ebook img

Voltage fluctuations on a superconductor grain attached to a quantum wire PDF

5 Pages·0.14 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Voltage fluctuations on a superconductor grain attached to a quantum wire

Voltage fluctuations on a superconductor grain attached to a quantum wire Masaki Oshikawa1 and Alexandre M. Zagoskin2 Department of Physics, Tokyo Institute of Technology1, Oh-oka-yama, Meguro, Tokyo 152-8511 JAPAN Department of Physics and Astronomy, University of British Columbia2, Vancouver, BC V6T1Z1 CANADA (October 15, 1998) 9 9 9 1 ations will translate into fluctuating Andreev boundary Whenafinitesuperconductorisincontactwitha1Dnor- condition [8]. n mal conductor, superconducting phase fluctuations lead to a First let us present a heuristic picture of the phe- power-lawresponseofthenormalsubsystem. Asaresult,the J nomenon at zero temperature. The effect of a sud- charge fluctuations on the superconductor at zero tempera- 3 den change in a boundary condition is a slow, power- ture have logarithmic correlator and a 1/ω power spectrum 1 law relaxation of the normal system to its new ground (1/f-noise). At higher temperatures 1/ω is pushed to the high frequency region, and 1/ω2 behavior prevails. state, that is, Fermi edge singularity. It is conveniently ] n described in terms of boundary conformal field theory, o which allows a unified description of FES and Ander- c son orthogonality catastrophe [11]. Both are character- - r ized by scaling dimension x of the boundary condition p Correlated character of transport in mesoscopic sys- changing operator , so that the correlation function u tems strongly influences character of current and charge †(τ) (0) τ−2Ox, and the dimension x itself is re- s fluctuations there. Examples are non-Poissonian shot O O ∼ t. (cid:10)latedtothe(cid:11)finite-size(O(1/L))correctiontotheground a noise in quantum point contacts and NS junctions( [1] state energy shift due to the action of : m and references therein) and giant current noise in super- O - conducting quantum point contacts [2–4]. Noise mea- L d surements may provide information about quantum cor- x= π∆E1/L. (1) n relations unattainable by any other means [5]. o Hereafterweset¯h=1andthevelocityoftheexcitations In this paper we will show that equilibrium fluctua- c (Fermi velocity in the non-interacting case) v =1 unless [ tions in a mesoscopic NS system reveal the many-body statedotherwise;Listhecharacteristicsizeofthesystem reactionofaconductingsystemtoasuddenexternalper- 3 (wire length in our case). This relation holds both for v turbation, thus linking the noise to such phenomena as usual potential scattering and Andreev scattering [9]. 3 Fermi edge singularity (FES) [6] and Anderson orthog- Theboundaryconditionchangingoperatorinourcase 0 onality catastrophe [7]. Unlike the standard situation 3 when the perturbation is due to creation of a scatterer, changes the phase of superconductor by Φ: OΦ = 0 exp[Φ ∂ ]=exp[iΦnˆ].Hereweusedtherelationbetween here the normal subsystem reacts to the change in the ∂φ 1 φ and the Cooper pair number operator on the grain, 8 off-diagonal pairing potential in the superconductor, to nˆ = 1 ∂ , whichholds if n 1[10]. Inagreementwith 9 which it is coupled via Andreev reflection processes on i∂φ h i≫ general CFT argument, the Green’s function of is / the NS boundary [8,9]. OΦ t ma ItMcoonrseisstpseociffiacamllye,socoscnospidicersuthpeerscyosntedmuctsihnogwgnraininFSig.i1n. GΦ(τ)=DOΦ†(τ)OΦ(0)E≡De−iΦnˆ(τ)eiΦnˆ(0)E=(τ/τ0)−2x, d- contact with a 1D normal wire N. We assume that the (2) grain contains a large number of Cooper pairs, n 1, n h i≫ o and its capacitance C is large enough (so we are in the with some cutoff parameter τ0. Though by itself GΦ has c oppositelimittotheCoulombblockaderegime[10]),but no direct physical meaning, it is related to the correla- v: the contact region is smaller than the Josephson screen- tor of charge fluctuations on the grain: KQ(τ)/(4e2) = i ing length λ , so that both the modulus and phase of [δnˆ(τ),δnˆ(0)] /2, where δnˆ =nˆ nˆ . Obviously, X J + −h i the order parameter ∆ = ∆eiφ can be taken constant (cid:10) (cid:11) ar tqhuearsei.paTrteimclpeeerxactiutraetiiosnlsowin|etnh|oeusguhp,eTrco≪nd|u∆ct|,ort.oFniengallelcyt, KQ(τ)4−e2KQ(0) ≡ 12 [δnˆ(τ),δnˆ(0)]+ − δnˆ2 = (cid:10) (cid:11) (cid:10) (cid:11) we assume that the 1D conductor is scattering free, and 1 [nˆ(τ),nˆ(0)] nˆ2 = lim 21(GΦ(τ)+GΦ(τ)∗)−1. (3) its length L ξ0, where ξ0 is the superconducting co- 2 + − Φ→0 Φ2 ≫ (cid:10) (cid:11) (cid:10) (cid:11) herence length. Chargefluctuationsaremeasuredbye.g. monitoringthe If isolated, the superconducting phase in the grain electrostatic potential of the grain. wouldfluctuate, the system being in the state with fixed The dimension x is easily found [9] if notice that the particle number (see e.g. [10], Ch.7). To the 1D conduc- energy correction entering (1) is simply the Josephson torbroughtincontactwiththegrain,thesephasefluctu- energy of a ballistic 1D SNS junction of length L, with 1 phase difference Φ [12,13]at zerotemperature: E (Φ)= √2e∂θ(x,t) J J(x,t)= . (5) Φ2/(4πL), so that x = LE (Φ)/π = [Φ/(2π)]2. Sub- − π ∂t J stituting this and (2) into (3), we find for the charge Therefore the electric charge of the superconductor at fluctuations on the grain time t is given by K (τ) 1 τ Q = δnˆ2 ln . (4) t √2e 4e2 − 2π2 τ Q(t)=Q( ) J(0,t)dt= θ(0,t), (6) (cid:10) (cid:11) 0 −∞ −Z π −∞ For the spectral density of charge fluctuations, S (ω)= Q ∞ where we have defined Q and θ so that Q = θ(0,t) = 0 2 K (τ)cos(ωτ)dτ, suchlogarithmicbehaviorofcor- 0 Q corresponds to the neutral superconductor grain. Simi- reRlator translates into SQ(ω) 1/ω (1/f-noise) in the ∝ lar identification was used in studies of Coulomb Block- corresponding frequency interval; it is often observed in ade [18], although our application is to quite different spin glasses, magnetic polycrystals and other systems regime. with wide distributionofrelaxationtimes [14,15]. Inour IncludingthechargingenergyassociatedwithQ,E = case the source of slow relaxation times is the normal Q system with its power-law approach to the new ground Q2C2 = πeC22θ(0,t)2, we arrive at the effective action of the state. Theideaofinvokinginfraredsingularitytoexplain problem: scaling that underlies current 1/f-noiseobservedin nor- 1 u malconductorswasexploitedine.g. modelof“quantum S = dtdx (∂ Θ)2 dt Θ(0,t)2. (7) µ Z Z 2 −Z 2 1/f-noise” due to Bremßtrahlung, but it encountered fa- tal difficulties (see [15] and references therein). In con- Here Θ = θ/√πK and u = 2e2K/(πC) (u = trast, we see no fundamental difficulty in the present 2e2K/(π¯hC) in conventionalunits). This is a free boson mechanism, which will be shown to be consistent with action with a mass term only on the boundary, and this fluctuation-dissipation theorem. effectivetheoryisexactlysolvable. Thecalculationofthe Expression (4) is obviously only an intermediate charge fluctuation in the superconductor is reduced to asymptotics for 1 τ/τ exp[2π2 δnˆ2 ]. Therefore ≪ 0 ≪ the calculation of the correlation function of the bound- we needamoreconsistentapproach,ta(cid:10)king(cid:11)intoaccount aryfieldΘ(0,t). Thephysicsofthe problemiscontained effectsofchargingenergyandfinitetemperature. Wewill in a proper choice of the boundary condition at x = 0 seethatthe regionof1/f-fluctuationsis pushedbyther- (on the NS boundary) [19]. mal effects to high frequencies, and 1/ω2-type spectrum The Matsubara Green’s function of Θ (for τ = it) prevails except at very low temperature. (x,τ)= T Θ(x,τ)Θ(0,0) , isthe Green’sfunctionof We will describe the wire in terms of Tomonaga- G −h τ i theLaplacianontheplane,withtheboundaryconditions Luttingerliquid(TLL),whichallowsustotreatsimulta- ∂ Θ(x=0,τ)=uΘ(x=0,τ)andΘ(x,τ+β)=Θ(x,τ), neously non-interacting (1D Fermi-liquid) and interact- x where β = 1/T is the inverse temperature. As a conse- ing case of such S-TLL system [16]. Though as we will quence, the Green’s function for the source at (x ,0) see interactions do not change qualitative picture of the G 0 satisfies phenomenon, they may lead to interesting side effects. Spin and charge degrees of freedom in TLL separate. (∂ 2+∂ 2) (x,τ)=δ(x x )δ(τ) (8) x τ 0 Since spin is totally reflected by NS boundary both in G − ∂ (x=0,τ)=u (x=0,τ). (9) the process of ideal Andreev reflection and ideal normal xG G reflection, it is irrelevant and will be dropped. For the Because we are interested in the correlation function of charge, the Lagrangiandensity is =1/(2πK)[(∂ θ)2 t the boundary, we take the limit x 0. Integrating L − 0 (∂ θ)2],whereK isthe couplingconstant,whichisunity → x (8) over x from 0 to x +0, we see that the boundary 0 for non-interacting electrons. condition (9) must be now modified: Physically, the boson field θ(x,t) corresponds to the phase of Charge Density Wave (CDW), while its dual ∂ (x=0,τ)=u (x=0,τ) δ(τ). (10) x G G − φ(x,t) to the superconducting phase of the Cooper pair [6]. For a perfect NS interface, the boundary value of For Fourier components we obtain an ordinary differen- the superconducting phase of the TLL should be equal tialequation: ∂2 (x,ω )=ω 2 withtheboundarycon- xG n n G to that of the superconductor, which in the absence of dition∂ (x=0,ω )=u (x=0,ω ) 1.Imposing the x n n G G − phase fluctuations gives a Neumann boundary condition asymptotic condition lim (x,τ) 0, we find x→∞ G → forthefieldθ(x,t)inthewire[17],equivalenttoaDirich- 1 let boundary condition on φ: φ(0,t)=Φ=const. (x,ω )= − e−|ωn|x. (11) n G u+ ω ThechargeQonthegraincanberelatedtothebound- n | | ary value of the “CDW” field θ. Indeed, the current in The Matsubara Green’s function for the boundary field TLL can be written as is (x=0,ω )= 1/(u+ ω ). n n G − | | 2 Thecorrelationfunctions inthe realtime areobtained istheGreen’sfunctionoftheLaplacianwiththesourceat by ananalytic continuationfromthe MatsubaraGreen’s x . The1/ωfluctuationofthegrainvoltagetranslatesto 0 function. In a standardway, we obtain the retarded and anω-linearfluctuation of the current,althougha similar advanced Green’s functions, GR(A)(ω) = 1/(u iω). behaviorpersists evenfor the currentfar fromthe grain. − ∓ Their mismatch at the real axis, which is pure imagi- So far, we have considered a semi-infinite wire con- nary, gives the spectral density of the system: ρ(ω) = nected to the grain. In the remainder of this paper, we 2ImGR(ω) = 2ω/(u2+ω2). The spectrum of electro- discuss how our results are affected by wire being finite. − static potential fluctuations on the grain is given by: Instead of manufacturing a very long wire, it would be easier to connect the other end of wire to a normal S (ω) ∞ 1 S (ω)= Q = Q(t)Q(0)+Q(0)Q(t) eiωtdt reservoir. Weassumethatthejunctiontothereservoiris V C2 Z−∞ 2C2h i adiabatic(nobackscattering),andthephaseinformation 2Ke2 1 βω 2Ke2 ω βω is immediately lost in the reservoir (perfect thermaliza- = ρ(ω) coth( )= coth( ). (12) πC2 2 2 πC2 u2+ω2 2 tion). Then the situation is not qualitatively different fromthecaseofthesemi-infinitewire,sincetheparticles This result has the form expected from fluctuation- can leave to infinity (reservoir), and the usual reason- dissipation theorem (FDT) [20]. ing of Landauer formalism applies [21]; the transport of In the low temperature/high frequency limit βω 1, ≫ the chargefrom/tothe graincontrolledbythe geometric the spectrum is proportional to ω /(u2 +ω2). At fre- | | restriction (narrow 1D channel) implies that our theory quency much higher than the “infrared cutoff” u, the basedon1Dholds,eventhoughthe systemis notpurely spectrum is indeed 1/ω as expected from the heuristic 1D. Thus we expect our results for the semi-infinite wire argument (4). On the other hand, at smaller ω fluctua- applies to the more realistic case of the finite wire con- tionsaresuppressed,becausefinitecapacitanceobviously nected to a normal reservoir, if the interaction effects in excludeslargefluctuationofthechargeinthegrain. The the wire can be neglected. 1/ω spectrum, which is independent of the interaction When the interaction in the wire is not negligible, the parameterK. Itautomaticallyarisesfromthefreeboson problem is more subtle, since a quasiparticle excitation fieldtheory oncethe bosonfielditself is identifiedwith a inside the wire is a collective motion of many electrons physical observable, because the correlation function of andcannotbe simply transferredto the reservoir. While thebosonfieldislogarithmic. Wehavethusshownanex- arigoroustheoryofjunction betweenaTLL andnormal ample of “quantum 1/f noise” naturally derived within reservoir is not yet established, it has been studied by the framework of equilibrium statistical mechanics and modelling the reservoir with a (semi-infinite) 1D Fermi fieldtheory. Althoughthe 1/ω behavioris onlyexpected liquid. As a simplest model, we study a finite wire of at very low temperature and high frequency, we believe lengthL,inwhichthecouplingconstantisK,attachedto this presentscertaininterestbothfromthe pointofview semi-infinite wire with coupling constant K′, K αK′. of translationof infrared divergences into 1/f-spectrum, ≡ Tomodelthereservoirbya1DFermiliquid,K′shouldbe and of its destruction by thermal fluctuations. unity but we shall obtain the solution for general values At higher temperature/lower frequency βω 1, the ≫ of K′ because it can be applied also to other interesting spectrum is proportional to T/(u2+ω2). The spectrum cases. It has been found that normal and Andreev-type is(restoringthefundamentalconstants)4kTKe2/(π¯hω2) reflections occur at the boundary if K =K′ [22]. for frequency larger than inverse capacitance ω u. In 6 ≫ Byasimilarlineofargumentsasbefore,weobtainthe the opposite limit of small frequency ω u, (but still ≪ Matsubara Green’s function of the boundary field Θ as ωβ 1),thefluctuationissimplyawhitenoisewiththe ≫ magnitude 2R kT/(4K) where R = h/e2, again in a H H (0,ω )= n complete agreementwithFDT.The transitionfromlow- G [cosh(ω L)+αsinh(ω L)] to high- temperature behavioroccurs at βω 1, namely − n | n| (14) ∼ (u+αω )cosh(ω L)+(ω +αu)sinh(ω L) kT hf (see Fig.2). For T = 10mK, the boundary is | n| n | n| n ∼ at f 200MHz. This would be within the capabilities Since the temperature effect is given by the same FDT ∼ of modern experiment; we expect the crossover between factor coth(βω/2)/2 as in the semi-infinite wire case, 1/ω2 and 1/ω behavior could be directly observable. we concentrate on how the spectral density ρ(ω) = The charge fluctuation in the grain means the current 2ImGR(ω) is affected. flowing through the wire also fluctuates. The correla- − For α = 1, the result reduces to the semi-infinite wire tion function of the currentcan be obtained by a similar case, as it should be. For α 1, the overall shape of method. Forexample,theMatsubaraGreen’sfunctionof ≈ thespectraldensityremainsthe same. However,avaria- ofthecurrentatJ(x )isgivenby(2e2ω2/π2) (x ,ω ), 0 n Gx0 0 n tion with the period 1/L is superimposed on the depen- where dence for semi-infinite wire (Fig.3). This is the resonant 1 u ω structuredue to (normalandAndreev)reflectionsatthe (x ,ω )= 1 −| n|e−2|ωn|x0 (13) Gx0 0 n −2ω (cid:20) − u+ ω (cid:21) boundarybetweentwowires. Ifthejunctionbetweenthe n n | | | | 3 wire and the reservoir is smooth and not abrupt, we ex- pectthattheresonantstructureissmearedout;thenthe observedspectrumwouldbequiteclosetotheprediction for the semi-infinite wire. If α far from 1, the resonant structure dominates. In [1] M.D.M. de Jong and C.W.J. Beenakker, in: ”Meso- particular,thecaseK′ 0(orK′ )isequivalentto scopic Electron Transport,” edited by L.L. Sohn, L.P. → →∞ a finite wire with open end (or a finite wire attached to Kouwenhoven, and G. Schoen, NATO ASI Series Vol. a very large superconductor reservoir), respectively. In 345(KluwerAcademicPublishers,Dordrecht,1997),pp. these limits α , (α 0) the spectral density devel- 225-258. →∞ → ops into a set of discrete δ-functions with peak spacings [2] A. Mart´ın-Rodero, A. Levy Yeyati, and F.J. Garc´ıa- of order 1/L, rather than a continuous function of ω. Vidal, Phys.Rev.B53, R8891 (1996). This reflects the finiteness of the system. In the L [3] D. Averin and H.T. Imam, Phys. Rev. Lett. 76, 3814 limit the δ-functions become dense and converge t→o th∞e (1996). [4] P. Dieleman et al., Phys. Rev.Lett. 79, 3486 (1997). smooth function we have obtained for the semi-infinite [5] M.Bu¨ttiker,Phys.Rev.Lett.68,843(1992);G.B.Leso- wire. vik and L.S. Levitov, Phys. Rev. Lett. 72, 538 (1994); It is alsoinstructiveto take the opposite limit ofshort T. Gramespacher and M. Bu¨ttiker, Phys. Rev.Lett. 81, wire L 0. For the open end case (K′ 0), the reso- 2763 (1998). → → nancesarefoundatω (2n 1)π/(2L)wherenisanat- [6] G.D. Mahan, Many-particle physics. New York: Plenum ∼ − uralnumber. Forthewireconnectedtothesuperconduc- Press (1990). tor reservoir, in the short-wire limit, the resonances are [7] P.W. Anderson,Phys.Rev.Lett. 18, 1049 (1967); Phys. found at ω nπ/L, n = 1,2,.... The resonance which Rev. 164, 352 (1967). would corres∼pond to n=0 is actually affected by the fi- [8] A.F. Andreev,Sov. Phys.JETP 19, 1228 (1964). nitecapacitance,anditslocationisgivenbyω =1/√LC. [9] A.M. Zagoskin and I.Affleck,to be published. This can be simply understood from the following ar- [10] M. Tinkham, Introduction to superconductivity (2nd ed.), New York: McGraw Hill (1996), Ch.7. gument: from the conjugate relation between the phase [11] I. Affleck and A.W.W. Ludwig, J. Phys. A: Math. Gen. and the charge, the effective Hamiltonian of the system 27, 5735 (1994). is givenby (see e.g.[10])Hˆ =−21C(∂∂Φ)2+21L(Φ−Φ0)2, [12] G. Ishii, Progr. Theor. Phys. 44, 1525 (1970). where Φ is the superconducting phase of the grain, the [13] A.V. Svidzinski˘ı, T.N. Antsygina, and E.N. Bratus’, firstandsecondtermrepresentthechargingenergyofthe Zh.E´ksp.Teor.Fiz. 61, 1612 (1971). grainand the Josephsonenergyof the wire,respectively. [14] Sh.M. Kogan, Sov.Phys. Usp. 28, 170 (1985). This is a Hamiltonian of a harmonic oscillator with the [15] P.DuttaandP.M.Horn,Rev.Mod.Phys.53,497(1981). eigenfrequency 1/√LC, exactly where the resonance is [16] D.L. Maslov, M. Stone, P.M. Goldbart, and D. Loss, found. Phys.Rev. B 53, 1548 (1996); R. Fazio, A.A. Odintsov Inconclusion,wehaveshownthatquantumphasefluc- and F.W.J. Hekking,ibid.,6653. tuations in a superconducting grain in contact with a [17] M. Oshikawa and A.M. Zagoskin, unpublished. [18] K.A. Matveev, Phys.Rev. B 51, 1743 (1994). normal wire are translated into equilibrium charge fluc- [19] When thereare more than one channels in thewire, the tuations in the system, with specific spectrum demon- action isasum offree boson ones, asfar asthewire can strating transition from 1/ω2 to 1/ω behavior at very be treated as collection of independent TLL’s. The volt- low temperature. The mechanism of fluctuations is re- agefluctuationisalsogivenbythesumofsingle-channel action of the normal subsystem to sudden change of the contributions.Inparticular,ifthereareN identicalchan- off-diagonalpairingpotentialontheNSboundary(Fermi nels, the fluctuation is multiplied by N, and it suffices edgesingularity),leadinginthelow-temperaturelimitto to consider the single-channel case. (If interchannel in- quantum1/f noise. Similareffectcanbeexpectedtoex- teractions cannotbeneglected ,thissimplepictureisno istalsoinsystemswithsuperconductinggrainincontact longerapplicableforwireswithN >1duetoappearance with 2D conductors under appropriate conditions. of collective interchannelexcitations [Ya.M. Blanter and While the experimental observation of this effect M.Bu¨ttiker, Europhys.Lett. 42, 535 (1998)].) [20] E.M. Lifshits and L.P. Pitaevski˘ı, Statistical Physics, presents a certain challenge, we hope it to be possible Part 2, Pergamon Press: Oxford et al. (1980). in near future. Concerning the experiments, impurity [21] Y.Imry,Introductiontomesoscopicphysics(OxfordUni- scatterings and physical effects of probing will require versity Press, 1997). further theoreticalstudy, whichis howeverbeyondscope [22] I.SafiandH.J.Schulz,cond-mat/9803326andreferences of the present paper. therein. WearegratefultoI.Affleck,N.KokuboandS.Okuma for valuable discussions andto M. Bu¨ttiker and R. Fazio for helpful comments on the manuscript. 4 FIG.1. SuperconductinggrainS incontactwitha1Dcon- ducting wire N. Tip P probes the electrostatic potential of thegrain. 103 3 102 2 10 1 1 _S 0 g l _ -1 T = 0.1 -2 1/w_2 1/w_ -3 -4 -1 0 1 lgw_ 2 3 4 FIG. 2. Reduced spectral density of voltage noise in a grain, S¯(ω) = SV(ω)/[Ke2/πc2] (Eq.(12)), as a function of ω¯ ≡ω/ufor different values of T¯≡T/u. 1 0.5 0 -0.5 -1 -1.5 -1 -0.5 0 0.5 1 FIG.3. Resonantstructureofspectraldensityforα=0.1 (dots), α= 1 (solid line), and α= 10. Dimensionless length ofthewireL=1; T¯=1. Dashedstraightlineillustrates1/ω¯ dependence. 5

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.