1: \documentclass[aps, preprint,amsfonts, amssymb,groupedaddress, floatfix]{revtex4}
2: %\documentclass[12pt]{article}
3: \usepackage[english]{babel}
4: %\usepackage{fullpage}
5: \usepackage{epsfig}
6: \usepackage{amsfonts, amssymb, amsmath}
7: \usepackage{times, mathptm}
8: \usepackage{psfrag}
9: \binoppenalty=10000 \relpenalty=10000
10: \newcommand{\Tr}{\mathop{\rm Tr}}
11: \newcommand{\sign}{\mathop{\rm sign}}
12: \newcommand{\Erf}{\mathop{\rm Erf}}
13: \newcommand{\arcsh}{\mathop{\rm arcsh}}
14: \begin{document}
15:
16:
17: \title{Coherent transport in Josephson-junction rhombi chain chain with quenched disorder}
18:
19: \author{Ivan V. Protopopov and Mikhail V. Feigel'man}
20:
21: \affiliation{L.D.Landau Institute for Theoretical Physics,Kosygin
22: str.2, Moscow, 119334, Russia}
23:
24:
25: \begin{abstract}
26: We consider a chain of Josephson-junction rhombi (proposed
27: originally in~\cite{Doucot}) in quantum regime. In a regular chain with no disorder in the maximally frustrated case
28: when magnetic flux through each rhombi $\Phi_r$ is equal to one half of superconductive flux quantum $\Phi_0$,
29: Josephson current is due to correlated transport of {\em pairs of Cooper pairs}, i.e. charge is quantized in units of $4e$.
30: Sufficiently strong deviation $\delta\Phi \equiv |\Phi_r-\Phi_0/2| > \delta\Phi^c$ from the
31: maximally frustrated point brings the system back to usual $2e$-quantized supercurrent.
32: For a regular chain $\delta\Phi^c$ was calculated in~\cite{my}. Here we present detailed analysis of the effect
33: of quenched disorder (random stray charges and random fluxes piercing rhombi) on the pairing effect.
34: \end{abstract}
35:
36: \maketitle
37: \section{Introduction}
38: Pairing of Cooper pairs in frustrated Josephson junction arrays
39: was theoretically proposed
40: recently~\cite{Doucot,IoffeFeigelman02,Doucot03,Doucot04} in the
41: search of topologically protected nontrivial quantum liquid
42: states. The simplest system where such a phenomenon could be
43: observed was proposed by Dou\c{c}ot and Vidal in~\cite{Doucot}. It
44: consists of a chain of rhombi (each of them being small ring of 4
45: superconductive islands connected by 4 Josephson junctions) placed
46: into transverse magnetic field. cf. Fig. ~1. It was shown
47: in~\cite{Doucot} that in the fully frustrated case (i.e. magnetic
48: flux through each rhombus $\Phi_r = \frac12\Phi_0 = \frac{hc}{4e}
49: $) usual tunnelling of Cooper pairs along the chain is blocked due
50: to destructive interference of tunneling going through two paths
51: within the same rhombus, while correlated 2-Cooper-pair transport
52: survives. Sufficiently strong deviation $ \delta\Phi \equiv
53: |\Phi_r-\Phi_0/2| > \delta\Phi^c$ from the maximally frustrated
54: point brings the system back to usual $2e$-quantized supercurrent.
55:
56: In ref. \cite{my} rhombi chain was studied for the experimentally
57: relevant situation when its Coulomb energy is determined by
58: capacitance of junctions and not by capacitance of superconductive
59: islands themselves. Expression for the critical deflection
60: $\delta\Phi^c$ was derived. However, ref. \cite{my} dealt with
61: regular chain with no disorder. In any real system two intrinsic
62: sources of disorder are always present: a) some weak randomness of
63: fluxes $\Phi_r^n$ penetrating different rhombi (due to unavoidable
64: differences in their areas), and b) random stray charges $q_n$
65: which produce, due to Aharonov-Casher effect, some random phase
66: factors to the phase slip tunnelling amplitudes leading to
67: suppression of quantum fluctuations in the chain (compare
68: analogous discussion in \cite{Larkin}).
69:
70: In this paper we adapt method used in ref. \cite{my} for the case
71: of rhombi chain with quenched disorder and study influence of
72: random stray charges and random fluxes in rhombi on the crossover
73: point $\delta\Phi=\delta\Phi^c$ between $4e$- and $2e$-regimes. It
74: turns out that in a long chain with large $E_J/E_C$ the pairing
75: effect is rather stable under influence of disorder.
76:
77: The rest of the paper is organized as follows: in Sec. II we
78: derive effective Schr\"{o}dinger equation describing rhombi chain with quenched random stray charges or random fluxes
79: piercing rhombi; in Sec. III we discuss influence of quenched stray charges on pairing effect, derive "phase diagram" of
80: the chain with fixed realization of disorder, calculate probability ${\cal P}_{4e}(E_J/E_C, N, \delta\Phi)$ to
81: find the chain in the regime with dominating $4e$-supercurrent and finally estimate critical deflection from the
82: maximally frustrated point $\delta\Phi^c$ destroying $4e$-supercurrent in a chain with stray charges; in Sec. IV
83: we consider modulation of the supercurrent in a clean and disordered chain by external capacitively coupled gates;
84: in Sec. V. we analyse influence of randomness of fluxes in rhombi on pairing of Cooper pairs. Finally, in Sec. VI.
85: we present our conclusions and suggestions.
86:
87: \section{Effective Hamiltonian in presence of disorder}
88: We study a chain of $N$ rhombi shown in Fig.~\ref{chain}. Each
89: rhombi consists of four superconductive islands connected by
90: tunnel junctions with Josephson coupling energy $E_J=\hbar
91: I_c^0/2e$; charging energy $E_C$ is determined by capacitance $C$
92: of junctions, $E_C = e^2/2C$ (we neglect self-capacitances of
93: islands which are assumed to be much smaller than $C$). Below we
94: consider Josephson current along the chain of $N \gg 1$ rhombi and
95: assume that the chain is of the ring shape, with total magnetic
96: flux $\Phi_c$ inside the ring. We also denote by $\Phi_r^n$ the
97: flux through $n$-th rhombus and define phases $\gamma$ and
98: $\varphi_n$:
99: \begin{equation}
100: \gamma=2\pi \frac{\Phi_c}{\Phi_0}\,, \qquad
101: \varphi_n=2\pi\frac{\Phi_r^n}{\Phi_0}\,,
102: \end{equation}
103: \begin{widetext}
104: \begin{figure}
105: \centering
106: \psfrag{Phir}[c][c]{\small $\Phi_r$}
107: \psfrag{Phic}[c][c]{\small $\Phi_c$}
108: \psfrag{N}[c][c]{\small $N$}
109: \psfrag{N1}[c][c]{\small $N$}
110: \psfrag{VG}[c][c]{\small $V_g$}
111: \psfrag{CG}[c][c]{\small $C_g$}
112: \psfrag{theta1N}[c][c]{\small $\theta_N^{(1)}$}
113: \psfrag{theta2N}[cl][c]{\small $\theta_N^{(2)}$}
114: \psfrag{theta3N}[c][c]{\small $\theta_N^{(3)}$}
115: \psfrag{theta4N}[c][c]{\small $\theta_N^{(4)}$}
116: \psfrag{theta11}[bl][c]{\small $\theta_1^{(1)}$}
117: \psfrag{theta21}[c][c]{\small $\theta_1^{(2)}$}
118: \psfrag{theta31}[c][c]{\small $\theta_1^{(3)}$}
119: \psfrag{theta41}[c][c]{\small $\theta_1^{(4)}$}
120: \psfrag{theta12}[c][c]{\small $\theta_2^{(1)}$}
121: \psfrag{theta22}[c][c]{\small $\theta_2^{(2)}$}
122: \psfrag{theta32}[c][c]{\small $\theta_2^{(3)}$}
123: \psfrag{theta42}[c][c]{\small $\theta_2^{(4)}$}
124: \psfrag{n1}[c][c]{\small $1$}
125: \psfrag{n2}[c][c]{\small $2$}
126: \psfrag{q11}[c][c]{\small $q_1^1$}
127: \psfrag{q12}[c][c]{\small $q_1^2$}
128: \psfrag{q13}[c][c]{\small $q_1^3$}
129: \psfrag{q21}[c][c]{\small $q_2^1$}
130: \psfrag{q22}[c][c]{\small \vspace{0.2cm} $q_2^2$}
131: \psfrag{q23}[c][c]{\small $q_2^3$}
132: \psfrag{qN2}[c][c]{\small $q_N^2$}
133: \psfrag{qN3}[c][c]{\small $q_N^3$}
134: \psfrag{qN1}[c][c]{\small $q_N^1$}
135: \includegraphics[width=400pt]{fig3.eps}
136: \caption{\small The chain of rhombi with random stray charges. Also shown is an external capacitively
137: coupled gate (c.f. Sec. IV).}
138: \label{chain}
139: \end{figure}
140: \end{widetext}
141: A regular chain with no disorder (no random stray charges and all
142: $\varphi_n\equiv \varphi$) is described by the imaginary-time
143: action
144: \begin{equation} S_E=\int
145: dt\sum_{n=1}^N\sum_{m=1}^4\left\{\frac{1}{16E_C}\left(\frac{d\theta^{(m)}_n}{dt}
146: \right)^2-E_J\cos\theta^{(m)}_n\right\}. \label{old_basic_action}
147: \end{equation}
148: and additional conditions
149: \begin{equation}
150: \sum_{m=1}^4\theta^{(m)}_n=\varphi\,, \qquad n=1,2,..,N\,,
151: \label{varphi}
152: \end{equation}
153: \begin{equation}
154: \sum_{n=1}^N\left(-\theta^{(3)}_n -
155: \theta^{(4)}_n\right)=\gamma\,. \label{gamma}
156: \end{equation}
157: Here the variable $\theta^{(m)}_n$ is the phase difference across
158: the $m$-th junction in the $n$-th rhombus (see Fig. \ref{chain}).
159:
160: This regular model was analyzed in details in ref. \cite{my}
161: within the limit $E_J\gg E_C$. At $E_C=0$ the phases
162: $\theta_n^{(m)}$ are classical variables which do not fluctuate.
163: Emerging classical states of the chain $\left|m,\{
164: \sigma_n^z\}\right>$ can be characterized by a set of binary
165: variables $\{\sigma_n^z\}$ (one for each rhombi) and an
166: integer-valued variable $m$. Note that each binary variable
167: $\sigma_n^z$ can be considered as $z$-projection of spin $\frac12$
168: ascribed to each rhombi. Energies of these classical states are
169: \begin{equation}
170: E_{m, \sigma}\approx \frac{E_J\sqrt2}{4N}(\widetilde{\gamma}-\pi
171: N/2-\pi S^z- 2\pi m)^2 -\sqrt{2}\delta S^z E_J+{\rm Const}.
172: \label{old E_under_flux_dif_from_pi}
173: \end{equation}
174: \begin{equation}
175: s_n^z=\frac{1}{2}\sigma_n^z\,, \qquad
176: S^z=\frac{1}{2}\sum_{n=1}^N\sigma_n^z\,, \qquad
177: \widetilde{\gamma}=\gamma+\frac{N\varphi}{2}\,.
178: \end{equation}
179:
180: Finite $E_J/E_C$ gives rise to quantum phase slips (QPS) in the
181: chain, which mix different classical states leading to formation
182: of truly quantum ground state. Let us denote as $\upsilon$ the
183: amplitude of a QPS in one contact. At large $N \gg 1$ this
184: amplitude does not differ from the "spin flip" amplitude for a
185: single rhombus at $\Phi_r \approx \Phi_0/2$. In this approximation
186: we can use result from Ref.~\cite{IoffeFeigelman02}:
187: \begin{equation}
188: \upsilon\approx k
189: \left(E_J^3E_C\right)^{1/4}\exp\left(-1.61\sqrt{\frac{E_J}{E_C}}\right).
190: \label{upsilon_gen}
191: \end{equation}
192: Here $k$ is a numerical factor of order one which slightly depends on $E_J/E_C$ (see ref. \cite{my} for details).
193:
194: In \cite{my} it was shown that calculation of the persistent
195: current in the large-$N$ limit can be reduced to the solution of a
196: Schr\"{o}dinger equation for a particle having a large spin
197: $S=\frac12 N$ moving in a periodic $\cos$-like potential, with
198: appropriate boundary condition.
199:
200: More precisely, the quantum ground-state energy $E$ of the chain
201: in the limit $E_J\gg E_C$ can be obtained from solution of
202: Schr\"{o}dinger equation
203: \begin{equation}
204: \frac{\partial^2\psi}{\partial x^2} +(\widetilde{E}- 2 w \cos
205: 2x\cdot
206: \sum_{n=1}^N\widehat{S}_n^x+2h\sum_{n=1}^{N}\widehat{S}_n^z)
207: \psi=0\,, \label{old_main_res_under_flux_dif_from_pi}
208: \end{equation}
209: where
210: \begin{equation}
211: \widetilde{E}=\frac{16NE}{\sqrt2 E_J \pi^2}\,, \qquad
212: w=\frac{64N\upsilon}{\sqrt2 E_J\pi^2}\,,\qquad
213: h=\frac{8N\delta}{\pi^2}\,. \label{bwh}
214: \end{equation}
215: Here $\upsilon$ --- amplitude for quantum phase slip in one rhombi
216: and $\delta=\pi-\varphi$. Function $\psi\equiv\psi(x,
217: \{\sigma_n\})$. $\widehat{S}_n^x$ and $\widehat{S}_n^z$ are
218: standard operators of $x$ and $z$ projections of spin $\frac 12$
219: acting on $\sigma_n$. The magnetic flux inside the whole ring
220: enters the problem through the twisted boundary condition
221: \begin{equation}
222: e^{i\pi\widehat{S}^z+i\pi N/2}\psi\left(x+\pi/2,\sigma\right)
223: =e^{i\widetilde{\gamma}}\psi(x,\sigma)\,, \qquad
224: \label{old_boundary} \widetilde{\gamma}=\gamma+\frac{N\varphi}{2}
225: \end{equation}
226: In \cite{my} on the basis of this formulation of the problem
227: critical deflection $\delta\Phi^c$ was derived
228: \begin{equation}
229: \left(\delta\Phi^c\right)_{reg}\approx
230: 0.2\left(\frac{\upsilon}{E_J}\right)^{2/3}\Phi_0\label{phi_c_reg}
231: \end{equation}
232:
233: The aim of this section is to derive effective Hamiltonian similar
234: to (\ref{old_main_res_under_flux_dif_from_pi}) for the chain with
235: random fluxes $\Phi_r^n$ through rhombi or random stray charges.
236:
237: To account for random stray charges we present the electrostatic
238: charging energy in the form
239: \begin{equation}
240: H_C=\sum_{n_1\,,\,n_2=1}^N\,\,\,\sum_{k_1\,,\,k_2=1}^3\frac12
241: \left[C^{-1}\right]^{k_1\,n_1}_{k_2\,n_2}
242: \left(Q_{n_1}^{(k_1)}-q_{n_1}^{(k_1)}\right)\left(Q_{n_2}^{(k_2)}-q_{n_2}^{(k_2)}\right)
243: \end{equation}
244: where $\left[C^{-1}\right]^{k_1\,n_1}_{k_2\,n_2}$ is the matrix of
245: inverse capacitances. Indices $n$ and $k$ numerates
246: superconducting islands ($3$ rows, $N$ islands in a row).
247: $Q_{n}^{(k)}$ --- charge of the $n$-th island in $k$-th row.
248: Parameters $q_{n}^{(k)}$ are determined by the random stray
249: charges.
250:
251: Starting from this charging energy we may derive an additional
252: term in action (\ref{old_basic_action}) emerging from presence of
253: random stray charges (compare with \cite{Larkin})
254: \begin{equation}
255: \delta S=-i\int
256: dt\sum_{n=1}^N\sum_{m=1}^4p_n^m\frac{d\theta_n^{(m)}}{dt}
257: \label{addition}
258: \end{equation}
259: Parameters $p_n^m$ can be expressed in terms of charges
260: $q^{(k)}_n$. Corresponding expressions are a bit cumbersome and we
261: do not present them here. Below we write down some special
262: combinations of $p_n^m$ which we will need in our paper.
263:
264: Additional term (\ref{addition}) has a form of total derivative.
265: Hence it does not change neither the classical states of the chain
266: nor the classical equations of motion and the real part of
267: classical action on a {\it single} tunneling trajectory. The only
268: effect of this term to give a tunneling amplitude along each path
269: its on phase factor. Since there are several QPS trajectories
270: between two classical states, all having the same real part of
271: tunneling action, these additional phase factors may give rise to
272: destructive interference of tunneling processes, leading to
273: reduction of total matrix element, connecting two classical
274: states.
275:
276: Following ref. \cite{my} and taking into account complications due
277: to random stray charges we write tight-binding Hamiltonian
278:
279: \begin{multline}
280: \widehat{H}|m, \sigma>= E_{m\sigma}|m,\sigma> + \frac{\upsilon}{2}
281: \sum_{n=1}^N\exp\left\{\frac{i\pi}{2}\left(3p_n^1-p_n^2-p_n^3-p_n^4
282: \right)\right\} \widehat{\sigma}_n^+|m,\sigma> +\\
283: \frac{\upsilon}{2}
284: \sum_{n=1}^N\exp\left\{\frac{i\pi}{2}\left(3p_n^2-p_n^1-p_n^3-p_n^4
285: \right)\right\} \widehat{\sigma}_n^+|m,\sigma> + \\
286: \frac{\upsilon}{2}
287: \sum_{n=1}^N\exp\left\{-\frac{i\pi}{2}\left(3p_n^3-p_n^1-p_n^2-p_n^4
288: \right)\right\} \widehat{\sigma}_n^+|m-1,\sigma> +\\
289: \frac{\upsilon}{2}
290: \sum_{n=1}^N\exp\left\{-\frac{i\pi}{2}\left(3p_n^4-p_n^1-p_n^2-p_n^3
291: \right)\right\} \widehat{\sigma}_n^+|m-1,\sigma> +h.\,c.
292: \label{H_acting_over_m_s}
293: \end{multline}
294: Performing Fourier transformation over variable $m$ according to
295: \begin{equation}
296: \left|x,\sigma\right>
297: =\sum_m\exp\left\{2i\left(2m-\frac{\widetilde{\gamma}}{\pi}+S^z+\frac{N}{2}\right)x\right\}
298: \left|m,\sigma\right>\,,\label{Fur'e}
299: \end{equation}
300: we obtain the effective Schr\"{o}dinger equation
301: \begin{equation}
302: \frac{\partial^2\psi}{\partial x^2} +\left(\widetilde{E}- 2 w \cos
303: 2x \sum_{n=1}^{N}a_n\widehat{S}_n^x - 2 w \sin 2x
304: \sum_{n=1}^{N}b_n\widehat{S}_n^y +2h \widehat{S}^z\right)\psi=0\,,
305: \label{main_res_random_charge}
306: \end{equation}
307: and twisted boundary condition
308: \begin{equation}
309: e^{i\pi\widehat{S}^z+i\pi N/2}\psi\left(x+\pi/2,\sigma\right)
310: =e^{i\widetilde{\gamma}}\psi(x,\sigma)\,. \label{boundary}
311: \end{equation}
312: Here parameters $\widetilde{E}$, $w$ and $h$ are described by
313: equation (\ref{bwh}); $a_n$ and $b_n$ are random coefficients
314: which can be expressed in term of random charges $q_n^{(k)}$
315: \begin{equation}
316: a_n=\frac{\cos\pi Q_n^1+\cos\pi Q_n^2}{2}\,, \qquad
317: b_n=\frac{\cos\pi Q_n^1-\cos\pi Q_n^2}{2}
318: \end{equation}
319: \begin{equation}
320: Q_n^1\equiv
321: p_n^1-p_n^2=q_n^{(2)}-\frac{1}{3N}\sum_{n=1}^N\sum_{k=1}^3q_n^{(k)}
322: \qquad Q_n^2\equiv
323: p_n^3-p_n^4=q_n^{(3)}-\frac{1}{3N}\sum_{n=1}^N\sum_{k=1}^3q_n^{(k)}
324: \label{Q}
325: \end{equation}
326: Here we measure charges $q_n^{(k)}$ in units $2e$.
327:
328: We turn now to generalization of (\ref{old_main_res_under_flux_dif_from_pi}) for the case of random
329: fluxes in rhombi. Imaginary-time action for this
330: problem is given by equation (\ref{old_basic_action}) but
331: additional conditions (\ref{varphi}) should be changed to
332: \begin{equation}
333: \sum_{m=1}^4\theta^{(m)}_n=\varphi_n\,, \qquad n=1,2,..,N\,,
334: \label{new_varphi}
335: \end{equation}
336: Following the same steps as before we see that to
337: account for disorder in flux piercing the rhombi one should just
338: replace the last term in
339: (\ref{old_main_res_under_flux_dif_from_pi}) by
340: \begin{equation}
341: 2\sum_{n=1}^{N}h_n\widehat{S}_n^z
342: \end{equation}
343: Here $h_n=8N(\pi-\varphi_n)/\pi^2$.
344:
345: Now we have effective Hamiltonians of the chain in presence of
346: disorder. In the rest of the paper we will analyse these
347: Hamiltonians in order to find out influence of disorder on the
348: crossover point between $4e$- and $2e$-regimes.
349:
350:
351:
352: \section{Influence of random charges on crossover point}
353: In this section we study influence of random stray charges on the
354: on pairing effect in rhombi chain. It is important to note that,
355: generally speaking, even at maximally frustrated point $h=0$ (in
356: contrast to regular chain) symmetry properties of Schr\"{o}dinger
357: equation (\ref{main_res_random_charge}) {\it do not} prohibit
358: $2e$-supercurrent. This is due to the fact that asymmetric
359: realizations of random charges with $q_n^{(2)}\neq q_n^{(3)}$ (and
360: thus $b_n\neq0$) break symmetry between two tunneling trajectories
361: of a Cooper pair within the same rhombus. Of course, random
362: charges preserve classical states of the rhombi chain and in a chain
363: with no QPS there would be no $2e$-supercurrent at maximally
364: frustrated point, but full quantum Hamiltonian of the chain does
365: not possess corresponding symmetry. Nevertheless, we will see
366: later that for typical realizations of random charges and at $\Phi_r=\Phi_0/2$
367: the $2e$-supercurrent is small as compared to
368: $4e$-supercurrent. The reason for that is as follows: if at least
369: in one rhombi $b_n=0$ ( or $a_n=0$) then $2e$-supercurrent is
370: prohibited. Here we start with analysis of general situation of
371: rhombi chain with random charges and $h\neq0$ and then make some
372: conclusion on disordered rhombi chain in maximally frustrated
373: point.
374:
375: We investigate the grand partition function for the system
376: described by equations (\ref{main_res_random_charge},
377: \ref{boundary})
378: \begin{equation}
379: Z=\int{\cal
380: D}x(\tau)\exp\left(-\int_0^\beta\frac{\dot{x}^2}{2}\right)\prod_{n=1}^N\Tr
381: \left[\widehat{U}_n(\beta)\right]\label{Z}
382: \end{equation}
383: with $\beta$ being the inverse temperature, $\beta\rightarrow
384: \infty$. Operators $\widehat{U}_n$ act in the space of spin
385: $\frac 12$ each and are functionals of $x(\tau)$ defined as
386: \begin{equation}
387: \frac{d \widehat{U}_n}{d\tau}=-\left(w f_n(\tau)
388: S^x+wg_n(\tau)\widehat{S}^y-h\widehat{S}^z\right)\widehat{U}_n
389: \label{defU}
390: \end{equation}
391: \begin{equation}
392: f_n(\tau)=a_n\cos 2x(\tau)\,, \qquad g_n(\tau)=b_n\sin 2x(\tau)
393: \end{equation}
394: As was shown in \cite{my}, in a regular chain with $E_J\gg E_C$
395: the borderline between $4e$- and $2e$-supercurrents is at rather
396: large $\Phi_r-\Phi_0/2$, in the sense that at the crossover point
397: $h\gg w$. So in this paper we will also consider this limit only.
398:
399: Under such condition equation (\ref{defU}) can be solved for
400: arbitrary functions $f(\tau)$ and $g(\tau)$. For
401: $\Tr\widehat{U}_n(\beta)$ we then find
402: \begin{multline}
403: \Tr \widehat{U_n}(\beta)=\exp\left(\frac{\beta
404: h}2+\frac{w^2}{4h^2}\left(f_n(0)-ig_n(0)\right)\left(f_n(\beta)+ig_n(\beta)\right)+\right.\\
405: \left. \frac{w^2}{8}\int_0^\beta d\tau_1 d\tau_2 \left(
406: f_n(\tau_1)f_n(\tau_2)+g_n(\tau_1)g_n(\tau_2)\right)e^{-h|\tau_1-\tau_2|}+
407: \right.
408: \\ \left.
409: i\frac{w^2}{4}\int_0^\beta d
410: \tau_1d\tau_2f_n(\tau_1)g_n(\tau_2)e^{-h|\tau_1-\tau_2|}\sign(\tau_1-\tau_2)\right)
411: \label{TRU}
412: \end{multline}
413: From equation (\ref{TRU}) we derive effective action for variable
414: $x$
415: \begin{equation}
416: Z=\int_0^\beta {\cal D}x(\tau) e^{-S[x(\tau)]}\qquad
417: S=S_{bound}+S_\tau
418: \end{equation}
419: \begin{equation}
420: S_{bound}=-\frac{w^2}{4h^2}\sum_{n=1}^N\left(a_n\cos
421: 2x(\beta)+ib_n\sin 2x(\beta)\right) \left(a_n\cos
422: 2x(0)-ib_n\sin2x(0)\right)
423: \end{equation}
424: \begin{multline}
425: S_\tau=\int_0^\beta d\tau\frac{\dot{x}^2(\tau)}{2}-
426: A\frac{w^2}{8}\int_0^\beta d\tau_1d\tau_2
427: \cos2x(\tau_1)\cos2x(\tau_2)\exp(|\tau_1-\tau_2|)-\\
428: B\frac{w^2}{8}\int_0^\beta
429: d\tau_1d\tau_2\sin2x(\tau_1)\sin2x(\tau_2)\exp(|\tau_1-\tau_2|)-
430: \\ iC\frac{w^2}{4h}\int_0^\beta d \tau_1d\tau_2\cos 2x(\tau_1)\sin2x(\tau_2)\exp(|\tau_1-\tau_2|)\sign(\tau_1-\tau_2)
431: \label{action_x}
432: \end{multline}
433: Where
434: \begin{equation}
435: A=\sum_{n=1}^N a_n^2\,, \qquad B=\sum_{n=1}^N b_n^2\,, \qquad
436: C=\sum_{n=1}^N a_n b_n \label{ABC}
437: \end{equation}
438:
439: Note that similar approximation was used previously in \cite{my} for a regular chain.
440: In a regular chain semiclassical analysis which do not rely on linearization with respect to $w/h$ is also
441: possible. It turns out that exact value of the critical deflection $\delta\Phi^c$ differs only by $12\%$ from the one
442: obtained by linearization even for the case when at the crossover point $w/h=1$.
443:
444: Action (\ref{action_x}) is nonlocal and looks a bit terrific, but
445: since
446: \begin{equation}
447: \left( \frac{\partial^2}{\partial
448: \tau_1^2}-h^2\right)e^{-h|\tau_1-\tau_2|}=-2h\delta(\tau_1-\tau_2)
449: \end{equation}
450: we can perform Hubbard-Stratonovich transformation and derive a
451: representation for grand partition function with local action
452: (after redefinition of time scale according to
453: $\tau\longrightarrow t/h$).
454: \begin{equation}
455: Z=\int_0^\beta {\cal D}x(\tau){\cal D}y(\tau){\cal D}z(\tau)
456: e^{-S[x(\tau)]}\qquad
457: \end{equation}
458: \begin{multline}
459: S=h\int_0^{\beta h}
460: d\tau\left(\frac{\dot{x}^2+\dot{y}^2+\dot{z}^2}2+\frac{y^2+z^2}{2}+
461: \alpha_1dy\cos2x+i\beta_1d\dot{y}\sin2x+\right.\\
462: \left. \alpha_2dz\sin2x+i\beta_2d\dot{z}\cos2x
463: -\frac{d^2}{2}(\beta_1^2\sin^2 2x+\beta_2^2 \cos^2 2x) \right)
464: \label{decoupling}
465: \end{multline}
466: Here $\alpha_1$, $\alpha_2$, $\beta_1$ and $\beta_2$ should
467: satisfy equations
468: \begin{equation}
469: \alpha_1^2+\beta_2^2=\frac{A}{N}\,, \qquad
470: \beta_1^2+\alpha_2^2=\frac BN\,, \qquad
471: \alpha_1\beta_1-\alpha_2\beta_2=\frac C N \label{abab}
472: \end{equation}
473: and
474: \begin{equation}
475: d=\sqrt{\frac{w^2N}{2h^3}}=\frac{\sqrt{2}\pi}{\delta^{3/2}}\frac{\upsilon}{E_J}\label{d}
476: \end{equation}
477:
478: As we see from equations (\ref{abab}) we have some freedom in
479: definitions of these parameters. Namely, if we set
480: \begin{equation}
481: \alpha_1=\sqrt{\frac{A}{N}}\sin\kappa_1\,, \qquad
482: \beta_2=\sqrt{\frac{A}{N}}\cos\kappa_1
483: \end{equation}
484: \begin{equation}
485: \alpha_2=\sqrt{\frac{B}{N}}\sin\kappa_2\,, \qquad
486: \beta_1=\sqrt{\frac{B}{N}}\cos\kappa_2
487: \end{equation}
488: then (\ref{abab}) reduces to
489: \begin{equation}
490: \sin(\kappa_1-\kappa_2)=\frac{C}{\sqrt{AB}}
491: \end{equation}
492: and we have only one equation for two parameters $\kappa_1$ and
493: $\kappa_2$. This freedom will not be important for us. In present
494: paper for the reasons described below we will mostly consider
495: action (\ref{decoupling}) under condition $C=0$ and chose
496: \begin{equation}
497: \alpha_1=\sqrt{\frac{A}{N}}\,, \qquad
498: \alpha_1=\sqrt{\frac{A}{N}}\,, \qquad \beta_1=0\,, \beta_2=0
499: \end{equation}
500:
501: The main feature of the result presented above is that we
502: have reduced original problem including large number of random
503: variables ($\sim N$) to a problem with only {\it three } random
504: parameters $A$, $B$, and $C$. More over, since $A$, $B$ and $C$
505: arise as sums of large number $N$ of independent random variables
506: (see equation (\ref{ABC})), it is natural to expect that they obey
507: Gaussian statistics. Let us assume that parameters $Q_n^1$ and
508: $Q_n^2$ in (\ref{Q}) are uniformly distributed on $[-1, 1]$. It
509: corresponds to strong fluctuations of stray charges from sample to
510: sample. Under such an assumption one can easily find
511: \begin{eqnarray}
512: \left<A\right>=\left<B\right>=\frac{N}{4}\nonumber\\
513: \left<A^2\right>=\left<B^2\right>=\frac{N^2}{16}+\frac{5N}{64}\,,
514: \qquad \left<AB\right>=\frac{N^2}{16}-\frac{3N}{64}
515: \label{stat_ABC}\\
516: \left<C\right>=0\,, \qquad \left<C^2\right>=\frac{N}{64}\nonumber
517: \end{eqnarray}
518:
519: Our strategy in the rest part of this section will be to analyse
520: properties of the system described by equations (\ref{decoupling},
521: \ref{boundary}) with {\it fixed } $A$, $B$ and $C$ and then make
522: some statistical analysis. Under fixed $A$, $B$ and $C$ the
523: main subject of our investigation will be whether the system is in
524: regime of dominating $4e$-supercurrent or not.
525:
526:
527: First of all let us analyze action (\ref{decoupling}) for a
528: trajectory where $x(\tau)$, $y(\tau)$ and $z(\tau$) are constant.
529: We have
530: \begin{equation}
531: S_{st}=\beta h\left(\frac{y^2+z^2}{2}+ \alpha_1dy\cos2x+
532: \alpha_2dz\sin2x-\frac{d^2}{2}(\beta_1^2\sin^2 2x+\beta_2^2 \cos^2
533: 2x) \right)
534: \end{equation}
535:
536: This static action has two groups of minima (we call them even and
537: odd, suppose $A>B$)
538: \begin{equation}
539: x=\pi n\,, \qquad y=-\alpha_1 d \,, \qquad z=0
540: \end{equation}
541: and
542: \begin{equation}
543: x=\frac{\pi}2+\pi n\,, \qquad y=\alpha_1 d \,, \qquad z=0
544: \end{equation}
545: where $n$ is an arbitrary integer. All these minima correspond to
546: the same value of $S_{st}$. So we have to consider two types of
547: tunnelling trajectories. Trajectories of the first type connect
548: minima of the same group, i.e. "even-even" and "odd-odd", and
549: corresponding variation of the variable $x$ between minima is $\pm
550: \pi$, whereas $y$ returns to its original value. Trajectories of
551: the second type connect minima of opposite parity (i.e. opposite
552: signs of $y$), and change $x$ variable by $\pm \frac{\pi}2 $. It
553: is not difficult to see from
554: Eqs.(\ref{Fur'e},\ref{main_res_random_charge},\ref{boundary} ),
555: that increment $\Delta x$ of the variable $x$ along tunnelling trajectory is in one-to-one
556: correspondence to the elementary charge transported along the
557: rhombi chain: $q_0 = \frac{4e}{\pi}\Delta x$. Therefore
558: trajectories of the first type lead to $4e$ - supercurrent,
559: whereas trajectories of the second type produces usual $2e$-quantized
560: supercurrent. In semiclassical approximation amplitudes of the
561: supercurrent components are determined primarily by the classical
562: actions on corresponding trajectories:
563: \begin{equation}
564: I(\gamma) = I_{2e}\sin\widetilde{\gamma} +
565: I_{4e}\sin(2\widetilde{\gamma}), \label{gen_cur}
566: \end{equation} where
567: \begin{equation}
568: I_{4e} = A_{4e} \exp(-S_E^{4e}), \quad I_{2e} = A_{2e}
569: \exp(-S_E^{2e}) \, , \label{components}
570: \end{equation}
571:
572: To find out whether $4e$-supercurrent dominates in the system, we
573: should compare classical actions on the trajectories of two types
574: described above. So we examine classical equations of motion for
575: action (\ref{decoupling}). For simplicity here we put coefficient
576: $C$ to zero. It can be shown that relatively small $C$ of order
577: $\sqrt{N}$ (compare to (\ref{stat_ABC})) is not important for our
578: future purpose.
579: \begin{equation}
580: \ddot{x}+2\alpha_1 d y \sin2x-2\alpha_2 d z \cos 2x =0
581: \label{equation_x}
582: \end{equation}
583: \begin{equation}
584: \ddot{y}-y=\alpha_1 d \cos 2x \label{equation_y}
585: \end{equation}
586: \begin{equation}
587: \ddot{z}-z=\alpha_2 d \sin 2x \label{equation_z}
588: \end{equation}
589:
590: Here we will analyse equations (\ref{equation_x},
591: \ref{equation_y}, \ref{equation_z}) and find analytic expression
592: for the borderline between $2e$- and $4e$-regimes under conditions
593: \begin{equation}
594: Ad^2/N\gg1\,,\qquad A-B\ll A\label{condAB}
595: \end{equation}
596: We also present results of numerical computation of the borderline
597: in general situation.
598:
599:
600: Let us start with $2e$-trajectory. Note that for variables $x$ and
601: $y$ characteristic frequency is 1. Let us suppose that on
602: $2e$-trajectory $x$ varies slowly so that characteristic frequency
603: $\omega_x\ll 1$. Than we can eliminate $y$ and $z$ from equations
604: of motion in adiabatic approximation and obtain
605: \begin{equation}
606: \ddot{x}-d_1^2\sin4x=0 \label{slow_x}
607: \end{equation}
608: \begin{equation}
609: d_1^2=\frac{(A-B)d^2}{N+2d^2(A+B)}
610: \end{equation}
611: We see that under conditions (\ref{condAB}) $\omega_x\sim d_1$ is
612: indeed small. Equation (\ref{slow_x}) has solution corresponding
613: to $2e$-trajectory.
614: \begin{equation}
615: x(\tau)=\frac{1}{2}\arccos{\left(-\tanh\left(2d_1\tau\right)\right)}\,,
616: \label{slow_x_sol}
617: \end{equation}
618: Classical action on this solution is
619: \begin{equation}
620: S_{2e}=h\sqrt{\frac{d^2(A-B)}{N}\left(1+2\frac{(A+B)d^2}{N}\right)}
621: \label{two_e_action}
622: \end{equation}
623:
624: Let us now turn to examination of $4e$-trajectories. Here we
625: suppose that $x$ is a fast variable i.e. $\omega_x\gg 1$. We can
626: neglect then variation of $y$ and $z$ on classical trajectory and
627: put $y=-\alpha_1d$, $z=0$. This is consistent with boundary
628: conditions for $4e$-trajectory. Equation for $x$ becomes
629: \begin{equation}
630: \ddot{x}-\frac{2A}{N}\, d^2 \sin2x =0 \label{fast_x}
631: \end{equation}
632: Again, due to (\ref{condAB}), we see that $x$ indeed varies fast
633: on $4e$-trajectory since $\omega_x\sim \sqrt{Ad^2/N}\gg1$. So we
634: find $4e$-trajectory and corresponding action
635: \begin{equation}
636: x(\tau)=\arccos{\left(-\tanh\left(2\sqrt{\frac{Ad^2}{N}}\tau\right)\right)}\,,
637: \label{x_0_reg2}
638: \end{equation}
639: \begin{equation}
640: S_{4e}=4h\sqrt{\frac{Ad^2}{N}} \label{four_e_action}
641: \end{equation}
642:
643: Comparing equations (\ref{two_e_action}) and (\ref{four_e_action})
644: we find the line of crossover between $4e$- and $2e$- regimes:
645: the set of points $(A, B)$ such that $S_{4e}=S_{2e}$.
646: \begin{equation}
647: \frac{Bd^2}{N}=\sqrt{\left(\frac{Ad^2}{N}\right)^2-8\frac{Ad^2}{N}}\label{cross_line}
648: \end{equation}
649:
650: The result presented above was derived under assumption $C=0$ but
651: proceeding in the same way with $C\lesssim \sqrt{N}$ one can show
652: that at large $N$ nonzero $C$ does not affect the crossover line
653: (\ref{cross_line}). Therefore in the rest part of this paper we
654: will completely ignore coefficient $C$.
655:
656: The borderline obtained from numerical solution of classical
657: equations of motion (\ref{equation_x}, \ref{equation_y},
658: \ref{equation_z}) and its asymptotic form (\ref{cross_line}) is
659: presented on figure \ref{crossover_fig}.
660: \begin{figure}[t] \centering
661: \psfrag{comb1}[cr][c][1.15][-90]{\tiny $\frac{\delta\Phi}{(\delta\Phi^c)_{reg}}\widetilde{B}$}
662: \psfrag{comb2}[tl][c][1.15]{\tiny $\frac{\delta\Phi}{(\delta\Phi^c)_{reg}}\widetilde{A}$}
663: \psfrag{subplotb}[bc][c]{\small b)}
664: \psfrag{u}[c][c]{\small $u$}
665: \psfrag{v}[c][c]{\small $v$}
666: \psfrag{a}[bc][c]{\small a)}\hspace{-1cm}
667: \includegraphics[width=200pt]{cross_xy.eps}\hspace{2cm}
668: \includegraphics[width=170pt]{cross_ug.eps}
669: \caption{\small "Phase diagram" of disordered chain. Solid lines on the both figures mark the crossover
670: $2e$- and $4e$-regimes. Note that these lines does not correspond to any phase transition. The crossover however is sharp
671: at large $N$ since actions $S_{4e}$ and $S_{2e}$ are proportional to the number of rhombi. Subplot a) presents crossover
672: line in variables $u$ and $v$ defined by (\ref{def_uv}) which are useful for calculations. Subplot b) depicts the same
673: in a bit more intuitive way. Here we introduce factors
674: $\widetilde{A}=\left(\frac{NA}{A^2+B^2}\right)^{1/3}$ and $\widetilde{B}=\left(\frac{NB}{A^2+B^2}\right)^{1/3}$
675: describing realization of disorder; $\left(\delta\Phi^c\right)_{reg}$ --- critical deviation for a clean chain with same
676: $E_J/E_C$.}
677: \label{crossover_fig}
678: \end{figure}
679: Instead of $A$ and $B$ we have used here more convenient variables
680: \begin{equation}
681: u=d^2\frac{(A-B)}{N}\,, \qquad v=\frac{d^2(A+B)}{2N}
682: \label{def_uv}
683: \end{equation}
684: Note that by definition $v>|u|/2$. From
685: (\ref{cross_line}) we see that in terms of $u$ and $v$ at large $v$ the borderline between $2e$- and
686: $4e$-regimes is given by
687: \begin{equation}
688: u=f(v)\approx\frac{4v}{v-2}\rightarrow4\,, \qquad v\rightarrow \infty
689: \end{equation}
690: On figure \ref{crossover_fig}b $\,$ "phase diagram" of the disordered chain is presented in a more intuitive manner:
691: we emphasize here that $4e$-supercurrent exist in a small vicinity of the maximally frustrated point.
692:
693: Equipped with this result, for any {\it given} set of quenched
694: random charges (characterized by coefficients $A$ and $B$) we can
695: (in principle) determine whether $4e$-supercurrent dominates in
696: the chain. Of course experimentally we have no access to such
697: quantities as $A$ and $B$. Hence we need some statistical
698: description of rhombi chain with random charges. Such a
699: description is provided by the probability ${\cal P}(E_J/E_C, N,
700: \delta)$ to find dominating $4e$-supercurrent in the system.
701:
702:
703: Assuming Gaussian statistics for $u$ and $v$ and taking into
704: account (\ref{stat_ABC}) one can derive probability distribution
705: for $u$ and $v$
706: \begin{equation}
707: P(u, v)=\frac{8 N}{\pi d^4} \exp\left(-\frac{2 N
708: u^2}{d^4}\right)\exp\left(-\frac{32N(v-v_0)^2}{d^4}\right)
709: \label{puv}
710: \end{equation}
711: \begin{equation}
712: v_0=\frac{d^2}{4}
713: \end{equation}
714:
715:
716: Required probability can be written as
717: \begin{equation}
718: {\cal P}_{4e}(E_J/E_C, N, \delta)=1-2\int_0^{+\infty} dv
719: \int_0^{f(v)} du P(u, v)\label{prob1}
720: \end{equation}
721:
722: Maximum of probability distribution (\ref{puv}) lies at $u=0$,
723: $v=d^2/4$. This point for $d=9$ is marked on figure
724: \ref{crossover_fig} with a cross. At sufficiently large $d$ we can
725: replace $f(v)$ in (\ref{prob1}) by $4$. We then get
726: \begin{equation}
727: {\cal P}_{4e}(E_J/E_C, N,
728: \delta)=1-\frac{2}{\sqrt{\pi}}\int_0^{\sqrt{\frac{32N}{d^4}}} du\,
729: \exp\left(-u^2\right)=1-\Erf\left(\sqrt{\frac{32N}{d^4}}\right)
730: \label{P4e_approx}
731: \end{equation}
732:
733: Let us introduce parameter $\kappa$ and {\it define} critical
734: deviation from the maximally frustrated point $\delta \Phi_c$ as
735: deviation under which the probability (\ref{P4e_approx}) equals
736: $\kappa$. Reasonable choice for $\kappa$ is $0.5$ or $0.75$ or
737: something else. From (\ref{P4e_approx}) we then get central result
738: of this section
739: \begin{figure}
740: \psfrag{P}[c][c]{${\cal P}_{4e}$}
741: \psfrag{delta}[tc][c]{$\frac{\delta\Phi^c}{\Phi_0}\times 10^3$}
742: \psfrag{mya}[c][c]{a)}
743: \psfrag{myb}[c][c]{b)}
744: \psfrag{myc}[c][c]{c)}
745: \psfrag{myd}[c][c]{d)}
746:
747: \hspace{-1cm}\includegraphics[width=200pt]{P1.eps}\hspace{2cm}
748: \includegraphics[width=200pt]{P2.eps}
749:
750: \hspace{-1cm}\includegraphics[width=200pt]{P3.eps}\hspace{2cm}
751: \includegraphics[width=200pt]{P4.eps}
752: \caption{\small Probability ${\cal P}_{4e}$ as function of deviation from maximally frustrated point.
753: Subplots a), b), c), d),
754: correspond to $N=10,\, 20,\, 30,\, 70$.}
755: \label{p4e}
756: \end{figure}
757:
758: \begin{equation}
759: \frac{\delta
760: \Phi_c}{\Phi_0}=\frac{\left(\Erf^{-1}\left(1-\kappa\right)\right)^{1/6}}{2^{3/2}\pi^{1/3}}\frac{1}{N^{1/6}}
761: \left(\frac{\upsilon}{E_J}\right)^{2/3}\label{phi_c_random_charges}
762: \end{equation}
763: Comparing (\ref{phi_c_random_charges}) and (\ref{phi_c_reg}) we
764: see that for any experimentally reasonable $N$ critical deviation
765: from the maximally frustrated point does not differ from the one
766: in a regular chain.
767: However one should remember that theory
768: presented above has several limitations. To clarify this
769: question we present here several graphics for ${\cal
770: P}_{4e}(\delta\Phi, E_J/E_C, N)$ at different $E_J/E_C$ and $N$
771: together with the boundaries of the validity region of the
772: approximation used (fig. \ref{p4e} ). Subplots a), b), c), d)
773: correspond to different number of rhombi $N=10, 20, 30, 70$
774: respectively. Solid and dashed lines $1, 2, 3, 4$ correspond to
775: different $E_j/E_C=10, 12, 15, 20$. Dashed lines were obtained
776: with the aid of equation (\ref{P4e_approx}) relying on the
777: condition $d^2>1$ whereas solid lines were produced by exploiting
778: equation (\ref{prob1}) with $f(v)$ determined from numerical
779: calculations. The validity region should be determined as
780: follows. First of all, we have used assumption $h>>w$. Taking
781: $h=w$ as a criterium for the edge of our validity region one finds
782: the dash-doted curves on figure \ref{p4e}. Strictly speaking only
783: those parts of solids curves are valid, which lie under the
784: corresponding dash-doted curves. Now we have to ensure that rhombi
785: chain is in quantum regime, i.e. currents are exponentially small.
786: Simple estimates show that this is true above the doted line on
787: figure \ref{p4e}. At large $N$ this limitation is very soft.
788: Finally, our approximations are not valid at probabilities
789: ${\cal P}_{4e}$ very small or very close to unity since they rely on
790: gaussian statistics for quantities $A$ and $B$ which certainly
791: does not describe rare events.
792:
793: Qualitatively, results presented above can be interpreted as
794: follows: at large $E_J/E_C$ phase fluctuations in a single rhombi
795: are weak and this makes random charges inefficient, however in a
796: long chain global supercurrent is still exponentially suppressed.
797: From the analysis presented above one concludes that we can
798: guarantee with high probability existence of dominating,
799: suppressed by quantum fluctuations, $4e$-supercurrent in long
800: chain ($N\sim 20$) with large $E_J/E_C\sim 20$ and for such a
801: chain critical deflection $\delta\Phi^c/\Phi_0\sim 10^{-3}$. If we
802: compare this result with critical deflection
803: $(\delta\Phi^c)_{reg}/\Phi_0\approx 1.2 \times 10^{-3}$ for a
804: regular chain with the same $E_J/E_C$ we conclude that the
805: influence of disorder on pairing effect is really small. On the
806: other hand in a regular chain it was possible to obtain critical
807: deflection of order $10^{-2}$ by choosing $E_J/E_C\sim 8$.
808: For this set of parameters we cannot, use our theory
809: quantitatively; we expect however that qualitatively the same
810: behaviour as described above for larger values $N$ and $E_J/E_C$ should be observed
811: here as well.
812:
813:
814: \section{Modulation of the supercurrent by capacitive gate}
815: In this section we will discuss briefly influence of regular gates on pairing effect
816: in maximally frustrated rhombi chain. We suppose that we have a gate, which is capacitively connected to all
817: superconducting islands in one row (fig. \ref{chain}).
818:
819: Let us consider first rhombi chain with no disorder. Our system still can be described by equations
820: (\ref{main_res_random_charge}, \ref{boundary}), but now $a_n$ and $b_n$ are no more random. This coefficients
821: can be expressed in terms of the gate voltage according to
822: \begin{equation}
823: a_n=\frac{1+\cos\pi C_gV_g}{2}\equiv a\,, \qquad
824: b_n=\frac{1-\cos\pi C_gV_g}{2}\equiv b
825: \end{equation}
826: We will show here that the supercurrent in the chain is rather sensitive to the gate voltage.
827: Since the gate voltage enters the problem in a particularly quantum way (via phases of virtual QPS), this
828: dependence may provide an experimental test for the quantum nature of chain's state near the maximally frustrated point.
829:
830: The problem we are considering now is much simpler as compared to the problem of random stray charges discussed above.
831: Now our Hamiltonian commutes with the total spin of the chain and we can apply semiclassical approximation directly
832: (e.g. by introducing spin coherent state path integral, c.f. \cite{my, Klauder}). It's easy to see that the ground state
833: corresponds to the maximal total spin $S=N/2$. After proper redefinition of the time scale we can write imaginary-time
834: action in the form
835: \begin{equation}
836: S_E=\frac{N}{2}\int d\tau\left(-i\cos\theta\dot{\phi}
837: +\frac{w\dot{x}^2}{N}+\left(a\cos2x\cos\phi+b\sin2x\sin\phi\right)\sin\theta\right)=N\widetilde{S}_E(E_J/E_C,\, C_gV_g)
838: \label{S_reg_gate}
839: \end{equation}
840: Here angles $\theta$ and $\phi$ parameterize coherent states of the spin.
841: Again one easily find two types of tunneling trajectories corresponding to $4e$- and $2e$-supercurrents respectively
842: (e.g. $4e$-trajectory which connects $\left|x=0,\, \theta=\pi/2,\, \phi=-\pi\right>$ with
843: $\left|x=\pi,\, \theta=\pi/2,\, \phi=-\pi\right>$ and $2e$-trajectory connecting
844: $\left|x=0,\, \theta=\pi/2,\, \phi=-\pi\right>$
845: with $\left|x=\pi/2,\, \theta=\pi/2,\, \phi=0\right>$).
846: Action $\widetilde{S}_E$ is a function of $E_J/E_C$ and $C_g V_g$ and does not depend on the number of rhombi.
847:
848: From (\ref{S_reg_gate}) we have classical equations of motion (it is convenient to change variables according to
849: $\theta \rightarrow \pi/2 +i\theta$)
850: \begin{eqnarray}
851: \frac{w}{N}\ddot{x}+(a \sin2x\cos\phi-b\cos2x\sin\phi)\cosh\theta=0\\
852: \dot{\theta}-(a\cos2x\sin\phi-b\sin2x\cos\phi)=0\\
853: \cosh\theta\, \dot{\phi}-(a\cos2x\cos\phi+b\sin2x\sin\phi)\sinh\theta=0
854: \end{eqnarray}
855:
856: Note that $w/N\sim\upsilon$ is a very good small parameter at $E_J\gg E_C$. It allow us to obtain actions
857: $\widetilde{S}_{4e}$ and
858: $\widetilde{S}_{2e}$ analytically. Characteristic frequency of variable $x\,$ is $\omega_x\sim \sqrt{N/w}\gg1\,$,
859: while for spins such a frequency is $\omega_s\sim 1$. This means that on
860: $4e$-trajectory, which does not require any change in spin, we can assume spin to be constant (at least while
861: coefficients $a$ and $b$ are not too close).
862: This immediately leads to
863: \begin{equation}
864: S_{4e}\approx \frac{16 N}{2^{1/4}\pi}\sqrt{\frac{\upsilon}{E_J}\left(1+\cos\pi C_g V_g\right)}
865: \label{S_4e_gate}
866: \end{equation}
867: On the other hand, on $2e$-trajectory smallness of $w/N$ allows us to omit $w\ddot{x}/N$ in equation for $x$.
868: Resulting equations can be integrated analytically and we obtain $2e$-action which, practically does not depend on
869: $E_J/E_C$
870: \begin{equation}
871: S_{2e}\approx N\arcsh\frac{\sqrt{\left|a^2-b^2\right|}}{b}=
872: N\arcsh\frac{2\sqrt{\left|\cos{\pi C_g V_g}\right|}}{1-\cos \pi C_g V_g}
873: \label{S_2e_gate}
874: \end{equation}
875:
876: Actions $\widetilde{S}_{4e}$ and $\widetilde{S}_2e$ can also be evaluated numerically. Results of numerical
877: calculations are presented on figure \ref{S4e2e_gate_fig}. Analytical results
878: (\ref{S_4e_gate}, \ref{S_2e_gate}) are in very good agrement with numerical data. Note that $2e$-supercurrent
879: dominates over the current of pairs of Cooper pairs only in small vicinity of the point
880: $C_g V_g/e=1$.
881:
882:
883: Note that in a regular chain change in $4e$-action upon applying external gate voltage is negative and proportional
884: to the number of rhombi, i.e. external gate leads to significant increase of otherwise suppressed by fluctuations
885: $4e$-supercurrent.
886: \begin{figure}
887: \psfrag{S4e}[c][c]{\tiny $\widetilde{S}_{4e}$}
888: \psfrag{S2e}[c][c]{\tiny $\widetilde{S}_{2e}$}
889: \psfrag{CV}[tc][c]{\tiny $C_g V_g/2e$}
890: \psfrag{ma}[bc][c]{\small a)}
891: \psfrag{myb}[bc][c]{\small b)}
892: \includegraphics[width=210pt]{action4e.eps}\hspace{1cm}
893: \includegraphics[width=210pt]{action2e.eps}
894: \caption{\small Modulation of $4e$- and $2e$-supercurrent in a clean chain by external gate. }
895: \label{S4e2e_gate_fig}
896: \end{figure}
897:
898:
899:
900: Let us now discuss influence of a regular gate on a chain with random stray charges.
901: We will treat this problem within the same limit as in the previous section: we consider chain which is rather
902: far from the maximally frustrated point in the sense $h\gg w$. From the preceding analysis we know that $4e$-action
903: is given by equation (\ref{four_e_action}). Coefficient $A$ acquires now the following form
904: \begin{equation}
905: A=\frac{1}{4}\sum_{n=1}^{N}\left(cos\pi Q_n^1+\cos\pi(Q_n^2+C_g V_g)\right)^2=A_0+\Delta A
906: \end{equation}
907: Here $A_0$ is the value of $A$ at zero external bias.
908: Assuming that fluctuations of random charges are strong one easily finds
909: \begin{equation}
910: \left<\Delta A\right> =0\, \qquad \left<\Delta A ^2\right>=\frac{N}{32}\left(\sin^2\pi C_g V_g+
911: 8\sin^2\frac{\pi C_g V_g}{2}\right)
912: \end{equation}
913: So typical change in $4e$-action can be estimated as
914: \begin{equation}
915: \delta S_{4e}\sim \frac{4\sqrt{2}\,d\,\delta}{\pi^2}\,\sqrt{N\left(\sin^2\pi C_g V_g+
916: 8\sin^2\frac{\pi C_g V_g}{2}\right)}
917: \end{equation}
918: This result differs from the analogous one for a regular (without offset charges)
919: chain in two important aspects: i) we see that
920: change in action is now proportional to $\sqrt{N}$ instead of $N$; ii) variation of the
921: tunnelling action can now be both negative or positive, i.e. applying external voltage we
922: can {\it decrease} $4e$-supercurrent as well as increase it, depending
923: on the realization of charge disorder.
924: From the above analysis we conclude that in a disordered chain current modulation
925: by external gates is random and much weaker than in a regular chain,
926: but it is still present and can be used to demonstrate quantum coherence of the
927: tunnelling processes which occur in the chain.
928:
929:
930:
931:
932: \section{Influence of "magnetic disorder" upon the crossover point}
933: We now consider the effect of randomness in the values of magnetix fluxes through
934: different rombi. Just as in the previous
935: section we start from the grand partition function given by
936: (\ref{Z}) with
937: \begin{equation}
938: \frac{d \widehat{U}_n}{d\tau}=-\left(w\, \cos2x(\tau)
939: S^x-h_n\widehat{S}^z\right)\widehat{U}_n \label{defU_random_h}
940: \end{equation}
941:
942: We presume that fluxes $\Phi_r^n$ are uniformly distributed on
943: $(\Phi_r-\Delta\Phi, \Phi_r+\Delta\Phi)$, i.e. probability
944: distribution for $h_n$ is
945: \begin{equation}
946: P(h)=\left\{
947: \begin{array}{c}
948: \frac{1}{2\sigma}\,, \qquad |h-h_0|<\sigma\\
949: 0\,, \qquad |h-h_0|>\sigma
950: \end{array}\right.
951: \end{equation}
952: where
953: \begin{equation}
954: h_0=\frac{16 N (\Phi_r-\Phi_0/2)}{\pi\Phi_0}\,, \qquad \sigma=
955: \frac{16 N \Delta\Phi}{\pi\Phi_0}\,.
956: \end{equation}
957: Actually particular form of $P(h)$ is not important for us as we
958: assume that $\sigma\ll h_0$ and use perturbation theory in
959: $\sigma/h_0$. We are interested in the critical deviation from the
960: maximally frustrated point $\Phi_r=\Phi_0/2$ destroying
961: $4e$-supercurrent. Therefore we presume that $h_0\gg w$. In such
962: an approximation we can use equation (\ref{TRU}) to calculate $\Tr
963: \widehat{U}_n(\beta)$ and get effective action for variable $x$
964: \begin{equation}
965: S=\int_0^\beta d\tau\frac{\dot{x}^2(\tau)}{2}- \frac{w^2 h
966: N}{4}\int \cos2x(\tau_1)D(\tau_1-\tau_2)\cos2x(\tau_2)
967: \label{random_h_action}
968: \end{equation}
969: \begin{equation}
970: D(\tau_1-\tau_2)=\frac{1}{2 h_0 N}\sum_{n=1}^N
971: \exp\left(-h_n|\tau_1-\tau_2|\right)\label{D}
972: \end{equation}
973:
974: At large $N$ we can replace $D(\tau_1-\tau_2)$ in
975: (\ref{random_h_action}) by its mean value
976: \begin{equation}
977: \overline{D}(\tau_1-\tau_2)=\frac{1}{2h_0}\int dh\,
978: e^{-h|\tau_1-\tau_2|}P(h)
979: \end{equation}
980: Fourier transformation of $\overline{D}$ for small $\sigma/h_0$
981: reads
982: \begin{equation}
983: \overline{D}(\omega)=\frac{1}{h_0^2+\omega^2}\left(1+ \sigma^2
984: \frac{h_0^2-3\omega^2}{3(h_0^2+\omega^2)^2}\right)
985: \end{equation}
986: Again, using the fact that
987: \begin{equation}
988: \overline{D}^{-1}(\omega)=\omega^2+h_0^2+\sigma^2-\frac{4}{3}\sigma^2
989: h_0^2 \frac{1}{h_0^2+\omega^2}
990: \end{equation}
991: we can perform Hubbard-Stratonovich transformation and find
992: representation for partition function with local action (after
993: redefinition of time scale)
994: \begin{equation}
995: S=h\int d\tau \left(
996: \frac{\dot{x}^2}{2}+\frac{\dot{y}^2}{2}+\frac{\dot{z}^2}{2}
997: +\left(1+\frac{\sigma^2}{h_0^2}\right)\frac{y^2}{2}+\frac{z^2}{2}+
998: dy\cos2x+\sqrt{\frac{4}{3}}\frac{\sigma}{h_0} y
999: z\right)\label{decoupling_random_h}
1000: \end{equation}
1001: \begin{figure}
1002: \psfrag{deldel0}[c][c]{$\frac{\delta\Phi^c}{(\delta\Phi^c)_{reg}}$}
1003: \psfrag{sh0}[c][c]{$\sigma/h_0$}
1004: \hspace{1cm}\includegraphics[width=300pt]{delta_cross.eps}
1005: \caption{\small Critical deviation from the maximally frustrated point in a chain with disordered flux in rhombi}
1006: \label{random_h}
1007: \end{figure}
1008: Note that our formulation of the problem is a bit specific: we fix relative diversity $\sigma/h_0$
1009: and look for $h_0$ which brings the chain to the point with equal $4e$- and $2e$-supercurrents.
1010:
1011: Discussion of the previous section concerning two types of
1012: tunneling trajectories and their connection to components of
1013: supercurrent can be literally applied to action
1014: (\ref{decoupling_random_h}). So we need to estimate classical
1015: action for $2e$- and $4e$-trajectories. Problem with $\sigma=0$
1016: was analysed in \cite{my}. It was shown that $2e$-supercurrent
1017: dominates at $d\ll 1$ while $4e$-supercurrent --- at $d\gg1$.
1018: Crossover takes place at $d\approx 3.24$. From
1019: (\ref{decoupling_random_h}) one can easily see that corrections to
1020: classical actions due to nonzero $\sigma$ should be of order
1021: $(\sigma/h)^2$. We analyse action (\ref{decoupling_random_h}) numerically.
1022: Results are presented on figure \ref{random_h}.
1023:
1024: It may come as a surprise, that with this formulation of the problem critical deviation $\delta\Phi^c$
1025: (determined through mean flux in rhombi) {\it grows} with $\sigma/h_0$. However, this is quit reasonable since
1026: zero deviation from the maximally frustrated point for {\it one} rhombi in a chain immediately prohibits
1027: $2e$-supercurrent. Allowing for diversity of fluxes in rhombi we allow some of the rhombi to be closer to the
1028: maximally frustrated point than a "mean" rhombi. This fact causes strong suppression of $2e$-supercurrent.
1029:
1030: So we conclude that randomness in fluxes is not important for the pairing effect,
1031: at least if the standard scatter of these fluxes $\Delta\Phi$
1032: is not too large in comparison with the critical value $\delta\Phi^{c}$ found for a regular chain,
1033: cf.Eq.(\ref{phi_c_reg}).
1034:
1035: \section{Conclusions}
1036: In this paper we provide detailed calculations of superconductive
1037: current in a long frustrated rhombi chain with quenched disorder.
1038: We have considered two types of disorder: random stray charges and random fluxes in rhombi.
1039: We found that small (as compared to the critical deflection $\left(\delta\Phi^c\right)_{reg}$ destroying
1040: $4e$-supercurrent in a clean chain) fluctuations of fluxes piercing rhombi are not that important for the pairing effect.
1041:
1042: Main results of our paper concern effect of quenched random stray charges on pairing of Cooper pairs.
1043: For a chain which is relatively far from the maximally frustrated point in the sense $h\gg w$ we managed to calculate
1044: probability to find the system in the regime with dominating $4e$-supercurrent. Stray charges, in principal, may
1045: significantly affect properties of the rhombi chain. In particular, in such a chain $2e$-supercurrent exist even at
1046: the maximally frustrated point. However, as we have demonstrated in Sec. III, at large $E_J/E_C$ and $\Phi_r=\Phi_0/2$
1047: probability to find large $2e$-supercurrent is small. This result itself is not a great surprise: in a perfectly
1048: classical chain stray charges have no effect at all. More important are two things: i) it is possible to combine
1049: low probability of finding significant $2e$-supercurrent at the maximally frustrated point with exponential
1050: suppression of the supercurrent by quantum fluctuations; ii) in a perfectly classical chain critical deflection
1051: $\delta\Phi^c$ scales with number of rhombi as $1/N$ (this can be easily seen from
1052: eq. (\ref{old E_under_flux_dif_from_pi}) describing energy spectrum of classical rhombi chain) while in a disordered quantum chain this dependence is much
1053: weaker $\delta\Phi^c\sim1/N^{1/6}$, c.f. (\ref{phi_c_random_charges}).
1054:
1055: In Sec. IV we have considered modulation of the supercurrent by capacitively coupled gate. In a regular chain, applying
1056: gate voltage one suppresses quantum fluctuations of rhombi and {\it increases} the supercurrent. Its dependence on the
1057: gate voltage is very strong: change of the {\it logarithm} of the supercurrent is linear in $N$. On the contrary, in a chain
1058: with strong random stray charges applying external gate can both {\it increase} or {\it decrease} supercurrent.
1059: Dependence of
1060: the supercurrent on the gate voltage is now much weaker: typical change of the logarithm of the
1061: supercurrent is now proportional to $\sqrt{N}$. Still we see that dependence of the current on gate voltage should be
1062: measurable. Such a dependence is one of the possible ways to demonstrate coherence of quantum phase slips in the chain.
1063:
1064: We are grateful to L.~B.~Ioffe and B.~Pannetier for many useful and inspiring
1065: discussions.
1066: This research was supported by the Program ``Quantum Macrophysics" of
1067: the Russian Academy of Sciences, by the Russian Ministry of Education and Science via the contract
1068: RI-112/001/417 and by Russian Foundation for Basic Research under the
1069: grant No.\ 04-02-16348.
1070: I.V.P. acknowledges financial support from the Dynasty Foundation and Landau-Juelich Scholarship.
1071:
1072:
1073:
1074: \begin{thebibliography}{10}
1075: \bibitem{Doucot} B. Dou\c{c}ot and J. Vidal, Phys. Rev. Lett. {\bf 88}, 227005
1076: (2002).
1077: \bibitem{IoffeFeigelman02} L. B. Ioffe and M. V. Feigel'man, Phys. Rev. B {\bf 66}, 224503 (2002)
1078: \bibitem{Doucot03} B. Dou\c{c}ot, M. V. Feigel'man and L. B. Ioffe, Phys. Rev. Lett.{\bf 90}, 107003 (2003)
1079: \bibitem{Doucot04} B. Dou\c{c}ot , L. B. Ioffe and J. Vidal, Phys. Rev. B {\bf 69}, 214501 (2004)
1080: \bibitem{my} I. Protopopov and M. Feigel'man, Phys. Rev. B
1081: {\bf 70}, 184519
1082: \bibitem{Larkin} K. A. Matveev, A. I. Larkin and L. I. Glazman,
1083: Phys. Rev. Lett. {\bf 89}, 096802 (2002).
1084: \bibitem{Klauder} J. R. Klauder, Phys. Rev. D {\bf 19}, 2349 (1979)
1085: \end{thebibliography}
1086:
1087: \end{document}
1088: