1: %% LyX 1.3 created this file. For more info, see http://www.lyx.org/.
2: %% Do not edit unless you really know what you are doing.
3: \documentclass[twocolumn,english,aps,prb,floatfix]{revtex4}
4: \usepackage[T1]{fontenc}
5: \usepackage[latin2]{inputenc}
6: \usepackage{amsmath}
7: \usepackage{graphicx}
8: \usepackage{amssymb}
9:
10: \makeatletter
11: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% User specified LaTeX commands.
12: \usepackage{bm}
13:
14: \usepackage{babel}
15: \makeatother
16: \begin{document}
17:
18: \title{Conductance of interacting Aharonov-Bohm systems}
19:
20:
21: \author{T. Rejec$^{1}$ and A. Ram\v{s}ak$^{1,2}$}
22:
23:
24: \affiliation{$^{1}$Jo\v{z}ef Stefan Institute, Jamova 39, SI-1000 Ljubljana, Slovenia\\
25: $^{2}$Faculty of Mathematics and Physics, University of Ljubljana,
26: Jadranska 19, SI-1000 Ljubljana, Slovenia }
27:
28: \begin{abstract}
29: A simple formula for the zero-temperature linear response conductance
30: of an interacting mesoscopic region, threaded by magnetic flux, and
31: attached to noninteracting single-channel leads is presented. The
32: formula is valid for a general interacting system exhibiting Fermi
33: liquid properties. As an example of the efficiency of the formula
34: the results for the conductance of a simple Aharonov-Bohm ring with
35: Kondo-Fano resonance physics are presented and compared with numerical
36: renormalization group results.
37: \end{abstract}
38:
39: \pacs{73.23.-b, 73.63.-b}
40:
41:
42: \date{13 March 2003}
43:
44: \maketitle
45: The electron-electron interaction often plays a crucial role in transport
46: through mesoscopic systems. The Kondo effect in quantum dots is e.g.
47: a prototype phenomenon, where correlations dominate the conductance~\cite{Glazman88,Goldhaber98}.
48: Another remarkable effect is the Fano resonance physics intertwined
49: with Kondo physics~\cite{Hofstetter01,Bulka01}. Recent advances
50: in nanofabrication of Aharonov-Bohm (AB) devices made possible a realization
51: and a systematic study of such phenomena~\cite{ABexp}.
52:
53: For systems where the interaction is absent, or very weak, transport
54: properties may be determined using the Landauer-B\"{u}ttiker formalism~\cite{Landauer57}.
55: On the other hand, if the interaction is important in the system being
56: studied, the use of a more general approach is essential. An appropriate
57: formalism, expressing the conductance in terms of non-equilibrium
58: Green's functions, was developed by Meir and Wingreen~\cite{Meir92}.
59: The formalism can in principle be used to treat systems at a finite
60: temperature, finite source-drain voltage, and can even be extended
61: to describe time-dependent transport phenomena~\cite{Jauho94}. Another
62: approach to describing the transport in interacting systems, applicable
63: only to systems in the linear-response regime, is the Kubo formalism~\cite{Oguri}.
64:
65: Recently, a simple method for calculating the conductance through
66: a region with electron-electron interaction (e.g., a molecule, a quantum
67: dot, a quantum dot array or a similar 'artificial molecule' system,
68: ...), connected to single-channel noninteracting leads, was presented
69: in Ref.~\onlinecite{Rejec03} (hereafter referred to as RR)~\cite{Favand98}.
70: The conductance is determined solely from the ground-state energy
71: of an auxiliary system, formed by connecting the ends of the leads
72: of the original system into a ring and threaded by magnetic flux.
73: The method is applicable to Fermi liquid systems at zero temperature
74: and in the linear response regime. The validity is additionally restricted
75: to the class of interacting systems obeying time reversal symmetry.
76: Several conductance formulae, becoming exact in the limit of a very
77: large ring, are derived, the most practical being the two-point formula,
78:
79: \begin{equation}
80: G=G_{0}\sin^{2}\left(\frac{\pi}{2}\frac{E\left(\pi\right)-E\left(0\right)}{\Delta}\right),\label{eq:gsin}\end{equation}
81: where $E\left(0\right)$ and $E\left(\pi\right)$ are the ground-state
82: energies of the auxiliary interacting system with periodic and antiperiodic
83: boundary conditions, respectively. The average level spacing at the
84: Fermi energy $\Delta=[N\rho\left(\epsilon_{F}\right)]^{-1}$ is determined
85: by the density of states in an infinite lead $\rho\left(\epsilon\right)$
86: and the circumference of the ring $N$, and $G_{0}=\frac{2e^{2}}{h}$
87: is the conductance quantum. The corresponding ground-state wave function
88: can be chosen real, which simplifies the determination of the ground-state
89: energy by, for example, variational or density matrix renormalization
90: group approaches.
91:
92: %
93: \begin{figure}[t]
94: \begin{center}\includegraphics[%
95: bb=210bp 0bp 610bp 652bp,
96: clip,
97: width=6cm,
98: keepaspectratio]{Fig1.eps}\end{center}
99:
100:
101: \caption{\label{cap:Fig1}(a) Interacting mesoscopic region (gray-shaded sites),
102: threaded by magnetic flux and coupled to noninteracting leads. (b)
103: Auxiliary ring system. (c) Behavior of energy levels as the flux threading
104: the ring is varied. (d) System from Ref.~\onlinecite{Hofstetter01}
105: with a quantum dot embedded in an AB ring.}
106: \end{figure}
107:
108:
109: In this paper we present a generalization of the conductance formula
110: Eq.~(\ref{eq:gsin}) to systems which explicitly exhibit time reversal
111: asymmetry, such as is e.g. an AB-type of system presented in Fig.~\ref{cap:Fig1}(a).
112: The formalism presented here is valid for general systems consisting
113: of interconnected sites with interaction, threaded with magnetic flux
114: and connected to noninteracting leads.
115:
116: First we consider the noninteracting case and single-electron states.
117: The transmission amplitudes $t_{k}$ and $t_{k}^{\prime}$ of an electron
118: with a wave vector $k$ and an energy $\varepsilon_{k}$ describe
119: the transmission from the left to the right lead and from the right
120: to the left lead of the system presented in Fig.~\ref{cap:Fig1}(a),
121: respectively. They are related by an expression derived in RR,
122:
123: \begin{equation}
124: t_{k}^{\prime}e^{i\phi}+t_{k}e^{-i\phi}=e^{ikN}+\frac{t_{k}}{t_{k}^{\prime\ast}}e^{-ikN},\label{eq:ttprime}\end{equation}
125: where $\Phi=\frac{\hbar}{e}\phi$ is the magnetic flux through the
126: auxiliary system shown in Fig.~\ref{cap:Fig1}(b). If there is no
127: AB flux threading the mesoscopic region, the time reversal symmetry
128: is restored, $t_{k}=t_{k}'$ and the energy is an even function of
129: $\phi$, as illustrated in Fig.~\ref{cap:Fig1}(c). In a general
130: case, the unitarity of the scattering matrix requires $\left|t_{k}\right|=\left|t_{k}^{\prime}\right|$
131: and the transmission amplitudes are related as $t_{k}=\tilde{t}_{k}e^{i\phi_{0k}}$
132: and $t_{k}'=\tilde{t}_{k}e^{-i\phi_{0k}}$. Expressing $\tilde{t}_{k}$
133: in terms of its modulus $\left|t_{k}\right|$ and a phase-shift $\varphi_{k}$,
134: $\tilde{t}_{k}=\left|t_{k}\right|e^{i\varphi_{k}}$, the eigenenergy
135: equation Eq.~(\ref{eq:ttprime}) reads\begin{equation}
136: \left|t_{k}\right|\cos(\phi-\phi_{0k})=\cos\left(kN-\varphi_{k}\right).\label{eigen}\end{equation}
137: Algebraic manipulation equivalent to that in RR gives an implicit
138: equation for $\left|t\left(\varepsilon_{k}\right)\right|$, exact
139: to the leading order in $\frac{1}{N}$, \begin{equation}
140: \frac{\partial\arccos\left(\mp\left|t\left(\varepsilon_{k}\right)\right|\cos\left[\phi-\phi_{0}\left(\varepsilon_{k}\right)\right]\right)}{\partial\cos\phi}=\pi N\rho(\varepsilon_{k})\frac{\partial\varepsilon_{k}}{\partial\cos\phi},\label{main}\end{equation}
141: where the sign $\mp$ depends on weather $k$ belongs to a decreasing
142: ($+$) or to increasing ($-$) branch of the cosine function in Eq.~(\ref{eigen})
143: and $\rho(\varepsilon)=\bigl(\pi\sqrt{4t_{0}^{2}-\varepsilon^{2}}\bigr)^{-1}$
144: with $t_{0}$ being the nearest neighbor hopping in the leads. By
145: using the principle of mathematical induction, as in RR but generalized
146: to a finite $\phi_{0}$, the transmission probability at the Fermi
147: energy $\varepsilon_{F}$ is expressed as\begin{equation}
148: \frac{1}{\pi}\frac{\partial\arccos^{2}\left(\mp\left|t\left(\varepsilon_{F}\right)\right|\cos\left[\phi-\phi_{0}\left(\varepsilon_{F}\right)\right]\right)}{\partial\cos\phi}=\pi N\rho(\varepsilon_{F})\frac{\partial E}{\partial\cos\phi},\label{ground}\end{equation}
149: where $E$ is the ground-state energy of a system containing an even
150: number of electrons. Neglecting the variation of $t\left(\varepsilon_{F}\right)$
151: and $\rho\left(\varepsilon_{F}\right)$ with $\phi$ (the error made
152: is of the order of $\frac{1}{N}$), Eq.~(\ref{ground}) leads to
153: a universal form of the ground-state energy of the auxiliary system
154: $E(\phi)=\pi^{-2}\Delta\arccos^{2}\left(\mp\left|t\left(\varepsilon_{F}\right)\right|\cos\left[\phi-\phi_{0}\left(\varepsilon_{F}\right)\right]\right)+\mathrm{const.}$\cite{odd}
155: From this form, the transmission probability can be extracted, and
156: the conductance is given by\begin{equation}
157: G=G_{0}\sin^{2}\left(\frac{\pi}{2}\frac{E\left(\phi_{0}+\pi\right)-E\left(\phi_{0}\right)}{\Delta}\right),\label{eq:gsin2}\end{equation}
158: where $\phi_{0}\equiv\phi_{0}\left(\epsilon_{F}\right)$ is determined
159: by the position of the minimum (or maximum) in the energy vs. flux
160: $\phi$ curve, schematically shown in Fig.~\ref{cap:Fig1}(c). The
161: conductance can also be calculated from the more convenient four-point
162: formula~\cite{3point}:
163:
164: \begin{alignat}{1}
165: G=G_{0}\biggl[ & \sin^{2}\left(\frac{\pi}{2}\frac{E\left(\pi\right)-E\left(0\right)}{\Delta}\right)+\nonumber \\
166: + & \sin^{2}\left(\frac{\pi}{2}\frac{E\left(\pi/2\right)-E\left(-\pi/2\right)}{\Delta}\right)\biggr],\label{eq:gsin3}\end{alignat}
167: and $\phi_{0}$ can be determined from the expression\begin{equation}
168: \phi_{0}=-\arctan\frac{\sin\left(\frac{\pi}{2}\frac{E\left(\pi/2\right)-E\left(-\pi/2\right)}{\Delta}\right)}{\sin\left(\frac{\pi}{2}\frac{E\left(\pi\right)-E\left(0\right)}{\Delta}\right)}.\label{eq:phi0}\end{equation}
169:
170:
171: We now consider the interacting case. If no AB flux is present in
172: the mesoscopic region and the ground state of the system exhibits
173: Fermi liquid behavior, i.e., the perturbation theory in the interaction
174: strength is valid, the imaginary part of the self-energy due to the
175: interaction vanishes quadratically at the Fermi energy, \begin{equation}
176: \mathrm{Im}\Sigma_{ij}\left(\omega+i\delta\right)\propto\left(\omega-\varepsilon_{F}\right)^{2},\label{Luttinger}\end{equation}
177: and a quasiparticle Hamiltonian can be constructed for each value
178: of $\varepsilon_{F}$. As shown in RR the conductance calculated from
179: the quasiparticle Hamiltonian reproduces the conductance of the interacting
180: system, and is exactly given by Eq.~(\ref{eq:gsin}) {[}or Eq.~(\ref{eq:gsin3})
181: since $E\left(\pi/2\right)=E\left(-\pi/2\right)$ in this case{]}.
182: If the time reversal symmetry is broken due to AB flux, Eq.~(\ref{Luttinger})
183: is not valid and the proof has to be reconsidered. Repeating the steps
184: as presented in detail in RR, the proof is restored and basically
185: unchanged if the exact self energy obeys the relation\begin{equation}
186: \frac{1}{2i}\left[\Sigma_{ij}\left(\omega+i\delta\right)-\Sigma_{ij}\left(\omega-i\delta\right)\right]\propto\left(\omega-\varepsilon_{F}\right)^{2}.\label{Luttinger2}\end{equation}
187: It follows that also the linear response conductance of an interacting
188: AB system at zero-temperature is given by the four-point formula Eq.~(\ref{eq:gsin3}).
189: This condition is fulfilled if the system is a Fermi liquid.
190:
191: In order to demonstrate the practical value of the method, we quantitatively
192: analyze the conductance through an Aharonov-Bohm ring with a quantum
193: dot placed in one of the arms as presented in Fig.~\ref{cap:Fig1}(d)~\cite{Hofstetter01,Bulka01}.
194: The quantum dot is described as an Anderson impurity with level position
195: $\epsilon_{d}$ and a charging energy $U$, coupled to each of the
196: leads with a tunneling matrix element $t_{1}$. Electrons can also
197: be transferred from one lead to the other directly through the upper
198: arm of the AB ring. This process is described by a tunneling matrix
199: element $t_{2}$. The ring is threaded by an AB flux $\Phi_{AB}=\frac{\hbar}{e}\phi_{AB}$
200: in such a way that only the direct tunneling matrix element is affected,
201: i.e. $t_{2}\rightarrow t_{2}e^{i\phi_{AB}}$. The Fermi energy is
202: set at the middle of the band, thus $\Delta=\frac{2\pi t_{0}}{N}$.
203: We perform the finite-size analysis of the four-point formula Eq.~(\ref{eq:gsin3}),
204: changing the circumference of the ring $N$. In order to be able to
205: compare our results with those of the NRG method, we choose the same
206: values of parameters as in Ref.~\onlinecite{Hofstetter01}.
207:
208: In Fig.~\ref{cap:Fig2} the results for the conductance of a noninteracting
209: ($U=0$) system with $\phi_{AB}=\frac{\pi}{4}$ are presented. Due
210: to a non-zero AB flux, the ground-state energy must be determined
211: at four $\phi$ points as required by Eq.~(\ref{eq:gsin3}). The
212: conductance exhibits a typical Fano resonance with a dip and a sharp
213: peak. Results calculated for various numbers of sites in the ring
214: $N$ are compared with the exact conductance curve. The inset shows
215: a convergence test of the method at the peak near $\epsilon_{d}=0$,
216: where due to a strong energy dependence of the transmission amplitude
217: the convergence is the most delicate.
218:
219: %
220: \begin{figure}[t]
221: \begin{center}\includegraphics[%
222: clip,
223: width=7cm,
224: keepaspectratio]{Fig2.eps}\end{center}
225:
226:
227: \caption{\label{cap:Fig2}The conductance of a noninteracting ($U=0$) AB
228: ring at $\phi_{AB}=\frac{\pi}{4}$. The inset shows convergence of
229: the conductance at the peak near $\varepsilon_{d}=0$. Parameters:
230: $t_{1}=0.177t_{0}$, $t_{2}=0.298t_{0}$.}
231: \end{figure}
232:
233:
234: In the interacting case, we determined the required ground-state energies
235: using an approach similar to the variational method of Gunnarson and
236: Sch\"{o}nhammer~\cite{Schonhammer75,Rejec03} for the Anderson model.
237: The variational basis set is generated from the ground state $\left|\tilde{0}\right\rangle $
238: of an auxiliary noninteracting Hamiltonian in which the energy level
239: of the dot $\varepsilon_{d}$ and the hopping matrix element $t_{1}$
240: between the dot and the leads are renormalized in such a way to minimize
241: the ground-state energy. Apart from the Hartree-Fock case where this
242: is the only variational wave function, we form two additional basis
243: sets. The first has three variational wave functions $P_{0}\left|\tilde{0}\right\rangle $,
244: $P_{1}\left|\tilde{0}\right\rangle $ and $P_{2}\left|\tilde{0}\right\rangle $,
245: where $P_{0}$, $P_{1}$ and $P_{2}$ are projectors onto unoccupied,
246: singly occupied and doubly occupied dot site. In the second basis
247: set we add four additional wave functions, $P_{0}VP_{1}\left|\tilde{0}\right\rangle $,
248: $P_{2}VP_{1}\left|\tilde{0}\right\rangle $, $P_{1}VP_{0}\left|\tilde{0}\right\rangle $
249: and $P_{1}VP_{2}\left|\tilde{0}\right\rangle $, where $V$ is the
250: operator describing hopping between the leads and the dot. In all
251: the three cases, the noninteracting limit is correctly reproduced.
252: Furthermore, except in the Hartree-Fock case, the method is also exact
253: in the limit where the dot is decoupled from the rest of the system.%
254: \begin{figure}[htbp]
255: \begin{center}\includegraphics[%
256: clip,
257: width=7cm,
258: keepaspectratio]{Fig3.eps}\end{center}
259:
260:
261: \caption{\label{cap:Fig3}Zero-temperature linear response conductance as
262: a function of level position $\varepsilon_{d}$ for various values
263: of the AB flux. Labels '3' and '7' correspond to variational wave
264: functions with three and seven terms, respectively. The dash-dotted
265: line is the result obtained using the HF approximation and the dotted
266: line is the NRG result from Ref.~\onlinecite{Hofstetter01}. Parameters:
267: $t_{1}=0.177t_{0}$ ($\Gamma=0.125t_{0}$), $t_{2}=0.298t_{0}$, $U=t_{0}=8\Gamma$.}
268: \end{figure}
269:
270:
271: An interacting AB ring with $U\neq0$ serves as a good nontrivial
272: test of the method. We choose a strong coupling regime as in Ref.~\onlinecite{Hofstetter01},
273: with $U=8\Gamma$, where $\Gamma=4\pi t{}_{1}^{2}\rho\left(\epsilon_{F}\right)$
274: is the line width of the dot level in the absence of the upper arm
275: of the AB ring. The results are presented in Fig.~\ref{cap:Fig3}
276: and compared with results obtained using the NRG method~\cite{Hofstetter01},
277: together with the corresponding Hartree-Fock (HF) curve. It should
278: be noted that with the parameter set chosen, the underlying physics
279: is in the strong correlation regime, where e.g. the width of the Kondo
280: peak is of the order $10^{-4}U$. The HF method therefore fails to
281: reproduce NRG results, while the variational results are very close
282: to the NRG curve. By extending the variational space from 3 to 7 terms,
283: the agreement with NRG improves even further and the small remaining
284: discrepancy would most probably be additionally reduced if an even
285: richer variational wave function is used. Interestingly, we find the
286: largest deviation in the 'empty orbital' regime, $\varepsilon_{d}\lesssim-U$.
287: The results of the variational and the HF methods here agree, both
288: deviating from the NRG result, while asymptotically all the curves
289: reach the correct limit $G=0.3G_{0}$~\cite{Hofstetter01}. We have
290: also checked the result in this regime using the second order perturbation
291: theory approach~\cite{Horvatic87} which gives a conductance in agreement
292: with variational and HF methods (not shown).
293:
294: %
295: \begin{figure}[t]
296: \begin{center}\includegraphics[%
297: clip,
298: width=7cm,
299: keepaspectratio]{Fig4.eps}\end{center}
300:
301:
302: \caption{\label{cap:Fig4}The same as in Fig.~\ref{cap:Fig3}(b), but calculated
303: for different numbers of sites in the auxiliary ring $N$. The inset
304: shows convergence of the conductance at $\epsilon_{d}=-U/2$.}
305: \end{figure}
306:
307:
308: In Fig.~\ref{cap:Fig4} a convergence test of the method for the
309: result from Fig.~\ref{cap:Fig3}(b) is shown. The convergence with
310: $N$ is fast in the empty orbital regime and becomes progressively
311: slower as $\epsilon_{d}$ shifts toward the Kondo regime. The reason
312: can clearly be attributed to a very strong coupling regime resulting
313: in an extremely narrow Kondo peak. A very fine energy resolution is
314: therefore required to resolve the peak and to obtain a converged conductance
315: curve.
316:
317: Broken time reversal symmetry in AB systems is signaled by $\phi_{0}\neq0$.
318: In Fig.~\ref{cap:Fig5} is shown the phase shift $\phi_{0}$ corresponding
319: to Figs.~\ref{cap:Fig2} and \ref{cap:Fig3}(b) as determined from
320: Eq.~(\ref{eq:phi0}). In contrast to the smooth noninteracting result,
321: the phase shift in the interacting case exhibits a well developed
322: plateau, corresponding to the Fano-suppressed Kondo plateau in the
323: conductance. The $\phi_{0}$ curve, which clearly cannot be correctly
324: reproduced in the HF approximation, can be explained as follows: In
325: the empty orbital regime, the current mainly flows through the upper
326: arm of the ring and therefore, electrons acquire an additional phase
327: shift $\phi_{0}\sim\phi_{AB}$ (note that $\phi_{0}$ and $\phi_{0}-\pi$
328: are physically equivalent). On the other hand, in the Kondo regime
329: almost all the current passes through the quantum dot and no additional
330: phase shift is present, $\phi_{0}\sim0$.
331:
332: %
333: \begin{figure}[b]
334: \begin{center}\includegraphics[%
335: clip,
336: width=7cm,
337: keepaspectratio]{Fig5.eps}\end{center}
338:
339:
340: \caption{\label{cap:Fig5}Phase shift $\phi_{0}$ as a function of $\varepsilon_{d}/t_{0}$
341: for the noninteracting AB ring of Fig.~\ref{cap:Fig2} (NI) and for
342: various ground-state energy methods applied to the interacting AB
343: ring of Fig.~\ref{cap:Fig3}(b).}
344: \end{figure}
345:
346:
347: In summary, we have derived a formula for zero-temperature linear
348: response conductance of an interacting electron region coupled to
349: single-channel leads where the electron-electron interaction is absent.
350: The interacting part of the system can be a general Aharonov-Bohm
351: type of interferometer with broken time reversal symmetry {[}Fig.~\ref{cap:Fig1}(a){]}.
352: The conductance of such an \emph{open} system is exactly determined
353: from an auxiliary \emph{closed} system {[}Fig.~\ref{cap:Fig1}(b){]},
354: where the leads of the original system are connected to form a ring
355: of $N$ noninteracting sites threaded by an auxiliary magnetic flux.
356: Eq.~(\ref{eq:gsin3}), which follows from the universal form of the
357: ground-state energy of the auxiliary system for a large but finite
358: $N$, expresses the conductance in terms of the ground-state energy
359: evaluated at four different values of the auxiliary flux. The proof
360: of validity of the formula for interacting systems relies on the mapping
361: of the system onto an effective quasiparticle problem and is therefore
362: valid for systems exhibiting Fermi liquid properties. We have demonstrated
363: the usefulness of the formula by applying it to a prototype system
364: exhibiting Kondo-Fano behavior. Results based on the four-point formula
365: and variational ground-state energies confirm results of the numerical
366: renormalization group method.
367:
368: The authors wish to acknowledge P. Prelov\v{s}ek and X. Zotos for helpful
369: discussions and W. Hofstetter for useful correspondence regarding
370: his results. We acknowledge J.~H. Jefferson for useful remarks and
371: the financial support of QinetiQ.
372:
373: \begin{thebibliography}{10}
374: \bibitem{Glazman88}L.~I. Glazman and M.~E. Raikh, Pis'ma Zh. Eksp. Teor. Fiz. \textbf{47},
375: 378 (1988) {[}Sov. Phys. JETP Lett. \textbf{47}, 452 (1988){]}; T.~K.
376: Ng and P.~A. Lee, Phys. Rev. Lett. \textbf{61}, 1768 (1988).
377: \bibitem{Goldhaber98}D. Goldhaber-Gordon \textit{et~al.}, Nature \textbf{391}, 156 (1998);
378: W.~G. van~der Wiel \textit{et~al.}, Science \textbf{289}, 2105
379: (2000).
380: \bibitem{Bulka01}B.~R. Bulka and P. Stefanski, Phys. Rev. Lett. \textbf{86}, 5128
381: (2001).
382: \bibitem{Hofstetter01}W. Hofstetter, J. K\"{o}nig and H. Schoeller, Phys. Rev. Lett. \textbf{87},
383: 156803 (2001).
384: \bibitem{ABexp}J. G\"{o}res \emph{et al}., Phys. Rev. B \textbf{62}, 2188 (2000); K.
385: Kobayashi \emph{et al.}, Phys. Rev. Lett. \textbf{88}, 256806 (2002)
386: \bibitem{Landauer57}R. Landauer, IBM J. Res. Dev. \textbf{1}, 233 (1957); Philos. Mag.
387: \textbf{21}, 863 (1970); M. B\"{u}ttiker, Phys. Rev. Lett. \textbf{57},
388: 1761 (1986).
389: \bibitem{Meir92}Y. Meir and N.~S. Wingreen, Phys. Rev. Lett. \textbf{68}, 2512 (1992).
390: \bibitem{Jauho94}A.-P. Jauho, N.~S. Wingreen, and Y. Meir, Phys. Rev. B \textbf{50},
391: 5528 (1994).
392: \bibitem{Oguri}A. Oguri, Phys. Rev B \textbf{56}, 13422 (1997); J. Phys. Soc. Jap.
393: \textbf{70}, 2666 (2001)
394: \bibitem{Rejec03}T. Rejec and A. Ram\v{s}ak, cond-mat/0301608 (unpublished).
395: \bibitem{Favand98}Similar approach, derived for noninteracting systems and applied to
396: systems with interaction, is found also in: J. Favand and F. Mila,
397: Eur. Phys. J. B \textbf{2}, 293 (1998); O.~P. Sushkov, Phys. Rev.
398: B \textbf{64}, 155319 (2001); R.~A. Molina \textit{et~al.}, cond-mat/0209552
399: (unpublished); V. Meden and U. Schollw\"{o}ck, cond-mat/0210515 (unpublished).
400: \bibitem{odd}For an odd number of electrons in a system, the universal form of
401: the ground-state energy is $E(\phi)=\pi^{-2}\Delta\arcsin^{2}\left(\left|t\left(\varepsilon_{F}\right)\right|\cos\left[\phi-\phi_{0}\left(\varepsilon_{F}\right)\right]\right)+\mathrm{const.}$
402: \bibitem{3point}It is also possible to calculate the conductance from the ground-state
403: energies at three distinct values of flux $\phi$ by solving Eq.~(\ref{ground})
404: numerically.
405: \bibitem{Schonhammer75}K. Sch\"{o}nhammer, Z. Phys. B \textbf{21}, 389 (1975); Phys. Rev. B \textbf{13},
406: 4336 (1976); O. Gunnarson and K. Sch\"{o}nhammer, Phys. Rev. B \textbf{31},
407: 4815 (1985).
408: \bibitem{Horvatic87}B. Horvati\'{c}, D. \v{S}ok\v{c}evi\'{c}, and V. Zlati\'{c}, Phys. Rev. B \textbf{36},
409: 675 (1987).\end{thebibliography}
410:
411: \end{document}
412: