian of the system is of the following form
105: \begin{equation}
106: H = H_L+H_R+H_0+V_p+H_T
107: \label{eqq1}
108: \end{equation}
109: where $H_L$ and $H_R$ describe the left and right electrodes 
110: \begin{equation}
111: H_L = \sum_{k\sigma} (\epsilon_{k L}+\sigma M) ~ c^{\dagger}_{kL\sigma}
112: c_{kL\sigma}
113: \label{eqq2}
114: \end{equation}
115: \begin{eqnarray}
116: H_R &=& \sum_{k\sigma} [(\epsilon_{k R}+\sigma M\cos\theta) ~
117: c^{\dagger}_{kR\sigma} c_{kR\sigma}  \nonumber \\
118: &+& \sum_{k\sigma} M\sin\theta ~ [c^{\dagger}_{kR\sigma} c_{kR\bar{\sigma}} ]
119: \ \ .
120: \label{eqq3}
121: \end{eqnarray}
122: In Eq. (\ref{eqq1}), $H_0$ describes the nonmagnetic (NM) scattering
123: region, 
124: \begin{equation}
125: H_0= \sum_{n\sigma} \epsilon_n d^{\dagger}_{n\sigma} d_{n\sigma}\ \ .
126: \label{hdot}
127: \end{equation}
128: $V_p$ is the time-dependent pumping potential and $H_T$ describes 
129: the coupling between electrodes and the NM scattering region 
130: with hopping matrix $T_{k\alpha n}$. To simplify the analysis, we 
131: assume the hopping matrix to be independent of spin index, hence 
132: \begin{equation}
133: H_T= \sum_{k \alpha n\sigma} [T_{k\alpha n} ~ 
134: c^{\dagger}_{k\alpha\sigma} d_{n\sigma} + c.c.]\ \ .
135: \label{ht}
136: \end{equation}
137: In these expressions $\epsilon_{k\alpha} = \epsilon^0_k + qV_\alpha$
138: with $\alpha = L,R$; $c^{\dagger}_{k\alpha\sigma}$ (with $\sigma=\uparrow,
139: \downarrow$ or $\pm 1$ and $\bar{\sigma}=-\sigma$) is the creation operator 
140: of electrons with spin index $\sigma$ inside the $\alpha$-electrode. 
141: Similarly $d^{\dagger}_{n\sigma}$ is the creation operator of electrons 
142: with spin $\sigma$ at energy level $n$ for the NM scattering region. In 
143: writing down Eqs.(\ref{eqq2}) and (\ref{eqq3}), we have made a 
144: simplification
145: that the value of molecular field $M$ is the same for the two electrodes, 
146: thus the spin-valve effect is obtained\cite{slon} by varying the angle 
147: $\theta$. Essentially, $M$ mimics the difference of density of states 
148: (DOS) between spin up and down electrons\cite{slon} in the electrodes. 
149: In this paper, we only consider the single electron behavior.
150: The charge quantization is not considered so that our system
151: is not in the Coulomb blockade regime. In addition, for the nonmagnetic
152: regions we are interested in, the Kondo effect can be neglected.
153: 
154: To proceed, we first apply the following Bogliubov transformation\cite{bog} 
155: to diagonalize the Hamiltonian of the right electrode\cite{wbg4}, 
156: \begin{equation}
157: c_{kR\sigma }\rightarrow \cos (\theta/2)C_{kR\sigma }-\sigma \sin (
158: \theta/2)C_{kR\bar{\sigma}}
159: \end{equation}
160: \begin{equation}
161: c_{kR\sigma }^{+}\rightarrow \cos (\theta/2)C_{kR\sigma
162: }^{+}-\sigma \sin (\theta/2)C_{kR\bar{\sigma}}^{+}
163: \end{equation}
164: from which we obtain the effective Hamiltonian
165: \begin{equation}
166: H_\alpha = \sum_{k\sigma} [(\epsilon_{k\alpha}+\sigma M)
167: C^{\dagger}_{k\alpha\sigma}C_{k\alpha\sigma}
168: \label{halpha}
169: \end{equation}
170: \begin{eqnarray}
171: H_T &=& \sum_{k n\sigma} [T_{kLn} ~ C^{\dagger}_{kL\sigma} 
172: d_{n\sigma} + T_{kRn} (\cos\frac{\theta}{2} ~ C^{\dagger}_{kR\sigma}  
173: \nonumber \\
174: &-& \sigma \sin\frac{\theta}{2} ~ C^{\dagger}_{kR\bar{\sigma}}) ~ 
175: d_{n\sigma} +c.c.]\ \ .
176: \label{eff}
177: \end{eqnarray}
178: In the following subsections, we will consider two cases: (1). parametric
179: pumping in the low frequency limit with finite pumping amplitude.
180: (2). pumping in the weak pumping limit with finite pumping frequency.
181: 
182: \subsection{Pumping in the low frequency limit}
183: %PUMPING
184: 
185: In the subsection, we examine the pumping current at low frequency
186: limit while maintaining pumping amplitude finite. In this limit,
187: the system is nearly in equilibrium and we will use the equilibrium
188: Green's function\cite{datta,jauho,wbg5} to characterize the pumping process.
189: %A parametric electron pump drives an electric current at zero bias 
190: %by cyclic deformations of two or more system 
191: %parameters. The system we considered is a
192: %quantum point of finite length which connected by two leads. Two gate
193: %voltages are applied to quantum point to form a double well potential with
194: %the well depth varying in a cyclic fashion. To analyze parametric quantum
195: %pumping, we make use of the nonequilibrium green's function method 
196: Using the distribution function, the total charge in the system during the
197: pumping is given by 
198: \begin{equation}
199: Q(x,t)=-iq\int (dE/2\pi )({\bf G}^{<}(E,\{V(t)\}))_{xx}
200: \end{equation}
201: where ${\bf G}^{<}$ is the lesser Green's function in real space, $x$
202: labels the position, and $\{V(t)\}$ describes a set of external parameters
203: which facilitates the pumping process. Within Hartree approximation, $
204: {\bf G}^{<}$ is related to the retarded and advanced Green's
205: functions ${\bf G}^{r}$ and ${\bf G}^{a}$, 
206: \begin{equation}
207: {\bf G}^{<}(E,\{V\})={\bf G}^{r}(E,\{V\})i\sum_\alpha 
208: {\bf \Gamma}_\alpha f_\alpha(E){\bf G}^{a}(E,\{V\})\ \ .
209: \end{equation}
210: where $f_\alpha(E) = f(E-qV_\alpha)$. In the low frequency limit, the 
211: retarded Green's function in real space is given by
212: 
213: \begin{equation}
214: {\bf G}^{r}(E,\{X\})=\frac{1}{E-H_0-{\bf V}_p-{\bf \Sigma}^{r}}
215: \end{equation}%
216: where ${\bf \Sigma}^{r}\equiv \sum_{\alpha} {\bf \Sigma}_{\alpha }^{r}$ 
217: is the self energy, and ${\bf \Gamma}_{\alpha}=
218: -2Im[{\bf \Sigma}_{\alpha}^{r}]$ is the linewidth function. 
219: In above equations, ${\bf G}^{r,a,<}$ denotes a $2\times 2$ matrix
220: with matrix elements $G_{\sigma, \sigma'}^{r,a<}$ and $\sigma=
221: \uparrow, \uparrow$. ${\bf V}_p = V_p {\bf I}$ where ${\bf I}$ is a
222: $2\times2$ unit matrix in the spin space.
223: In real space representation, $V_{p}$ is a diagonal matrix describing 
224: the variation of the potential landscape due to the external pumping 
225: parameter $V$. The self-energies are given\cite{wbg4}
226: 
227: \begin{equation}
228: {\bf \Sigma}^r_\alpha(E) = \hat{R}_\alpha \left( 
229: \begin{array}{cc}
230: \Sigma^r_{\alpha\uparrow} & 0 \\ 
231: 0 & \Sigma^r_{\alpha\downarrow}
232: \end{array}
233: \right) \hat{R}^\dagger_\alpha
234: \label{self}
235: \end{equation}
236: and 
237: \begin{equation}
238: {\bf \Sigma}^<_\alpha(E) = i f_\alpha \hat{R}_\alpha \left( 
239: \begin{array}{cc}
240: \Gamma^0_{\alpha\uparrow} & 0 \\ 
241: 0 & \Gamma^0_{\alpha\downarrow}
242: \end{array}
243: \right) \hat{R}_\alpha^\dagger
244: \label{ss}
245: \end{equation}
246: with the rotational matrix $\hat{R}_\alpha$ for electrode $\alpha$ 
247: defined as 
248: \begin{equation}
249: \hat{R} = \left( 
250: \begin{array}{cc}
251: \cos\theta_\alpha/2 ~~ & \sin\theta_\alpha/2 \\ 
252: -\sin\theta_\alpha/2 ~~ & \cos\theta_\alpha/2
253: \end{array}
254: \right)\ \ .
255: \end{equation}
256: Here angle $\theta_{\alpha}$ is defined as $\theta_L=0$ and 
257: $\theta_R=\theta$ and $\Sigma^r_{\alpha \sigma}$ is given by 
258: \begin{equation}
259: \Sigma^r_{\alpha \sigma mn}= \sum_k \frac{T^*_{k\alpha m} T_{k\alpha n}} 
260: {E-\epsilon^0_{k \alpha \sigma} +i\delta}
261: \end{equation}
262: and $\Gamma^0_{\alpha\sigma}= -2Im(\Sigma^r_{\alpha\sigma})$ is the
263: linewidth function when $\theta=0$. 
264: 
265: In order for a parametric electron pump to function at low
266: frequency, we need simultaneous variation of two or more 
267: system parameters controlled by gate voltages: $V_{i}(t)=V_{i0}+
268: V_{ip}\cos (\omega t +\phi_i)$. 
269: Hence, in our case, the potential due to the gates can be written as 
270: $V_{p}=\sum_i V_{i}{\bf \Delta }_{i}$, where ${\bf \Delta }_{i}$ 
271: is potential profile due to each pumping potential. For simplicity we 
272: assume a constant gate potential
273: such as that $({\bf \Delta }_{1})_{xx}$ is one for $x$ in the first gate
274: region and zero otherwise. If the time variation of these parameters are
275: slow, i.e. for $V(t)=V_{0}+\delta V\cos (\omega t)$, then the
276: charge of the system coming from all contacts due to the infinitesimal
277: change of the system parameter ($\delta V\rightarrow 0$) is
278: 
279: \begin{equation}
280: dQ_p(t)=\sum_{i}\partial_{V_{i}}Tr[Q(x,t)]~\delta V_{i}(t)
281: \end{equation}
282: it is easily seen that the total charge in the system in a period is zero
283: which is required for the charge conservation. To calculate the pumped
284: current, we have to find the charge $dQ_{p\alpha }$ passing through contact 
285: $\alpha $ due to the change of the system parameters. Using 
286: %the fact that ${\bf G}^{r}{\bf \Gamma} {\bf G}^{a}=
287: %i({\bf G}^{r}-{\bf G}^{a})$ and
288: the Dyson equation $\partial_{V_{i}}{\bf G}^{r}={\bf G}^{r}\Delta _{i}
289: {\bf G}^{r}$, the above equation becomes,
290: 
291: \begin{eqnarray}
292: && dQ_p(t) = q\sum_j \int \frac{dE}{2\pi} \sum_\beta {\rm Tr}[ {\bf G}^r 
293: {\bf \Delta}_j {\bf G}^r {\bf \Gamma}_\beta {\bf G}^a
294: \nonumber \\
295: &&+{\bf G}^r {\bf \Gamma}_\beta {\bf G}^a {\bf \Delta}_j {\bf G}^a 
296: ] f_\beta(E) \delta V_j(t)
297: \nonumber \\
298: &&=-q\int \frac{dE}{2\pi} \sum_j \sum_\beta f_\beta
299: {\rm Tr} [ \partial_E[{\bf G}^r {\bf \Gamma}_\beta
300: {\bf G}^a {\bf \Delta}_j] ] \delta V_j(t) \nonumber
301: \end{eqnarray}
302: where the wideband limit has been taken. Integrating by part, we
303: obtain
304: \begin{equation}
305: dQ_p(t)=q\int dE \sum_\beta (\partial_E f_\beta) 
306: \sum_j \frac{dN_\beta}{dV_j} \delta V_j(t)
307: \end{equation}
308: where we have used the injectivity\cite{but5}
309: \begin{equation}
310: \frac{dN_\beta}{dV_j} = \frac{1}{2\pi} {\rm Tr} 
311: [{\bf G}^r {\bf \Gamma}_\beta {\bf G}^a {\bf \Delta}_j] 
312: \label{dndx}
313: \end{equation} 
314: Using the partial density of states $dN_{\alpha \beta}/dV_j$
315: defined as\cite{but6}
316: 
317: \begin{eqnarray}
318: &&\frac{dN_{\alpha \beta}}{dV_j} = \frac{1}{4\pi} {\rm Tr} [
319: {\bf G}^r {\bf \Gamma}_\alpha {\bf G}^r \Delta_j +c.c.]
320: \delta_{\alpha \beta} \nonumber \\
321: &&+\frac{1}{4\pi} {\rm Tr} [i {\bf G}^r {\bf \Gamma}_\beta
322: {\bf G}^a {\bf \Gamma}_\alpha {\bf G}^r \Delta_j +c.c.]
323: \end{eqnarray}
324: with $\sum_\alpha dN_{\alpha \beta}/dV_j = dN_{\beta}/dV_j$,
325: we obtain 
326: 
327: \begin{equation}
328: dQ_{p\alpha}(t)=-q\int dE \sum_\beta (-\partial_E f_\beta) 
329: \sum_j \frac{dN_{\alpha \beta}}{dV_j} \delta V_j(t)
330: \end{equation}
331: If we include the charge passing through contact $\alpha$ due the
332: external bias, then
333: 
334: \begin{eqnarray}
335: &&dQ_{\alpha}(t)=-q\int dE \sum_\beta (-\partial_E f_\beta) 
336: \sum_j \frac{dN_{\alpha \beta}}{dV_j} \frac{dV_j(t)}{dt} dt
337: \nonumber \\
338: &&-q\int dE \sum_\beta {\rm Tr} [{\bf \Gamma}_\alpha {\bf G}^r
339: {\bf \Gamma}_\beta {\bf G}^a] (f_\alpha - f_\beta) dt
340: \end{eqnarray}
341: Furthermore, the total current flowing through contact $\alpha$ due to 
342: both the variation of parameters $V_j$ and external bias, 
343: in one period, is given by
344: \begin{equation}
345: J_\alpha = \frac{1}{\tau} \int_0^{\tau} dt ~ dQ_\alpha/dt
346: \label{current}
347: \end{equation}
348: where $\tau=2\pi/\omega$ is the period of cyclic variation. If there
349: are two pumping parameters, Eq.(\ref{current}) can be written as
350: when $\alpha=1$,
351: 
352: \begin{eqnarray}
353: &&J^{(1)}_1 =\frac{q}{2\tau} \int dt \int dE ~ \partial_E (f_1-f_2)
354: \sum_j \nonumber \\
355: &&[ \frac{dN_{11}}{dV_j} \frac{dV_j(t)}{dt} - 
356: \frac{dN_{12}}{dV_j} \frac{dV_j(t)}{dt} ] dt \nonumber \\
357: &&-\frac{q}{\tau}\int dt \int dE {\rm Tr} [{\bf \Gamma}_1 {\bf G}^r
358: {\bf \Gamma}_2 {\bf G}^a] (f_1 - f_2) 
359: \label{cur1}
360: \end{eqnarray}
361: and 
362: \begin{eqnarray}
363: &&J^{(2)}_1 =\frac{q}{2\tau} \int dt \int dE ~ \partial_E (f_1+f_2)
364: \sum_j \nonumber \\
365: &&[ \frac{dN_{11}}{dV_j} \frac{dV_j(t)}{dt} + 
366: \frac{dN_{12}}{dV_j} \frac{dV_j(t)}{dt} ] dt 
367: \label{cur2}
368: \end{eqnarray}
369: where $J_1=J^{(1)}_1+J^{(2)}_2$. In Ref.\onlinecite{levinson1},
370: $J^{(1)}_1$ has been identified as the current due to the
371: external bias and $J^{(2)}$ as pumping current. For $\theta=0,\pi$,
372: all the $2\times 2$ matrix are diagonal. In this case, it is easy to
373: show that Eq.(\ref{current}) agrees with the result of 
374: Ref.\onlinecite{levinson1}. If the external bias is zero, 
375: Eq.(\ref{current}) reduces to the familiar formula\cite{brouwer}
376: when there are two pumping potentials,
377: 
378: \begin{equation}
379: J_\alpha = \frac{q\omega}{2\pi} \int_0^{\tau} dt \left[\frac{dN_\alpha}
380: {dX_1} \frac{dX_1}{dt} + \frac{dN_\alpha}{dX_2} \frac{dX_2}{dt}\right]
381: \label{pump}
382: \end{equation}
383: 
384: \subsection{Finite frequency pumping in the weak pumping limit}
385: %FINITE
386: 
387: In this subsection, we will calculate the pumping current at finite
388: frequency. The Keldysh nonequilibrium Green's function approach used 
389: here is in the standard tight-binding representation\cite{datta}.
390: We could not use the momentum space version because the time dependent 
391: perturbation (pumping potential) inside the scattering region is position 
392: dependent. Hence it is most suitable to use a tight-binding real space 
393: technique. In contrast, in previous investigations\cite{pre} on 
394: photon-assisted processes the time-dependent potential is uniform 
395: throughout the dot and therefore a momentum space method is easier to 
396: apply. 
397: 
398: Assuming the time-dependent perturbations located at the different sites, 
399: $j=i_0, j_0$, and $k_0$ etc, with 
400: \begin{equation}
401: V_j(t)=V_j\cos (\omega t+\phi_j)
402: \end{equation}
403: When there is no interaction between electrons in the ideal leads $L$ 
404: and $R$, the standard nonequilibrium Green's function theory gives the 
405: following expression for the time dependent current\cite{jauho},
406: 
407: \begin{eqnarray}
408: J_{\alpha }(t)&=&-q\int_{-\infty }^{t}dt_{1} 
409: {\rm Tr}[{\bf G}^{r}(t,t_{1}){\bf
410: \Sigma }_{\alpha }^{<}(t_{1},t) \nonumber \\
411: &+&{\bf G}^{<}(t,t_{1}){\bf \Sigma }
412: _{\alpha }^{a}(t_{1},t)+c.c.]
413: \end{eqnarray}
414: and transmission coefficient
415: \begin{equation}
416: T(E) ={\rm Tr} [{\bf \Gamma}_{L}^{r}(E)){\bf G}^{r}(E)  {\bf 
417: \Gamma}_{R}^{r}(E)){\bf G}^{a}(E)] 
418: \end{equation}
419: where the scattering Green's functions and self-energy are
420: defined in the usual manner: 
421: \begin{equation}
422: {\bf G}_{ij \sigma \sigma'}^{r,a}(t_{1},t_{2})=\mp 
423: i\theta (\pm t_{1}\mp t_{2})\langle
424: \{d_{i,\sigma}(t_{1}),d_{j,\sigma'}^{+}(t_{2})\}\rangle
425: \end{equation}
426: 
427: \begin{equation}
428: {\bf G}_{ij,\sigma,\sigma'}^{<}(t_{1},t_{2})=
429: i\langle d_{j,\sigma'}^{+}(t_{2})d_{i,\sigma,}(t_{1})\rangle
430: \end{equation}
431: 
432: \begin{equation}
433: {\bf \Sigma }_{\alpha ij}^{r,a,<}(t_{1},t_{2})=\sum_{k}T_{\alpha
434: ki}^{\ast }T_{\alpha kj}{\bf g}_{\alpha }^{r,a,<}(t_{1},t_{2})
435: \end{equation}
436: 
437: The average current $J_L(t)$ from the left lead can be written as 
438: \begin{eqnarray}
439: <J_{L}(t)>&=&-\frac{q}{\tau}\int_0^\tau dt \int_{-\infty }^{t}dt_{1}
440: Tr[{\bf G}_{11}^{r}(t,t_{1}){\bf \Sigma}_{L}^{<}(t_{1},t) \nonumber \\
441: &+&{\bf G}_{11}^{<}(t,t_{1}){\bf \Sigma}
442: _{L}^{a}(t_{1},t)+c.c.]
443: \label{average}
444: \end{eqnarray}
445: 
446: In the absence of pumping, the retarded Green's function
447: is defined in terms of the Hamiltonian $H_{0}$, 
448: \begin{equation}
449: {\bf G}^{0r}(E)=\frac{1}{E-H_0-{\bf \Sigma }^{r}}
450: \end{equation}
451: and ${\bf G}^{0<}$ is related to the retarded and advanced 
452: Green's functions ${\bf G}^{0r}$ and ${\bf G}^{0a}$ by,
453: 
454: \begin{equation}
455: {\bf G}^{0<}(E)={\bf G}^{0r}(E){\bf \Sigma}^{<}(E){\bf G}^{0a}(E)
456: \end{equation}
457: Now we make use of the time-dependent pumping potentials $V_j(t)$ 
458: as the perturbations to calculate all kinds of Green's functions up to 
459: the second order, and corresponding average current.
460: 
461: First, we calculate the current corresponding to the term 
462: ${\bf G}_{11}^{r}(t,t_{1}) {\bf \Sigma}_{L}^{<}(t_{1},t)$ in
463: Eq.(\ref{average}).  Dyson equation for ${\bf G}_{11}^{r}(t,t_{1})$ 
464: gives the second order contribution:
465: \begin{eqnarray}
466: &&{\bf G}_{11}^{(2)r}(t,t_{1}) =\sum_{jk} 
467: \int \int dxdy  {\bf G}_{1j}^{0r}(t-x) \nonumber \\
468: &&{\bf V}_{j}(x){\bf G}_{jk}^{0r}(x-y){\bf V}_{k}(y)
469: {\bf G}_{k 1}^{0r}(y-t_{1}) \nonumber \\
470: && \equiv \sum_{jk} {\bf G}_{1j}^{0r}
471: {\bf V}_{j} {\bf G}_{jk}^{0r} {\bf V}_k
472: {\bf G}_{k 1}^{0r}
473: \label{g2}
474: \end{eqnarray}
475: 
476: Substituting Eq.(\ref{g2}) into Eq.(\ref{average}) and completing
477: the integration over time $x,y,t_1,t$, it is not difficult to calculate 
478: the average current $<J_{L1}>$ due to the first term in Eq.(\ref{average}) 
479: (see appendix for details),
480: \begin{eqnarray}
481: && <J_{L1}(t)>=-\sum_{jk} \frac{qV_{j}V_{k}}{4}
482: \int \frac{dE}{2\pi }Tr\left[{\bf \Sigma }_{L}^{<}
483: {\bf G}_{1j}^{0r} \right. \nonumber\\
484: &&\left. [{\bf G}_{jk}^{0r}(E_{-} )e^{i\Delta_{kj}} 
485: +{\bf G}_{jk}^{0r}(E_{+})e^{-i\Delta_{kj}}] {\bf G}_{k 1}^{0r} \right]  
486: \label{eq1}
487: \end{eqnarray}
488: where $\Delta_{kj}=\phi_{k}-\phi_{j}$ is the phase difference,  
489: $E_{\pm}=E\pm\omega$, and $G^{r,a,<}\equiv G^{r,a,<}(E)$. 
490: 
491: Now we calculate the second term ${\bf G}_{11}^{<}(t,t_{1}) 
492: {\bf \Sigma }_{L}^{a}(t_{1},t)$ in Eq.(\ref{average}). Using Keldysh 
493: equation, ${\bf G}^<={\bf G}^r{\bf \Sigma}^< {\bf G}^a$ , we have 
494: \begin{equation}
495: {\bf G}_{jk}^{<} ={\bf G}_{j 1}^{r} {\bf \Sigma }_{L}^{<} 
496: {\bf G}_{1k}^{a}  
497: +{\bf G}_{j N}^{r} {\bf \Sigma }_{R}^{<} {\bf G}_{N k}^{a}
498: \label{less}
499: \end{equation}
500: where ${\bf \Sigma}_\alpha^<=i{\bf \Gamma}_\alpha f_\alpha$.
501: Expanding $G^{r,a}$ up to the second order in pumping parameters, 
502: we obtain the second order contribution from ${\bf G}^<_{11}$,
503: 
504: \begin{eqnarray}
505: &&{\bf G}_{11}^{(2)<}={\bf G}_{11}^{r} {\bf \Sigma }_{L}^{<} 
506: {\bf G}_{11}^a +{\bf G}_{1N}^r {\bf \Sigma}_{R}^< {\bf G}_{N1}^a 
507: \nonumber \\
508: &&= {\bf G}_{11}^{(2)r}  {\bf \Sigma }_{L}^{<} {\bf G}_{11}^{0a}
509: +{\bf G}_{11}^{(1)r} {\bf \Sigma }_{L}^{<} {\bf G}_{11}^{(1)a}
510: +{\bf G}_{11}^{0r} {\bf \Sigma }_{L}^{<} {\bf G}_{11}^{(2)a}
511: \nonumber \\
512: &&+{\bf G}_{1N}^{(2)r}{\bf \Sigma}_{R}^{<} {\bf G}_{N1}^{0a} 
513: +{\bf G}_{1N}^{(1)r} {\bf \Sigma }_{R}^{<} {\bf G}_{N1}^{(1)a} 
514: +{\bf G}_{1N}^{0r} {\bf \Sigma }_{R}^{<} {\bf G}_{N1}^{(2)a}
515: \nonumber \\
516: &=& \sum_{jk} [{\bf G}_{1j}^{0r}{\bf V}_{j}{\bf G}_{jk}^{0r}
517: {\bf V}_{k} {\bf G}_{k 1}^{0<} +{\bf G}_{1j}^{0r}{\bf V}_j
518: {\bf G}_{jk}^{0<} {\bf V}_k {\bf G}_{k 1}^{0a} \nonumber \\
519: &&+ {\bf G}_{1j}^{0<} {\bf V}_j {\bf G}_{jk}^{0a} 
520: {\bf V}_k {\bf G}_{k 1}^{0a}]
521: \label{g22}
522: \end{eqnarray}
523: where we have used Eq.(\ref{less}) to simplify the expression.
524: After some algebra, we have the following three expressions  
525: corresponding to each term in Eq.(\ref{g22}), 
526: %\begin{eqnarray}
527: %&&-\sum_{\alpha \beta} \frac{qV_{\alpha}V_{\beta}}{4}
528: %\int \frac{dE}{2\pi }Tr \left[{\bf \Sigma }_{L}^{a}
529: %{\bf G}_{1\alpha}^{0r}
530: %\right. \nonumber \\
531: %&&\left. 
532: %[{\bf G}_{\alpha \beta}^{0r}(E_{-})e^{i\Delta_{\beta \alpha}}
533: %+e^{-i\Delta_{\beta \alpha}}{\bf G}_{\alpha \beta}^{0r}(E_+)]
534: %{\bf G}_{\beta 1}^{0<} \right. \\
535: %&& \left. + {\bf \Sigma }_{L}^{a} {\bf G}_{1\alpha}^{0r}[
536: %{\bf G}_{\alpha \beta}^{0<}(E_{-}) e^{i\Delta_{\beta \alpha}}
537: %+e^{-i\Delta_{\beta \alpha}} 
538: %{\bf G}_{\alpha \beta}^{0<}(E_+)] {\bf G}_{\beta 1}^{0a}
539: %\right. \\
540: %&&\left. + {\bf \Sigma}_{L}^{a} {\bf G}_{1\alpha}^{0<}
541: %[{\bf G}_{\alpha \beta}^{0a}(E_{-})
542: %e^{i\Delta_{\beta \alpha}} 
543: %+e^{-i\Delta_{\beta \alpha}}{\bf G}_{\alpha \beta}^{0a}(E_+)]
544: %{\bf G}_{\beta 1}^{0a}
545: %\right]
546: %\end{eqnarray}
547: 
548: \begin{eqnarray}
549: &&-\sum_{jk} \frac{qV_j V_k}{4} \int \frac{dE}{2\pi} {\rm Tr} 
550: \left[{\bf \Sigma }_{L}^{a} {\bf G}_{1j}^{0r} \right. \nonumber \\
551: &&\left. [{\bf G}_{jk}^{0r}(E_{-})e^{i\Delta_{kj}}
552: +e^{-i\Delta_{kj}}{\bf G}_{jk}^{0r}(E_+)] {\bf G}_{k 1}^{0<} \right] 
553: \label{eq2}
554: \end{eqnarray}
555: 
556: \begin{eqnarray}
557: &&-\sum_{jk} \frac{qV_j V_k}{4} \int \frac{dE}{2\pi }Tr \left[
558: {\bf \Sigma }_{L}^{a} {\bf G}_{1j}^{0r}  \right. \nonumber \\
559: && \left. [ {\bf G}_{jk}^{0<}(E_{-}) e^{i\Delta_{kj}}
560: +e^{-i\Delta_{kj}} {\bf G}_{jk}^{0<}(E_+)] {\bf G}_{k 1}^{0a}
561: \right]
562: \label{eq3}
563: \end{eqnarray}
564: 
565: \begin{eqnarray}
566: &&-\sum_{jk} \frac{qV_j V_k}{4} \int \frac{dE}{2\pi }Tr \left[
567: {\bf \Sigma}_{L}^{a} {\bf G}_{1j}^{0<} \right. \nonumber \\
568: && \left. [{\bf G}_{jk}^{0a}(E_{-}) e^{i\Delta_{kj}} 
569: +e^{-i\Delta_{kj}}{\bf G}_{jk}^{0a}(E_+)] {\bf G}_{k 1}^{0a} \right]
570: \label{eq4}
571: \end{eqnarray}
572: The final pumped current is the sum of Eq.(\ref{eq1}), (\ref{eq2}), 
573: (\ref{eq3}), (\ref{eq4}) and their complex conjugates, in addition to 
574: the current directly due to the external bias (see second term of
575: Eq.(\ref{cur1})). If the external bias is zero, the expression of the 
576: pumped current can be simplified significantly.
