1: %\documentstyle[aps,prl,twocolumn,epsf]{revtex}
2: \documentclass[aps,prl,twocolumn,showpacs]{revtex4}
3:
4: \usepackage{graphicx}
5:
6: %\renewcommand{\baselinestretch}{2.0}
7:
8: \begin{document}
9:
10: %\draft
11:
12: \title{Proximity-induced sub-gaps in Andreev billiards}
13:
14: \author{ J. Cserti\thanks{e-mail: cserti@galahad.elte.hu},
15: A. Korm\'anyos, Z. Kaufmann, J. Koltai}
16: \affiliation{Department of Physics of Complex Systems,
17: E{\"o}tv{\"o}s University
18: \\ H-1117 Budapest, P\'azm\'any P{\'e}ter s{\'e}t\'any 1/A, Hungary}
19: \author{C. J. Lambert\thanks{e-mail: c.lambert@lancaster.ac.uk}}
20: %\author{C. J. Lambert\thanks{e-mail: c.lambert@lancaster.ac.uk}}
21: \affiliation{Department of Physics, Lancaster University,
22: Lancaster, LA1 4YB, UK}
23: %\date{\today}
24:
25:
26: %\wideabs{
27:
28: \begin{abstract}
29:
30: We examine the density of states of an Andreev billiard and show
31: that any billiard with a finite upper cut-off in the path length
32: distribution $P(s)$ will possess an energy gap on the scale of the
33: Thouless energy. An exact quantum mechanical calculation for
34: different Andreev billiards gives good agreement with the
35: semiclassical predictions when the energy dependent phase shift
36: for Andreev reflections is properly taken into account. Based on
37: this new semiclassical Bohr-Sommerfeld approximation of the
38: density of states, we derive a simple formula for the energy gap.
39: We show that the energy gap,
40: in units of Thouless energy, may exceed the value predicted
41: earlier from random matrix theory for chaotic billiards.
42:
43: \end{abstract}
44:
45: \pacs{74.80.Fp 03.65.Sq 05.45.Mt 74.50.+r}
46:
47: \maketitle
48:
49: %}
50:
51:
52: Studies of sub-gap transport in hybrid superconductors are an
53: important starting point for the design and simulation of
54: superconducting nanoscale devices. For a given inhomogeneous
55: structure, a fundamental question is the existence or otherwise of
56: a finite density of states at the Fermi energy. This is important
57: experimentally, since the sub-gap spectrum determines the
58: tunneling conductance of an N-S contact. The ability to address
59: this question is also important theoretically, since a well-posed
60: problem of this kind provides a testing ground for complementary
61: (and occasionally competing) theoretical techniques.
62:
63: One important class of structures, for which a general analysis
64: might be forthcoming, are known as Andreev billiards. These are
65: formed when a classically-chaotic normal dot is placed in contact
66: with a superconductor
67: \cite{Kosztin,Melsen,Altland,Lesovik,Nazarov,Heny,Richter1,Richter2}.
68: Consider a ballistic two dimensional normal dot of area $A$, with
69: the mean level spacing of the isolated normal system $\delta
70: =\frac{2\pi \hbar^2}{mA}$ at the Fermi energy $E_F$. If a
71: superconductor of width $W$ and bulk order parameter $\Delta$ is
72: placed in contact with such a billiard (see Fig.~\ref{geo-fig}),
73: then the question of interest is whether or not an energy gap on
74: the scale of the Thouless energy
75: $E_T=\frac{M\delta}{4\pi}$~\cite{note_1}, exists in the sub-gap
76: spectrum $E <\Delta$, when $\delta << E_T < \Delta$. The number
77: of open channels in the S region is the integer part of
78: $M=\frac{k_{\rm F}W}{\pi}$, and the energy levels of the Andreev
79: billiards are the positive eigenvalues $E$ (measured from the
80: Fermi energy) of the Bogoliubov-de Gennes equation\cite{BdG-eq}.
81: \begin{figure}[hbt]
82: \includegraphics[scale=0.5]{wedge-geo.eps}
83: %\centerline{\leavevmode \epsfxsize=7cm \epsffile{wedge-geo.eps }}
84: \caption{Geometry. \label{geo-fig}}
85: \end{figure}
86:
87: An initial study of this problem based on random matrix
88: theory\cite{Melsen}, concluded that a classically chaotic
89: billiard possesses an energy gap, whereas an integrable system is gapless.
90: As a consequence it was suggested that the existence of such a gap could be
91: used to distinguish between integrable and chaotic systems.
92: Later it was shown that billiards with classically mixed phase space
93: do possess a smaller gap
94: comparing to the fully chaotic systems\cite{Heny}.
95: Studying the existence of the gap for chaotic billiards
96: Ihra et al.\ \cite{Richter1} have drawn the attention to the role of the
97: non-diagonal terms of the scattering matrix which could explain the
98: disagreement between the random matrix theory and the Bohr-Sommerfeld
99: semiclassical approximation. The non-universal feature of the
100: excitation spectra of the Andreev billiards has been recently studied
101: by Ihra and Richter in Ref.\ \cite{Richter2}.
102:
103: The aim of this Letter is to identify for the first time, a
104: pseudo-integrable billiard which does possess an energy gap.
105: Moreover, this gap can be larger than that predicted for
106: chaotic billiards on the scale of $E_T$. To this end,
107: we study the quasi-particle spectrum of the ballistic structure shown
108: in Fig.~\ref{geo-fig}, for different values of the angle $\alpha$
109: and length $d$. Our key result is illustrated
110: in Fig.~\ref{N-zero_d-fig}, which shows that for $d=0$,
111: the counting function or integrated density of states
112: $N(E)=\int_0^E\, n(E^\prime)\, dE^\prime$ (where $n(E)$ is the density of
113: states for the Andreev billiard) exhibits an
114: energy gap $E_{\rm gap}$ varying between
115: $E_{\rm gap} \approx 0.5 E_T$ and $E_{\rm gap}\approx 1.5 E_T$
116: as $\alpha$ varies from $80^{\circ}$ to $45^{\circ}$
117: (for the details of the calculation see below).
118: \begin{figure}[hbt]
119: \includegraphics[scale=0.5]{fig2.eps}
120: %\centerline{\leavevmode \epsfxsize=7cm \epsffile{fig2.eps }}
121: \caption{The integrated density of states, $N(E)$ as a function of $E$ (in
122: units of $\Delta$) for $d=0$ and angle
123: $\alpha = 80^{\circ}, 60^{\circ}, 45^{\circ}$.
124: For each angle $\alpha$ the stair type lines correspond
125: to the exact diagonalization of the Bogoliubov-de Gennes equation,
126: while the solid lines are obtained by using Eq.~(\ref{N-eq}).
127: The parameters are $M=55.5$, $\Delta/E_F = 0.015$.
128: \label{N-zero_d-fig}}
129: \end{figure}
130: In contrast, as shown in Fig.~\ref{N-nonzero_d-fig}, for $d \ne
131: 0$, no such energy gap exists.
132: \begin{figure}[hbt]
133: \includegraphics[scale=0.45]{fig3.eps}
134: %\centerline{\leavevmode \epsfxsize=7cm \epsffile{fig3.eps }}
135: \caption{The integrated density of states, $N(E)$ as a function
136: of $E$ (in units of $\Delta$) for $d \ne 0$ and angle $\alpha =
137: 80^{\circ}, 60^{\circ}, 45^{\circ}$. In each case, $d$ varies
138: such that the area of the normal dot is fixed at $A=5W^2$ for the
139: angles $\alpha=80^{\circ}, 60^{\circ}, 45^{\circ}$. This choice
140: results in approximately the same number of states below $\Delta$
141: for different values of $\alpha$. The meaning of the lines and the
142: parameters are the same as those in Fig.~\ref{N-zero_d-fig}.
143: \label{N-nonzero_d-fig}}
144: \end{figure}
145: To understand this result, we start from the formula relating
146: the integrated density of states $N(E)$ to the path length distribution
147: $P(s)$ (which is the classical probability
148: that an electron entering the billiard at the N-S contact exits
149: after a path length $s$) derived by using the Bohr-Sommerfeld approximation,
150: \begin{equation}
151: N\left(E \right) = M \sum_{n=0}^\infty \,
152: \int_{s_n\left(E \right)}^\infty \, P(s)\, ds,
153: \label{N-eq}
154: \end{equation}
155: where
156: \begin{equation}
157: s_n\left(E \right) = \frac{\left(n+
158: \frac{1}{\pi}\arccos\frac{E}{\Delta}\right)\pi}{E/\Delta}\,\, \xi_0.
159: \label{s_eps-eq}
160: \end{equation}
161: Here $\xi_0=\frac{\hbar v_{\rm F}}{\Delta}
162: = W\, \frac{2}{\pi M}\, \frac{E_F}{\Delta}$
163: is the coherence length and
164: the distribution $P(s)$ is normalized to one,
165: $\int_0^\infty\, P(s)\, ds =1$.
166: The density of states $n(E)=\frac{dN(E)}{dE}$ can be found
167: from Eq.~(\ref{N-eq}) yielding
168: \begin{equation}
169: n\left(E \right) =
170: \frac{M}{E} \sum_{n=0}^\infty \,
171: P\left(s_n\left(E \right) \right)
172: \left[\frac{\xi_0}{\sqrt{1-{\left(E/\Delta\right)}^2}} +
173: s_n\left(E \right)\right].
174: \label{n-eq}
175: \end{equation}
176: This expression reduces to that of derived in a different way
177: by Melsen et al.\ \cite{Melsen,Heny}, Lodder and Nazarov\cite{Nazarov},
178: Ihra et al.\ \cite{Richter1}, in the limit
179: $E << \Delta$ and $\xi_0^2 << A$,
180: but more generally incorporates an energy dependent path length
181: correction~\cite{note_2}.
182:
183: To test the semiclassical expression, we perform an
184: exact (numerical) diagonalization of the Bogoliubov-de Gennes
185: equation by matching the wave functions at the N-S interface.
186: This results in a secular equation including the scattering matrix,
187: $S_0(E)$ of the normal billiard opened at the N-S interface~\cite{elsewhere}:
188: \begin{equation}
189: \det \left[1-e^{-2i\arccos\frac{E}{\Delta}}\,
190: S_{\rm eff}(E)\, S_{\rm eff}^*(-E) \right] = 0,
191: \label{secular-eq}
192: \end{equation}
193: where
194: \begin{eqnarray}
195: S_{\rm eff}(E) &=&
196: {\left[Q(E)+K(E)D(E)\right]}^{-1} \times
197: \nonumber \\
198: && \left[Q^*(E)-K(E)D(E)\right],
199: \label{Seff-eq}\\[2ex]
200: D(E)&=&\left[1-S_0(E)\right]{\left[1+S_0(E)\right]}^{-1}. \nonumber
201: \end{eqnarray}
202: Here $Q$ and $K$ are diagonal matrices with elements
203: $Q_{nm}(E)=i\, \delta_{nm}\,q_n(E)$ and $K_{nm}(E)=\delta_{nm}\,
204: k_n(E)$, where $q_n (E)=k_F \sqrt{1 + i\,
205: \frac{\sqrt{\Delta^2-E^2}}{E_F}-\frac{n^2}{M^2}}$ are the
206: transverse wavenumbers of the electron in the S region and
207: $k_n(E)=k_F\sqrt{1+\frac{E}{E_F}-\frac{n^2}{M^2}}$ are the
208: transverse wavenumbers of the electron in the S region when
209: $\Delta=0$. It is assumed that the Fermi wavenumber, $k_F =
210: \sqrt{2mE_F/\hbar^2}$ is the same in the S and N regions. All
211: the matrices are $M$ by $M$ dimensional. Finally, $S_0$ was
212: calculated using Bessel functions in the wedge part of the
213: normal region, and including evanescent modes. In a different
214: context, the same type of normal billiard was studied by Kaplan
215: and Heller~\cite{Heller}. A secular equation similar to
216: Eq.~(\ref{secular-eq}) was derived by Beenakker~\cite{Carlo_2_3}
217: for SNS systems but there instead of $S_{\rm eff}(E)$ the
218: scattering matrix $S_0(E)$ of the normal billiard appears. In
219: this sense, the matrix $S_{\rm eff}(E)$ can be regarded as an
220: effective scattering matrix. Note that in Andreev
221: approximation\cite{Colin-review} $q_n\approx k_n$, and one then
222: finds $S_{\rm eff}(E)=S_0(E)$. In contrast the secular
223: equation~(\ref{secular-eq}) for Andreev billiards is valid outside
224: Andreev approximation. Using the unitarity of $S_0(E)$, one can
225: show that the following equation
226: \begin{eqnarray}
227: \det \left(\rm Im \left\{
228: e^{-i\arccos\frac{E}{\Delta}} \,
229: \left[Q(E)+K(E)D(E)\right] \times \right. \right.
230: \nonumber && \\
231: \left. \left.
232: {\left[Q(E)+K(E)D^*(-E)\right]}^{-1}
233: \right\}
234: \right) &=& 0,
235: \label{secular-real-eq}
236: \end{eqnarray}
237: has the same zeros as those of Eq.~(\ref{secular-eq}), and is a
238: more suitable form for numerical calculations (the left hand side
239: of Eq.~(\ref{secular-real-eq}) is a real function instead of a
240: complex one).
241:
242: Figures~\ref{N-zero_d-fig} and \ref{N-nonzero_d-fig} show both
243: exact results and an evaluation of the semiclassical expression
244: of $N(E)$ given by Eq.~(\ref{N-eq}) for Andreev billiards shown
245: in Fig.~\ref{geo-fig} with different parameters~\cite{note_3}.
246: One can see that the results of the two methods are in good
247: agreement over the entire energy range $E<\Delta$. It is important
248: to note that inclusion of the energy dependence of the phase shift
249: was essential to obtain such a good agreement. A small deviation
250: can be seen only at energies close to the value of $\Delta$ and
251: for $d\ne 0$.
252:
253: Both the semiclassical and the exact calculations reveal the
254: presence of an energy gap when $d=0$, while no gap appears in
255: case of $d\ne 0$. The origin of this result can be traced to the
256: existence of a finite upper cut-off $s_{\rm max}$ in the path
257: length distribution $P(s)$. From Eq.~(\ref{s_eps-eq}) it is
258: obvious that if the path length of the electron has a maximum
259: value, then the energy spectrum will possess a gap. The condition
260: for determining $E_{\rm gap}$ is $s_n(E) > s_{\rm max}$ for $n\ge
261: 0$. Expanding the arccos term in Eq.~(\ref{s_eps-eq}) to first
262: order in $E/\Delta$, one finds that the energy gap is given by the
263: following simple equation:
264: \begin{equation}
265: \frac{E_{\rm gap}}{E_{\rm T}} = \pi^2 \,
266: \frac{A}{Ws_{\rm max}}\, \frac{1}{1+\xi_0/s_{\rm max}},
267: \label{gap-eq}
268: \end{equation}
269: where $A$ is the area of the billiard and $W$ is the width of the
270: superconductor lead.
271:
272: As examples, Fig.~\ref{Ps-fig} shows $P(s)$ for
273: $\alpha=80^{\circ}, 60^{\circ}, 45^{\circ}$ along with the
274: corresponding semiclassical density of states $n(E)$ given by
275: Eq.~(\ref{n-eq}). From the figure it is seen that the upper
276: cut-off $s_{\rm max}$ of $P(s)$ equals to $2W$ for $\alpha
277: =80^{\circ}, 60^{\circ}$, and $s_{\rm max} = 2\sqrt{2}\,W$ for
278: $\alpha = 45^{\circ}$. It is interesting to note that for $\alpha
279: =45^\circ$ (more generally for $\alpha = 90^{\circ}/k$ with an
280: integer $k>0$) the density of states $n(E)$ has a
281: singularity at some
282: energy $E<\Delta$ while for other values of $\alpha$ no
283: pronounced singularities exist. In Bohr-Sommerfeld approximation,
284: it can be shown that this singularity is related to the
285: singularity of the path length distribution $P(s)$. The classical
286: trajectories of the electron resulting in a singularity
287: for $\alpha =45^\circ$ are
288: similar to those moving in a lead with length $d+W$ and
289: $\alpha =90^\circ$. The nature of the singularities of the
290: density of states in this rectangular shape of Andreev billiard
291: was studied in Ref.~\onlinecite{box-disk-paper}.
292: The reason for the disappearance
293: of these singularities for $\alpha \ne 90^\circ/k$ ($k=1,2,3,\ldots$) is
294: consistent with the points made in this reference.
295: \begin{figure}[hbt]
296: \includegraphics[scale=0.45]{fig4ff.eps}
297: %\centerline{\leavevmode \epsfxsize=7cm \epsffile{fig4ff.eps }}
298: \caption{The path length distribution $P(s)$ for Andreev billiard of
299: Fig.~\ref{geo-fig} when $d=0$
300: for $\alpha=80^{\circ}, 60^{\circ}, 45^{\circ}$.
301: Insets show the corresponding semiclassical density of states $n(E)$
302: (in units of $1/\delta$) using Eq.~(\ref{n-eq}).
303: The same parameters was used as those in Fig.~\ref{N-zero_d-fig}.
304: \label{Ps-fig}}
305: \end{figure}
306:
307: To check the expression (\ref{gap-eq}) we have determined the
308: energy gap from exact calculations for several angles $\alpha$.
309: For the billiard of
310: Fig.~\ref{geo-fig} with $d=0$, one can show analytically that the
311: upper cut-off $s_{\rm max}$ is
312: \begin{equation}
313: \frac{s_{\rm max}}{W}= \left\{
314: \begin{array}{ll}
315: -2\, \frac{\sin k\alpha \sin \left[\left(k+1\right)\alpha\right]}
316: {\sin\alpha \cos \left[\left(2k+1\right)\alpha\right]}
317: %\frac{1-\cos \alpha/\cos \left[\left(2k+1\right)\alpha\right]}{\sin\alpha}
318: & \mbox{if \,\,
319: $\frac{\pi/2}{k+1} < \alpha \le \frac{\pi/2}{k+\frac{2}{3}}$}, \\[1.2ex]
320: %\frac{\sqrt{2\left(1-\cos 2k\alpha \right)}}{\sin \alpha}
321: 2\, \frac{\sin k\alpha}{\sin \alpha}
322: & \mbox{if \,\,
323: $\frac{\pi/2}{k+\frac{2}{3}} < \alpha \le \frac{\pi/2}{k}$},
324: \end{array}
325: \right.
326: \label{smax-exp}
327: \end{equation}
328: where $k=1,2,3,\ldots .$
329: Using Eqs.~(\ref{gap-eq}) and (\ref{smax-exp})
330: the gap is plotted as a function of the angle $\alpha$
331: in Fig.~\ref{gap-fig} together with the exact results.
332: \begin{figure}[hbt]
333: \includegraphics[scale=0.45]{gap-BS.eps}
334: %\centerline{\leavevmode \epsfxsize=7cm \epsffile{gap-BS.eps }}
335: \caption{The gap $E_{\rm gap}$ (in units of Thouless energy)
336: as a function of angle $\alpha$ for Andreev billiards with $d=0$.
337: The solid and the dashed curves are the semiclassical results
338: from Eq.~(\ref{gap-eq}) when $M=55.5$, and from Eq.~(\ref{gap-semi-eq})
339: i.e.\ when $M\rightarrow \infty$, respectively.
340: The circles are the results of the exact diagonalization when $M=55.5$.
341: \label{gap-fig}}
342: \end{figure}
343: The numerical results (full circles in Fig.~\ref{gap-fig}) agree
344: very well with the predictions of Eq.~(\ref{gap-eq}). One can see
345: that decreasing $\alpha$ the energy gap in units of Thouless
346: energy increases and tends to a finite value as $\alpha
347: \rightarrow 0$. From Eq.~(\ref{smax-exp}) and Eq.~(\ref{s_eps-eq})
348: it can be seen that $s_{\rm max}$ increases with decreasing
349: $\alpha$, whereas $E_{\rm gap}/\Delta$ decreases. The Thouless
350: energy $E_T$ also decreases, since the mean level spacing $\delta$
351: becomes smaller for larger area $A$ of the normal region. The two
352: effect together result in a finite value of $E_{\rm gap}/E_T$ at
353: $\alpha = 0$.
354:
355: In the limit $\xi_0 << s_{\rm max}$ (for example, in the semiclassical
356: limit of large $M$), the expression~(\ref{gap-eq}) of the gap
357: reduces to
358: \begin{equation}
359: \frac{E_{\rm gap}}{E_{\rm T}} = \pi^2 \,
360: \frac{A}{Ws_{\rm max}}.
361: \label{gap-semi-eq}
362: \end{equation}
363: In Fig.~\ref{gap-fig} the dashed curve is the result from this
364: formula. One can see that for all angles $\alpha$ this limiting
365: result for the energy gap is larger than the finite $M$ value.
366: Note that the area is $A=\frac{1}{2} \,W^2\,\cot \alpha $, and
367: therefore, $E_{\rm gap}/E_{\rm T}$ is {\em only} a function of
368: $\alpha$ and independent of $W$ when $\xi_0 << s_{\rm max}$. It
369: can also be shown that $E_{\rm gap}/E_{\rm T} \rightarrow \pi^2/4
370: \approx 2.47$ as $\alpha \rightarrow 0.$ Finally, it is worth
371: mentioning that for Andreev billiard of Fig.~\ref{geo-fig} when
372: $d=0$ the energy gap
373: can be much larger for small angle $\alpha$ than the value of
374: $0.6 E_T$ predicted by Melsen et al.~\cite{Melsen} for the chaotic
375: billiard.
376:
377: It is interesting to mention that besides the billiard studied
378: in this work (see Fig.~\ref{geo-fig} for $d=0$)
379: there are a number of different shapes of
380: normal dots in which $P(s)$ exhibits finite upper cut-off resulting in a
381: sub-gap. Examples for such dots are shown in
382: Fig.~\ref{other_dots-fig} (provided $\alpha +\beta < 180^{\circ}$
383: in the case of quadrangle).
384: \begin{figure}[hbt]
385: \includegraphics[scale=0.45]{other-dot.eps}
386: %\centerline{\leavevmode \epsfxsize=85mm \epsffile{other-dot.eps }}
387: \caption{A triangle (a), and quadrangle (b) shape of normal billiard
388: having a finite upper cut-off of the length distribution $P(s)$.
389: \label{other_dots-fig}}
390: \end{figure}
391:
392: In conclusion, we have shown that a billiard with a finite upper
393: cut-off in the path length distribution $P(s)$ will possess an
394: energy gap in the density of states on the scale of the Thouless
395: energy. By including the energy dependent phase shift of the
396: Andreev reflection, a new expression for the density of states
397: has been given within the frame work of the semiclassical
398: Bohr-Sommerfeld approximation. We have also derived a formula for
399: the energy gap. To check these results we have performed an exact
400: diagonalization of the Bogoliubov-de Gennes equation for
401: different Andreev billiards. The results of the two methods
402: agrees very well both for the integrated
403: density of states (even for energy levels close to the value of
404: $\Delta$) and for the energy gap. Finally we have shown that the
405: energy gap on the scale of the Thouless energy can be much larger
406: than the value $0.6 E_T$ predicted from random matrix theory\cite{Melsen}
407: for chaotic billiards.
408:
409: One of us (J. Cs.) gratefully acknowledges very helpful discussions
410: with C. Beenakker.
411: This work was supported by the EU.\ RTN within the programme
412: ``Nanoscale Dynamics, Coherence and Computation'',
413: the Hungarian Science Foundation OTKA TO25866 and TO34832.
414: One of us (Z. K.) would like to thank the Hungarian
415: Academy of Sciences for support as a J\'anos Bolyai fellowship.
416:
417:
418:
419: %\begin{references}
420: \begin{thebibliography}{10}
421:
422:
423: \bibitem{Kosztin}
424: I. Kosztin, D.~L. Maslov, and P.~M. Goldbart, Phys. Rev. Lett. {\bf 75}, 1735
425: (1995).
426:
427: \bibitem{Melsen}
428: J.~A. Melsen, P.~W. Brouwer, K.~M. Frahm, and C.~W.~J. Beenakker, Eur. Phys.
429: Lett. {\bf 35}, 7 (1996);
430: J.~A. Melsen, P.~W. Brouwer, K.~M. Frahm, and C.~W.~J. Beenakker, Physica
431: Sripta {\bf T69}, 223 (1997).
432:
433: \bibitem{Altland}
434: A. Altland and M.~R. Zirnbauer, Phys. Rev. Lett. {\bf 76}, 3420 (1996).
435:
436: \bibitem{Lesovik}
437: G.~B. Lesovik, A.~L. Fauch\`ere, and G. Blatter, Phys. Rev. B {\bf 55}, 3146
438: (1997).
439:
440: \bibitem{Nazarov}
441: A. Lodder and Y.~V. Nazarov, Phys. Rev. B {\bf 58}, 5783 (1998).
442:
443: \bibitem{Heny}
444: H. Schomerus and C.~W.~J. Beenakker, Phys. Rev. Lett. {\bf 82}, 2951
445: (1999).
446:
447: \bibitem{Richter1}
448: W. Ihra, M. Leadbeater, J.~L. Vega, and K. Richter,
449: Eur.\ Phys.\ J.\ B {\bf 21}, 425 (2001)
450:
451: \bibitem{Richter2}
452: W. Ihra and K. Richter, Physica E {\bf 9}, 362 (2001).
453:
454: \bibitem{note_1}
455:
456: The same definition for the Thouless energy is used
457: as that for example, in Refs.~\onlinecite{Melsen,Heny}.
458:
459: \bibitem{BdG-eq}
460: P.~G. de~Gennes, {\em Superconductivity of Metals and Alloys} (Benjamin, New
461: York, 1996).
462:
463: \bibitem{note_2}
464: The $\arccos$ term in Eq.~(\ref{s_eps-eq}) due to
465: the phase shift $-\arccos(E/\Delta)$
466: for Andreev reflection at the N-S interface. This energy dependent
467: correction was approximated by $\pi/2$
468: in Refs.~\cite{Melsen,Nazarov,Heny,Richter1}
469: which is strictly valid only at $E=0$.
470: In this case the first term in the square bracket of Eq.~(\ref{n-eq})
471: is missing.
472:
473: \bibitem{elsewhere}
474: Details of the analytical calculations will be given in a later publication.
475:
476: \bibitem{Heller}
477: L. Kaplan, E. J. Heller, Physica D {\bf 121}, 1 (1998).
478:
479: \bibitem{Carlo_2_3}
480: C.~W.~J. Beenakker, Phys. Rev. Lett. {\bf 67}, 3836 (1991), [{\bf 68}, 1442,
481: (1992) (Erratum)];
482: C.~W.~J. Beenakker, in {\em Transport Phenomena in Mesoscopic Systems},
483: edited by H. Fukuyama and T. Ando (Springer-Verlag, Berlin, 1992),
484: pp.\ 235--253.
485:
486: \bibitem{Colin-review}
487: In this case $\Delta/E_{\rm F} \ll 1$ and quasi-particles
488: whose incident/reflected directions are approximately
489: perpendicular to the N-S interface, see e.g.\
490: C.~J. Lambert and R. Raimondi, J. Phys. Condens. Matter {\bf 10}, 901 (1998).
491: O. \v{S}ipr and B.~L. Gy{\"o}rffy, J. Phys.: Condens. Matter {\bf 8}, 169
492: (1996);
493: O. \v{S}ipr and B.~L. Gy{\"o}rffy, J. Low. Temp. Phys. {\bf 106}, 315 (1997).
494:
495: \bibitem{note_3}
496: $P(s)$ was calculated analytically for $d=0$ and
497: numerically using Monte Carlo methods for $d\ne0$.
498:
499: \bibitem{box-disk-paper} J. Cserti, A. Bodor, J. Koltai, G. Vattay,
500: preprint, cond-mat/0105472.
501:
502:
503: %\end{references}
504: \end{thebibliography}
505:
506:
507:
508:
509: \end{document}
510:
511: %%%%%%%%%%%%%%%%% junk
512:
513: Figure \ref{N-nonzero_d-fig} shows that the two are in excellent agreement,
514: even for large energies close to $\Delta$. For comparison the dashed line
515: in figure 4 shows the formula ???, which
516: does incorporate the energy dependent phase shift at the N-S interface.
517: As expected, this agrees with the exact result at low energies,
518: but deviates from it at higher energies.
519:
520:
521: For the billiard of Fig.~\ref{geo-fig} with $d=0$, one can show
522: analytically that
523: \begin{equation}
524: \frac{s_{\rm max}}{W}= \left\{
525: \begin{array}{ll}
526: \frac{1-\cos \alpha/\cos 5\alpha}{\sin \alpha} & \mbox{if \,\,
527: $30^\circ \le \alpha <33.75^\circ$} \\[1.2ex]
528: \frac{\sqrt{2\left(1-\cos 4\alpha \right)}}{\sin \alpha} & \mbox{if \,\,
529: $33.75^\circ \le \alpha < 45^\circ$} \\[1.2ex]
530: -2\,\frac{\sin 2\alpha}{\cos 3\alpha} & \mbox{if \,\,
531: $45^\circ \le \alpha < 54^\circ$} \\[1.2ex]
532: 2 &\mbox{if \,\, $54^\circ \le \alpha < 90^\circ$}
533: \end{array}
534: \right.
535: \label{smax-exp}
536: \end{equation}
537:
538:
539:
540:
541: Figures~\ref{N-zero_d-fig} and \ref{N-nonzero_d-fig} show a
542: comparisons between the semiclassical expression of $N(E)$ given
543: by Eq.~(\ref{N-eq})~\cite{note_3}, and
544: an exact (numerical) diagonalization of the Bogoliubov-de Gennes
545: equation.
546:
547:
548: It can be shown that if $P(s)$ possesses an upper cut-off at
549: $s=s_{\rm max}$, then as $M\rightarrow \infty$ the density of states
550: $n(E)$ possesses an energy gap.
551:
552:
553: Given that no ballistic dot has yet been identified which exhibits a true
554: energy gap on the scale of $E_T$, it is of interest to ask
555: if such dots exist.
556: