1: %\documentstyle[preprint,aps]{revtex}
2: \documentstyle[aps,prbbib,twocolumn,epsf]{revtex}
3:
4: \begin{document}
5: \draft
6: \title{Non-adiabatic charge pump: an exact solution}
7:
8: \author{Baigeng Wang$^1$, Jian Wang$^{1,2,a}$, and Hong Guo$^3$}
9: \address{1. 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: 3. Department of Physics, McGill University, Montreal, Quebec,
14: Canada H3A 2T8}
15: \maketitle
16:
17: \begin{abstract}
18: We derived a general and exact expression of current for quantum parametric
19: charge pumps in the non-adiabatic regime at finite pumping frequency and finite
20: driving amplitude. The non-perturbative theory predicts a remarkable plateau
21: structure in the pumped current due to multi-photon assisted processes in
22: a double-barrier quantum well pump involving only a {\it single} pumping
23: potential. It also predicts a current reversal as the resonant level of the
24: pump crosses the Fermi energy of the leads.
25: \end{abstract}
26:
27: \pacs{73.23.Ad,73.40.Gk,72.10.Bg}
28:
29: Parametric charge pump is a device that drives the flow of a DC electric
30: current by a time dependent variation of {\it two} or more device parameters.
31: The quantum parametric pump has received considerable attention both
32: experimentally and theoretically\cite{brouwer,switkes,zhou,shutenko,wei1,levinson,avron,buttiker1,vavilov1,wang1,wbg1,buttiker2}.
33: A very useful and clear transport theory on quantum parametric pump was
34: provided by Brouwer\cite{brouwer} for the adiabatic regime, {\it i.e.} in the
35: $\omega\rightarrow 0$ limit, where $\omega$ is the frequency of driving force
36: of the pump. Since then there have been several attempts devoted towards the
37: understanding of parametric pumping at finite frequency, {\it i.e.} for the
38: more general non-adiabatic regime\cite{buttiker1,vavilov1,wang1}.
39: These theories\cite{buttiker1,vavilov1,wang1} were all based perturbative
40: schemes either in terms of pumping frequency or in pumping amplitude. Recently,
41: a Floquet scattering matrix theory was developed\cite{buttiker2} and an exact
42: solution obtained for an oscillating double-barrier pump structure. Despite
43: the progress, so far a general and non-perturbative theory for quantum
44: parametric pump has not been reported. Such a theory will be very important
45: to establish a general and unambiguous physical picture of quantum charge
46: pumps for the non-adiabatic regime, and provide new features not accessible
47: by perturbative theories. It is the purpose of this Letter to present the
48: exact solution for the current delivered by a quantum parametric pump at
49: {\it both} finite driving frequency and amplitude.
50:
51: Our theory is based on Keldysh non-equilibrium Green's functions
52: (NEGF)\cite{wbg1} and is non-perturbative. It reproduces not only all
53: previous
54: theories in their perspective limits, but also reveals qualitative new and
55: interesting transport properties of quantum parametric pump. Our theory is
56: also applicable to the case where the driving potential is in the interior
57: of the pump. Because of the generality of our theory, it is now possible
58: to investigate situations beyond the established physics of adiabatic
59: pump. As an example, we investigate a peculiar case of a quantum-well charge
60: pump driven by only a {\it single} parameter: such a pump is only possible
61: in the non-adiabatic regime. In particular, if the pumping potential is
62: located at the barriers of the quantum well, the pumped current exhibits
63: remarkable plateaus due to a multi-photon assisted pumping process. The width
64: of the plateaus is determined by $\hbar \omega$ while their height vary
65: depending on the pumping amplitude and the coupling strength between the
66: pump and the leads. By tuning a gate voltage, the resonant level of the
67: quantum well crosses Fermi energy of the leads and a pumped current reversal
68: is observed. On the other hand, if the pumping potential is at the center
69: point of the quantum well, due to the spatial symmetry the pumped current
70: vanishes exactly. Moving away from this symmetric point, the pumped current
71: increases rapidly by several orders of magnitude and similar plateaus
72: in pumped current are observed.
73:
74: Now we derive the pumped current in the non-adiabatic regime. Assuming that
75: the time-dependent driving potential has the following form
76: $H'(t_1)=(V/2)\exp(i\omega t_1)+(V^*/2)\exp(-i\omega t_1)$ where $V$ is the
77: effective pumping potential profile. Neglecting interaction between electrons
78: in the ideal leads labeled by $L$ and $R$, we start from the expression of
79: time averaged current\cite{jauho} ($\hbar=1$) in standard NEGF theory,
80: \begin{eqnarray}
81: I_{\alpha}&=&-\frac{q}{\tau}\int_0^\tau dt \int dt_{1}
82: Tr[{\bf G}^{r}(t,t_{1}){\bf \Sigma}_{\alpha}^{<}(t_{1},t) \nonumber \\
83: &+&{\bf G}^{<}(t,t_{1}){\bf \Sigma}
84: _{\alpha}^{a}(t_{1},t)+c.c.]
85: \label{X4average}
86: \end{eqnarray}
87: where $\tau=2\pi/\omega$ is the period of pumping cycle. Using the following
88: double time Fourier transform
89: \begin{equation}
90: G^{\gamma}(t_1, t_2) = \int \frac{dE_1}{2\pi} \frac{dE_2}{2\pi}
91: e^{-i E_1 t_1 + i E_2 t_2} G^{\gamma}(E_1, E_2)
92: \label{X2fourier2}
93: \end{equation}
94: with $\gamma=r,a,<$, Eq.(\ref{X4average}) becomes,
95: \begin{eqnarray}
96: &&I_\alpha = -\frac{q}{2N\tau} \int \frac{dE_1}{2\pi} \frac{dE_2}{2\pi}
97: {\rm Tr} \{ [G^r (E_1,E_2) -G^a(E_1,E_2)]
98: \nonumber \\
99: &&\Sigma^<_\alpha(E_2,E_1) + G^<(E_1,E_2) [\Sigma^a_\alpha(E_2,E_1)
100: -\Sigma^r_\alpha(E_2,E_1)]\} \nonumber \\
101: &&\equiv -\frac{q}{8N\tau \pi^2} {\rm Tr} [({\bf G}^r-{\bf G}^a)
102: {\bf \Sigma}^<_L + {\bf G}^< ({\bf \Sigma}^a_L - {\bf \Sigma}^r_L)]
103: \label{new}
104: \end{eqnarray}
105: where we have extended the integration range for $dt$ in Eq.(\ref{X4average})
106: to $[-N\tau,N\tau]$ with $N->\infty$. Note that in Eq.(\ref{new}) the matrix
107: ${\bf G}^\gamma$ has element $G^\gamma(E_1,E_2)$. It is also easy to show that
108: ${\bf G}^r - {\bf G}^a = -i {\bf G}^r {\bf \Gamma} {\bf G}^a$.
109: Using the fact that ${\bf G}^< = {\bf G}^r {\bf \Sigma}^< {\bf G}^a$
110: and $\Sigma^\gamma(E_2,E_1) = 2\pi \delta(E_1-E_2)
111: \Sigma^\gamma(E_1)$ (no external bias), we obtain
112: \begin{eqnarray}
113: I_\alpha &=& \frac{q}{2N\tau}\int \frac{dE_1}{2\pi} \frac{dE_2}{2\pi}
114: {\rm Tr} \left[ \Gamma_\alpha(E_1) G^r(E_1,E_2) \right. \nonumber \\
115: && \left. \Gamma(E_2) G^a(E_2,E_1) \right] (f(E_2)-f(E_1))
116: \label{X4general}
117: \end{eqnarray}
118: Using the identity
119: \begin{equation}
120: \sum_E = 2N\tau \int dE/(2\pi)
121: \label{iden}
122: \end{equation}
123: we finally arrive at
124: \begin{eqnarray}
125: I_\alpha &=& \frac{q}{(2N\tau)^2}\int \frac{dE}{2\pi}
126: \sum_{n=-\infty}^{+\infty}
127: {\rm Tr} \left[ \Gamma_\alpha(E) G^r(E,E+n\omega) \right. \nonumber \\
128: && \left. \Gamma_n(E) G^a(E+n\omega,E) \right] (f_n(E)-f(E))
129: \label{final}
130: \end{eqnarray}
131: where $\Gamma_n(E)=\Gamma(E+n\omega)$ and $f_n(E)=f(E+n\omega)$. Now we
132: proceed to calculate $G^r(E,E+n\omega)$. From the Dyson equation, we have
133: \begin{equation}
134: G^r(t,t')=G^{0r}(t,t')+\int G^r(t,t_1) H'(t_1) G^{0r}(t_1,t') dt_1
135: \label{eq2}
136: \end{equation}
137: where $G^{0r}$ is the equilibrium Green's function. Taking a double-time
138: Fourier transform of Eq.(\ref{eq2}), we find
139: \begin{eqnarray}
140: G^r(E_1,E_2)&=&2\pi G^{0r}(E_1) \delta(E_1-E_2)\nonumber \\
141: &+&\int \frac{dE}{2\pi} G^r(E_1,E_2+E) H'(E) G^{0r}(E_2) \ .
142: \end{eqnarray}
143: Since $H'(E)= \pi [V \delta(E+\omega)+ V^*\delta(E-\omega)]$, we obtain
144: \begin{eqnarray}
145: &&G^r(E_1,E_2)=2\pi G^{0r}(E_1) \delta(E_1-E_2) \nonumber \\
146: &+&[G^r(E_1,E_2-\omega) V+ G^r(E_1,E_2+\omega) V^*] G^{0r}(E_2)/2\ .
147: \label{X4gr1}
148: \end{eqnarray}
149: Using this equation, the general expression of $G^r(E,E+n\omega)$ can
150: be obtained. To simplify notation, we use the following abbreviations
151: $G^r_n(E) \equiv G^r(E,E+n\omega)$ and $G^{0r}_n(E) \equiv G^{0r}(E+n\omega)$.
152: We then have
153: \begin{eqnarray}
154: && G^r_0 = 2\pi G^{0r}_0 \delta(0) + (G^r_1 V^* + G^r_{-1} V) G^{0r}_0/2
155: \nonumber \\
156: && G^r_1 = (G^r_2 V^* + G^r_0 V) G^{0r}_1/2
157: \nonumber \\
158: && G^r_{-1} = (G^r_0 V^* + G^r_{-2} V) G^{0r}_{-1}/2
159: \label{eq1}
160: \end{eqnarray}
161: If we restrict ourselves for single photon process only, we can neglect
162: $G^r_2$ in Eq.(\ref{eq1}) and obtain
163: \begin{equation}
164: G^r_0(E) = \frac{2\pi \delta(0)}{[G^{0r}_0(E)]^{-1}-\Sigma^r_1(E)}
165: \label{X4gr1a}
166: \end{equation}
167: where $2\pi\delta(0)=\int dE=2N\tau$. Here the self-energy
168: $\Sigma^r_1(E) \equiv (V G^{0r}_1 V^*+ V^* G^{0r}_{-1} V)/4$.
169:
170: The exact solution of $G^r_n$ by including all photon processes can be
171: obtained by iterating the following equation obtained from Eq.(\ref{X4gr1}),
172: \begin{equation}
173: G^r_n = (G^r_{n+1} V^* + G^r_{n-1} V) G^{0r}_n/2
174: \end{equation}
175: we obtain
176: \begin{equation}
177: G^r_n = G^r_{n-1} V G^{0r}_n \frac{1}{2\alpha^r_n}
178: \label{grn}
179: \end{equation}
180: where $\alpha^r_n=[(a_n,\bar{a}_n),(a_{n+1},\bar{a}_{n+1}),
181: ...]$ with $a_n=iV G^{0r}_{n+1}/2$ and
182: $\bar{a}_n=iV^* G^{0r}_n/2$. The continued fraction
183: $[(a_1,\bar{a}_1),(a_{2},\bar{a}_2),...]$ is defined as
184: \begin{equation}
185: [(a_1,\bar{a}_1),(a_{2},\bar{a}_2),...]
186: =1+a_1\frac{1}
187: {\displaystyle 1+a_2\frac{1}{\displaystyle
188: 1+... }\bar{a}_2}\bar{a}_1
189: \end{equation}
190: From Eq.(\ref{grn}), we obtain $G^r_1=G^r_0 V G^{0r}_1/(2\alpha^r_1)$
191: and $G^r_{-1}=G^r_0 V^* G^{0r}_{-1}/(2\beta^r_{-1})$. Here
192: $\beta^r_{-n}=[(b_{-n},\bar{b}_{-n}),(b_{-n-1},\bar{b}_{-n-1}),...]$
193: with $b_{-n}=i V^* G^{0r}_{-n-1}/2$ and $\bar{b}_{-n}=i V G^{0r}_{-n}/2$.
194: Substituting the expressions of $G^r_{\pm 1}$ into Eq.(\ref{eq1}), we have
195: \begin{equation}
196: G^r_0(E) = \frac{2\pi \delta(0)}{[G^{0r}_0(E)]^{-1}-\Sigma^r(E)}
197: \label{X4gr1b}
198: \end{equation}
199: with
200: \begin{equation}
201: \Sigma^r= V G^{0r}_1 \frac{1}{4\alpha^r_1} V^*
202: +V^* G^{0r}_{-1} \frac{1}{4\beta^r_{-1}} V
203: \label{self}
204: \end{equation}
205: Once $G^r_0$ is obtained, $G^r_n$ can be calculated recursively from
206: Eq.(\ref{grn}). Eqs.(\ref{final}), (\ref{grn}), (\ref{X4gr1b}) and
207: (\ref{self}) form the central result of this paper\cite{foot2}.
208: Note that $G^{0r}_n$ is the Green's function for the process of absorbing
209: (emitting) n-photons whereas $G^r_n$ is the renormalized n-photon Green's
210: function containing the contribution of multi-photon process. From
211: Eq.(\ref{grn}), we see that $G^r_n$ is at least of the order of $V^n$.
212: For this reason, as we will see below that in the resonant tunneling regime,
213: the summation in Eq.(\ref{final}) converges rapidly even in the strong pumping
214: regime. Using the fact that ${\bf G}^r {\bf \Gamma} {\bf G}^a
215: ={\bf G}^a {\bf \Gamma} {\bf G}^r$, it is easy to show $\sum_\alpha
216: I_\alpha=0$, {\it i.e.}, current conservation for parametric pumping.
217:
218: Eq.(\ref{final}) reproduces all previously known results of parametric
219: pumping. We give two examples here. First, in Ref.\onlinecite{wbg1}, the
220: pumped current is obtained using perturbation theory for finite pumping
221: frequency up to the second order in pumping amplitude\cite{foot1}. This
222: result is recovered by only keeping the terms $n=\pm 1$ in Eq.(\ref{final})
223: to obtain $G^r_1 = G^{0r}_0 V G^{0r}_1/2$. Hence Eq.(\ref{final}) becomes,
224: \begin{equation}
225: I_\alpha = q\int \frac{dE}{8\pi} \sum_{n=\pm 1} {\rm Tr} \left[
226: \Gamma_\alpha G^{0r}_0 V G^{0r}_n \Gamma_n G^{0a}_n V^* G^{0a}_0 \right]
227: (f_n-f)
228: \label{wbg1}
229: \end{equation}
230: which is exactly the same as that of Ref.\onlinecite{wbg1,foot1}. Second, we
231: consider the well-studied adiabatic regime $\omega\rightarrow 0$ where
232: the instantaneous approximation for the Green's function is appropriate.
233: We transform Eq.(\ref{final}) to the Wigner representation using
234: \begin{equation}
235: G^{\gamma}(t_1, t_2) = \int \frac{dE}{2\pi}
236: e^{-i E ( t_1 - t_2)} {\cal G}^{\gamma}(E, T)
237: \end{equation}
238: where $T=(t_1+t_2)/2$ and ${\cal G}^\gamma$ is the NEGF in the Wigner
239: representation. We then have\cite{foot3}
240: \begin{equation}
241: G^r_n=\int dT {\cal G}^r(E,T) e^{-i n \omega T}
242: \label{wigner}
243: \end{equation}
244: Substituting Eq.(\ref{wigner}) into Eq.(\ref{final}) and keep only the
245: $O(\omega)$ term, we obtain
246: \begin{eqnarray}
247: &&I_\alpha = \frac{q}{(2N\tau)^2}\int \frac{dE}{2\pi}
248: \int_{-\infty}^\infty dT dT' \sum_{n=-\infty}^{+\infty} n\omega ~
249: {\rm Tr} \left[ \Gamma_\alpha(E) \right. \nonumber \\
250: && \left. {\cal G}^r(E,T) \Gamma(E) {\cal G}^a(E,T') \right]
251: \exp(-i n \omega (T-T')) \partial_E f(E)
252: \label{temp1}
253: \end{eqnarray}
254: Note that
255: \begin{eqnarray}
256: &&\int dT \sum_{n=-\infty}^{+\infty} n\omega ~ {\cal G}^r(E,T)
257: \exp(-i n \omega (T-T')) \nonumber \\
258: &&= -2i N\tau ~ \partial_{T'} {\cal G}^r(E,T')
259: \end{eqnarray}
260: where we have used Eq.(\ref{iden}). Eq.(\ref{temp1}) now becomes,
261: \begin{eqnarray}
262: I_\alpha = \frac{-iq}{\tau} \int_0^\tau dT \int \frac{dE}{2\pi}
263: {\rm Tr} \left[ \Gamma_\alpha \partial_T {\cal G}^r(T)
264: \Gamma {\cal G}^a(T) \right] \partial_E f\ ,
265: \label{temp2}
266: \end{eqnarray}
267: this is exactly the Brouwer's formula\cite{brouwer,buttiker2} for adiabatic
268: pumping.
269:
270: In the following we calculate pumped current using Eq.(\ref{final}) for
271: a symmetric double $\delta$-barrier structure given by
272: $U(x)=V_0 \delta (x+a)+V_0 \delta (x-a)$. For this system the Green's function
273: $G(x,x')$ can be calculated exactly\cite{yip}. In the adiabatic theory, two
274: pumping potentials are needed in order to give a nonzero pumping current.
275: At finite frequency, a {\it single} pumping potential was reported to be
276: enough to pump a current\cite{wbg1} which is peculiar, and our exact
277: non-perturbative theory derived above allows us to investigate this
278: situation clearly and unambiguously. We chose the single pumping potential
279: to be sinusoidal $V(x,t)=V_p \cos(\omega t) \delta(x-x_0)$, and we
280: calculate pumped current from the left lead at zero temperature and set
281: $V_0=200$ unless specified otherwise. A gate voltage $v_g$ is applied to the
282: double barrier structure so that the resonant single particle energy level
283: is controlled by it. We fix the Fermi level of leads in line with the resonant
284: level at $v_g=0$. Finally the unit is set by $\hbar=2m=q=2a=1$. For GaAs
285: material with well width $a=1000\AA$, the energy unit is $E=0.056meV$
286: which corresponds to $\omega=13.2$ GHz. The unit for pumped current is
287: $5\times 10^{-10}$ A.
288:
289: Fig.1 shows the pumped current versus the gate voltage in the strong
290: pumping regime for $x_0=-a$, $V_p=160$, and $\omega=1$. Following observations
291: are in order. (1) The pumped current displays a series of remarkable
292: plateaus. The width of a plateau is equal to $\hbar\omega$ whereas the height
293: of the plateau depends on $V_p$ and $V_0$ in a nonlinear fashion.
294: (2) The pumped current reverses its direction when the resonant level across
295: the Fermi level of the lead, {\it i.e.}, when $v_g$ changes sign. As a result,
296: we see from Fig.1 that the pumped current is quenched near $v_g=0$. (3) The
297: height of the current plateaus for negative $v_g$ is considerablely
298: larger than those of positive $v_g$. To understand these results, we rewrite
299: Eq.(\ref{final}) as
300: \begin{equation}
301: I_L = q \int \frac{dE}{2\pi} \sum_{n=1}^{+\infty}
302: F_n(E) (f_n(E)-f(E))
303: \label{final1}
304: \end{equation}
305: with
306: \begin{eqnarray}
307: F_n(E)&=& \frac{q}{(2N\tau)^2} {\rm Tr} \left[ \Gamma_L(E)
308: G^r(E,E+n\omega) \Gamma_n(E) \right. \nonumber \\
309: && \left. G^a(E+n\omega,E) - \Gamma_{L n}(E) G^r(E+n\omega,E)
310: \right. \nonumber \\
311: && \left. \Gamma(E) G^a(E,E+n\omega) \right]
312: \end{eqnarray}
313: We notice that kernel $F_n$ consists of two terms: one due to photon
314: absorption process\cite{foot6} and the other to photon emission process.
315: These two contributions differ by a sign indicating the competition
316: between the photon absorption and emission processes similar to that
317: discussed in Ref.\onlinecite{wbg1}. In Fig.2 we plot the integrand of
318: Eq.(\ref{final1}),
319: $F_n(E)$, for $n=1,2,3$ (solid line, dotted line, dashed line,
320: respectively). The sidebands due to photon absorption (when $v_g<0$) and
321: photon emission are clearly seen (for an n-photon emission process, the
322: energy has been shifted by $n\omega$ due to the transformation
323: from Eq.(\ref{final}) to Eq.(\ref{final1})). To obtain pumped current,
324: one integrates these sidebands over energy. For the n-th sideband ($n>0$),
325: the integral range is $[v_g-n\omega, ~v_g]$: it is therefore clear that the
326: plateau structure is a direct result of multi-photon assisted processes.
327: The positive current is due to photon absorption processes and the
328: current reverses its sign if photon emission process dominates. The behavior
329: of current reversal has been observed before for pumping in Carbon
330: nanotubes\cite{wei2}, charge quantization\cite{aharony}, and heat current
331: generated during the pumping\cite{buttiker2}, although from completely
332: different origins. Due to the energy dependence of coupling between the
333: pump and the leads, the sideband is asymmetric with a larger peak for
334: photon absorption process. As a result, the height of the current plateau
335: is larger for photon absorption process. Our numerical plots show that this
336: plateau structure persists for different frequencies and pumping amplitudes.
337: Our analysis show that there are two effects which affect the current
338: plateau structure. The first is barrier heights of the quantum well:
339: current plateaus can only be observed in the strong tunneling regime
340: and they disappear for low barrier height. The second is temperature:
341: the plateau structure is rounded off at finite temperature and destroyed
342: when temperature reaches $\sim 200 mK$ for $\omega=10$GHz.
343:
344: When the pumping potential is inside of the double-barrier structure, we
345: found a similar plateau structure due to the same photon assisted processes.
346: In Fig.3, the pumped current as a function of pumping position $x_0$ is
347: plotted at a fixed $v_g=-0.67$ for a much smaller pumping amplitude $v_p=6$.
348: We observe that the pumped current $I_p$ is {\it antisymmetric} about the
349: center of the pump. As expected for a symmetric structure, $I_p$
350: vanishes identically at $x_0=0$. At $x_0=-0.5$, we found $I_p=2\times 10^{-6}$.
351: Fig.3 shows that $I_p$ can increase by three orders of magnitude by varying
352: the position of the pumping potential away from the barriers.
353:
354: In summary, we have derived a non-perturbative and exact current expression
355: for parametric pumping at finite frequency and finite pumping amplitude.
356: This theory can also account for the case of pumping potential located in
357: the {\it interior} of the scattering region. Our theory reproduces all
358: previously known results in both adiabatic regime and in perturbative theory.
359: For a double-barrier quantum well pump, we predicted current plateaus to
360: appear due to multi-photon assisted processes with the plateau width
361: given by the pumping frequency. As the gate voltage is varied and the
362: resonance level sweeps through the Fermi energy of leads, the pumped current
363: is found to reverse its direction.
364:
365: \section*{Acknowledgments}
366: We gratefully acknowledge support by a RGC grant from the SAR Government of
367: Hong Kong under grant number HKU 7113/02P and from NSERC of Canada and FCAR
368: of Quebec (H.G).
369:
370: \bigskip
371:
372: \noindent{$^{a)}$ Electronic mail: jianwang@hkusub.hku.hk}
373:
374: \begin{thebibliography}{00}
375:
376: \bibitem{brouwer}
377: P.W. Brouwer, Phys. Rev. B {\bf 58}, R10135 (1998).
378:
379: \bibitem{switkes}
380: %M. Switkes, C. Marcus, K. Capman, and A.C. Gossard, Science {\bf 283},
381: %1905 (1999).
382: M. Switkes et al, Science {\bf 283}, 1905 (1999).
383:
384: \bibitem{zhou}
385: %F. Zhou, B. Spivak, and B.L. Altshuler, Phys. Rev. Lett. {\bf 82},
386: %608 (1999).
387: F. Zhou et al, Phys. Rev. Lett. {\bf 82}, 608 (1999).
388:
389: \bibitem{shutenko}
390: %T.A. Shutenko, I.L. Aleiner, and B.L. Altshuler, Phys. Rev. B {\bf
391: %61}, 10366 (2000).
392: T.A. Shutenko et al, Phys. Rev. B {\bf 61}, 10366 (2000).
393:
394: \bibitem{wei1}
395: %Y.D. Wei, J. Wang, and H. Guo, Phys. Rev. B {\bf 62}, 9947 (2000);
396: Y.D. Wei et al, Phys. Rev. B {\bf 62}, 9947 (2000).
397:
398: \bibitem{levinson}
399: %Y. Levinson, O. Entin-Wohlman, and P. Wolfle, Physica A
400: %{\bf 302}, 335 (2001).
401: Y. Levinson et al, Physica A {\bf 302}, 335 (2001).
402:
403: \bibitem{avron}
404: %J.E. Avron, A. Elgart, G.M. Graf, and L. Sadun, Phys. Rev. Lett.
405: %{\bf 87}, 236601 (2001).
406: J.E. Avron et al, Phys. Rev. Lett. {\bf 87}, 236601 (2001).
407:
408: \bibitem{buttiker1}
409: M. Moskalets and M. Buttiker, Phys. Rev. B {\bf 66}, 035306 (2002).
410:
411: \bibitem{vavilov1}
412: %M.L. Polianski, M.G. Vavilov, and P.W. Brouwer, Phys. Rev. B {\bf 65},
413: %245314 (2002).
414: M.L. Polianski et al, Phys. Rev. B {\bf 65}, 245314 (2002).
415:
416: \bibitem{wang1}
417: B.G. Wang and J. Wang, Phys. Rev. B {\bf 66}, 125310 (2002);
418: cond-mat/0207495.
419:
420: \bibitem{wbg1}
421: %B.G. Wang, J. Wang, and H. Guo, Phys. Rev. B {\bf 65}, 073306 (2002).
422: B.G. Wang et al, Phys. Rev. B {\bf 65}, 073306 (2002).
423:
424: \bibitem{buttiker2}
425: M. Moskalets and M. Buttiker, cond-mat/0208356.
426:
427: \bibitem{jauho}
428: %A.P. Jauho, N.S. Wingreen, and Y. Meir, Phys. Rev. B {\bf 50},
429: %5528 (1994).
430: A.P. Jauho et al, Phys. Rev. B {\bf 50}, 5528 (1994).
431:
432: \bibitem{foot2}
433: The factor $(2\pi\delta(0))^2$ from $G^r$ and $G^a$ cancels $(2N\tau)^2$
434: in Eq.(\ref{final}).
435:
436: \bibitem{foot1}
437: The result in Ref.\onlinecite{wbg1} has also been obtained using scattering
438: matrix approach in Ref.\onlinecite{buttiker1}.
439:
440: \bibitem{foot3}
441: Similar to Ref.\onlinecite{buttiker2}, we have used the adiabatic
442: approximation so that the ${\cal G}^r(E+n\omega/2,T)= {\cal G}^r(E,T)$.
443:
444: \bibitem{yip}
445: %M.K. Yip, J. Wang, and H. Guo,
446: %Z. Phys. B: Condens. Matter {\bf 104}, 463 (1997).
447: M.K. Yip et al, Z. Phys. B: Condens. Matter {\bf 104}, 463 (1997).
448:
449: \bibitem{foot6}
450: Since $G^r_n$ is the renormalized Green's function, it also has the photon
451: emission component through the dependence of $\beta^r_{-n}$.
452:
453: \bibitem{wei2}
454: Y.D. Wei et al, Phys. Rev. B {\bf 64}, 115321 (2001).
455:
456: \bibitem{aharony}
457: O. Entin-Wohlman and A. Aharony, Phys. Rev. B {\bf 66}, 035329 (2002).
458:
459: \end{thebibliography}
460:
461: \begin{figure}
462: \caption{
463: The pumped current versus the gate voltage for $x_0=-a$, $\omega=1$, and
464: $v_p=160$.
465: }
466: \end{figure}
467:
468: \begin{figure}
469: \caption{
470: The integrand of Eq.(23) versus gate voltage for $n=1$ (solid line),
471: $n=2$ (dotted line), and $n=3$ (dash-dotted line). The system
472: parameters are the same as that of Fig.1. For illustrating purpose, we have
473: offset the curve by $0.02$.
474: }
475: \end{figure}
476:
477: \begin{figure}
478: \caption{
479: The pumped current versus position of the pumping potential. Here
480: $v_g=-0.67$, $\omega=1$, and $v_p=6$.
481: }
482: \end{figure}
483:
484:
485:
486: \end{document}
487:
488: