1: \documentclass[prb,aps,twocolumn,showpacs,superscriptaddress,floatfix]{revtex4}
2: \usepackage{epsfig}
3: \usepackage{amsmath}
4: \usepackage{amsfonts}
5: \usepackage{mathptmx}
6: \usepackage{eucal}
7: \usepackage{bm}
8:
9: \DeclareMathOperator{\Tr}{\mathop{\mathrm{Tr}}}
10: \DeclareMathOperator{\sgn}{\mathop{\mathrm{sgn}}}
11: \DeclareMathOperator{\re}{\mathop{\mathrm{Re}}}
12: \DeclareMathOperator{\im}{\mathop{\mathrm{Im}}}
13: \DeclareMathOperator{\arctanh}{arctanh}
14: \DeclareMathOperator{\const}{\mathop{\mathrm{const}}}
15: \newcommand{\Eq}[1]{Eq.~(\ref{#1})}
16: \newcommand{\Eqs}[1]{Eqs.~(\ref{#1})}
17:
18: \hyphenpenalty=2
19:
20: \begin{document}
21:
22: \title{Subgap current in superconducting tunnel junctions with diffusive electrodes}
23:
24: \author{E.~V.~Bezuglyi}
25: %
26: \affiliation{Institute for Low Temperature Physics and
27: Engineering, Kharkov 61103, Ukraine}
28: %
29: \affiliation{Chalmers University of Technology, S-41296
30: G\"oteborg, Sweden}
31:
32: \author{A.~S.~Vasenko}
33: %
34: \affiliation{Chalmers University of Technology, S-41296
35: G\"oteborg, Sweden}
36: %
37: \affiliation{Department of Physics, Moscow State University,
38: Moscow 119992, Russia}
39:
40: \author{E.~N.~Bratus'}
41: %
42: \affiliation{Institute for Low Temperature Physics and
43: Engineering, Kharkov 61103, Ukraine}
44: %
45:
46: \author{V.~S.~Shumeiko}
47: %
48: \affiliation{Chalmers University of Technology, S-41296
49: G\"oteborg, Sweden}
50:
51: \author{G.~Wendin}
52: %
53: \affiliation{Chalmers University of Technology, S-41296
54: G\"oteborg, Sweden}
55:
56: \date{\today}
57:
58: \begin{abstract}
59: We calculate the subgap current in planar superconducting tunnel
60: junctions with thin-film diffusive leads. It is found that the
61: subharmonic gap structure of the tunnel current scales with an
62: effective tunneling transparency which may exceed the junction
63: transparency by up to two orders of magnitude depending on the
64: junction geometry and the ratio between the coherence length and
65: the elastic scattering length. These results provide an
66: alternative explanation of enhanced values of the subgap current
67: in tunneling experiments often ascribed to imperfection of the
68: insulating layer. We also discuss the effect of finite lifetime of
69: quasiparticles as the possible origin of additional enhancement of
70: multiparticle tunnel currents.
71: \end{abstract}
72:
73: \pacs{74.45.+c, 74.40.+k, 74.25.Fy, 74.50.+r}
74:
75: \maketitle
76:
77: Subgap quasiparticle current in superconducting junctions at small
78: applied voltages $eV<2\Delta$ is the subject of persistent
79: theoretical interest and experimental research. Recently, the
80: problem has attracted new attention, and a number of measurements
81: of the subgap current in high-quality tunnel junctions have been
82: performed,\cite{Gubrun2001,Lang2003} motivated by the problem of
83: decoherence in Jo\-seph\-son-junction-based superconducting
84: qubits.\cite{Makhlin} The subgap current at zero temperature is
85: due to multiparticle tunneling (MPT) processes,\cite{MPT} whose
86: intensities strongly depend on the quality of the insulating
87: layer, being enhanced by disorder, localized electronic states,
88: pinholes, etc.\cite{KBT} The effect of disorder in the junction
89: electrodes on the subgap current has never been questioned.
90:
91: According to the MPT theory,\cite{MPT} the subgap tunnel current
92: depends on the transparency $D$ of the tunnel barrier: it
93: decreases with decreasing voltage in a steplike fashion with step
94: heights proportional to $(D/2)^n$ at voltages $eV=2\Delta/n$,
95: $n=1,2...$ [subharmonic gap structure (SGS)]. Similar results have
96: been obtained for junctions with ballistic
97: electrodes,\cite{Bratus95} and mesoscopic point contacts with
98: diffusive electrodes\cite{Scheer} on the basis of the theory of
99: multiple Andreev reflections (MAR).\cite{KBT}
100:
101: Experimentally, the SGS scaling parameter in atomic size junctions
102: nicely agrees with the theory;\cite{Jan2000} however, in
103: macroscopic tunnel junctions it is usually much larger
104: \cite{Gubrun2001,Lang2003} (see also earlier data
105: \cite{Cristiano}); moreover, there is a smooth residual current at
106: a very low voltage.\cite{Gubrun2001} Although enhanced SGS in
107: high-trans\-mission junctions could be explained by assuming
108: randomly distributed resonant levels within the tunnel
109: barrier,\cite{Likharev} enhanced subgap current in
110: low-transmission junctions with presumably good insulating layers
111: remains an open question.
112:
113:
114: In this paper we reexamine the problem of the subgap current in
115: {\em macroscopic} tunnel junctions, and consider the effects of
116: diffusive electrodes and planar junction geometry common for the
117: experiment (see Fig.~\ref{model}). Our main result is that the SGS
118: scaling parameter for such junctions significantly exceeds the
119: junction transparency: for the sandwich-type junction with
120: thin-film leads shown in Fig.~\ref{model}(b), the scaling is
121: determined by the effective transparency defined as
122: %
123: \begin{equation}\label{Deff}
124: D_{\textit{eff}} = (3\xi_0^2/\ell d)D,
125: \end{equation}
126: %
127: where $\xi_0=\sqrt{\mathcal{D}/2\Delta}$ is the diffusive
128: coherence length (we assume $\hbar = k_B =1)$, $\ell$ is the
129: elastic mean free path, $d \ll \xi_0$ is the thickness of the
130: leads, and $\mathcal{D}$ is the diffusion coefficient. For Al
131: junctions with $\ell \sim d = 50$ nm and $\xi_0 = 300$ nm, the
132: enhancement factor approaches 100. For the junctions with
133: one-dimensional (1D) geometry of Fig.~\ref{model}(a),
134: $D_{\textit{eff}} = (3\xi_0/\ell)D$.\cite{Jos} This result also
135: applies to nonhomogeneous tunnel barriers as soon as the size of
136: pinholes (more transparent spots) exceeds the elastic mean free
137: path, otherwise the ballistic scaling\cite{Bratus95} will be
138: valid.
139:
140: %
141: \begin{figure}[bt]
142: \epsfxsize=8.5cm\epsffile{model.eps} \vspace{-2mm}
143: %
144: \caption{One-dimensional (a) and planar (b) models of the tunnel
145: junction with diffusive leads; equilibrium reservoirs are far from
146: the contact, $L \gg\xi_0$.} \label{model} \vspace{-4mm}
147: \end{figure}
148: %
149:
150: The enhancement effect can be qualitatively understood by
151: considering a short point contact with the reservoirs located very
152: close to the contact, $L\ll\xi_0$ [cf. Fig.~\ref{model}(a)]. In
153: this case, the current can be calculated within the mesoscopic
154: approach,\cite{Averin97} by integrating over contributions of
155: normal conducting eigenmodes with randomly distributed
156: transparencies. The relevant distribution is known to be spread
157: over the interval $\sim (L/\ell)D \gg D$.\cite{Beenakker} The most
158: transparent modes dominate the subgap current, giving
159: $D_{\textit{eff}} \sim (L/\ell)D$. In our case of junctions with
160: large distance to the reservoirs, the scale of the spatial
161: variation of the Green's function $\xi_0$ plays the role of the
162: effective junction length giving qualitatively our result,
163: $D_{\textit{eff}} \sim (\xi_0/\ell)D$.\cite{note} We note that for
164: the long junctions under consideration the statistics of the
165: eigenmode transparencies is not known, and a quantitative result
166: has to be derived from the quasiclassical theory for diffusive
167: superconductors.
168:
169: Our analysis is based on the diffusive equations \cite{LOnoneq}
170: for the quasiclassical two-time Keldysh-Green functions
171: $\check{G}(\bm{r},t_1,t_2)$,
172: %
173: \begin{align} \label{Keldysh}
174: [\check{H},\circ\check{G}]=i {\mathcal D}\nabla \check{J}, \quad
175: \check{G}\circ \check{G}=1,\quad
176: %
177: \quad \check{G} =\begin{pmatrix}
178: \hat{g}^R & \hat{G}^K \\
179: 0 & \hat{g}^A \end{pmatrix}.
180: \end{align}
181: %
182: Here $\check{H}(t_1,t_2)=[i\sigma_z\partial_{t_1} + \Delta
183: \exp(i\sigma_z \phi)i\sigma_y]\delta(t_1-t_2)$, $\phi$ is the
184: superconducting phase, the sign $\circ$ denotes time convolution,
185: and $\check{J}=\check{G} \circ \nabla \check{G}$. Equation
186: \eqref{Keldysh} can be decomposed into the Us\-a\-del equation for
187: the retar\-ded or advan\-ced Green's functions $\hat{g}^{R,A}$ and
188: the equation for the Keldysh function $\hat{G}^K = \hat{g}^R
189: \circ\hat{f} - \hat{f}\circ \hat{g}^A$, where $\hat{f} = f +
190: \sigma_z f_-$ is the distribution function.
191:
192: We present detailed calculations for the simpler, 1D geometry of
193: Fig.~\ref{model}(a). At the left electrode, $x = -L$, the Fourier
194: transformations of the two-time functions $\hat{g}$ and $\hat{f}$
195: with respect to the variable $t_1-t_2$ are given by the
196: equilibrium expressions
197: %
198: \begin{align}
199: \hat{g}(E) &= \sigma_z u(E) + i\sigma_y v(E), \quad
200: \hat{f}(E)=\tanh({E}/{2T}), \label{g0}\\
201: (u,v)&=(E,\Delta)/\xi(E), \quad \xi^{R,A} = [{(E\pm
202: i0)^2-\Delta^2}]^{1/2}.
203: \end{align}
204: %
205: At the right voltage-biased electrode, $x=L$, the function
206: $\check{G}$ is defined through the gauge
207: transformation\cite{Artemenko}
208: %
209: \begin{align} \label{gauge}
210: \check{G}(L) = \overline{\check{G}}(-L) \equiv S(t_1)
211: \check{G}(-L) S^\dagger(t_2),
212: \end{align}
213: %
214: with a unitary operator $S(t) =\exp[i\sigma_z\phi(t)/2]$, where
215: the phase $\phi$ satisfies the Josephson relation $\phi(t) =
216: 2eVt$.
217:
218: The boundary conditions \cite{KL} for the functions $\check{G}$
219: and $\check{J}$ at the tunnel barrier ($x=\pm 0$) are given by the
220: relations
221: %
222: \begin{equation}
223: \check{J}_{-0} = \check{J}_{+0} = \frac{W}{\xi_0}
224: \bigl[\check{G}_{-0},\circ \check{G}_{+0}\bigr],\quad W=
225: \frac{R_0}{2R} =\frac{3\xi_0}{4\ell}D,\label{Boundary2}
226: \end{equation}
227: %
228: where $R$ is the resistance of the tunnel barrier, $R_0= \xi_0/g$
229: is the resistance of a piece of the lead with length $\xi_0$, and
230: $g$ is the conductance of the leads per unit length. Assuming a
231: small value of the tunneling parameter $W$, we neglect the charge
232: imbalance function $f_-$ and the superfluid momentum within the
233: leads, as well as the variation of $\Delta$. In such an
234: approximation, \Eq{gauge} extends to the whole right lead,
235: $\check{G}(x) = \overline{\check{G}}(-x)$ for $0 < x < L$. The
236: problem is therefore reduced to the solution of a static equation
237: for the function $\check{G}(x,t_1,t_2)$ at $-L<x<0$ with the
238: time-dependent boundary condition \eqref{Boundary2} at $x=-0$. The
239: electric current is related to the Keldysh component $\hat{J}^K$
240: of the matrix current $\check{J}$ as $I(t)=({\pi g}/{4e}) \Tr
241: \sigma_z \hat{J}^K(x,t,t)$.\cite{LOnoneq} Using \Eqs{gauge} and
242: \eqref{Boundary2}, it can be expressed as
243: %
244: \begin{equation}
245: I(t)=(\pi /8eR) \Tr \sigma_z
246: [\check{G},\circ\overline{\check{G}}]^K(t,t). \label{Curr2}
247: \end{equation}
248: %
249: In this and following equations, the functions are taken at the
250: boundary $x=-0$. Expanding all functions over harmonics of the
251: Josephson frequency, $A(E,t)=\sum\nolimits_m A(E,m)e^{-2ieVm t}$
252: [$t=(t_1+t_2)/2$], we arrive at the equation for the dc current
253: $I$,
254: %
255: \begin{align}
256: I &= \frac{1}{16eR}\int_{-\infty }^\infty dE\, \Tr
257: \sum\nolimits_m \bigl[\hat{h}(E,m)\overline{\hat{G}^K}(E,-m)\nonumber
258: \\
259: &-\overline{\hat{h}}(E,m) \hat{G}^K(E,-m)\bigr], \quad \hat{h}=
260: \sigma_z \hat{g}^R - \hat{g}^A \sigma_z. \label{I0}
261: \end{align}
262: %
263:
264: In the tunneling limit $W\ll 1$, the amplitude of the $m$th
265: harmonic is proportional to $W^m$; thus the zero harmonic $m=0$ of
266: the functions $\hat{g}$ and $\hat{G}^K$ in \Eq{I0} plays the key
267: role, while the high-order harmonics can be neglected. The effect
268: of these harmonics will be discussed later. Within this
269: approximation, the Green's function matrix structure in \Eq{g0}
270: holds, and the current \Eq{I0} {\em exactly} transforms to the
271: form
272: %
273: \begin{equation}\label{I00}
274: I = \frac{1}{eR}\int_{-\infty }^\infty dE\, N(E) N(E-eV)
275: [n(E-eV)-n(E)].
276: \end{equation}
277: %
278: Here $N(E)=\re u^R$ is the density of states (DOS) normalized to
279: its value in the normal state, and the distribution function $n=
280: (1/2)(1-f)$ approaches the Fermi function $n_F$ in equilibrium.
281: Furthermore, we split the integral in \Eq{I00} into pieces of
282: length $eV$, denoting $A_k(E) \equiv A(E+keV)$,
283: %
284: \begin{align}\label{I001}
285: I &= \frac{1}{eR}\int_{0}^{eV}dE \, J(E), \quad J(E)=
286: \sum\nolimits_{k=-\infty}^\infty j_k(E), \\
287: j_k &= ({n_{k-1} - n_k}){\rho_k^{-1}}, \quad \rho_k^{-1}= N_k
288: N_{k-1}.\label{I002}
289: \end{align}
290: %
291: The distribution function $n(E,x=-0)$ satisfies the recurrence
292: relation following from the kinetic equation
293: $\partial_x\left(D_+\partial_x n\right) = 0$,
294: %
295: \begin{equation}\label{recurr}
296: \Theta(|E_k|-\Delta)[n_F(E_k) -n_k] = r (j_{k+1} - j_k),
297: \end{equation}
298: %
299: where $\Theta(x)$ is the Heaviside step function, $r = R_N/R \ll
300: 1$, and $R_N$ is the normal resistance of the lead. To justify
301: \Eq{recurr}, we note that the diffusion coefficient $D_+= (1/2)(1
302: + |u|^2 - |v|^2)$ is approximately constant, $D_+ \approx 1$, at
303: $|E|>\Delta$, which leads to the linear function $n(E,x)=n_{-0} +
304: (x/L)(n_{-0}-n_F)$. At $|E|<\Delta$, $D_+$ turns to zero at $|x|
305: \gg \xi_0$, which reflects complete Andreev reflection in the
306: leads and results in zero probability current, $D_+\partial_x n
307: =0$. Then, using the boundary condition at the tunnel barrier
308: following from \Eq{Boundary2}, $D_+\partial_x n =
309: (2W/\xi_0)\sum\nolimits_{k=\pm 1}NN_k (n_k - n)$, we arrive at
310: \Eq{recurr}.
311:
312: %
313: \begin{figure}[tb]
314: \epsfxsize=8.5cm\epsffile{circuit.eps}\vspace{-3mm}
315: %
316: \caption{Circuit representation of charge transport in a diffusive
317: tunnel junction, $eV = 2.5\Delta$.} \label{circuit}\vspace{-5mm}
318: \end{figure}
319: %
320:
321: A convenient interpretation of \Eqs{I002} and \eqref{recurr} in
322: terms of circuit theory \cite{Bezugly} is given by an infinite
323: network in the energy space with the period $eV$, graphically
324: presented in Fig.~\ref{circuit}. The electric current spectral
325: density\ $J(E)$ consists of partial currents $j_k$, which flow
326: through the tunnel ``resistors'' $\rho_k$ connected to adjacent
327: nodes having ``potentials'' $n_k$ and $n_{k-1}$. At $|E|>\Delta$,
328: the nodes are also attached to the distributed ``equilibrium
329: source'' $n_F(E)$ through equal resistors $r$. Below we impose the
330: equilibrium quasiparticle distribution at $|E|>\Delta$, $n(E)=
331: n_F(E)$, neglecting the effect of small resistors $r$.
332:
333: The currents flowing between the nodes outside the gap are related
334: to the thermal current; at $T=0$, these nodes have equal
335: populations ($n_k=1$ at $E_k<-\Delta$, $n_k=0$ at $E_k>\Delta$),
336: thus the corresponding partial currents are zero, and the thermal
337: current vanishes. As a result, only the current $j_0$ flowing
338: across the gap through the resistor $\rho_0$ survives at $T=0$.
339:
340:
341: Taking the DOS in the BCS form $N=N_S \equiv \re (E/\xi^R)$, we
342: see that if any node falls into the gap, the adjacent resistances
343: turn to infinity, and the current vanishes. For this reason, the
344: network period must exceed the gap, $eV>2\Delta$, and the
345: integration in \Eq{I001} is confined to the region $\Delta < E <
346: eV - \Delta$. This recovers the tunneling model result for the
347: single-particle tunnel current:\cite{Werthamer} the current
348: appears above the threshold, $eV = 2\Delta$, having the threshold
349: value $I_1(2\Delta)= \pi\Delta/2eR $.
350:
351: To evaluate the subgap current, $eV <2\Delta$, the DOS must be
352: calculated to next order in the parameter $W$, which requires
353: solution of the equations for $\hat{g}$ following from
354: \Eqs{Keldysh} and (\ref{Boundary2}). Using the standard
355: parametrization $\hat{g}= \sigma_z e^{\sigma_x\theta}$, we obtain
356: the equation and the boundary condition for the spectral angle
357: $\theta$,
358: %
359: \begin{align}
360: &\sinh(\theta - \theta_S) = i \partial_{zz}\theta\sinh\theta_S ,
361: \;\;\;z = x/\xi_0,
362: \label{Eqtheta}\\
363: &\partial_z\theta +W\sinh\theta (\cosh\theta_1 +
364: \cosh\theta_{-1})=0\quad (z=-0). \label{Boundtheta}
365: \end{align}
366: %
367: With exponential accuracy, the solution of \Eq{Eqtheta} for $z<0$
368: can be approximated by the formula for a semi-infinite wire,
369: %
370: \begin{align}
371: \tanh\{[\theta(z) - \theta_S]/4\} = \tanh[({\theta_{-0} - \theta_S
372: })/{4}] \exp(kz), \label{Soltheta}
373: \end{align}
374: %
375: where $k^{-1}(E) ={\sqrt{i\sinh\theta_S}}$. Equation
376: \eqref{Soltheta} describes the decay of perturbations of the
377: spectral functions at distances $\gtrsim \xi_0$ from the barrier,
378: where the spectral angle approaches its bulk value $\theta_S=
379: \arctanh (\Delta/E)$. The boundary value of $\theta$ is to be
380: found from the equation following from \Eqs{Boundtheta} and
381: \eqref{Soltheta},
382: %
383: \begin{equation}
384: 2k\sinh [({\theta_S - \theta})/{2}] =W\sinh\theta (\cosh\theta_1 +
385: \cosh\theta_{-1}). \label{Eqtheta0}
386: \end{equation}
387: %
388: A direct expansion of $\theta$ with respect to $W$ in
389: \Eq{Eqtheta0} leads to the following expression for the DOS within
390: the BCS gap,
391: %
392: \begin{equation}\label{DOSSPT}
393: N(E)=W({1 - E^2/\Delta^2})^{-5/4} [N_S(E+eV)+N_S(E-eV)].
394: \end{equation}
395: %
396: The DOS divergencies at $|E|= \Delta, \Delta -eV$ in \Eq{DOSSPT}
397: are potentially dangerous (cf. Refs.~\onlinecite{MPT}), but they
398: can be eliminated by improving the perturbation proce\-dure by
399: solving a set of recurrences in \Eq{Eqtheta0} in the vicinity of
400: these points.
401:
402: %
403: \begin{figure}[tb]
404: \epsfxsize=8.5cm\epsffile{curr23.eps}\vspace{-2mm}
405: %
406: \caption{DOS and subgap circuits at the applied voltages
407: $eV=1.2\Delta$ (a) and $0.7\Delta$ (b), for the tunneling
408: parameter $W =10^{-3}$.} \label{DOS}\vspace{-4mm}
409: \end{figure}
410: %
411:
412: As follows from \Eq{DOSSPT}, the tunneling processes transfer the
413: DOS in the energy space into the BCS gap at the distances $\pm eV$
414: from the regions $|E|>\Delta$, thus forming an effective spatial
415: {\em potential well} of the width $\sim \xi_0$ at the tunnel
416: barrier. At $eV > \Delta$ the BCS gap is entirely filled with the
417: quasiparticle states with a small local DOS $\sim W$, as shown in
418: Fig.~\ref{DOS}(a). The appearance of localized states enables the
419: quasiparticles to overcome the BCS gap at $eV < 2\Delta$ via two
420: steps involving intermediate Andreev reflection at energies
421: $|E|<\Delta$. The population of the intermediate state cannot be
422: taken to be in equilibrium because the subgap quasiparticles
423: cannot access the equilibrium electrodes. In the circuit terms,
424: the node $k=0$ is disconnected from the equilibrium source, and
425: the subgap current flows through the two large resistances
426: $\rho_0,\rho_1 \sim W^{-1}$ (two-particle current), see
427: Fig.~\ref{DOS}(a). The corresponding partial currents are equal,
428: $j_0 = j_1 = [{n_F(E_1)-n_F(E_{-1})}]/({\rho_0 + \rho_1})$, and
429: their contribution to $I(V)$ is confined to the energy region $0<
430: E < eV-\Delta$ (a similar contribution at $\Delta < E < eV$ comes
431: from $j_0$ and $j_{-1}$). Thus the two-particle current appears
432: above the threshold $eV = \Delta$, having the threshold value
433: $I_2(\Delta) = \pi W\Delta/eR = 2WI_1(2\Delta)$. At $eV= 2\Delta$,
434: the two-particle current exhibits a sharp peak with the height
435: $I_2(2\Delta) \approx 2.3 {W^{2/5}\Delta }/{eR}$; at larger
436: voltages, it approaches a constant value giving rise to the excess
437: current $I_{\textit{exc}}\approx 6.2 {W^{2/3}\Delta }/{eR}$.
438:
439: At $eV < \Delta$, a minigap opens in the DOS around the zero
440: energy [see Fig.~\ref{DOS}(b)], however, since the number of
441: subgap resistors increases up to three (three-particle current),
442: the current across the minigap will persist as long as the network
443: period exceeds the minigap size, $eV> 2(\Delta-eV)$, i.e., at $eV
444: >2\Delta/3$. The central resistance $\rho_0$ is large,
445: $\rho_0 \sim W^{-2}$, and dominates the net subgap resistance.
446: This leads to a smaller charge current with the threshold value
447: $I_3(2\Delta/3) \approx 2WI_2(\Delta)$. At $eV<2\Delta/3$ the
448: network period becomes smaller than the minigap, and further
449: correction to the DOS is required.
450:
451: Similar results were found for the planar junction
452: Fig.~\ref{model}(b), using the equation for the functions
453: $\check{G}_{\pm 0}$ at the top ($+0$) and bottom ($-0$) sides of
454: the tunnel barrier, $i[\sigma_z E + i\sigma_y
455: \Delta,\check{G}_{-0}] = 2\Delta W[\check{G}_{-0},
456: \check{G}_{+0}]$, with the modified tunneling parameter
457: $W=(3\xi_0^2/4\ell d)D$. This equation is derived by averaging
458: \Eq{Keldysh} over the thickness of overlapping leads and using
459: \Eq{Boundary2} (cf. Ref.~\onlinecite{VolkovSIN}). From this
460: equation we obtain a relation for the spectral angle that does not
461: significantly differ from \Eq{Eqtheta0},
462: %
463: \begin{equation}
464: k^2\sinh ({\theta_S - \theta}) =W\sinh\theta (\cosh\theta_1 +
465: \cosh\theta_{-1}), \label{Eqtheta1}
466: \end{equation}
467: %
468: thus giving results which are close to those for the 1D model with
469: the same magnitude of the parameter $W$.
470:
471: The presented calculation scheme, combining circuit theory
472: arguments with DOS iteration procedures, suggests an appealingly
473: simple explanation for the diffusive SGS: the decreasing applied
474: voltage results in a shrinking period of the network in
475: Fig.~\ref{circuit}; hence a stepwise increase of the number of
476: subgap resistors involved; simultaneously, the number of DOS
477: steps, scaled as $W^n$, increases, as shown in Fig.~\ref{iv}(a).
478: This results in the current staircase with the height of the steps
479: given by $I_n\sim (2W)^{n-1}I_1$, at $2\Delta/n<eV<2\Delta/(n-1)$.
480: The quantitative result for the current at arbitrary voltages and
481: temperatures is
482: %
483: \begin{align}
484: I(V) &= \int_0^{eV} \frac{dE}{eR} \frac{N_+ + N_-}{\rho_\Delta}
485: (n_- -n_+) + \int_\Delta^{\infty} \frac{2dE}{eR\rho_1} (n_F -
486: n_{F1}),\nonumber \\ N_\pm &= \text{Int}\,[(\Delta \mp
487: E)/eV]+1,\quad n_\pm(E) = n_F(E_{\pm N_\pm}).\label{Inet}
488: \end{align}
489: %
490: In this equation, the second term represents the thermal current,
491: the integers $\pm N_\pm$ are the indices of the nodes closest to
492: the gap edges outside the gap, Int$(x)$ denotes integer part of
493: $x$, and the quantity $\rho_\Delta(E) = \sum\nolimits_{k= 1-N_-
494: }^{N_+} \rho_k$ has the meaning of net subgap resistance. The
495: subgap distribution function reads
496: %
497: \begin{align} \label{nnet}
498: n(E)=n_+ + (n_- -
499: n_+)\rho_\Delta^{-1}\sum\nolimits_{k=1}^{N_+}\rho_k.
500: \end{align}
501: %
502:
503: Equations \eqref{Inet} and \eqref{nnet} are the main technical
504: results of the paper. The $I$-$V$ characteristic (IVC) of the
505: planar tunnel junction calculated from \Eqs{Inet} and
506: \eqref{Eqtheta1} at $T=0$ and shown in Fig.~\ref{iv}(b), was found
507: to be very close to the result for a ballistic point contact
508: \cite{Bratus95} with the effective transparency $D_{\textit{eff}}
509: = 4W= (3\xi_0^2/\ell d)D$. This justifies our statement made in
510: the introduction, and is the main conclusion of this paper.
511:
512: In low-transmissive junctions, enhanced subgap current at
513: $eV<\Delta$ has been observed (see, e.g.,
514: Ref.~\onlinecite{Gubrun2001}). This anomaly might be due to
515: many-body interaction effects which introduce a finite lifetime
516: (damping) of the quasiparticles. The damping effect can be
517: qualitatively modeled by a small imaginary addition to the energy
518: in the spectral functions, $E \to E+i\gamma$. This would lead to a
519: small residual DOS within the BCS gap and cut the DOS staircase at
520: the level of the order of $\gamma/\Delta$, see Fig.~\ref{iv}(a).
521: This will result in the smearing of the tunneling SGS and
522: crossover to a linear IVC at low voltages, $I=2.2(\gamma/\Delta)^2
523: V/R$, similar to the incoherent MAR regime.\cite{Bezugly} The IVC
524: calculated from \Eq{Inet} for $\gamma = 0.003\Delta$ and shown in
525: Fig.~\ref{iv}(b) by a dashed line confirms these considerations.
526:
527: %
528: \begin{figure}[tb]
529: \epsfxsize=8.5cm\epsffile{iv.eps}\vspace{-3mm}
530: %
531: \caption{DOS at $eV=0.4\Delta$ (a) and $I$-$V$ characteristics (b)
532: for the tunneling parameter $W=10^{-3}$ and two values of the
533: damping parameter: $\gamma =0$ (solid line) and $\gamma=
534: 0.003\Delta$ (dashed line).} \label{iv}\vspace{-5mm}
535: \end{figure}
536: %
537:
538: We conclude our analysis with the estimation of the contribution
539: of higher harmonics of the functions $\hat g$ and $\hat G^K$ to
540: the dc current. At $T=0$, the contribution $\delta I$ of the first
541: harmonics $|m|=1$ (the higher harmonics, $|m|>1$, are smaller,
542: $\sim W^m$) is
543: %
544: \begin{equation}\label{}
545: \delta I = \frac{2W}{eR} \int_0^{eV} dE\,\im v
546: \im \Bigl(\frac{v}{p} \cosh^2\frac{\chi}{2} + \frac{v}{q}
547: \sinh^2\frac{\tilde\chi}{2}\Bigr),
548: \end{equation}
549: %
550: where $\chi = \theta_{1}+ \theta_{-1}$, $\tilde\chi = \theta_{1}+
551: \theta_{-1}^\ast$, $(p,q)^2 = ({\xi_{1}^R +
552: \xi_{-1}^{R,A}})/{2i\Delta}$, and $v=\sinh\theta$. At $eV <
553: \Delta$, the energy $E_{-1}$ appears in the subgap region, where
554: $\theta^*_{-1} = \theta_{-1}+\pi i$ and $\xi_{-1}^A =\xi_{-1}^R$;
555: for this reason, $\delta I$ turns to zero at $eV < \Delta$,
556: similar to $I_2$. Numerical calculations show that the
557: contribution of the first harmonics to the IVC does not exceed
558: $30\%$. From this we conclude that the adopted quasistatic
559: approach gives a rather good approximation to a complete solution.
560:
561: In our treatment, we have neglected inelastic scattering, which
562: might affect the quasiparticle distribution at subgap energies.
563: Analysis shows that this effect becomes essential under the
564: condition $W\tau_\epsilon \Delta \ll 1$, where $\tau_\epsilon$ is
565: the relaxation time. However, this does not affect the estimate of
566: the effective scaling factor and only changes the details of the
567: IVC shape.
568:
569: In conclusion, we have developed a theory of subgap charge
570: transport and subharmonic gap structure in superconducting tunnel
571: junctions with planar geometry and diffusive thin-film electrodes.
572: We found that the role of scaling factor in the sub\-har\-mo\-nic
573: gap structure is played by the effective tunneling transparency
574: $D_{\textit{eff}}=(3\xi_0^2/\ell d)D$, which may greatly exceed
575: the bare transparency $D$ of the junction tunnel barrier.
576:
577: Support from the Swedish grant agencies SSF OXIDE, VR, and KVA is
578: gratefully acknowledged.
579: %
580: \vspace{-5mm}
581:
582: \begin{thebibliography}{99}
583:
584: \bibitem{Gubrun2001}
585: M.~A.~Gubrud {\it et al}, IEEE\ Trans.\ Appl.\ Supercond.\ {\bf
586: 11}, 1002 (2001).
587:
588: \bibitem{Lang2003}
589: K.~M.~Lang {\it et al.}, IEEE Trans.\ Appl.\ Supercond.\ {\bf 13},
590: 989 (2003); S.~Oh {\it et al.}, Supercond.\ Sci.\ Techn.\ {\bf
591: 18}, 1396 (2005).
592:
593: \bibitem{Makhlin}
594: Yu.~Makhlin, G.~Sch\"on, and A.~Shnirman, Rev.\ Mod.\ Phys.\ {\bf
595: 73}, 357 (2001); G.~Wendin and V.~S.~Shumeiko, in {\em Handbook in
596: Theoretical and Computational Nanoscience}, edited by M.~Rieth\
597: and\ W.~ Schommers (American Scientific Publishers, Los Angeles,
598: 2006) Chap. 129 (cond-mat/0508729).
599:
600: \bibitem{MPT}
601: J.~R.~Schrieffer and J.~W.~Wilkins, Phys.\ Rev.\ Lett.\ {\bf 10},
602: 17 (1963); L.~E.~Hasselberg, M.~T.~Levinsen, and M.~R.~Samuelsen,
603: Phys.\ Rev.\ B {\bf 9}, 3757 (1974).
604:
605: \bibitem{KBT}
606: T.~M.~Klapwijk, G.~E.~Blonder, and M.~Tinkham, Physica B+C {\bf
607: 109-110}, 1657 (1982).
608:
609: \bibitem{Bratus95}
610: E.~N.~Bratus', V.~S.~Shumeiko, and G.~Wendin, Phys.\ Rev.\ Lett.\
611: {\bf 74}, 2110 (1995); J.~C.~Cuevas, A.~Martin-Rodero, and
612: A.~L.~Yeyati, Phys.\ Rev.\ B \ {\bf 54}, 7366 (1996).
613:
614: \bibitem{Scheer}
615: E.~Scheer, {\em et al.}, Phys.\ Rev.\ Lett. {\bf 86}, 284 (2001).
616:
617: \bibitem{Jan2000}
618: B.~Ludoph {\em et al.}, Phys.\ Rev.\ B {\bf 61}, 8561 (2000).
619:
620: \bibitem{Cristiano}
621: R.~Cristiano {\it et al}, Phys.\ Rev.\ B {\bf 49}, 429 (1994).
622:
623: \bibitem{Likharev}
624: Y.~Naveh {\it et al}, Phys.\ Rev.\ Lett.\ {\bf 85}, 5404 (2000).
625:
626: \bibitem{Jos}
627: A similar scaling factor appears in the theory of high-order
628: corrections to the Josephson current: M.~Yu.~Kupriyanov, JETP
629: Lett.\ {\bf 56}, 399 (1992); E.~V.~Bezuglyi, E.~N.~Bratus', and
630: V.~P.~Galaiko, Low Temp.\ Phys.\ {\bf 25}, 167 (1999);
631: A.V.~Galaktionov and Chang-Mo Ryu, J.\ Phys.:\ Condens.\ Matter
632: {\bf 12}, 1351 (2000).
633:
634: \bibitem{Averin97}
635: A.~Bardas and D.~V.~Averin, Phys.\ Rev.\ B {\bf 56}, R8518 (1997).
636:
637: \bibitem{Beenakker}
638: C.~W.~J. Beenakker, B.~Rejaei, and J.~A.~Melsen, Phys.\ Rev.\
639: Lett.\ {\bf 72}, 2470 (1994).
640:
641: \bibitem{note}
642: Additional factor $\xi_0/d$ in \Eq{Deff} reflects the
643: ``anti-point-con\-tact'' geometry of the junctions with
644: overlapping leads and the junction cross section exceeding the one
645: of the leads (see Ref.~\onlinecite{VolkovSIN}).
646:
647: \bibitem{LOnoneq} A.~I.~Larkin and Yu.~N.~Ovchinnikov, Sov.\ Phys.\
648: JETP {\bf 41}, 960 (1975); {\bf 46}, 155 (1977).
649:
650: \bibitem{Artemenko}
651: S.~N.~Artemenko, A.~F.~Volkov, and A.~V.~Zaitsev, Sov.\ Phys.\
652: JETP {\bf 49}, 924 (1979).
653:
654: \bibitem{KL}
655: M.~Yu.~Kupriyanov and V.~F.~Lukichev, Sov.\ Phys.\ JETP {\bf 67},
656: 1163 (1988).
657:
658: \bibitem{Bezugly}
659: E.~V.~Bezuglyi {\it et al}, Phys.\ Rev.\ B {\bf 62}, 14439 (2000).
660:
661: \bibitem{Werthamer}
662: N.~R.~Werthamer, Phys.\ Rev.\ {\bf 147}, 255 (1966).
663:
664: \bibitem{VolkovSIN}
665: A.~F.~Volkov,\ Physica B {\bf 203}, 267 (1994).
666:
667: \end{thebibliography}
668:
669: \end{document}
670: