Synchronized Andreev Transmission in Chains of SNS Junctions
aa r X i v : . [ c ond - m a t . m e s - h a ll ] J un N. M. Chtchelkatchev , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur Institute for High Pressure Physics, Russian Academy of Sciences, Troitsk142190, Moscow region, Russia L.D. Landau Institute for Theoretical Physics, Russian Academy of Sciences,Akademika Semenova av. 1-A, Chernogolovka 142432, Moscow Region, Russia Materials Science Division, Argonne National Laboratory, Argonne, Illinois60439, USA [email protected] Institute of Semiconductor Physics, 13 Lavrentjev Avenue, Novosibirsk 630090,Russia
Abstract.
We construct a nonequilibrium theory for the charge transfer througha diffusive array of alternating normal (N) and superconducting (S) islands com-prising an SNSNS junction, with the size of the central S-island being smaller thanthe energy relaxation length. We demonstrate that in the nonequilibrium regimethe central island acts as Andreev retransmitter with the Andreev conversions atboth NS interfaces of the central island correlated via over-the-gap transmissionand Andreev reflection. This results in a synchronized transmission at certain reso-nant voltages which can be experimentally observed as a sequence of spikes in thedifferential conductivity.
An array of alternating superconductor (S) - normal metal (N) islands is a fun-damental laboratory representing a wealth of physical systems ranging fromJosephson junction networks and layered high temperature superconductorsto disordered superconducting films in the vicinity of the superconductor-insulator transition. Electronic transport in these systems is mediated by An-dreev conversion of a supercurrent into a current of quasiparticles and viceversa at interfaces between the superconducting and normal regions [1]. A fas-cinating phenomenon benchmarking this mechanism is the enhancement of theconductivity observed in a single SNS junction at matching voltages constitut-ing an integer ( m ) fraction of the superconducting gap, V = 2 ∆/ ( em ) [2–11]due to the effect of multiple Andreev reflection (MAR) [12–14]. The current-voltage characteristics of diffusive SNS junctions were discussed in great detail N. M. Chtchelkatchev , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur eV = 3 /7 ∆ eV = 4 /7 ∆ ∆ N N S L S R S C ∆ N N S L S R S C ∆ N N S L S R S C ∆ N N S L S R S C eV = 6 /7 ∆ eV = 8 /7 ∆ eV tot = 2 ∆ ∆ N N S L S R S C A B C D e e ∆ N N S L S R S C A ' B ' C ' D ' h h n = (a) (b) eV tot = ∆ n = (c)(e) (d)(f) Fig. 1.1.
Diagrams of the SAT processes for the first, n = 1, (a)-(b), and second, n = 2 (c)-(f), subharmonics of the resonant singularities in dI/dV described byEq. (1.1) for the SNSNS junction with the normal resistances ratio R /R = 3 / ε = ± ∆ at the electrodesS L and S R . The dashed solid lines show paths starting and/or ending at the edgesof the gap of the central island S C . The circle denotes the over-the-gap Andreevreflections at the electrodes. The paths ABCD [panel (a)] and D ′ C ′ B ′ A ′ [panel (b)]correspond to the electron- and hole trajectories, respectively. Synchronization ofthe energies of the incident and emitted quasiparticles at points B and C (B ′ andC ′ ) is shown by arrows. SAT is realized by trajectories passing through the singularpoints ε = ± ∆ of the central island S C and including over-the-gap transmissionsand Andreev reflections. Trajectories synchronizing other transmissions across S C and those of higher orders are not shown. Note, that voltage drops eV = 6 ∆/ eV = 8 ∆/ eV = 3 ∆/ eV = 4 ∆/ L N S C or S C N S R parts. Synchronized Andreev Transmission in Chains of SNS Junctions 3 in Refs. [15, 16]. Further developments were obliged to studies of large arrayscomprised of many SNS junctions [17–21]. Experimental results, especiallythose obtained on the multiconnected arrays [17,18,21], indicated clearly thatsingularities in transport characteristics cannot be explained by MARs at in-dividual SNS junctions only and that there is evidently a certain coherenceof the Andreev processes that occur at different NS interfaces. These findingscall for a comprehensive theory of transport in large SNS arrays.In this article we develop a nonequilibrium theory of electronic transport ina series of two diffusive SNS junctions, i.e. an SNSNS junction and derive thecorresponding current-voltage characteristics. We demonstrate that splittingthe normal part of the SNS junction into two normal islands that have, in gen-eral, different resistances and are coupled via a small superconducting granule,S C leads to the nontrivial physics and emergence of a new distinct resonantmechanism for the current transfer: the Synchronized Andreev Transmission (SAT). The main component of our consideration is a nonequilibrium circuittheory of the charge transfer across S C . [The symmetric case with the equalresistances of the normal parts was discussed in detail in [22]. Unfortunatelythe technique developed there does not allow straightforward generalizationonto a nonsymmetric case.]In the SAT regime the processes of Andreev conversion at the boundariesof the central superconducting island are correlated: as a quasiparticle withthe energy ε hits one NS C interface, a quasiparticle with the same energyemerges from the other S C N interface and enters the bulk of the normal is-land (and vice versa, see Fig. 1). This energy synchronization is achieved viaover-the-gap Andreev processes [19], which couple MARs occurring within theeach of the normal islands and make the quasiparticle distribution at the cen-tral island essentially nonequilibrium. Effectiveness of the synchronization iscontrolled by the value of the energy relaxation lengths of both, the quasipar-ticles crossing S C with energies above ∆ , and the quasiparticles experiencingMAR in the normal parts. The SAT processes result in spikes in the differ-ential conductivity of the SNSNS circuit, which appear at resonant values ofthe total applied voltage V tot defined by the condition V tot = 2 ∆en (1.1)with integer n , irrespectively of the details of the distribution of the partialvoltages at the two normal islands.The article is organized as follows. In the Section 1.2 we define the system,a diffusive SNSNS junction which will be a subject of our study. Section 1.3 isdevoted to introduction and description of the employed theoretical tools: theelectronic transport of the system in the resistive state is given by the Larkin-Ovchinnikov equation in a form of matrix equations for the Green’s functionstaken in Keldysh representation. In Sections 1.4-1.8 we construct an equivalentcircuit theory for an SNSNS junction resulting in the recurrent relations forthe spectral current flow in the energy space. In Section 1.9 we present the N. M. Chtchelkatchev , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur original numerical method enabling us to solve the recurrent relations for thespectral current and obtain the I - V characteristics for the SNSNS junction.The obtained results are discussed in Section 1.10, where we demonstrate, inparticular, that the SAT-induced features become dominant in large arraysconsisting of many SNS junctions. We consider charge transfer across an S L N S C N S R junction, where S L , S C ,and S R are mesoscopic superconductors with the identical gap ∆ ; the ‘edge’superconducting granules S L and S R play the role of electrodes, and S C is thecentral island separating the two normal parts with, in general, different nor-mal resistances. We discuss the common experimental situation of a diffusiveregime where the most of the energy scales are smaller than ~ /τ , where τ isthe impurity scattering time. We assume the size, L C , of the central islandto be much larger than the superconducting coherence length ξ , hence pro-cesses of subgap elastic cotunneling and/or direct Andreev tunneling [23] donot contribute much to the charge transfer. In general, this condition ensuresthat L C is large enough so that charges do not accumulate in the centralisland and Coulomb blockade effects are irrelevant for the quasiparticle trans-port. At the same time L C is assumed to be less than the charge imbalancelength, such that we can neglect the coordinate dependence of the quasipar-ticle distribution functions across the island S C . Additionally, the condition ℓ ε ≫ L C , where ℓ ε is the energy relaxation length, implies that quasiparticleswith energies ε > ∆ traverse the central superconducting island S C withoutany noticeable loss of energy. The normal parts N and N are the diffusivenormal metals of length L , > ξ , and L , > L T , L T = p ~ D N /ε , where D N is the diffusion coefficient in the normal metal. We assume the Thoulessenergy, E Th = ~ D N /L , , to be small, E Th ≪ ∆ , and not to exceed the char-acteristic voltage drops, E Th < eV , . If these conditions that define the socalled incoherent regime [15] are met, the Josephson coupling between the su-perconducting islands is suppressed. And, finally, we let the energy relaxationlength in the normal parts N and N be much larger than their sizes, so thatquasiparticles may experience many incoherent Andreev reflections inside thenormal regions. The current transfer across the SNSNS junction is described by quasiclassicalLarkin-Ovchinnikov (LO) equations for the dirty limit [24, 25]: − i [ ˇ H eff ◦ , ˇG ] = ∇ ˇJ , ˇ J · n = 12 σ S R [ ˇ G S , ˇ G N ] , (1.2) Synchronized Andreev Transmission in Chains of SNS Junctions 5 where ˇ H eff = ˇ1( i ˆ σ z ∂ t − ϕ ˆ σ + ˆ ∆ ), ˇJ = D ˇG ◦ ∇ ˇG is the matrix current, thesubscripts “S” and “N” denote the superconducting and normal materials,respectively, “ ◦ ” stands for the time-convolution, ˆ σ i (i = x,y,z) are the Paulimatrices, operating in the Nambu space of 2 × ∆ = i ˆ σ x Im ∆ + i ˆ σ y Re ∆ , and R is the resistance of an NS interface. Thediffusion coefficient D assumes the value D N in the normal metal and thevalue D S in the superconductor, and ϕ is the electrical potential which wecalculate self-consistently. The unit vector n is normal to the NS interfaceand is assumed to be directed from N to S. The momentum averaged Green’sfunctions ˇG ( r , t, t ′ ) are 2 × × ˇG = (cid:18) ˆ G R ˆ G K G A (cid:19) ; ˆ G R(A) = (cid:18) G R(A) F R(A) ˜ F R(A) ˜ G R(A) (cid:19) , (1.3) r is the spatial position, t and t ′ are the two time arguments. The Keldyshcomponent of the Green’s function is parametrized as [24]: ˆ G K = ˆ G R ◦ ˆ f − ˆ f ◦ ˆ G A , where ˆ f is the distribution function matrix, diagonal in Nambu space,ˆ f ≡ diag [1 − n e , − n h ], n e(h) is the electron (hole) distribution function.In equilibrium n e(h) becomes the Fermi function. And, finally, the Green’sfunction satisfies the normalization condition ˇG = ˇ1.The edge conditions closing Eqs. (1.2) are given by the expressions for theGreen’s functions in the bulk of the left (L) and right (R) superconductingleads: ˇG L(R) ( t, t ′ ) = e − iµ L(R) t ˆ τ / ~ ˇG ( t − t ′ ) e iµ L(R) t ′ ˆ τ / ~ , the chemical potentials are µ L = 0 and µ R = eV . Here, ˇG ( t ) is the equilib-rium bulk BCS Green’s function.The current density is expressed through the Keldysh component of ˇJ as I ( t, r ) = πσ N σ z ˆ J K ( t, t ; r ) = 12 Z dε [ I e ( ε ) + I h ( ε )] , (1.4)where the spectral currents I e and I h representing the electron and hole quasi-particle currents, respectively, are the time Wigner-transforms of top- andbottom diagonal elements of the matrix current ˇJ (K) .On the normal side of the superconductor-normal metal interface, theKeldysh component of Eqs. (1.2) yield the conservation conditions: ∇ I e(h) = 0 , (1.5) I e ( ε ) = σ N { D p ( ε + u ) ∇ n e ( ε ) − D m ( ε + u ) ∇ n h ( ε + 2 u ) } , (1.6) I h ( ε ) = σ N { D p ( ε − u ) ∇ n h ( ε ) − D m ( ε − u ) ∇ n e ( ε − u ) } , (1.7)where u is the electrical potential of the adjacent superconductor, D p(m) =( D − ± D + ) / N. M. Chtchelkatchev , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur n (1) ( 2 ) ε + u h − ε R P ( C ) ( ) + u n + u (C) ( 2 ) ε h − ε R P ( C ) ( ) + u R (1C) ( ) ε +u Q R P ( C ) () ε + u I + u ( 2 ) ε h I ( ) ε e S C n (C) ( ε) e n (1) ( ε) e R (1C) ( ) ε +u Q n (1) ( 2 ) ε + u h n +u (C) ( ) ε F R ( C ) ( ) ε + u + R P ( C ) () ε + u I + u ( 2 ) ε h I ( ) ε e S C n (1) ( ε) e R ( C ) ( ) ε + u + (a) (b) Fig. 1.2.
Effective circuit for the boundary between the normal metal and thesuperconductor. Kirchhoff laws where the role of the potential in the nodes is takenby the electron- and hole distribution functions give the boundary conditions forthe LO equations. (a) A general nonequilibrium case. (b) Equivalent circuit for anequilibrium case where quasiparticle distribution functions in the superconductorare the Fermi-functions, n F , then n e ( ε ) = n F ( ε + u ) = n h ( ε + 2 u ), used in [15]. Thesuperconductor electrical potential at the boundary is equal to u . D + ( ε ) = 14 Tr[ˆ1 − ˆ G R ( ε ) ˆ G A ( ε )] ,D − ( ε ) = 14 Tr[ˆ1 − σ z ˆ G R ( ε ) σ z ˆ G A ( ε )] , (1.8)and the trace is taken over components in the Nambu-space. In the bulk ofa normal metal, ˆ G R( A ) ( ε ) → ± ˆ σ z and D + ≈ D − ≈
1, so I e = σ N ∇ n e and I h = σ N ∇ n h . We start the construction of the circuit theory with the corresponding for-mulation of the boundary conditions for the distribution functions at the in-terface between the normal parts and the central superconducting island. Weconsider a stationary situation where the applied voltage does not depend ontime. Then the Green’s functions can be parameterized near an NS interfaceas follows: [ ˆ G R ] j ( ε, ε ′ ) = ˆ σ z δ ε − ε ′ cosh θ j ( ε )+ˆ σ + δ ε − ε ′ +2u sinh θ j ( ε ) − ˆ σ − δ ε − ε ′ − u sinh θ j ( ε ) ,G A = − ˆ σ z ( G R ) † ˆ σ z , (1.9)where ˆ σ ± = ˆ σ x ± i ˆ σ y , j = S, N, and u is the electrochemical potential.The effective diffusion coefficients are correspondingly D + = cos Im θ and D − = cosh Re θ . When deriving Eq.(1.9), we have used the condition thatthe Josephson coupling between the superconducting islands in the junctionis suppressed. The proximity effect results in an additional term in Eq.(1.9)proportional to δ ( ε − ε ′ − u ′ − u )), where u ′ is the potential of the adjacentsuperconductor involved. Synchronized Andreev Transmission in Chains of SNS Junctions 7
Taking the Keldysh component of the boundary term in Eq. (1.2) we derivethe boundary conditions for the currents I e(h) at the NS interface, whichassume the form of Kirchhoff’s laws for the circuit shown in Fig. 1.2(a). Theelectron and hole distribution functions take the role of voltages at the nodes.The equation for an electronic spectral current flowing into the lower leftcorner node: I e ( ε ) = n (C) e ( ε ) − n (1) e ( ε ) R (1C)Q ( ε + u ) + n (C) h ( ε + 2 u ) − n (1) e ( ε )[ − R (1C)P ( ε + u )] + n (1) h ( ε + 2 u ) − n (1) e ( ε ) R (1C)P ( ε + u ) . (1.10)The equation for the hole current going into the top left node of the circuitof the Fig. 1.2(a) is easily obtained analogously to (1.10) with the aid of theadditional transformation ε → ε − u i.e. by shifting all the energies over − u : I h ( ε ) = n (C) h ( ε ) − n (1) h ( ε ) R (1C)Q ( ε − u ) + n (C) e ( ε − u ) − n (1) h ( ε )[ − R (1C)P ( ε − u )] + n (1) e ( ε − u ) − n (1) h ( ε ) R (1C)P ( ε − u ) . (1.11)In an equilibrium the quasiparticles in the superconductor follow the Fermidistribution, then n (C) e ( ε ) = n F ( ε + u ) = n (C) h ( ε + 2 u ) and Eqs.(1.10)-(1.11)reduce to I e ( ε ) = n F ( ε − u ) − n (1) e ( ε ) R (1C) + ( ε + u ) + n (1) h ( ε + 2 u ) − n (1) e ( ε ) R (1C)P ( ε + u ) , (1.12)where the interjacent resistances are defined as R − Q ( P ) ( ε ) = { R − − ( ε ) ± ¯ R − ( ε ) } /
2; here 1 /R ± ( ε ) = [ N N ∓ M ± M ± ] /R , N j ( ε ) = Re cosh θ j , M + j ( ε ) + i M − j ( ε ) = sin θ j and j = 1 , We consider the normal metal between the left superconducting lead S L and the superconducting island, S C , see Fig. 1.3. The boundary conditions,Eqs.(1.10)-(1.11), relate electron and hole distribution functions at the rightNS and left NS interfaces. Below we relate electron and hole distribution func-tions at x = 0 and x = d building the effective circuit, where d is the length N. M. Chtchelkatchev , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur ε d x I ( ) ε e I ( ) ε h I + u ( 2 ) ε h I + u' ( 2 ) ε h I u' ( 2 ) ε− e I u ( 2 ) ε− e u' u S C S L Fig. 1.3.
Illustration of the spectral currents flow in the normal metal [white area]surrounded by the superconductors [grey area]. The black boxes at the interfacesencode the boundary conditions picture like it is shown in Fig.1.2. of the normal layer. At the first step we neglect the proximity effect changeof the junction resistance and take D ± = 1 everywhere in the normal layer.Then, I e(h) = σ ∇ n e(h) , △ n e(h) = 0 and therefore I e(h) ( ε ) = [ n e(h) ( d, ε ) − n e(h) ( x = 0 , ε )] /R , (1.13)where R = σ /d is the normal resistance of the N -layer. Eq.(1.13) resemblesthe Ohm law for the resistor R , but where the role of the voltages play thedistribution functions at the ends of the resistor. Eq.(1.13) is approximatebecause it neglects the proximity renormalization of the normal layer conduc-tivity [26]. It was shown in Ref. [15] for a SNS junction that the replacement of n e(h) ( x = { , d } ) by the properly chosen proximity renormalized distributionfunctions makes Eq.(1.13) accurate. We show below that this idea is appli-cable when electron and hole distribution functions in the superconductorsessentially deviate from the Fermi functions and when the electron and holecurrents can not be in general related by a shift of the energy like in SNSjunction.At the left NS-interface the spectral currents, I e ( ε ) and I h ( ε + 2 u ) arerelated by the Andreev process, see Fig.1.2a. It follows from Eqs.(1.5)-(1.7)that the combination of the quasiparticle currents, I (1C) ± ( ε ) = I e ( ε ) ± I h ( ε +2 u ) = σ D (1C) ± ( ε + u ) ∇ n (1C) ± ( ε + u ), conserve in the normal metal: ∇ I (1C) ± = 0.Integrating the last equation over x we get, I (1C) ± Z d dxD (1C) ± ( x ) = σ [ n ± ( d ) − n ± ( x )] , (1.14)where n (1C) ± ( ε ) = n e ( ε ) ± n h ( ε + 2 u ). Eq.(1.14) can be equivalently rewritten: I (1C) ± ( d − x ) = σ [¯ n (1C) ± ( d ) − n (1C) ± ( x )] , (1.15)¯ n (1C) ± ( d ) ≡ n (1C) ± ( d ) − m (1C) ± I (1C) ± ( ε + u ) , (1.16)where Synchronized Andreev Transmission in Chains of SNS Junctions 9 m (1C) ± = 1 σ Z d (cid:18) D (1C) ± ( x ) − (cid:19) dx. (1.17)Here the variable x occupies the domain ξ N ≪ x ≪ d − ξ N where the Cooperpair wave functions from the left and right superconductors, see Fig.1.3, donot overlap. At these values of x , the angle θ ( x ) → D ± ( x ) → x by 0 in the integral written in Eq.(1.14).Taking into account that I (1C) ± ( ε + u ) = I e ( ε ) ± I h ( ε + 2 u ) we finally getthe following important result:( d − x ) I e ( ε ) = σ [¯ n (1C) e ( d ) − n (1C) e ( x )] , (1.18)where¯ n (1C) e ( d ) = n (1C) e ( d ) − I e ( ε ) m (1C) e ( ε + u ) − I h ( ε + 2 u ) m (1C) h ( ε + u ) . (1.19)Here m (1C) e(h) = [ m (1C) + ± m (1C) − ] / ξ N ≪ x ≪ d − ξ N : x I e ( ε ) = σ [ n (1L) e ( x ) − ¯ n (1L) e ( x = 0)] , (1.20)where¯ n (1L) e ( x = 0) = n (1L) e ( x = 0) + I e ( ε ) m (1L) e ( ε + u ′ )+ I h ( ε + 2 u ′ ) m (1L) h ( ε + u ′ ) . (1.21)Here m (1L) e(h) = [ m (1L) + ± m (1L) − ] / m (1L) ± = 1 σ Z d (cid:18) D (1L) ± ( x ) − (cid:19) dx. (1.22)Eqs.(1.18)-(1.20) show how n (1)e depends on x in the central part of thenormal layer in Fig.1.3. Eq.(1.18) must be consistent with Eq.(1.20). The onlyway to satisfy this condition is the following one: I e ( ε ) = ¯ n (1C) e ( d ) − ¯ n (1L) e ( x = 0) R , (1.23)where R = d/σ is the normal resistance of the N-layer and we used that n (1L) e ( x ) = n (1C) e ( x ). The condition Eq.(1.23) resembles the Ohm law. It allowsto relate the distribution functions at x = 0 and x = d .Similar condition holds for I h : I h ( ε ) = ¯ n (1C) h − ¯ n (1L) h R , (1.24) , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur where¯ n (1C) h = n (1C) h ( d ) − m (1C) h ( ε − u ) I e ( ε − u ) − m (1C) e ( ε − u ) I h ( ε ) , (1.25)¯ n (1L) h = n (1L) h + m (1L) h ( ε − u ′ ) I e ( ε − u ′ ) + m (1L) e ( ε − u ′ ) I h ( ε )] . (1.26)The last step is the formulation of the boundary conditions at the NS inter-faces in terms of the distribution functions with bars. Using the I ± notationswe can rewrite the boundary conditions, Eqs.(1.10)-(1.11), in the compactform I (1C) ± ( ε ) = n (C) ± ( ε ) − n (1C) ± ( ε ) R (1C) ± ( ε + u ) . (1.27)Then it follows from Eq.(1.15) that we can write: I (1C) ± ( ε + u ) = n (C) ± − ¯ n (1C) ± ¯ R (1C) ± ( ε + u ) , (1.28)¯ R (1C) ± ( ε ) = m (1C) ± ( ε ) + R (1C) ± ( ε ) . (1.29)The same form has the boundary condition at x = 0: I (1L) ± ( ε + u ′ ) = ¯ n (1L) ± − n (L) ± ¯ R (1L) ± ( ε + u ′ ) , (1.30)¯ R (1L) ± ( ε ) = m (1L) ± ( ε ) + R (1L) ± ( ε ) . (1.31)It follows that the physical meaning of m ± terms is the proximity effect con-tribution to the NS interface resistance, see [15].It is more convenient to work with the boundary conditions for I e(h) ratherthen with those for I ± . Then one can use Eqs.(1.10),(1.11) but with n (1L) e(h) → ¯ n (1L) e(h) and R (1C) Q(P) → ¯ R (1C) Q(P) , where, for example, ¯ R (1C) Q(P) = 2 ¯ R (1C) − ¯ R (1C) + / [ ¯ R (1C) + ± ¯ R (1C) − ]. Having formulated the boundary conditions for the distribution functions weturn now to advanced and retarded Greens functions behaviors.Normal layers in experimental SNS junctions and SNS arrays, see Ref.[17, 19], connect with superconductors like it is shown in Figs.1.4. The junc-tions of this type are usually referred to as “weak-links”. [27, 28] Boundaryconditions for retarded and advanced Greens functions, Eq.(1.2), can be sim-plified in this case: retarded and advanced Greens functions at superconduct-ing sides of NS boundaries can be substituted by Greens function from the
Synchronized Andreev Transmission in Chains of SNS Junctions 11 bulk of the superconductors. These “rigid” boundary conditions approxima-tion is reasonable because the magnitude of the current is much smaller thanthe critical current of the superconductor [this is assumed] and the currententering the superconductor from narrow normal metal wire with the widthcomparable with the Cooper pair size. There are also other cases when therigid boundary conditions are correct, for example, if the NS boundary hasthe small transparency due to, e.g., an insulator layer at the NS interface.The recipe telling how one should evaluate θ ( x ) in the normal metal nearthe NS boundary, where the rigid boundary conditions hold, can be takenfrom e.g. Ref. [15], and we reproduce briefly their result for the completeness.We will write down θ ( x ) near the right NS boundary (see Fig.1.3 ) taking u ′ = 0. In the superconductor, θ S = atanh ( ∆/ε ), where ∆ is the gap. Thevalue of θ N = θ ( x = 0) in the normal metal side should be found from theequation: W s i∆ε + i/ τ σ sinh( θ N − θ S ) + 2 sinh θ N , (1.32)where W = R ∆ /R NS , R ∆ = ξ ∆ /σ N , ( ξ ∆ = p D/∆ ) is the resistance of thenormal metal layer with the width ξ ∆ . Here τ σ is the pair breaking rate [29][e.g., induced by electron-phonon or electron-electron interactions] and R NS isthe normal resistance of the interface. Then the solution for θ ( x >
0) is thefollowing: tanh θ (cid:18) − xξ ε √ i (cid:19) tanh θ N , (1.33)where ξ ε = p D/ [2( ε + i/ τ σ )]. The effective conductances ¯ g ± should be ex-pressed through θ found from Eqs.(1.32)-(1.33). S N
I Iw
Fig. 1.4.
Typical experimental array of SNS junctions [17, 19]. This type of thelink enables us to use the rigid boundary conditions for the retarded and advancedGreens functions.2 N. M. Chtchelkatchev , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur I + u( 2 )ε h I + u( 2 )ε h ( )I ε e ( )I ε e n (1) ( 2 )ε+ u h − ε R P ( C ) ( ) + u n + u (C) ( 2 )ε h − ε R P ( C ) ( ) + u R (1C) ( )ε+u Q R P ( C ) () ε + u R P ( C ) () ε + u − ε R P ( C ) ( ) + u − ε R P ( C ) ( ) + u n (2) ( 2 )ε+ u h R (2C) ( )ε+u Q I + u( 2 )ε h I + u( 2 )ε h I ( )ε e I ( )ε e S C n (2) (ε) e n (C) (ε) e n (1) (ε) e R (1C) ( )ε+u Q R (2C) ( )ε+u Q (a) n (1) ( 2 )ε+ u h I + u( 2 )ε h I ( )ε e n (1) (ε) e I ( )ε e n (2) (ε) e n (2) ( 2 )ε+ u h I + u( 2 )ε h R ( )ε+u D R ( )ε+u D − ε R ( ) + u B () ε + u B R () ε + u B R − ε R ( ) + u B (b) (с) Fig. 1.5. (a) Effective circuit representing current conversion at the interfaces ofthe central superconducting island S C . Resistors, R P and R Q stand for an Andreev-and a normal processes respectively. The role of voltages at the nodes is played bythe electron and hole distribution functions. (b) An illustration of the boundaryconditions Eqs.(1.34)-(1.36) in terms of a pyramid-circuit is given in this figure.Electron and hole currents entering the left side of the pyramid flow in one normallayer, the right currents flow in the other normal layer. The effective resistance ¯ R D describes the “direct” quasiparticle transmission from one normal layer to the otherthrough the superconductor and the resistance ¯ R B describes Andreev processes. c)Equivalent 3D-sketch of the circuit (b). The circuit shown in Fig.1.5a is the graphic representation of the boundaryconditions to Eq.(1.2) at the edges of the superconducting island. It is con-structed from the circuit units shown in Fig.1.2a. We consider the case wherethe size of the superconducting island is less than the charge imbalance length,and therefore the coordinate dependence of the quasiparticle distribution func-tions at the island can be neglected. Solving the Kirchhoff equations for thecircuit shown in Fig.1.5a we exclude the quasiparticle distribution functionscorresponding to the superconducting island and express the spectral currentsthrough the quasiparticle distribution functions in the normal layers: I e ( ε ) = ¯ n (2)e ( ε ) − ¯ n (1)e ( ε )¯ R D ( ε + u ) + ¯ n (2)h ( ε + 2 u ) − ¯ n (1)h ( ε + 2 u )¯ R B ( ε + u ) , (1.34) I h ( ε ) = ¯ n (2)h ( ε ) − ¯ n (1)h ( ε )¯ R D ( ε − u ) + ¯ n (2)e ( ε − u ) − ¯ n (1)e ( ε − u )¯ R B ( ε − u ) , (1.35)¯ R D(B) = 2 (cid:20) R (1C) + + ¯ R (2C) + ± R (1C) − + ¯ R (2C) − (cid:21) − . (1.36) Synchronized Andreev Transmission in Chains of SNS Junctions 13
The effective resistance ¯ R D describes the “direct” quasiparticle transmissionfrom one normal layer to the other through the superconductor and the re-sistance ¯ R B describes the Andreev processes, see Fig.1.5b. Note that the di-rect and indirect transmissions here are different from the so-called “elasticco-tunneling” and “crossed Andreev tunneling” [23, 30] processes where Bo-goliubov quasiparticles tunnel below the gap through a thin (with the widthof the order of the Cooper pair size) superconducting layer. The probability ofthese tunneling processes decreases exponentially if the width of the supercon-ducting layer exceeds the Cooper pair size. and they occur without generatingsupercurent across a superconductor (the supercurrent flows “virtually”). Thesize of superconucting islands of the SNS arrays that we consider here exceedwell the Cooper pair size, and the current of the quasiparticles with the ener-gies below the hap converts at the NS interface into the supercurrent acrossthe S-islands and then transforms again into the quasiparticle current at theopposite SN-interface. We have demonstrated that there is a direct correspondence between the effec-tive electric circuit and the solution of the Usadel equations with the appropri-ately chosen boundary conditions. The effective circuit describing transport inSNSNS-array is shown in Fig.1.6. We choose the direction of the current flowin such a way that the electron, I e , and the hole, I h , currents go in oppositedirections. The expression for the total current then assumes the form: I ( V ) = − e Z dε ( I e + I h ) . (1.37)The spectral currents I e and I h satisfy in general the relation: I e ( ε ) = − I h ( ε ) | V →− V . Similarly, n e ( ε ) = n h ( ε ) | V →− V , ensuring the identity I ( − V ) = −I ( V ).Writing down the Kirchofs equations for potential distribution at the cir-cuit in Fig. 1.6 we arrive at the recurrent relations, see Appendix A: R ( ε, − u , − V ) I h ( ε ) − ρ ( ◦ ) ( ε − u) I e ( ε − − ρ ( ⊲ ) ( ε ) I e ( ε ) − ρ ( ⊳ ) ( ε − V ) I e ( ε − V ) = n F ( ε ) − n F ( ε − V ) , (1.38) R ( ε, u , V ) I e ( ε ) − ρ ( ◦ ) ( ε + u) I h ( ε + 2u) − ρ ( ⊲ ) ( ε ) I h ( ε ) − ρ ( ⊳ ) ( ε + V ) I h ( ε + 2 V ) = n F ( ε + V ) − n F ( ε ) . (1.39)Here the effective resistance R = R + R + ρ ( ⊲ ◦ ⊳ ) , where , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur R + (1L) ( ) ε n (1L) ( ε) e n ( ) ε F n (1L) ( ) ε h − ε R ( ) +u B − ε− R ( ) u B I h I e R n (1) ( ε) e n (1) ( 2 ) ε + u h I + u ( 2 ) ε h I u ( 2 ) ε− e n (1) ( ) ε h n (1) ( 2 ε− ) u e R ( ) ε +u B R ( ) ε− u B R ( ) ε− u B R ( ) ε +u B R ( ) ε− u D R ( ) ε +u D − ε R ( ) +u B − ε− R ( ) u B n (2) ( 2 ) ε + u h n (2) ( ε) e n (2) ( ) ε h n (2) ( 2 ε− ) u e I V ( 2 ) ε− e I h R R I e I u ( 2 ) ε− e I + u ( 2 ) ε h I + V ( 2 ) ε h n V ( ) ε− F n V ( ) ε+ F n (2R) ( ) ε h n (2R) ( 2 ε− ) V e n (2R) ( 2 ε+ ) V h n (2R) ( ε) e R P (2R) ( ) ε− V R + (1L) ( ) ε R + (2R) ( ) ε− V R + (2R) ( ) ε− V R + (2R) ( ) ε+ V R + (2R) ( ) ε+ VR R P (2R) ( ) ε+ VR P (1L) ( ) ε Fig. 1.6.
MAR in a SNSNS array. The graph shows the effective circuit for quasi-particle currents I e and I h in the energy space. The role of voltages here play quasi-particle distribution functions. Boxes, triangles and ovals play the role of effectiveresistances that come from Usadel equations and their boundary conditions. ρ ( ⊲ ◦ ⊳ ) = (1 / X α = ± { ¯ R ( ) α,ε + ¯ R ( ) α,ε +u + ¯ R ( ) α,ε +u + ¯ R ( ) α,ε + V } , (1.40) ρ ( ◦ ) = (1 / { ¯ R ( )+ + ¯ R ( )+ − ¯ R ( ) − − ¯ R ( ) − } , (1.41) ρ ( ⊳ ) = (1 / { ¯ R ( )+ − ¯ R ( ) − } , (1.42) ρ ( ⊲ ) = (1 / { ¯ R ( )+ − ¯ R ( ) − } . (1.43)In the normal state of the array (or if | ε | ≫ ∆ ) R reduces to a normal resis-tance of the array whereas ρ ( ⊳ ) and ρ ( ⊲ ) vanish. Then we find from Eqs.(1.38)-(1.39) that I h ( ε ) = [ n F ( ε ) − n F ( ε − V )] / R , and I e ( ε ) = [ n F ( ε + V ) − n F ( ε )] / R that with Eq.(1.37) reproduces the Ohm’s law, I = V / R .It is easy to find the island potential in the case of symmetrical array whenthe transmitivities of the island-normal metal interfaces are equal as well asthe transmitivities of the lead-normal metal interfaces and R = R . Thenthe resistances ¯ R ( ) ± = ¯ R ( ) ± , ¯ R ( ) ± = ¯ R ( ) ± and for the symmetry reasons, u = V /
2. At the same time the recurrent relations Eqs.(1.38)-(1.39) become
Synchronized Andreev Transmission in Chains of SNS Junctions 15 invariant under the substitution I e ( ε − V ) = I h ( ε ) and reduce to the relation: R ( ε, V ) I e ( ε ) − ρ ( ⊲ ) ( ε ) I e ( ε − V ) − ρ ( ⊳ ) ( ε + V ) I e ( ε + V ) = n F ( ε + V ) − n F ( ε ) , (1.44)where R ( ε, V ) ≡ R ( ε, V / , V ) − ρ ( ◦ ) ( ε + V /
2) == R N ( ε ) + (1 / X α = ± { ¯ R ( ) α,ε + ¯ R ( ) α,ε + V } , (1.45)where R N ( ε ) = R + R + ¯ R ( ) − ,ε + V / + ¯ R ( ) − ,ε + V / .The recurrent relation, Eq.(1.44), is similar to that of a (symmetric) SNSjunction, see Ref. [15, 22] and Appendix B, but in our case the normal re-sistance R N ( ε ) becomes energy dependent [22]. In other words, a symmetricSNSNS array has the same transport properties as a single (symmetric!) SNSjunction, but with the energy dependent resistance of the normal layer. Theimbalance resistance ¯ R (1C) − ,ε has singularities at the energy corresponding tothe gap edges of the superconducting island in the center of our SNSNS ar-ray. This is the origin of the subharmonic singularities in the current-voltagecharacteristics at voltages 2 ∆/V = n/ n = 1 , , . . . , contrasting the “conven-tional values” in an SNS junction determined by the relations 2 ∆/V = n . Itfollows from Eq.(1.45) the unusual subharmonic singularities should disappearif the resistance of the normal layer greatly exceeds the resistance of the SNinterfaces. Then R ≫ ¯ R ( ) − ,ε + V / + ¯ R ( ) − ,ε + V / and the central superconductingisland of the SNSNS array effectively “disappears” and the array completelytransforms into a SNS junction [22]. Calculation of the current-voltage characteristics I ( V ) requires numericalsolving of the recurrent relations, Eqs. (1.38)-(1.39). To accomplish the numer-ical task, we have developed a computational scheme allowing to bypass insta-bilities caused by the non-analytic behavior of the spectral currents I e(h) ( ε ),which poses the major computational challenge. The procedure is as follows:first, we fix certain chosen energy ε and identify the set of energies connectedthrough the equations in the given energy interval, solving afterwards theresulting subsystem of equations. We then repeat the procedure, until therequired energy resolution of δε = 10 − ∆ is achieved. Typically, up to 10 linear equations had to be solved for every given voltage, but the complexityof the coupled subsystem depends on the commensurability of u and V .Figure 1.7 shows the comparative results for the SNSNS junction andtwo SNS junctions in series. The latter corresponds to the case where thesize of the central island well exceeds the energy relaxation length, L C > ℓ ε . , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur SN SR N R N R N eV tot / m =2m =2m =2 m =2 n=1 m =2 m =29/10 4/53/4 d V / d I R N R N R N m =2m =2m =2 m =2m =2m =2 9/104/53/4 SN S+SN S Fig. 1.7.
Left panel: Differential resistances as functions of the applied voltage V tot (around n = 1 in Eq. (1)) for the SN SN S junction. The fractions 3/4, and 4/5, and9/10 represent the ratios of resistances of the normal regions, R /R . The differen-tial resistance dV /dI of the SN SN S junction demonstrates the pronounced SATspike at V tot = 2 ∆/e , irrespectively to the partial voltage drops. The SAT spike issandwiched between the two additional spikes corresponding to individual MAR pro-cesses occurring at junctions SN S and SN S for m , m = 2. The voltage positionsof these features depend on R /R . Right panel: The corresponding dV /dI ( V + V )for the two SN S and SN S junctions in series as they would have appeared in theabsence of the synchronization process, i.e. in the case where L C > ℓ ε . These dV /dI dependencies were calculated following [15] (with transmissivity W=1). We display the differential resistances as functions of the applied voltage,which demonstrate the singularities in Andreev transmission more profoundlythan the I - V curves. There is a pronounced SAT spike in the dV /dI foran SNSNS junction at V tot = 2 ∆/e . The spike appears irrespectively to thepartial voltage drops in the normal regions and is absent in the correspondingcurves representing two individual MAR processes at the junctions SN S andSN S.The resonant voltages of the SAT singularities can be found from theconsideration of the quasiparticle trajectories in the space-energy diagrams.
Synchronized Andreev Transmission in Chains of SNS Junctions 17
Such a diagram for the first subharmonic, n = 1 and ratio R /R = 3 / ε = − ∆ to traverse N , and the quasiparticle that starts fromthe central island S c with the same energy as the incident one to take up uponthe current across the island N , and hit S R with the energy ε = ∆ (the ABCDpath, the corresponding path for the hole is D ′ C ′ B ′ A ′ ). In general, relevanttrajectories yielding resonant voltages of Eq. (1.1) have the following structure:they start and end at the BCS quasiparticle density of states singular points( ε = ± ∆ ), contain the closed polygonal path, which include MAR staircasesin the normal parts and over-the-gap transmissions and Andreev reflections,and pass the density of states singular points at the central island. Apartfrom the main singularities [Eq. (1.1)], additional SAT satellite spikes appearat V = (2 ∆/e )( p + q ) /n , where p/q is the irreducible rational approximationof the real number r = R /R , (we take R < R ), and n > ( p + q ).The achieved qualitative understanding enables us to observe that themanifestations of the SAT mechanism in an experimental situation becomeseven more pronounced with the growth of the number of SNS junctions in thesystem. To see this, let us assume that the resistances of the normal islands ina chain of SNS junctions are randomly scattered around their average value R and follow Gaussian statistics with the standard deviation σ R = σR ,where σ is dimensionless. Accordingly, the dispersion of the distribution ofthe MAR resonant voltages is characterized by the same σ , and the MARfeatures get smeared. Let us distribute the voltage drop 2 ∆/e among the n successive islands. Then the quasiparticle SAT path starts at the loweredge of the superconducting gap at island j , traverses n − j + n -th islandin the chain. The standard deviation of the voltage drop on the n islandsgrows as √ n resulting in a voltage deviation per one island ∝ / √ n , i.e. thedispersion of the distribution of V n drops with increasing n : σ SAT = σ/ √ n . Incontrast to the MAR-induced features, with an increase of n , the subharmonicspikes at voltages V n per junction due to SAT processes become more sharpand pronounced. In conclusion, we have developed a nonequilibrium theory of the charge trans-fer across a central superconducting island in an SNSNS array and found thatthis island acts as Andreev retransmitter . We have shown that the nonequi-librium transport through an SNSNS array is governed by the synchronizedAndreev transmission with the correlated conversion processes at the oppositeNS interfaces of the central island. The constructed theory is a fundamentalbuilding unit for a general quantitative description of a large array consistingof many SNS junctions. , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur We thank A. N. Omelyanchuk for helpful discussions. The work was supportedby the U.S. Department of Energy Office of Science under the Contract No.DE-AC02-06CH11357, and partually by the RFBR 10-02-00700, Russian Pres-ident Science Support foundation mk-7674.2010.2, the Dynasty, Russian Fed-eral Programs and the Programs of the Russian Academy of Science.
A Recurrent relations for the quasiparticle currents anddistribution functions
The Kirchhoff’s laws applied for the circuit in Fig. 1.6 generate the followinglinear system of equations: I e ( ε − V ) = ¯ n (2R) h ,ε − ¯ n (2R) e ,ε − V ¯ R (2R)P ,ε − V + n F ( ε − V ) − ¯ n (2R) e ,ε − V ¯ R (2R) + ,ε , (1.46) I h ( ε ) = ¯ n (2R) h − ¯ n (2R) e ,ε − V ¯ R (2R)P ,ε − V + ¯ n (4)h − n F ( ε − V )¯ R (2R) + ,ε − V , (1.47) I h ( ε ) = ¯ n (2)h − ¯ n (2R) h R , (1.48) I h ( ε ) = ¯ n (1)h ,ε − ¯ n (2)h ,ε ¯ R D ,ε − u + ¯ n (1)e ,ε − u − ¯ n (2)e ,ε − u ¯ R B ,ε − u , (1.49) I e ( ε − u ) = ¯ n (2)e ,ε − u − ¯ n (1)e ,ε − u ¯ R D ,ε − u + ¯ n (2)h ,ε − ¯ n (1)h ,ε ¯ R B ,ε − u , (1.50) I h ( ε ) = ¯ n (1L) h − ¯ n (1C) h R (1.51) I h ( ε ) = ¯ n (1L) e − ¯ n (1L) h ¯ R (1L)P ,ε + n F ( ε ) − ¯ n (1L) h ¯ R (1L) + ,ε , (1.52)and Synchronized Andreev Transmission in Chains of SNS Junctions 19 I h ( ε + 2 V ) = ¯ n (2R) h ,ε +2 V − ¯ n (2R) e ¯ R (2R)P ,ε + V + ¯ n (2R) h ,ε +2 V − n F ( ε + V )¯ R (2R)P ,ε + V , (1.53) I e ( ε ) = ¯ n (2R) h ,ε +2 V − ¯ n (2R) e ,ε ¯ R (2R)P ,ε + V + n F ( ε + V ) − ¯ n (2R) e ,ε ¯ R (2R)P ,ε + V , (1.54) I e ( ε ) = ¯ n (2R) e ,ε − ¯ n (2C) e ,ε R , (1.55) I e ( ε ) = ¯ n (2C) e ,ε − ¯ n (1C) e ,ε ¯ R (12) D ,ε + u + ¯ n (2C) e ,ε +2 u − ¯ n (1C) e ,ε +2 u ¯ R (12) B ,ε + u , (1.56) I h ( ε + 2 u ) = ¯ n (1C) h ,ε +2 u − ¯ n (2C) h ,ε +2 u ¯ R (12) D ,ε + u + ¯ n (1C) e ,ε − ¯ n (2C) e ,ε ¯ R (12) B ,ε + u , (1.57) I e ( ε ) = ¯ n (1C) e ,ε − ¯ n (1L) e ,ε R , (1.58) I e ( ε ) = ¯ n (1L) e ,ε − ¯ n (1L) h ,ε ¯ R (1L)P ,ε + ¯ n (1L) e ,ε − n F ( ε )¯ R (1L)P ,ε . (1.59)Eqs.(1.46)-(1.59) are the recurrent relations (i.e. the relations coupling thefunctions at energy ε with the functions at ε ± V ) for the currents and thedistribution functions.It follows from Eqs.(1.48),(1.51) that I h ,ε [ R + R ] = ¯ n (2C) h − ¯ n (2R) h + ¯ n (1L) h − ¯ n (1C) h . (1.60)The distributions functions entering Eq.(1.60) we can express below throughthe currents. Combining Eqs.(1.49)-(1.50) we get,¯ n ( )h − ¯ n ( )h = (cid:2) I h ,ε ¯ R I ,ε − u + I e ,ε − u ¯ R D ,ε − u (cid:3) ¯ R D ,ε − u ¯ R B ,ε − u (cid:0) ¯ R B ,ε − u (cid:1) − (cid:0) ¯ R D ,ε − u (cid:1) . (1.61)At the same time from Eqs.(1.52),(1.59) follows that¯ n (1L) h ,ε = n F ( ε ) + ¯ R (1L) + ,ε I e ¯ R (1L) + ,ε − I h (cid:0) ¯ R (1L)P ,ε + ¯ R (1L) + ,ε (cid:1) R (1L) + ,ε + ¯ R (1L)P ,ε , (1.62)and finally from Eqs.(1.46)-(1.47) we get¯ n (2R) h ,ε = n F ( ε − V ) + ¯ R (2R) + ,ε − V − I e ,ε − V ¯ R (2R) + ,ε − V + I h (cid:16) ¯ R (2R) + ,ε − V + ¯ R (2R)P ,ε − V (cid:17) R (2R) + ,ε − V + ¯ R (2R)P ,ε − V . (1.63)Combining Eq.(1.60) and Eqs.(1.61)-(1.63) we find the recurrent relationfor the currents, Eq.(1.38). Similar procedure helps to derive Eq.(1.39). , , , T. I. Baturina , , A. Glatz , and V. M. Vinokur B Charge transport in SNS junctions
We discuss below the transport properties of SNS and SNN’ junctions to makea mapping between our technique and the well-known results obtained beforeus. The recurrent relations, Eqs.(1.38)-(1.39), solve the transport problem ina SNS junction in the incoherent regime. Then there is no island, so ρ ( ◦ ) = 0and we should remove the island resistances with the indices (1 C ) and (2 C )from the coefficient functions of the recurrent relations. So, R ( ε, − V ) I h ( ε ) − ρ ( ⊲ ) ( ε ) I e ( ε ) − ρ ( ⊳ ) ( ε − V ) I e ( ε − V ) = n F ( ε ) − n F ( ε − V ) , (1.64) R ( ε, V ) I e ( ε ) − ρ ( ⊲ ) ( ε ) I h ( ε ) − ρ ( ⊳ ) ( ε + V ) I h ( ε + 2 V ) = n F ( ε + V ) − n F ( ε ) . (1.65)where, for example, R ( ε, V ) = R + R + (1 / X α = ± { ¯ R ( ) α,ε + ¯ R ( ) α,ε + V } . (1.66)Eqs.(1.64)-(1.65) are invariant under the following transformation, I e ( ε − V ) → I h ( ε ), if at the same time we exchange the resistances, ¯ R ( ) ± ,ε ↔ ¯ R ( ) ± ,ε .Thus the relation, I e ( ε − V ) = I h ( ε ) and the reduction of the recurrent rela-tions to the one equation for I e or for I h as it was done in Ref. [15]: R ( ε, V ) I e ( ε ) − ρ ( ⊲ ) ( ε ) I e ( ε − V ) − ρ ( ⊳ ) ( ε + V ) I e ( ε + V ) = n F ( ε + V ) − n F ( ε ) , (1.67)holds only for a symmetric SNS junction with ¯ R ( ) ± ,ε = ¯ R ( ) ± ,ε .To summarize here our consideration [summarized by the recurrent rela-tions Eqs.(1.64)-(1.65)], reduces to that presented in Ref. [15] only in thecase where the contacts are symmetric and the assumption I e ( ε − V ) = I h ( ε )holds. References
1. A. F. Andreev, Zh. Eksp. Teor. Fiz. (1964) 1823 [Sov. Phys. JETP (1964)1228].2. J. M. Rowell and W. E. Feldmann, Phys. Rev. , 393 (1968).3. P. E. Gregers-Hansen, E. Hendricks, M. T. Levinsen, and G. R. Pickett, Phys.Rev. Lett. , 524 (1973).4. W. M. van Huffelen, T. M. Klapwijk, D. R. Heslinga, M. J. de Boer, andN. van der Post, Phys. Rev. B , 5170 (1993).5. A. W. Kleinsasser, R. E. Miller, W. H. Mallison, and G. B. Arnold, Phys. Rev.Lett. , 1738 (1994). Synchronized Andreev Transmission in Chains of SNS Junctions 216. E. Scheer, P. Joyez, D. Esteve, C. Urbina, and M. H. Devoret, Phys. Rev. Lett. , 3535 (1997).7. J. Kutchinsky, R. Taboryski, T. Clausen, C. B. Sørensen, A. Kristensen,P. E. Lindelof, J. Bindslev Hansen, C. Schelde Jacobsen, and J. L. Skov, Phys.Rev. Lett. , 931 (1997).8. A. Frydman and R. C. Dynes, Phys. Rev. B , 8432 (1999).9. T. Hoss, C. Strunk, T. Nussbaumer, R. Huber, U. Staufer, andC. Sch¨onenberger, Phys. Rev. B , 4079 (2000).10. T. I. Baturina, Z .D. Kvon, R. A. Donaton, M. R. Baklanov, E. B. Olshanetsky,K. Maex, A. E. Plotnikov, J. C. Portal, Physica B , 1860 (2000).11. Z. D. Kvon, T. I. Baturina, R. A. Donaton, M. R. Baklanov, K. Maex, E. B. Ol-shanetsky, A. E. Plotnikov, J. C. Portal, Phys. Rev. B , 11340 (2000).12. T. M. Klapwijk, G. E. Blonder, and M. Tinkham, Physica B+C (Amsterdam) , 1657 (1982).13. M. Octavio, M. Tinkham, G. E. Blonder, and T. M. Klapwijk, Phys. Rev. B ,6739 (1983).14. K. Flensberg, J. Bindslev Hansen, and M. Octavio, Phys. Rev. B , 8707(1988).15. E. V. Bezuglyi, E. N. Bratus’, V. S. Shumeiko, G. Wendin, H. Takayanagi, Phys.Rev. B , 14439 (2000). In this work superconductors were considered to bein a local equilibrium and the relation I e ( ε ) = I h ( ε − V ) was satisfied. Thisapproach was further developed by N. M. Chtchelkatchev, [22]; however in caseof geometrically non-symmetric SNS arrays, it results in an equivalent circuitwith the ennumerable number of elements.16. J. C. Cuevas, J. Hammer, J. Kopu, J. K. Viljas, and M. Eschrig, Phys. Rev. B , 184505 (2006).17. T. I. Baturina, Z. D. Kvon, and A. E. Plotnikov, Phys. Rev. B , 180503(R)(2001).18. T. I. Baturina, Yu. A. Tsaplin, A. E. Plotnikov, and M. R. Baklanov, JETP Lett. , 10 (2005).19. T. I. Baturina, D. R. Islamov, and Z. D. Kvon, JETP Lett. , 326 (2002).20. J. Fritzsche, R. B. G. Kramer, and V. V. Moshchalkov, Phys. Rev. B , 094514(2009).21. T. I. Baturina, A. Yu. Mironov, V. M. Vinokur, N. M. Chtchelkatchev,A. Glatz, D. A. Nasimov, A. V. Latyshev, Physica C (2009)doi:10.1016/j.physc.2009.11.10722. N. M. Chtchelkatchev, JETP Lett. , 250 (2005).23. G. Deutscher and D. Feinberg, Appl. Phys. Lett. , 487 (2000).24. A.I. Larkin and Yu.N. Ovchinnikov, Sov. Phys. JETP , 960 (1975); ibid , ,155 (1977).25. M. Yu. Kupriyanov and V. F. Lukichev, Zh. Eksp. Teor. Fiz. , 139 (1988)[Sov. Phys. JETP , 1163 (1988)].26. A.F. Volkov and T.M. Klapwijk, Phys. Lett. A , 217 (1992).27. K.K. Likharev, Rev. Mod. Phys. , 101 (1979).28. C.W.J. Beenakker, Phys. Rev. Lett. , 3836 (1991).29. M.Tinkham, Introduction to superconductivity, Mc.Graw-Hill Inc., 1996.30. N.M. Chtchelkatchev, JETP Lett. , 230 (2003). a ( E , V = . ))