577: Note that in the equilibrium, the lesser Green's function satisfys
578: the fluctuation-dissipation theorem, 
579: \begin{equation}
580: {\bf G}^{0<}(E)=-f(E)\left[ {\bf G}^{0r}(E)-{\bf G}^{0a}(E)\right]
581: \end{equation}
582: and 
583: \begin{equation}
584: {\bf G}^{0<}(E_{\pm})=-f(E_{\pm})\left[ {\bf G}^{0r}(E_{\pm})
585: -{\bf G}^{0a}(E_{\pm})\right]
586: \end{equation}
587: %In the following, we will use these two equations to simplify the e
588: 
589: (\ref{eq1})+(\ref{eq2})+(\ref{eq4})$^*$ leads to 
590: 
591: \begin{eqnarray}
592: &&-i\sum_{jk} \frac{qV_j V_k}{4} \int \frac{dE}{2\pi } {\rm Tr} 
593: \left[{\bf \Gamma}_L {\bf G}_{1j}^{0r} f(E) \right. \nonumber \\
594: && \left. [{\bf G}_{jk}^{0r}(E_{-})e^{i\Delta_{kj}}
595: +e^{-i\Delta_{kj}}{\bf G}_{kj}^{0r}(E_+)] {\bf G}_{k 1}^{0a} \right] 
596: \label{eq5}
597: \end{eqnarray}
598: 
599: while (\ref{eq1})$^*$+(\ref{eq2})$^*$+(\ref{eq4}) gives to
600: \begin{eqnarray}
601: &&i\sum_{jk} \frac{qV_j V_k}{4} \int \frac{dE}{2\pi } {\rm Tr} 
602: \left[{\bf \Gamma}_L {\bf G}_{1j}^{0r} f(E) \right. \nonumber \\
603: && \left. [{\bf G}_{jk}^{0a}(E_{-})e^{i\Delta_{kj}}
604: +e^{-i\Delta_{kj}}{\bf G}_{jk}^{0a}(E_+)] {\bf G}_{k 1}^{0a} \right] 
605: \label{eq6}
606: \end{eqnarray}
607: 
608: furthermore, (\ref{eq3})+c.c. becomes 
609: \begin{eqnarray}
610: &&i\sum_{jk} \frac{qV_j V_k}{4} \int \frac{dE}{2\pi} {\rm Tr} 
611: \left[{\bf \Gamma}_L {\bf G}_{1j}^{0r} \right. \nonumber \\
612: && \left. [f_{-} ({\bf G}_{jk}^{0r}(E_{-})
613: -{\bf G}_{jk}^{0a}(E_{-})) e^{i\Delta_{kj}} \right. \nonumber \\
614: &&\left.+e^{-i\Delta_{kj}} f_+ ({\bf G}_{jk}^{0r}(E_+)- 
615: {\bf G}_{jk}^{0a}(E_+))] {\bf G}_{k 1}^{0a} \right] 
616: \label{eq7}
617: \end{eqnarray}
618: where $f_{\pm}=f(E_{\pm})$. 
619:  
620: Combining Eqs.(\ref{eq5}), (\ref{eq6}) and (\ref{eq7}), we finally
621: obtain
622: \begin{eqnarray}
623: &&J_L=i\sum_{jk} \frac{qV_j V_k}{4} \int \frac{dE}{2\pi} {\rm Tr} 
624: \left[{\bf \Gamma}_L {\bf G}_{1j}^{0r} \right. \nonumber \\
625: && \left. [(f_{-}-f) ({\bf G}_{jk}^{0r}(E_{-})
626: -{\bf G}_{jk}^{0a}(E_{-})) e^{i\Delta_{kj}} \right. \nonumber \\
627: &&\left.+e^{-i\Delta_{kj}} (f_+-f) ({\bf G}_{jk}^{0r}(E_+)- 
628: {\bf G}_{jk}^{0a}(E_+))] {\bf G}_{k 1}^{0a} \right] 
629: \label{eq8}
630: \end{eqnarray}
631: In the limit of small frequency, we expand Eq.(\ref{eq8}) up to the
632: first order in frequency and use the fact that
633: 
634: \begin{equation}
635: {\bf G}^{0r}-{\bf G}^{0a}=-i{\bf G}^{0r} {\bf \Gamma} {\bf G}^{0a}
636: \end{equation}
637: and Dyson equation,
638: \begin{equation}
639: {\bf G}_{ji}^{0r} {\bf G}_{ik}^{0r}=\frac{\partial 
640: {\bf G}_{jk}^{0r}}{\partial {\bf V_{i}}}
641: \end{equation}
642: we obtain, 
643: \begin{eqnarray}
644: &&J_L=\sum_{jk} \frac{q\omega V_j V_k \sin (\Delta_{kj})}{2} 
645: \int \frac{dE}{2\pi } \partial_{E} f(E) \nonumber \\
646: && {\rm Tr} \left\{ {\bf \Gamma }_{L} {\bf G}_{1j}^{0r}(E)
647: \left[ {\bf G}_{jk}^{0r}(E)-{\bf G}_{jk}^{0a}(E)
648: \right] {\bf G}_{k 1}^{0a}(E)\right\} \nonumber \\
649: &&=-\sum_{jk} \frac{iq\omega V_j V_k \sin (\Delta_{jk})}{2}
650: \int \frac{dE}{2\pi}\partial_{E}f(E) \nonumber \\
651: &&{\rm Tr} \left[{\bf \Gamma }_{L}\frac{\partial 
652: {\bf G}_{11}^{0r}}{\partial {\bf V}_j} {\bf \Gamma }_{L}
653: \frac{\partial {\bf G}_{11}^{0a}}{\partial {\bf V}_k}
654: +{\bf \Gamma }_{L} \frac{\partial {\bf G}_{12}^{0r}}
655: {\partial {\bf V}_j} {\bf \Gamma }_{R}
656: \frac{\partial {\bf G}_{21}^{0a}}{\partial {\bf V}_{k}}\right]
657: \end{eqnarray}
658: which is the same as Ref.\onlinecite{brouwer} when $\theta=0$. 
659: 
660: \section{Results}
661: %RESULTS
662: We now apply our formula Eqs.(\ref{cur1}), (\ref{cur2}) and (\ref{eq8})
663: to a TMR junction. For current perpendicular
664: to the plane geometry, the TMR junction can be modeled by an 
665: one-dimensional quantum structure with a double barrier potential 
666: $U(x)=X_1 \delta (x+a)+X_2 \delta (x-a)$ where $2a$ is the well width. 
667: For this system the Green's function $G(x,x')$ can be calculated 
668: exactly\cite{yip}. The adiabatic pump that we consider 
669: is operated by changing barrier heights adiabatically and periodically: 
670: $X_1=V_0+V_p\sin(\omega t)$ and $X_2=V_0+V_p\sin(\omega t+\phi)$.  
671: %This can be achieved by microfabricating metallic gates at the barrier 
672: %region and applying a time dependent gate potential. 
673: In the following, we will study zero temperature behavior of the pumped 
674: current. In the calculation, we have chosen $M=37.0$ and $V_0=79.2$.
675: Finally the unit is set by $\hbar=2m=1$\cite{foot2}.
676: 
677: We first study the pumped current with two pumping potentials in the
678: adiabatic regime.  Fig.1 depicts the transmission coefficient $T$
679: versus Fermi energy at several angles $\theta$. As expected, among different
680: angles, $T$ is the largest at $\theta=0$ and smallest at $\theta=\pi$
681: with the ratio $T_{max}(0)/T_{max}(\pi) \sim 4$. This gives the usual 
682: spin valve effect\cite{slon}. Fig.2 plots the pumped current versus
683: Fermi energy at different $\theta$. Here we have set the phase difference
684: of two pumping potentials to be $\pi/2$. Similar to the transmission 
685: coefficient, we obtain largest pumped current at $\theta=0$ and smallest
686: current at $\theta=\pi$. We found that the ratio $I_{max}(0)/I_{max}(\pi)$
687: is about the same as that of transmission coefficient. As the pumping 
688: amplitude doubles, the peak of pumped current is broadened and the maximum
689: pumped current is doubled (see inset of Fig.2). This is understandable
690: since at large pumping amplitude the instantaneous resonant level 
691: oscillates with a large amplitude and hence can generate heat current 
692: in a broad range of energy. The spin valve effect of 
693: pumped current is illustrated in Fig.3 where the pumped current versus
694: $\theta$ is shown when the system is at resonance. We see that the pumped
695: current is maximum at $\theta=0$ and decreases quickly as one increases
696: $\theta$ from $0$ to $\pi$. For larger pumping amplitude, we have similar
697: behavior (inset of Fig.3). In Fig.4, we plot the pumped current as a 
698: function of phase difference $\phi$ between two pumping potentials. 
699: We see that the pumped current is antisymmetric about the $\phi=\pi$. 
700: The  nonlinear behavior is clearly seen which deviates from the 
701: sinusoidal behavior at small pumping amplitude. In Fig.5, we show the 
702: pumped current in the presence of external bias. In the calculation we 
703: assume that $V_L=-\omega V/2$ and $V_R=\omega V/2$ so that the external 
704: bias is against the pumped current when $V$ is positive. Due to the 
705: external bias, the total pumped current (dashed line) decreases near
706: resonant energy and 
707: reverses the direction at other energies. Now we turn to the case of finite
708: frequency pumping. We first present our results (Fig.6 to Fig.8) at 
709: small pumping frequency $\omega=0.002$. In Fig.6 we plot the pumped 
710: current as a function of phase difference $\phi$ near the resonant energy. 
711: At $\theta=0$, the magnitude of pumped current is much larger than 
712: that at $\theta=\pi/2$ or $\pi$. We notice that at $\phi=0$ 
713: and $\phi=\pi$, the pumped current is nonzero similar to the experimental 
714: anomaly observed experimentally for nonmagnetic system\cite{switkes}. The 
715: pumped current away from resonant energy is shown in Fig.7. At
716: $\theta=0$, we see that the pumped current is sharply peaked at resonant 
717: energy. The pumped current is positive for $\phi=\pi/2$ and negative 
718: for $\phi=0$ and $\pi$. At $\theta=\pi/2$ or $\pi$, the pumped current at
719: $\phi=\pi/2$ (dashed line) is much larger than that at other angles. 
720: Fig.8 displays the pumped current as a function of $\theta$ near resonant 
721: energy.  For $\phi=\pi/2$ (dotted line), we see the usual behavior 
722: that large pumped current 
723: occurs at $\theta=0$ and it decreases to the minimum at $\theta=\pi$. 
724: For $\phi=0$ or $\pi$, however, we see completely different behavior. 
725: The pumped current is the still the largest at $\theta=0$ but the
726: direction of the pumped current is reversed.  As one increases 
727: $\theta$, the pumped current decreases and reaches a flat region with
728: almost zero pumped current. Now 
729: we study the effect of frequency to the pumped current versus
730: $\theta$ (see Fig.9). We will fix the phase difference to be 
731: $\phi=\pi/2$ and energy near resonance. At small frequency $\omega=0.002$
732: (dashed line), the pumped current versus $\theta$ show usual behavior. 
733: When the frequency is increased to $\omega=0.004$ (dot-dashed line), two 
734: peaks show up symmetrically near $\theta=\pm \pi/4$ while the minimum 
735: is still at $\theta=\pi$. As frequency is increased further to 
736: $\omega=0.006$ (dotted line), the pumped current near 
737: $\theta=0$ reverses the direction and the new peak position shifts to 
738: $\theta \sim 0.35\pi$.  Upon further increasing $\omega$, the curve of 
739: pumped current versus $\theta$ develops a flat region between 
740: $\theta=\pm 0.35\pi$ with positive current while the magnitude of the 
741: negative pumped current at $\theta=0$ becomes larger (see Fig.9b). 
742: Finally, at even 
743: larger frequency $\omega=0.1$, all the pumped currents are negative. 
744: This behavior can be understood from the photon assisted process\cite{wbg1}. 
745: The quantum interference between contributions due to photon emission 
746: (or absorption) near two pumping potentials is essential to understand 
747: the nature of pumped current. In addition to the interference effect, 
748: the pumped current is also affected by a competition of between the photon 
749: emission and absorption processes which tend to cancel to each other. It 
750: is the interplay between this competition and interference that gives rise 
751: to the interesting spin valve effect for the pumped current.
752: Finally, we show in Fig.10 the pumped current versus pumping frequency. 
753: We see that at small frequency the current is positive and small, at large
754: frequency the current is much large and is negative. 
755: 
756: In summary, we have developed a general theory for parametric pumping in
757: the presence of ferromagnetic leads. Our theory is based on the
758: nonequilibrium Green's function method and is valid for multi-modes 
759: (in two or three dimensions) and can be easily extended to the case of 
760: multi-probes (although most of
761: calculations are for two probes). In the parametric pumping, 
762: two kinds of driving forces are present: multiple pumping potentials 
763: inside the scattering system as well as the external bias in the 
764: multi-probes. Two cases are considered. In the adiabatic regime, the
765: system is in near equilibrium.  In this case our theory is for general 
766: pumping amplitude. At finite frequency, the system is away from
767: equilibrium. Our theory is up the quadratic order in pumping amplitude.
768: This theory allows us to examine the pumped current in broader
769: parameter space including pumping amplitude, pumping frequency, phase
770: difference between two pumping potentials, the angle between
771: magnetization of two leads.
772: 
773: \section*{Acknowledgments}
774: We gratefully acknowledge support by a RGC grant from the SAR Government of 
775: Hong Kong under grant number HKU 7091/01P.
776: 
777: \section{Appendix}
778: %APPENDIX
779: 
780: Now we show that 
781: \begin{eqnarray}
782: &&B\equiv \frac{1}{\tau}\int_0^\tau dt
783: \int_{-\infty }^{t}dt_{1} {\rm Tr} [{\bf F}_0(t_1,t)
784: {\bf G}(t,t_1)]
785: \nonumber \\
786: &&=\frac{V_{\alpha} V_\beta}{4}
787: \int \frac{dE}{2\pi}{\rm Tr} \left[{\bf F}_0(E)
788: {\bf F}_1(E)[{\bf F}_2(E_+) e^{i\Delta_{\beta \alpha}} \right. 
789: \nonumber \\
790: &&\left.+e^{-i\Delta_{\beta \alpha}} {\bf F}_2(E_-)] 
791: {\bf F}_3(E) \right] 
792: \label{show}
793: \end{eqnarray}
794: where
795: \begin{equation}
796: {\bf G} \equiv {\bf F}_1 {\bf V}_{\alpha} {\bf F}_2 {\bf V}_{\beta}
797: {\bf F}_3
798: \end{equation}
799: and $F_i$'s ($i=0,1,2,3$) satisfy $F_i(t_1,t_2)=F_i(t_1-t_2)$.
800: 
801: Taking the Fourier transform, 
802: \begin{equation}
803: {\bf F}(t)=\frac{1}{2\pi }\int dE  e^{-iE t}{\bf F}(E)
804: \end{equation}
805: we obtain
806: \begin{eqnarray}
807: &&B=\frac{1}{\tau}\int_0^\tau dt \int_{-\infty }^{t}dt_{1}\int 
808: \frac{dE}{2\pi }{\bf F}_0(E) e^{-iE(t_{1}-t)}
809: \int dxdy \nonumber \\
810: &&[{\bf F}_1(t-x)V_{\alpha}(x) {\bf F}_2(x-y) V_{\beta}(y)
811: {\bf F}_3(y-t_{1})] 
812: \nonumber \\
813: &&=\frac{1}{\tau}\int_0^\tau dt \int \prod_{i=1,5} 
814: \frac{dE_{i}}{2\pi} \int \frac{dE}{2\pi}
815: \int_{-\infty }^{t}dt_{1}\int dx dy \nonumber \\
816: &&{\bf F}_0(E){\bf F}_1(E_1) {\bf V}_\alpha (E_{2})
817: {\bf F}_2(E_3) {\bf V}_\beta (E_4) {\bf F}_3(E_{5})  \nonumber \\
818: &&e^{i(E-E_1)t}e^{i(E_5-E)t_1}e^{i(E_1-E_2-E_3)x}e^{i(E_3-E_4-E_5)y}
819: \nonumber
820: \end{eqnarray}
821: where 
822: \begin{equation}
823: {\bf V}_{\alpha}(E)=\pi V_{\alpha}[e^{i\phi_{\alpha}}\delta (E_+)
824: +e^{-i\phi_{\alpha}}\delta (E_-)]
825: \label{fourier}
826: \end{equation}
827: 
828: Integrating over x and y yields,
829: \begin{eqnarray}
830: &&B=\frac{1}{\tau}\int_0^\tau dt \int \prod_{i=1,5} \frac{dE_{i}}{2\pi} 
831: \int \frac{dE}{2\pi} \int_{-\infty }^{t}dt_{1}{\bf F}_0(E) 
832: \nonumber \\
833: && {\bf F}_1(E_{1}) {\bf V}_\alpha (E_2) {\bf F}_2(E_{3})
834: {\bf V}_\beta (E_4) {\bf F}_3(E_{5}) e^{i(E-E_{1})t}
835: \nonumber \\
836: && e^{i(E_{5}-E)t_{1}}(2\pi )^{2}  
837: \delta (E_{1}-E_{2}-E_{3})\delta (E_{5}-E_{4}-E_{5})  \nonumber
838: \end{eqnarray}
839: 
840: Integrating over $t_1,t$ and using Eq.(\ref{fourier}), we have 
841: \begin{eqnarray}
842: &&B=\int \frac{dE_{2}}{2\pi } \frac{dE_{4}}{2\pi} \frac{dE_{5}}{2\pi}
843: \int \frac{dE}{2\pi }\frac{{\bf F}_0(E)}{i[E_5-E-i\delta]} 
844: \nonumber \\
845: && {\bf F}_1(E_{2}+E_{4}+E_{5}) {\bf V}_{\alpha}(E_{2})
846: {\bf F}_2(E_{4}+E_{5}) \nonumber \\
847: && {\bf V}_{\beta}(E_{4}) {\bf F}_3(E_5)
848: \delta(E_2+E_4)
849: \nonumber \\
850: &&=\frac{V_\alpha V_\beta}{4}\int \frac{dE_{5}}{2\pi } \frac{dE}{2\pi }
851: \frac{{\bf F}_0(E)}{i[E_{5}-E-i\delta ]}
852: {\bf F}_1(E_{5}) \nonumber \\
853: && [{\bf F}_2(E_{5}+\omega) e^{i\Delta_{\beta
854: \alpha}}+e^{-i\Delta_{\beta \alpha}} {\bf F}_2(E_{5}-\omega )]
855: {\bf F}_3(E_{5})  \nonumber 
856: \end{eqnarray}
857: Using the theorem of residue, we have 
858: \begin{equation}
859: \int dE \frac{F_0(E)}{i[E_5-E-i\delta]}=2\pi F_0(E_5)
860: \end{equation}
861: we thus obtain Eq.(\ref{show}).
862: 
863: \bigskip
864: \bigskip
865: \bigskip
866: \noindent{$^{a)}$ Electronic mail: jianwang@hkusub.hku.hk}
867: \begin{thebibliography}{00} 
868: 
869: \bibitem{brouwer}
870: P.W. Brouwer, Phys. Rev. B {\bf 58}, R10135 (1998). 
871: 
872: \bibitem{aleiner}
873: I.L. Aleiner and A.V. Andreev, Phys. Rev. Lett. {\bf 81}, 1286 (1998).
874: 
875: \bibitem{switkes}
876: M. Switkes, C. Marcus, K. Capman, and A.C. Gossard, Science {\bf 283}, 
877: 1905 (1999). 
878: 
879: \bibitem{zhou}
880: F. Zhou, B. Spivak, and B.L. Altshuler, Phys. Rev. Lett. {\bf 82}, 
881: 608 (1999).
882: 
883: \bibitem{shutenko}
884: T.A. Shutenko, I.L. Aleiner, and B.L. Altshuler, Phys. Rev. B {\bf
885: 61}, 10366 (2000).
886: 
887: \bibitem{wei1}
888: Y.D. Wei, J. Wang, and H. Guo, Phys. Rev. B {\bf 62}, 9947 (2000).
889: 
890: \bibitem{aleiner1}
891: I.L. Aleiner, B.L. Altshuler, and A. Kamenev, Phys. Rev. B {\bf 62},
892: 10373 (2000).
893: 
894: \bibitem{avron1}
895: J.E. Avron, A. Elgart, G.M. Graf, and L. Sadun, Phys. Rev. B {\bf
896: 62}, R10618 (2000).
897: 
898: \bibitem{levinson}
899: Y. Levinson, O. Entin-Wohlman, and P. Wolfle, Physica A {\bf 302},
900: 335 (2001). 
901: 
902: \bibitem{brouwer2}
903: P.W. Brouwer, Phys. Rev. B {\bf 63}, 121303 (2001); 
904: M.L. Polianski and P.W. Brouwer, Phys. Rev. B {\bf 64}, 075304
905: (2001).
906: 
907: \bibitem{buttiker1}
908: M. Moskalets and M. Buttiker, Phys. Rev. B {\bf 64}, 201305 (2001).
909: 
910: \bibitem{avron2}
911: J.E. Avron, A. Elgart, G.M. Graf, and L. Sadun, 
912: Phys. Rev. Lett. {\bf 87}, 236601 (2001).
913: 
914: \bibitem{wang1}
915: J. Wang et al, Appl. Phys. Lett. {\bf 79}, 3977 (2001). 
916: 
917: \bibitem{vavilov}
918: M.G. Vavilov, V. Ambegaokar, and I.L. Aleiner, Phys. Rev. B {\bf
919: 63}, 195313 (2001). 
920: 
921: \bibitem{wei2}
922: Y.D. Wei, J. Wang, H. Guo, and C. Roland, Phys. Rev. B {\bf 64}, 
923: 115321 (2001).
924: 
925: \bibitem{renzoni}
926: F. Renzoni and T. Brandes, Phys. Rev. B {\bf 64}, 245301 (2001).
927: 
928: \bibitem{makhlin}
929: Y. Makhlin and A.D. Mirlin, Phys. Rev. Lett. {\bf 87}, 276803 (2001).
930: 
931: \bibitem{chu}
932: C.S. Tang and C.S. Chu, Solid State Comm. {\bf 120}, 353 (2001).
933: 
934: \bibitem{wbg1}
935: B.G. Wang, J. Wang, and H. Guo, Phys. Rev. B {\bf 65}, 073306
936: (2002). 
937: 
938: \bibitem{wang2}
939: J. Wang and B.G. Wang, Phys. Rev. B {\bf 65}, 153311 (2002). 
940: 
941: \bibitem{levinson2}
942: Y. Levinson, O. Entin-Wohlman, and P. Wolfle, cond-mat/0104408.
943: 
944: \bibitem{levinson1}
945: O. Entin-Wohlman, A. Aharony, and Y. Levinson, cond-mat/0201073. 
946: 
947: \bibitem{buttiker2}
948: M. Moskalets and M. Buttiker, cond-mat/0201259.
949: 
950: \bibitem{vavilov1}
951: M.L. Polianski, M.G. Vavilov, and P.W. Brouwer, cond-mat/0202241.
952: 
953: \bibitem{entin1}
954: O. Entin-Wolman and A. Aharony, cond-mat/0202289. 
955: 
956: \bibitem{wbg3}
957: B.G. Wang and J. Wang, cond-mat/0203581.
958: 
959: \bibitem{wbg2}
960: B.G. Wang and J. Wang, cond-mat/0204067. 
961: 
962: \bibitem{blaauboer}
963: M. Blaauboer, cond-mat/0204340.
964: 
965: \bibitem{wu}
966: J.L. Wu, B.G. Wang, and J. Wang, unpublished. 
967: 
968: \bibitem{review}
969: R. Meservey and P.M. Tedrow, Phys. Rep. {\bf 238}, 173 
970: (1994); M. Baibich {\it et.al.}, Phys. Rev. Lett. {\bf 61}, 2472 (1988);
971: G.A. Prinz, Science, {\bf 282}, 1660 (1998);
972: D.J. Monsma, J.C. Lodder, Th.J.A. Popma, and B. Dieny, Phys. Rev. Lett.
973: {\bf 74}, 5260 (1995); D.J. Monsma, R. Vlutters, J.C. Lodder, Science, {\bf
974: 281}, 407 (1998); A. Wolf et al, Science {\bf 294}, 1488 (2001).
975: 
976: \bibitem{slon}
977: J.C. Slonczewski, Phys. Rev. B {\bf 39}, 6995 (1989). 
978: 
979: \bibitem{moodera}
980: J.S. Moodera, L.R. Kinder, T.M. Wong, and R. Meservey, Phys. Rev. Lett.
981: {\bf 74}, 3273 (1995);  S. Zhang, P.M. Levy, A.C. Marley and S.S.P. 
982: Parkin, Phys. Rev. Lett. {\bf 79}, 3744 (1997); X.D. Zhang, B.Z. Li, G.
983: Sun, F.C. Pu, Phys. Rev. B {\bf 56}, 5484 (1997); J.S. Moodera, J. Nowak 
984: and R.J.M. van de Veerdonk, Phys. Rev. Lett. {\bf 80}, 2941 (1998);
985: J. Barna\'s and A. Fert, Phys. Rev. Lett. {\bf 80}, 1058 (1998);
986: L. Sheng, Y. Chen, H.Y. Teng, and C.S. Ting, Phys. Rev. B {\bf 59}, 
987: 480 (1999); K. Tsukagoshi, B.W. Alphenaar, and H. Ago,
988: Nature, {\bf 401}, 572 (1999); H. Mehrez et al, Phys. Rev. Lett.
989: {\bf 84}, 2682 (2000). 
990: 
991: \bibitem{bog}  
992: N.N. Bogoliubov, J. Phys. USSR, {\bf 11}, 23(1947).
993: 
994: \bibitem{wbg4}
995: B.G. Wang, J. Wang, and H. Guo, J. Phys. Soc. Jpn. {\bf 70}, 2645 (2001).
996: 
997: \bibitem{datta}  M. P. Anantram and S. Datta, Phys. Rev. B {\bf 51}, 7632
998: (1995).
999: 
1000: \bibitem{jauho}  
1001: A.P. Jauho, N.S. Wingreen, and Y. Meir, Phys. Rev. B {\bf 50},
1002: 5528 (1994).
1003: 
1004: \bibitem{wbg5}
1005: B.G. Wang, J. Wang, and H. Guo, Phys. Rev. Lett. {\bf 82}, 398 (1999);
1006: J. Appl. Phys. {\bf 86}, 5094 (1999).
1007: 
1008: \bibitem{pre}
1009: C.A. Stafford and N.S. Wingreen, Phys. Rev. Lett. {\bf 76}, 1916(1996).
1010: Q.F. Sun, J. Wang, T.H. Lin, 
1011: Phys. Rev. B {\bf 58}, 13007 (1998); Phys. Rev. B {\bf 59}, 13126 (1999); 
1012: Phys. Rev. B {\bf 61}, 12643, (2000).
1013: H.K. Zhao and J. Wang, Europhys. J. B {\bf 9}, 513 (1999). 
1014: 
1015: \bibitem{but5}
1016: M. B\"uttiker, J. Phys. Condens. Matter {\bf 5}, 9361 (1993). 
1017: 
1018: \bibitem{but6}
1019: T. Gramespacher and M. B\"uttiker, Phys. Rev. B {\bf 56}, 13026 (1997). 
1020: 
1021: \bibitem{foot2}
1022: For the system of Fe/Ge/Fe with $a=100A$, the energy uint is $E=4.64 meV$
1023: which corresponds to frequency $\omega=1.1 \times 10^{12}$ Hz. The
1024: unit of the pumped current in the adiabatic regime depends on the
1025: pumping frequency $\omega$. If $\omega=100$MHz, then the pumped current
1026: is $1.6 \times 10^{-11}$A. For pumped current at finite frequency, the 
1027: unit of current is $2\times10^{-6}$A. 
1028: 
1029: \bibitem{yip}
1030: M.K. Yip, J. Wang, and H. Guo, 
1031: Z. Phys. B: Condens. Matter {\bf 104}, 463 (1997). 
1032: 
1033: \end{thebibliography}
1034: 
1035: \begin{figure}
1036: \caption{
1037: %%% note that in caption, \ref is not allowed.
1038: The transmission coefficient as a function of Fermi energy at different 
1039: angle $\theta$ between the magnetizations of two leads:  $\theta=0$
1040: (dashed line), $\theta=\pi/2$ (solid line), and $\theta=\pi$ (dotted line).
1041: }
1042: \end{figure}
1043: 
1044: \begin{figure}
1045: \caption{
1046: The pumped current as a function of Fermi energy at different 
1047: $\theta$:  $\theta=0$ (solid line), $\theta=\pi/2$ (dot-dashed line), 
1048: and $\theta=\pi$ (dotted line). Other parameters are $\phi=\pi/2$ and 
1049: $V_p=0.05V_0$. Inset: the same as the main figure except $V_p=0.1 V_0$.
1050: }
1051: \end{figure}
1052: 
1053: \begin{figure}
1054: \caption{
1055: The pumped current as a function of $\theta$. Here $\phi=\pi/2$, 
1056: $E_F=37.55$ and $V_p=0.05 V_0$. Inset: the same as main figure except
1057: $V_p=0.1 V_0$. 
1058: }
1059: \end{figure}
1060: 
1061: \begin{figure}
1062: \caption{
1063: The pumped current as a function of phase difference $\phi$ at different
1064: $\theta$. $\theta=0$ (dashed line), $\theta=\pi/2$ (dotted line), 
1065: and $\theta=\pi$ (solid line). Other parameters are $E_F=37.55$ and 
1066: $V_p=0.05 V_0$. 
1067: }
1068: \end{figure}
1069: 
1070: \begin{figure}
1071: \caption{
1072: The pumped current as a function of Fermi energy with an external bias 
1073: $V_L-V_R=-0.02\omega$ at different $\theta$: (a). $\theta=\pi$. 
1074: (b). $\theta=\pi/2$. (c). $\theta=0$. 
1075: In (a), (b) and (c), solid line: pumped current, 
1076: dotted line: current due to external bias, dashed line: total current. 
1077: Here $\phi=\pi/2$ and $V_p=0.05V_0$.
1078: }
1079: \end{figure}
1080: 
1081: \begin{figure}
1082: \caption{
1083: The pumped current as a function of $\phi$ at finite frequency. 
1084: Dashed line: $\theta=0$, dotted line: $\theta=\pi/2$, solid line: 
1085: $\theta=\pi$.  Here $E_F=37.55$ and $\omega=0.002$. 
1086: }
1087: \end{figure}
1088: 
1089: \begin{figure}
1090: \caption{
1091: The pumped current as a function of Fermi energy at finite frequency at
1092: different $\theta$: (a). $\theta=0$. (b). $\theta=\pi/2$. (c). 
1093: $\theta=\pi$. In (a), (b) and (c),  dotted line: $\phi=0$, dashed line: 
1094: $\phi=\pi/2$, solid line: $\phi=\pi$.  Here $\omega=0.002$. 
1095: }
1096: \end{figure}
1097: 
1098: \begin{figure}
1099: \caption{
1100: The pumped current as a function of $\theta$ at finite frequency. 
1101: Here dashed line: $\phi=0$, dotted line: $\phi=\pi/2$, solid line: 
1102: $\phi=\pi$. Other parameters are $E_F=37.55$ and $\omega=0.002$.
1103: }
1104: \end{figure}
1105: 
1106: \begin{figure}
1107: \caption{
1108: The pumped current as a function of $\theta$ at different frequencies.
1109: (a). $\omega=0.002$ (short dashed line), $0.004$ (dot-dashed line), $0.006$
1110: (dotted line), $0.008$ (solid line). (b). $\omega=0.01$ (solid line), 
1111: $0.02$ (dot-dashed line), $0.05$ (short dashed line), $0.1$ (dotted
1112: line).  Other parameters are $E_F=37.55$ and $\phi=\pi/2$.
1113: }
1114: \end{figure}
1115: 
1116: \begin{figure}
1117: \caption{
1118: The pumped current as a function of frequency. Here $\theta=0$, 
1119: $\phi=\pi/2$ and $E_F=37.55$.
1120: }
1121: \end{figure}
1122: 
1123: \end{document}
1124: