1: %\documentstyle[preprint,aps]{revtex}
2: \documentstyle[aps,prbbib,twocolumn,epsf]{revtex}
3: \begin{document}
4: \draft
5: \title{Quantization of adiabatic pumped charge
6: in the presence of superconducting lead}
7:
8: \author{Jian Wang and Baigeng Wang}
9:
10: \address{Department of Physics, The University of Hong Kong,
11: Pokfulam Road, Hong Kong, China}
12: \maketitle
13:
14: \begin{abstract}
15: We investigate the parametric electron pumping of a double barrier
16: structure in the presence of a superconducting lead. The parametric
17: pumping is facilitated by cyclic variation of the barrier heights $x_1$
18: and $x_2$ of the barriers. In the weak coupling regime, there exists
19: a resonance line in the parameter space $(x_1,x_2)$ so that the energy
20: of the
21: quasi-bound state is in line with the incoming Fermi energy. Levinson
22: et al found recently that the pumped charge for each pumping cycle is
23: quantized with $Q=2e$ for normal structure when the pumping contour
24: encircles the resonance line. In the presence of a superconducting lead,
25: we find that the pumped charge is quantized with the value $2e$.
26: \end{abstract}
27:
28: \pacs{73.23.Ad,73.40.Gk,72.10.Bg,74.50.+r}
29:
30: Physics of parametric electron pump has attracted great attention
31: recently\cite{brouwer,switkes,zhou,wagner,Avron,aleiner1,wei1,vavilov,brouwer2,sharma,kravtsov,wei2}.
32: A classical example of electron pump is the Thouless pump facilitated
33: by a traveling wave potential\cite{thouless}. The pumped charge is
34: quantized\cite{thouless} and can be used as a quantum standard for
35: electric charge.\cite{niu}.
36: The quantization of pumped charge has also been studied for a large,
37: almost open quantum dot\cite{aleiner,shutenko} and a small, strongly
38: pinched quantum dot\cite{levinson}. In the latter case, there exists
39: a resonance line along which the transmission through the quantum dot
40: is at resonance. The pumped charge is quantized if the pumping contour
41: in parameter space is properly chosen to encircle the resonance
42: line\cite{levinson}. Recently, we have studied the parametric pumping in
43: presence of a superconducting lead\cite{apl}. At the normal
44: conductor-superconductor
45: (NS) interface, an incoming electron-like excitation can be Andreev
46: reflected as a hole-like excitation\cite{andreev}. In contrast to the
47: current doubling effect\cite{beenakker}, we found that due to the quantum
48: interference of direct reflection and the multiple Andreev reflection,
49: the pumped current is four times of the value when the leads are normal
50: in the {\it weak pumping regime}. In this paper, we explore the effect of
51: superconducting lead on electron pumping in the opposite limit, i.e., we
52: study the pumped charge during the pumping cycle in the the strong pumping
53: regime. Here the pumped charge equal to the pumped current multiplied by
54: the period of pumping cycle. Similar to the Ref.\onlinecite{levinson}, we
55: examine the behavior of pumped charge near the resonance line. We find
56: that the pumped charge in one pumping cycle is quantized with the
57: value of $Q=2e$ when one of the leads is superconducting.\cite{foot2}
58:
59: We consider a parametric pump which consists of a double barrier
60: %tunneling structure\cite{foot9} attached to a normal left lead and a
61: tunneling structure attached to a normal left lead and a
62: superconducting right lead.
63: Due to the cyclic variation of external parameters $x_1$ and $x_2$,
64: the adiabatic charge transfer in the presence of
65: a superconducting lead is\cite{brouwer,foot4,foot1}
66: \begin{equation}
67: Q^{NS}=2e\int_0^\tau dt [\frac{dN_L}{dx_1}
68: \frac{dx_1}{dt} + \frac{dN_L}{dx_2} \frac{dx_2}{dt}]
69: \label{q1}
70: \end{equation}
71: where $\tau$ is the period of cyclic variation and the quantity $dN_L/dx$
72: is the injectivity\cite{buttiker,jwang} given, at zero temperature, by
73: \begin{eqnarray}
74: \frac{dN_L}{dx_j}= \frac{1}{2\pi} Im [{\cal S}^*_{ee}
75: \frac{\partial {\cal S}_{ee}}{\partial x_j}
76: - {\cal S}^*_{he} \frac{\partial {\cal S}_{he}}{ \partial x_j}]
77: \label{inj}
78: \end{eqnarray}
79: where the first term is the injectivity of electron due to the
80: variation of the external parameter\cite{buttiker,jwang}, {\it i.e.} the
81: partial density of states (DOS) for an electron coming from left lead
82: and exiting the system as an electron,
83: and the second term is the injectivity of a hole,
84: {\it i.e.} the DOS for a hole coming from left lead and exiting the
85: system as an electron. Using the Green's theorem, the pumped charge
86: can be expressed as surface integral over area A enclosed
87: by the path $(x_1(t), x_2(t))$ in the parameter space\cite{brouwer}
88: \begin{equation}
89: Q^{NS}=\frac{2e}{\pi}\int_A dx_1 dx_2 \Pi^{NS}(x_1,x_2)
90: \label{q2}
91: \end{equation}
92: with
93: \begin{equation}
94: \Pi^{NS}(x_1,x_2)=
95: Im [\frac{\partial {\cal S}^*_{ee}}
96: {\partial x_1} \frac{\partial {\cal S}_{ee}}{\partial x_2}
97: - \frac{\partial {\cal S}^*_{he}}{\partial x_1} \frac{\partial
98: {\cal S}_{he}}{\partial x_2}]
99: \end{equation}
100: Note that the area $A$ is a measure of variation of pumping parameters
101: $x_1$ and $x_2$. $A$ is very small in the weak pumping limit while
102: remains finite in the strong pumping regime.
103:
104: For the NS structures, the scattering matrix is described by
105: $2\times 2$ matrix $\hat{{\cal S}}$ when the Fermi energy is within
106: the superconducting gap $\Delta$.
107:
108: \begin{equation}
109: \hat{{\cal S}} = \left( \begin{array}{ll}
110: {\cal S}_{ee} & {\cal S}_{eh} \\
111: {\cal S}_{he} & {\cal S}_{hh}
112: \end{array}
113: \right)
114: \end{equation}
115: where ${\cal S}_{ee}$ (or ${\cal S}_{he}$) is the scattering
116: amplitude of the incident electron reflected as an electron (or a
117: hole). Using Andreev approximation\cite{andreev}, we
118: have\cite{beenakker,lesovik}
119: \begin{equation}
120: \hat{{\cal S}} = \hat{S}_{11} + \hat{S}_{12} (1 - \hat{R}_I
121: \hat{S}_{22})^{-1} \hat{R}_I \hat{S}_{21}
122: \label{lesovik}
123: \end{equation}
124: where $\hat{S}_{\beta \gamma}(E)$ ($\beta, \gamma=1,2$) is a diagonal
125: $2 \times 2$ scattering matrix for the double barrier structure with
126: matrix element $S_{\beta \gamma}(E)$ and $S^*_{\beta \gamma}(-E)$.
127: For instance, we have
128:
129: \begin{equation}
130: \hat{S}_{11} = \left( \begin{array}{ll}
131: S_{11}(E) & ~~~ 0 \\
132: 0 & ~~~ S^*_{11}(-E)
133: \end{array}
134: \right)
135: \end{equation}
136: In Eq.(\ref{lesovik}) $\hat{R}_I=\alpha \sigma_x$ is the $2\times 2$
137: scattering matrix at NS interface due to the Andreev reflection
138: with off diagonal matrix element
139: $\alpha$. Here $\alpha = (E-i\nu \sqrt{\Delta^2-E^2})/\Delta$ with
140: $\nu=1$ when $E>-\Delta$ and $\nu=-1$ when $E<-\Delta$. In
141: Eq.(\ref{lesovik}), the energy $E$ is measured relative to the chemical
142: potential $\mu$ of the superconducting lead. Eq.(\ref{lesovik})
143: has a clear physical meaning\cite{lesovik}. The first term is the
144: direct reflection from the normal scattering structure and the second
145: term can be expanded as $\hat{S}_{12} \hat{R}_I \hat{S}_{21} + \hat{S}_{12}
146: \hat{R}_I \hat{S}_{22} \hat{R}_I \hat{S}_{21} + ...$ which is
147: clearly the sum of the multiple Andreev reflection in the hybrid structure.
148: It is the quantum interference of these two terms which gives rise
149: the enhancement of pumped current in the weak pumping regime for NS
150: system\cite{apl}. From Eq.(\ref{lesovik}) we obtain the well known
151: expressions for
152: the scattering matrix ${\cal S}_{ee}$ and ${\cal S}_{he}$\cite{beenakker}
153: \begin{equation}
154: {\cal S}_{ee}(E) = S_{11}(E) + \alpha^2 S_{12}(E) S_{22}^*(-E) M_e
155: S_{21}(E)
156: \label{see}
157: \end{equation}
158:
159: \begin{equation}
160: {\cal S}_{he}(E) = \alpha S_{12}^*(-E) M_e S_{21}(E)
161: \label{she}
162: \end{equation}
163: and
164: \begin{equation}
165: M_e = [1- \alpha^2 S_{22}(E) S_{22}^*(-E)]^{-1}
166: \label{me}
167: \end{equation}
168: The double barrier structure which we consider is modeled by potential
169: $U(y)=V_1 \delta (y+a/2)+V_2 \delta (y-a/2)$ where $V_1$ and $V_2$ are
170: barrier heights which varys in a cyclic fashion to allow the charge
171: pumping. For this system the retarded Green's function $G^r(y,y')$ can be
172: calculated exactly\cite{yip}. This is done by applying the Dyson's equation
173: regarding that any one of the $\delta$-barrier is just a perturbation of
174: the remaining system. This way $G^r(y,y')$ is obtained by applying
175: Dyson's equation twice starting from the Green's function of the
176: one-dimensional free space. With $G^r(y,y')$ we can calculate
177: scattering matrix exactly from the Fisher-Lee relation\cite{lee}
178: \begin{equation}
179: S_{\beta \gamma}=-\delta_{\beta \gamma}+i v G^r_{\beta \gamma}
180: \label{fisher}
181: \end{equation}
182: where $G^r_{\beta \gamma} = G^r(y_\beta,y_\gamma)$ and $v=2k$ is the
183: electron velocity in the normal lead. For normal structure, we
184: have\cite{levinson}
185: \begin{equation}
186: S_{11} = [1-i x_2 - (1+i x_1) \sigma^2]/D
187: \end{equation}
188: \begin{equation}
189: S_{22} = [1-i x_1 - (1+i x_2) \sigma^2]/D
190: \end{equation}
191: and
192: \begin{equation}
193: S_{12} = S_{21} = x_1 x_2 \sigma /D
194: \end{equation}
195: where $D = -(1-i x_1)(1-i x_2)+\sigma^2$, $x_{1,2}=2k V_{1,2}$, and
196: $\sigma=\exp(ika)$. For the double barrier structure, the resonant
197: tunneling is mediated by the quasi-bound state. When the energy of the
198: incident electron is in line with the energy of the quasi-bound
199: state the transmission coefficient reaches maximum. The energy of
200: quasi-bound states can be determined either by looking at the pole of
201: the scattering matrix\cite{levinson} which works well in one
202: dimension or by calculating the dwell time of the incident
203: electron for two or three dimensional systems\cite{apl1}. In the case
204: of double $\delta$ barriers structure, the energy of
205: quasi-bound state is given by\cite{levinson} $E = E_r + \Delta E$
206: with $\Delta E = -(k_r/a)(x_1+x_2)$ where $E_r = k_r^2 = (n\pi/a)^2$
207: is energy of the bound state when the system is isolated. This
208: defines a resonance line $x_1+x_2=-\delta$ in parameter space
209: $(x_1,x_2)$ along which the transmission is at
210: resonance\cite{levinson}. Here $\delta<0$ is the detuning of the Fermi
211: energy from the bound state.
212:
213: To show the quantization of charge transfer in the NS system, it is
214: useful to recall the calculation of the normal case and make the
215: comparison. In the normal case the charge transfer is given
216: by\cite{brouwer,foot2}
217: \begin{equation}
218: Q^{N}=\frac{2e}{\pi}\int_A dx_1 dx_2 \Pi^{N}(x_1,x_2)
219: \label{normal}
220: \end{equation}
221:
222: \begin{equation}
223: \Pi^{N}(x_1,x_2)=
224: Im [\frac{\partial {\cal S}^*_{11}}
225: {\partial x_1} \frac{\partial {\cal S}_{11}}{\partial x_2}
226: + \frac{\partial {\cal S}^*_{12}}{\partial x_1} \frac{\partial
227: {\cal S}_{12}}{\partial x_2}]
228: \label{normal1}
229: \end{equation}
230: The pumped charge in this case has been calculated in
231: Ref.\onlinecite{levinson}. In the {\it weak pumping limit}, it is
232: easy to show that only $\partial_x S_{11}$ contributes to the
233: pumped charge. In the strong pumping regime, we will show in the
234: following that the contribution from $\partial_x S_{12}$ to
235: the pumped charge in normal structure is zero.
236: As discussed in detail in Ref.\onlinecite{levinson}, we neglect
237: the smooth energy dependence of $x_1$ and $x_2$.
238: From Eq.(\ref{normal1}), we obtain the contribution due to
239: $\partial_x S_{12}$
240: \begin{equation}
241: \Pi_1^{N}(x_1,x_2)=F_1(x_1,x_2)/F_2^2(x_1,x_2)
242: \label{pi1}
243: \end{equation}
244: with
245: \begin{equation}
246: F_1(x_1,x_2)= -2x_1 x_2 (x_1-x_2) \sin^2(\delta/2)
247: \label{f1}
248: \end{equation}
249: \begin{eqnarray}
250: F_2(x_1,x_2)&=&x_1^2 x_2^2 +(x_1+x_2)^2 +2(x_1+x_2)\sin\delta
251: \nonumber \\
252: &+&2(1-x_1x_2)(1-\cos\delta)
253: \label{f2}
254: \end{eqnarray}
255: To compute the surface integral of $\Pi^N_1$ in Eq.(\ref{pi1}), it
256: is convenient to change the variables from $x_{1,2}$ to $p$ and
257: $z$:
258:
259: \begin{equation}
260: x_1 = -p ~ \delta ~ (1+z)/2
261: \label{x1}
262: \end{equation}
263: and
264: \begin{equation}
265: x_2 = -p~ \delta ~ (1-z)/2
266: \label{x2}
267: \end{equation}
268: with $0<p<\infty$ and $-1<z<1$. Substituting Eqs.(\ref{x1}) and
269: (\ref{x2}) into Eqs.(\ref{f1}) and (\ref{f2})
270: and expanding Eqs.(\ref{f1}) and (\ref{f2}) in terms of small
271: $\delta$, we have
272: \begin{equation}
273: F_1=z(1-z^2)\delta^5 p^3/8
274: \end{equation}
275: \begin{equation}
276: F_2=\delta^2[(1-p)^2+\delta^2 g(p,z)]
277: \end{equation}
278: where $g(p,z)$ (an even function of $z$) is given in Eq.(8) of
279: Ref.\onlinecite{levinson}. Since $F_1$ is an odd function of $z$, the
280: contribution due to $\partial_x S_{12}$ to the pumped charge is zero.
281:
282: Now we follow the same procedure to calculate the pumped charge for the
283: NS system. For the parametric pumping at zero temperature, we only need
284: the scattering matrix at the Fermi level, i.e., at $E=0$. From
285: Eqs.(\ref{she}) and (\ref{me}), we see that ${\cal S}_{he}$ is a real
286: quantity and hence makes no contribution to the pumped charge in
287: Eq.(\ref{q2}). It is straightforward to calculate $\Pi^{NS}$ using
288: Eq.(\ref{see}), from which we obtain,
289: %\begin{equation}
290: $\Pi^{NS}(x_1,x_2)=F_3(x_1,x_2)/F_4^3(x_1,x_2)$
291: %\end{equation}
292: where
293: %\begin{equation}
294: $F_3 = 4x_1^4 x_2^3 (2-2\cos\delta+x_2\sin\delta)$
295: %\end{equation}
296: and
297: %\begin{eqnarray}
298: %F_4 &=& x_1^2 x_2^2 +2(x_1+x_2)^2+4(x_1+x_2)\sin\delta \nonumber \\
299: %&+&4(1-x_1x_2)(1-\cos\delta)
300: %\end{eqnarray}
301: $F_4 = x_1^2 x_2^2 +2(x_1+x_2)^2+4(x_1+x_2)\sin\delta
302: +4(1-x_1x_2)(1-\cos\delta)$.
303:
304: In Fig.1 we plot both $\Pi^{NS}$ and $\Pi^N$ as well as their
305: cross-sections
306: along and perpendicular to the resonance line. We see that $\Pi^{NS}$
307: and $\Pi^N$ are peaked around the resonance line. Two features are worth
308: noticing. First of all, the peak of $\Pi^{NS}$ is much sharper than that
309: of $\Pi^N$. This is understandable and is due to the resonance nature
310: of NS structures near the resonance line. In the Breit-Wigner form, the
311: transmission coefficients for normal and NS structures are, respectively
312: %\begin{equation}
313: %|S_{21}|^2 = \frac{\Gamma_1 \Gamma_2}{(E-E_r)^2+\Gamma^2/4}
314: %\end{equation}
315: $|S_{21}|^2 = \Gamma_1 \Gamma_2/[(E-E_r)^2+\Gamma^2/4]$
316: and\cite{beenakker}
317: %\begin{equation}
318: %|{\cal S}_{he}|^2 = \frac{4\Gamma_1^2 \Gamma_2^2}{[4(E-E_r)^2
319: %+\Gamma_1^2+\Gamma_2^2]^2}
320: %\end{equation}
321: $|{\cal S}_{he}|^2 = 4\Gamma_1^2 \Gamma_2^2/[4(E-E_r)^2
322: +\Gamma_1^2+\Gamma_2^2]^2$,
323: where $E_r$ is the resonant level, $\Gamma_1$ and $\Gamma_2$ are
324: the decay widths into the left and right lead.
325: Hence $|{\cal S}_{he}|^2$ decays much faster away from $E_r$ than
326: $|S_{21}|^2$. The scattering matrix $S_{21}$ and ${\cal S}_{he}$ will
327: appear, respectively, in Eqs.(\ref{q2}) and (\ref{normal}) implicitly as
328: can be seen from Fisher-Lee relation Eq.(\ref{fisher}) and the Dyson
329: equation $\partial_{X_2} G^r_{11} = G^r_{12} G^r_{21}$\cite{gasparian}.
330: Secondly, the peak height of $\Pi^{NS}$ is four times larger than that of
331: $\Pi^N$. This is precisely due to the constructive interference of direct
332: reflection and multiple Andreev reflection\cite{apl}.
333: Now the physics of pumping at resonance is clear. For the resonance
334: pumping in the weak pumping regime, we are looking at the small
335: neighborhood of the peak. The area of the neighborhood
336: has to be small since it is the weak pumping. The neighborhood has to be
337: around the peak with $x_1 \sim x_2$ since only around the peak the
338: transmission coefficient is approximately one. As a result, we obtain
339: immediately the pumped charge or pumped current of NS structure near
340: the resonance is four times of that of corresponding normal structure.
341: In the other extreme, for strong pumping, we take a large contour
342: enclosing entire resonance line. Since $\Pi^{NS}$ decreases much
343: faster than $\Pi^N$ away from the peak, it is understandable that the
344: pumped charges (the integral of $\Pi$ over the area enclosed by the
345: contour) for both normal and NS structures are equal, which will be
346: shown analytically below.
347:
348: After the expansion in powers of $\delta$ in Eqs.(\ref{f1}) and
349: (\ref{f2}) and keep the leading orders of $\delta$, we have
350: \begin{equation}
351: F_3=p^7[2+p(-1+z)](-1+z)^3 (1+z)^4\frac{\delta^9}{64}
352: \end{equation}
353:
354: \begin{eqnarray}
355: F_4&=&2(1-p)^2 \delta^2 + [-\frac{1}{6}+\frac{2p}{3}
356: \nonumber \\
357: &+&\frac{1}{2}p^2(-1+z^2) +\frac{1}{16}p^4(-1+z^2)^2] \delta^4
358: \end{eqnarray}
359:
360: So Eq.(\ref{q2}) becomes,
361: \begin{equation}
362: Q^{NS} = \frac{e}{\pi} \int_0^{\infty} p dp \int_{-1}^1 dz
363: \frac{F_3} {F_4^3} ~ \delta^2
364: \label{eq1}
365: \end{equation}
366: using the fact that $\lim_{\delta->0} \delta^5/(x^2+\delta^2)^3 =
367: (3/8)\pi \delta(x)$, Eq.(\ref{eq1}) becomes
368:
369: \begin{equation}
370: Q^{NS} = 3\sqrt{2}e\int_{-1}^1 dz \frac{(1-z^2)^3(1+z)^2}{(1
371: +6z^2+z^4)^{5/2}} = 2e
372: \end{equation}
373: Hence the pumped charge for NS system is quantized at the same
374: value of that of the normal structure.
375:
376: Now we have a better physical picture for the transport properties
377: of the NS structure. For the conductance or the I-V curve, we need
378: $S_{21}$ or ${\cal S}_{he}$. For normal structure, the current is
379: given by $I^N = 2e/h \int dE [f(E-eV_1) - f(E-eV_2)] |S_{12}|^2$
380: and hence at resonance and at zero temperature $G^N = I^N/(V_1-V_2)
381: = 2e^2/h$. For NS structure, we
382: have\cite{beenakker,sun} $I^{NS} = 2e/h \int dE
383: (f(E-eV_1)-f(E+eV_1)) |{\cal S}_{he}|^2$ and at resonance $G^{NS} =
384: I^{NS}/V_1 = 4e^2/h$ which is the well known doubling of the conductance.
385: For pumped charge or pumped current at resonance, however, it depends
386: only on $\partial_{x_i} S_{11}$ or $\partial_{x_i} {\cal S}_{ee}$ ($E=0$
387: is assumed). Because of the constructive interference between direct
388: reflection and multiple Andreev reflection in the {\it weak pumping
389: regime}, the charge transfer increases by a factor of four when one of
390: the lead becomes superconducting. In the strong pumping regime,
391: however, the charge transfer
392: is quantized at the value equal to that of normal structure, if the
393: pumping contour is chosen such that the resonance line is enclosed.
394: The physics behind this can be understood as follows. In the normal
395: case, the contour enclosing the resonance line in the parameter space
396: passes through the resonance line at two points $(x_1,x_2)=
397: (0,-\delta)$ when the left contact is almost closed and $(-\delta,0)$
398: when the right contact is almost closed. When passing through those
399: two points, the resonance level of the dot crosses the Fermi
400: energy. At each crossing, the occupation of the level changes, and
401: two electrons with opposite spin enter or exit the region between the
402: barriers. Since one of the tunnel barriers has zero
403: conductance at those points, it is clear that the electrons must
404: have tunneled through the other contact upon entering or leaving
405: the quantum dot. Hence, in the pumping cycle, electrons are
406: shuttled pairwise through the dot. In the presence of
407: superconducting lead, the resonance level (both the energy and the
408: width) is exactly the same as that of normal case since the scattering
409: matrix is given by ${\cal S}_{he} = i |S_{12}|^2/(1+|S_{22}|^2)$
410: when $E=0$. Therefore the same argument applies to the
411: superconducting case and the quantization unit is $2e$. Note that our
412: statement is only valid when the electron interaction is neglected.
413: For the case of two normal-metal contacts, if interactions are included
414: the quantization will remain, but now the quantum is only $e$: Only one
415: electron at a time can enter the region between the barriers; addition of a
416: second electron is forbidden by Coulomb blockade. In the presence of
417: superconducting lead, since the Andreev reflection requires two
418: electrons with opposite spin in order to produce the supercurrent, it
419: seems that the pumping is not allowed in the strong pumping regime due
420: to Coulomb blockade.
421: In this paper, we have also neglected the effect of temperature and the
422: effect of inelastic scattering. As discussed in Ref.\onlinecite{levinson}
423: the temperature will destroy the quantization of the pumped charge. When
424: inelastic channel is present an additional physical mechanism for an
425: incoherent pump effect will show up\cite{buttiker1}.
426:
427: \section*{Acknowledgments}
428: We gratefully acknowledge support by a RGC grant from the SAR Government of
429: Hong Kong under grant number HKU 7215/99P.
430:
431: \medskip
432:
433: \begin{thebibliography}{00}
434:
435: \bibitem{brouwer}
436: P.W. Brouwer, Phys. Rev. B {\bf 58}, R10135 (1998).
437:
438: \bibitem{switkes}
439: M. Switkes, C. Marcus, K. Capman, and A.C. Gossard, Science {\bf 283},
440: 1905 (1999).
441:
442: \bibitem{zhou}
443: F. Zhou, B. Spivak, and B.L. Altshuler, Phys. Rev. Lett. {\bf 82},
444: 608 (1999).
445:
446: \bibitem{wagner}
447: M. Wagner, Phys. Rev. Lett. {\bf 85}, 174 (2000).
448:
449: \bibitem{Avron}
450: J.E. Avron, A. Elgart, G.M. Graf, and L. Sadun, Phys. Rev. B {\bf
451: 62}, R10618 (2000).
452:
453: \bibitem{aleiner1}
454: I.L. Aleiner, B.L. Altshuler, and A. Kamenev, Phys. Rev. B {\bf 62},
455: 10373 (2000).
456:
457: \bibitem{wei1}
458: Y.D. Wei, J. Wang, and H. Guo, Phys. Rev. B {\bf 62}, 9947 (2000).
459:
460: \bibitem{vavilov}
461: M.G. Vavilov, V. Ambegaokar, and I.L. Aleiner, Phys. Rev. B {\bf
462: 63}, 195313 (2001).
463:
464: \bibitem{brouwer2}
465: P.W. Brouwer, Phys. Rev. B {\bf 63}, 121303 (2001);
466: M.L. Polianski and P.W. Brouwer, Phys. Rev. B {\bf 64}, 075304
467: (2001).
468:
469: \bibitem{sharma}
470: P. Sharma and C. Chamon, Phys. Rev. Lett. {\bf 87}, 096401 (2001)
471:
472: \bibitem{kravtsov}
473: X.B. Wang and V.E. Kravtsov, Phys. Rev. B {\bf 64}, 033313 (2001).
474:
475: \bibitem{wei2}
476: Y.D. Wei, J. Wang, H. Guo, and C. Roland, Phys. Rev. B {\bf 64},
477: 115321 (2001).
478:
479: \bibitem{thouless}
480: D.J. Thouless, Phys. Rev. B {\bf 27}, 6083 (1983).
481:
482: \bibitem{niu}
483: Q. Niu, Phys. Rev. Lett. {\bf 64}, 1812 (1990).
484:
485: \bibitem{aleiner}
486: I.L. Aleiner and A.V. Andreev, Phys. Rev. Lett. {\bf 81}, 1286 (1998).
487:
488: \bibitem{shutenko}
489: T.A. Shutenko, I.L. Aleiner, and B.L. Altshuler, Phys. Rev. B {\bf
490: 61}, 10366 (2000).
491:
492: \bibitem{levinson}
493: Y. Levinson, O. Entin-Wohlman, and P. Wolfle, cond-mat/0010494.
494:
495: \bibitem{apl}
496: J. Wang et al, Appl. Phys. Lett. {\bf
497: 79}, 3977 (2001).
498: %J. Wang, Y.D. Wei, B.G. Wang, and H. Guo, Appl. Phys. Lett. {\bf
499: %79}, 3977 (2001).
500:
501: \bibitem{andreev}
502: A.F. Andreev, Zh. Eksp. Teor. Fiz. {\bf 46}, 1823 (1964) [Sov. Phys.
503: JETP {\bf 19}, 1228 (1964)].
504:
505: \bibitem{beenakker}
506: C.W.J. Beenakker, Rev. Mod. Phys. {\bf 69}, 731 (1997).
507:
508: \bibitem{foot2}
509: A factor of two has been included for spin.
510:
511: %\bibitem{foot9}
512: %Our analysis can also be applied to quantum dot if strong
513: %electron-electron interaction can be neglected. For example, one can
514: %consider a 2D quantum dot similar to the experiment of
515: %Ref.\onlinecite{switkes}, but with leads connected to the dot through
516: %narrow constrictions controlled by a gate voltage. Since the threshold of
517: %electron propagation in the constriction maybe lower than that in the
518: %lead, the constrictions act like a double barrier whose height can be
519: %tuned by the gate voltage. However, to analyze such structure one needs
520: %to perform two dimensional calculations.
521:
522: \bibitem{foot4}
523: The units are fixed by setting $\hbar=2m=1$ in the following
524: analysis. For the GaAs system with $a=1000A$, the energy uint is
525: $E=56 \mu eV$.
526:
527: \bibitem{foot1}
528: This formula can be derived using the Keldysh nonequilibrium Green's
529: function method (see Ref.\onlinecite{wbg1}) and we have assumed that
530: the Coulomb blockade effect can be neglected.
531:
532: \bibitem{wbg1}
533: B.G. Wang, J. Wang, and H. Guo, cond-mat/0107078.
534:
535: \bibitem{buttiker}
536: T. Gramespacher and M. Buttiker, Phys. Rev. B {\bf 61}, 8125 (2000).
537:
538: \bibitem{jwang}
539: J. Wang, Y.D. Wei, H. Guo, Q.F. Sun, and T.H. Lin, Phys. Rev. B {\bf
540: 64}, 104508 (2001).
541:
542: \bibitem{lesovik}
543: G.B. Lesovik, A.L. Fauchere, and G. Blatter, Phys. Rev. B {\bf 55},
544: 3146 (1997).
545:
546: \bibitem{lee}
547: D. S. Fisher and P. A. Lee, Phys. Rev. B {\bf 23}, 6851 (1981).
548:
549: \bibitem{apl1}
550: J. Wang, Y.J. Wang, and H. Guo, Appl. Phys. Lett. {\bf 65}, 1793
551: (1994).
552:
553: \bibitem{yip}
554: M.K. Yip, J. Wang, and H. Guo,
555: Z. Phys. B: Condens. Matter {\bf 104}, 463 (1997).
556:
557: \bibitem{gasparian}
558: V. Gasparian, T. Christen, and M. Buttiker, Phys. Rev. A {\bf 54}, 4022
559: (1996).
560:
561: \bibitem{sun}
562: Q.F. Sun, J. Wang, and T.H. Lin, Phys. Rev. B {\bf 59}, 3831 (1999).
563:
564: \bibitem{buttiker1}
565: M. Moskalets and M. Buttiker, cond-mat/0108061.
566:
567: \end{thebibliography}
568:
569: \begin{figure}
570: \caption{
571: %%% note that in caption, \ref is not allowed.
572: %(a). The integrand $\Pi$ of Eqs.(\ref{q2}) and (\ref{normal}) in the
573: (a). The integrand $\Pi$ of Eqs.(3) and (15) as a function of $x_1$
574: and $x_2$ for $\delta=-0.2$. For illustrating purpose, the origin
575: of $\Pi^N(x_1,x_2)$ has been shifted by $(0.1,0.1)$. (b). The
576: cross-section of $\Pi$ along the resonance line $x_1+x_2=-\delta$.
577: Solid line: $\Pi^{NS}$; dotted line: $\Pi^N$. Inset: the cross-section
578: of $\Pi$ along the direction $x_1-x_2=c_0$ which is perpendicular to
579: the resonance line. Left inset: $c_0=0.01$; right inset: $c_0=-0.042$.
580: }
581: \end{figure}
582:
583: \end{document}
584:
585: