Kondo e(cid:27)e t with non ollinear polarized leads: a numeri al renormalization group analysis 1 2 3 4 P.1Simon , P. S. Cornaglia , D. Feinberg , and C. A. Balseiro Laboratoire de Physique et Modélisation des Milieux Condensés, 7 2 CNRS and Université Joseph Fourier, BP 166, 38042 Grenoble, Fran e 0 Centre de3Physique Théorique, É ole Polyte hnique, CNRS, 91128 Palaiseau edex, Fran e 0 Laboratoire d'Etudes des Propriétés Ele troniques des Solides, CNRS, 2 asso iated to U4niversité Joseph Fourier, BP 166, 38042 Grenoble, Fran e and n Instituto Balseiro and Centro Atómi o Barilo he, a Comisión Na ional de Energía Atómi a, 8400 Barilo he, Argentina J (Dated: February 6, 2008) 0 The Kondo e(cid:27)e t in quantum dots atta hed to ferromagneti leads with general polarization 3 dire tions is studied ombining poor man s aling and Wilson's numeri al renormalization group methods. Weshowthatpolarizedele trodeswillleadingeneraltoasplittingoftheKondoresonan e ] l inthequantumdotdensityofstatesex eptforasmallrangeofangles losetotheantiparallel ase. l a We also show that an external magneti (cid:28)eld is able to ompensate this splitting and restore the h unitarylimit. Finally,westudytheele troni transportthroughthedevi einvariouslimiting ases. - s PACSnumbers: 75.20.Hr,72.15.Qm,72.25.-b,73.23.Hk e m t. I. INTRODUCTION Mostofthepreviousworkshavefo usedquitesurpris- a inglyonlyonparallelandantiparallelpolarizations. Ser- m 10 gueev et al. onsidered non ollinear polarizations but One of the most remarkable a hievements of re ent - didnottakeintoa ountthespinsplittingofthedoten- d progress in nanoele troni s has been the observation of ergy level whi h naturally gives rise to spurious results. n the Kondo e(cid:27)e t in a single semi ondu tor quantum dot 19 o (QD).1,2 Sin e these early experiments, the Kondo ef- In a re ent preprint, Swirkowi z et al. analyze the non c 3,4,5 ollinear situation using the equation of motion method fe thasalsobeenobservedin arbonnanotubeQDs, [ 6 supplemented by an e(cid:27)e tive ex hange (cid:28)eld that has to and mole ular transistors onne ted to normal metalli bedeterminedself- onsistently. Su hmethodgivesqual- 2 leads. The su ess of these experiments has opened the T TK TK v way to a systemati analysis, both theoreti ally and ex- itative results for temperature ∼ , where is the 4 Kondotemperature. Themainresultisthatthesplitting perimentally,ofthee(cid:27)e tsoftheleads orrelationsonthe 9 of the zero bias Kondoanomalyde reasesmonotoni ally Kondoe(cid:27)e t. Ofparti ularinterestaretheexperimental 2θ 7 when in reasingthe angle between the two lead mag- 7 realizations of the Kondo e(cid:27)e t onne ted to super on- 7 8 netizations. In this paper, we go beyond the equation 0 du tingleads ortoferromagneti ele trodes. Thelatter of motion method by ombining a poor man s aling ap- 6 isparti ularlyimportantsin eitis onne tedto thenew 0 emerging(cid:28)eld alled(cid:16)spintroni s(cid:17).9 Onegoalofspintron- proa h and the numeri al renormalization group (NRG) t/ i s is to provide devi es where the transport properties approa h. The poormans alingprovidesus withan an- a are ontrolled by magneti degrees of freedom. In this alyti alexpression forthe Kopndotemθperature asa fun - m tion of the lead polarization and [see Eq.(12)℄ while sense, aKondoquantum dot onne tedto ferromagneti - ele trodes may be regarded as an interesting setup to NRG allows us to ompute dot spe tral densities and d the near equilibrium transport properties like the on- study intera tion e(cid:27)e ts in spintroni s devi es. n du tan e. The planofthepaperisthefollowing: InSe . o From the theoreti al point of view, it has been pre- II we des ribe our model Hamiltonian. In Se . III, we c di ted, despite some early ontroversialresults,10,11,12,13 : use poor man s aling to estimate the Kondo tempera- v that ferromagneti ele trodes with parallel polarization ture. In Se . IV, we present the NRG resultsfor the dot 14,15,16 i produ easplittingoftheKondozerobiasanomaly X density of states and for the ondu tan e. Finally, Se . whereas polarized ele trodes with antiparallel polariza- V ontains a brief summary of the results. r 14,16,17 tion generally do not. Some of these predi tions a 60 have been he ked experimentally by onta ting C 8 mole uleswithferromagneti ni kelele trodes. The un- derlying me hanism responsible of the splitting of the Kondo resonan e is quite lear: virtual quantum harge II. MODEL HAMILTONIAN (cid:29)u tuations between the dot and the leads be ome spin dependent. As a result, an e(cid:27)e tive ex hange magneti In order to des ribe this system we use the following (cid:28)eld appears in the dot whi h generally lifts the spin general Hamiltonian degenera y on it. For antiparallel on(cid:28)gurations, the ontributions of ea h ele trode to the e(cid:27)e tive (cid:28)eld may H =H +H +H , leads tun dot eventually an el ea h other for symmetri ouplings. (1) 2 H leads 14 where denotes the Hamiltonian of the ondu tion in the leads. The spin degenera y of the Anderson en- ele trons in the leads and is de(cid:28)ned by ergy levels is therefore lifted with an e(cid:27)e tive Zeeman Hleads = εk±c†k,γ,±ck,γ,±. (2) splitting δε =ε ε pΓcos(θ)ln(D/Λ). k,γ,± d d↑ d↓ X − ∼ (6) c† Iεn this equaγtio=n Lk,,Rγ,± reates an ele tro~nn with en~nergy This splitting prevents from rea hing of the strong ouB- k± in lead with spin along ± γ where γ is pling regime. Nonetheless, an external magneti (cid:28)eld a unitarγy ve tor parallel to the~nmagnet~nization2θmoment an be applied lo ally to ompensate tεh↑is inteεrn↓al (cid:28)eld in lead . The angle between L and R is . Sin e and restore the degenera y between and . One the leads are assumed to be polarized, there is a strong an then perform a S hrie(cid:27)er-Wol(cid:27) transformation. The ρ (ω) spin asymmetry in the density of states γ± . The Kondo Hamiltonian so obtained is similar to the usual tunneling jun tions between the leads and the dot may Kondo Hamiltonian: 1 be des ribed by a standard tunneling Hamiltonian H = (Jz↑c† c Jz↓c† c )Sz Htun = (tγc†k,γ,σdσ+H.c.), (3) K 2k,kX′,α,β kα↑ kβ↑− kα↓ kβ↓ k,γX,σ=↑,↓ + J⊥(c†kα↓ckβ↑S++c†kα↑ckβ↓S−). (7) d σ = σ where destroys an ele tron in the dot with spin ↑ , The bare values of the Kondo ouplings are ↓. Herethepossiblevaluesforthedotspinarequantized ~e = (~n +~n )/~n +~n calong the axis z L R | L R|. The ocperator J⊥ =Jz↑ =Jz↓ 2t2 U =J, k,γ,σ is a linear ombination of the operators k,γ,s=± ∼ (cid:18)(|εd|(εd+U)(cid:19) (8) and reads: (cid:18)ccαα,,kk,,↑↓ (cid:19)=(cid:18)csoins((θθ22αα)) −cosisn((θ2αθ2α))(cid:19)(cid:18)ccαα,,kk,,+− (cid:19), (4) pwehreproesεitdio∼n εodf↑ψ∼L+εda↓n.dNψoLte−.alOsonteh aatnψtLh↑enispaerlifnoermartshue- se ond stage of the RG analysis dire tly on the Kondo θ = θ H where α ± here. dot is the usual Anderson Hamil- Hamiltonian. The RG equations an be obtained by in- tonian and reads tegratingΛo0utthεedhighΛenergydegreesoffreedombetween H = ε d†d +Ud†d d†d gµ BSz, a ut-o(cid:27) ∼ and and read: dot d,σ s s ↑ ↑ ↓ ↓− B (5) dλ 1+p σX=↑,↓ z↑ = 2 cos2(θ/2)+ sin2(θ/2) λ2, dlnl 1 p ⊥ (9) where Sz = (d†↑d↑ − d†↓d↓). The last term of Eq. (5) dλz↓ = 2(cid:20)cos2(θ/2)+ 1−−psin2(θ/2)(cid:21)λ2, des ribes the Zeeman energy splitting of the dot. In dlnl 1+p ⊥ (10) (cid:20) (cid:21) what follows we mainly onsider a symmetri situation dλ 1 p in wΓh=i hΓLtγ+=ΓtR. We deΓnLot=e tπhte2(tρo+ta+l tρu−n)neling strength dln⊥l = (cid:20)cos2(θ/2)+ 1+−psin2(θ/2)(cid:21)λ⊥λz↑ by with . The e(cid:27)e tive 1+p magneti (cid:28)eldsgeneratedbytheferromagneti ele trodes + cos2(θ/2)+ sin2(θ/2) λ λ , ~e =(~n +~n )/~n +~n ⊥ z↓ z L R L R 1 p (11) isthenalongtheaxis | |. Wenext (cid:20) − (cid:21) perform a poor man s aling analysis. where wλ⊥e h=av√eρi+ntρr−oJdu⊥ eλdzt↑h/↓e=dimρ±enJszi,o↑/n↓less Klon=doΛ0 /oΛu- plings , and . III. POOR MAN SCALING ANALYSIS The dimensionless Kondo ouplings are all driven to T K strong oupling at the same energy s ale whi h an The analysis extends in a straightforward manner the bedeterminedexpli itlybyintegratingouttheaboveRG one developed in Ref. [14℄. In this part we negle t the equations. The Kondo temperature an be expressed as: energyρdependρen e in the density of states and also sup- arctanh(pcos(θ)) ppo=se(ρ+γ± ρ≈−)/±(ρ.+I+nρt−h)e ferromagneti leads, the ratio TK(p,θ)=Λ0exp − 2λ0pcos(θ) , (12) − isin generaldi(cid:27)erentfrom zero (cid:18) (cid:19) and as shown bellow it is one of the relevant parame- λ0 =J0(ρ++ρ−) ters to des ribe the e(cid:27)e t of the spin polarization. For where . Thisexpressiongeneralizesthe θ an energy dependent density of states, the situation is one obtained in Ref. [14℄ to all values of . The depen- p,θ more omplex as will be shown in Se . IV. One per- den e of the Kondo temperature with appears as a pcos(θ) forms a two stage renormΛalizationDgroup (RG) analy- fun tion of the singleθparTameter and istherefTor0e sis by redu ing the uto(cid:27) from the bandwidth to auniformfun tionof . K takesitsmaximumvalue K ε¯ =ε +ε p θ = π/2 d d↑ d↓ whi h renormalizesthe parameters of the independent of at whi h orresponds to an- ε ε T θ = 0 ↑ ↓ K Andersonmodel. Ingeneral, and renormalizedi(cid:27)er- tiparallel magnetizations and is minimum for ently due to the spin dependent density of states (DOS) as expe ted. 3 f(ω) IV. NUMERICAL RENORMALIZATION where istheFermifun tion. Notethatthedensities GROUP ANALYSIS ofstatesintheleadsarenolonger onstantwhi hismore 20 realisti to des ribe ferromagneti leads. For the sake In order to ompute dot spe tral densities and trans- ofsimpli ity, from hereonwetakethe hemi alpotential portpropertieswepro eedwithaNRGanalysis. Atthis equal to zepro. Note that for this parti ular hoi e the pointitis onvenienttoperformtwounitarytransforma- parameter de(cid:28)ned in the previous se tδioǫn is zero while tions on the initial Hamiltonian de(cid:28)ned in Eq. (1). We the spin polarization is not. For small anPd tem4δpǫera- (cid:28)rst make an even/odd transformation introdu ing the tures the magnetization is simply given by ≃ πD. In operators the present ase, the most important e(cid:27)e t of Bthe leads int c =(c c )/√2. magnetismistheo urren eofaninternal(cid:28)eld that k,e/o,± k,L,± k,R,± ± (13) reatesthe e(cid:27)e tive Zeeman splitting alreadymentioned above. We next introdu e a new basis de(cid:28)ned by To al ulatethedotdensityofstatesweneedthelo al bk,e,↑/↓ =ck,e,± , bk,o,↑/↓ =ck,e,∓. Green fun tion. In the standard form, we have (14) G−1(ω)=G−1(ω) Σ(ω), σ 0σ − (19) The tunneling Hamiltonian an be written in the new basis as: Σσ(ω) G0σ(ω)−1 where is the self-energyand isthe non- H = (t b† d +H.c.) tun η kη,σ σ (15) intera ting Green fun tion k,ηX=e/o,σ 1 G0σ(ω)= . te = t√2cos(θ/2) to = t√2sin(θ/2) ω−ǫd,σ−gµBσ−∆σ(ω) (20) where , and . The θ ru ial point is that the angle arising due to the non ∆σ(ω) The hybridizationfun tion isspin-dependentand ollinear lead magnetizations is now hidden in the tun- te to has in general a non-zero real part that is given by neling amplitudes and . [∆ (0)]=σ4t2δǫcos(θ)/D2, Note that in this formulation the dot is oupled with R σ (21) a single e(cid:27)e tive lead with an angle and spin dependent ω 0 bulk spin DOS given by for our hoi e of the DOS and ∼ . This is equivalent ρeff =ρ cos2(θ/2)+ρ sin2(θ/2), to an angle-dependent e(cid:27)e tive magneti (cid:28)eld ↑/↓ ± ∓ (16) µ B =8t2δǫcos(θ)/D2, B int (22) whi h is mixture of the initial bulk spin densities of states. This allows us to treat this problem in a sim- that polarizes the dot's spin and destroys the Kondo ef- µ B &T B int K pli(cid:28)ed way using the numeri al renormalization group fe t for . 21,22,23 method with the generalizationsto treat arbitrar- The (cid:28)rst thing we want to he k is whether the polar- θ 24 ily shaped densities of states. ized leads with a given angle are able to split the dot ρ (ω)= [G (ω)]/π σ σ We insist that su h formulation has been made possi- DOS[ −I ℄. Ithasbeenshownthatpo- bleonlybe ausewe onsidersymmetri tunnelingampli- larized leads with antiparallel polarized dire tions ( or- t = t θ =π/2 L R (tθud=esπ/2) we h.aNveotρee↑ftfha=t ρfoe↓rffanatnidpatrhaelleel(cid:27)pe otliavreizmatoiodnesl orefssptoantdesin.1g3,1t4o Is there hae(cid:28)ren)itdeoinntoetrvspallitartohuenddotθd=enπsi/ty2 has spin symmetry. where this property holdρs? We have plotted in Fig. 1 ↑ the dot spe tral density in the spin up se tor for dif- θ ferent values of . As expe ted the low energy peak of ρ ω =0 θ =0 ↑ A. Study of the dot density of states is pinned at for but is shifted away from ω = 0 θ as soon as is de reased. This is mu h learer in From hereon we will onsider the DOS along the po- the right inset of Fig. 1 where the peak is zoomed. ρ = larizationdire tion in ea hlead to be given by the tight- ρ T+hρe splitting of the total dot density of states ↑ ↓ binding expressions is shownθin the left inset of Fig. 1 for four difθ- ferent values of . The splitting gradually in rease as 2 θ = π/2 θ = 0 2 ω+δǫ ρ±(ω)= 1 ± , is de reased from aθnd is maxπim/2um for . πDs − D (17) For very small deviations of around , the e(cid:27)e tive (cid:18) (cid:19) µ B .T B int K magneti (cid:28)eld andnosplittingisexpe ted. where δǫ+ = −δǫ− = δǫ is the spin-dependent shift of For ollinearpolarizationdire tions,ithasbeenshown the bands with respe t to the Fermi level. The quantity that this splitting an be ompensated by an external δǫ P 15 magneti (cid:28)eld restoring the Kondo e(cid:27)e t. For general hara terizes the magnetization of the leads given by dire tionsoftheleadpolarizations,ane(cid:27)e tivemagneti ∞ (cid:28)eld is also generated. This e(cid:27)e tive magneti (cid:28)eld an P =Z−∞dωf(ω)[ρ+(ω)−ρ−(ω)], (18) binet hoemppleannseat(e~ndLb,y~nRa)ppwlyitihngaandireex tteironnalopmpaogsniteeti to(cid:28)e~elzd. 4 25 20 20 50 25 20ρ(ω) 234000 ρ(ω)↑121005 θ=θ0.=405.54θπ8=5ππ/2 15 ρ(ω)↑1105 µµµΒΒΒBBB === 148.. δ26ε6 δ tεδ2 εt 2t2 15 10 5 5 ω) θ=0 ω) θ=0 (↑ -00.01 -0.005 0 0.005 0.01 -00.01 -0.005 0 0.005 0.01 (↑10 0 ρ 10 ω/D ω/D ρ -0.005-0.0025ω0/D0.00250.005 θ=0 5 5 θ=π/2 0 0 -0.2 -0.1 0 0.1 0.2 -0.2 -0.1 0 0.1 0.2 ω/D ω/D FIG.1: Dotdensityofstatesforthθe=spi0n,-0u.p45se5 πt,o0r.a4s85fuπn, πti/o2n FIG.2: Dotdensityofstatesforthespin-upse torasfun tion of energy for di(cid:27)erent values ofδǫ=0.5D . ofenergyfor di(cid:27)erentvaluesoθft=he0magneti (cid:28)eldinthe ase ThesUhif=to0f.2theǫdb=an−dsUi/s2(cid:28)xedtot = 0.071 . Otherparameter with parallel magnetizations ( ). Otherparametersas in are: , , and . Right inset: zoom Fig. 1. The spin polarization in the dot and the splitting of ρ↑ θ of for thρe↑same values of . Wθe learly seeθth=esπh/if2t of the theKondopeak anbe ompensatedwith a magneti (cid:28)eld. maximum when de reasing away from . Left 20 inset: Total densityof states as fun tion of frequen y for the 20 θ tafoesumsropoaenfroaratesumtrhee.entei(cid:27)one etdivveamluaegsnoefti .(cid:28)TehldeidselnarsgiteyrothfastnattheseiKsosnpdliot 15 (ω) 1105 µµµµµµΒΒΒΒΒΒBBBBBB ==== == 0303 55....8484..995 5 δδ δδεεδδεε εεtt tt 22tt2222 ↑ ρ 5 We therefore expe t that the splitting of the Kondozero ω) θ=π/8 ( 10 bias anomaly indu ed by the polarized ele trodes an be ρ↑ -00.005-0.0025 0 0.00250.005 ω/D orre ted by an adequate magneti (cid:28)eld for all values of θ . We have he kedthatthis propertyholdsusingNRG. ρ ↑ 5 We have presented in Fig. 2 the DOS as fun tion of θ =0 energyfor (wherethesplittingismaximum)andfor di(cid:27)erentvalueoftheexternalmagneti (cid:28)eld. Foragiven value of the external magneti (cid:28)eld, the splitting of the 0 -0.2 -0.1 0 0.1 0.2 Kondo peak vanishes and the Kondo peak is restored at ω/D ω =0 θ . Thisisalsotrueforageneralangle (seeFig.3). Thesepropertiesofthedot'sspe traldensity will learly θ=π/8 FIG. 3: Same as Fig. 2 with . re(cid:29)e t in the transport properties of the system as we will see. ρ (0) = ρ (0) d↑ d↓ see. In our NRG formulation we have . Sin etheDOSintheleadsareenergydependent(seeEq. B. Transport properties (17)), this equality does not prevent to have a non zero p = 0 polarization following Eq. (18) though . For this Γσσ′ = (Σσσ′) Using the Keldysh formalism, it has been shown that σa,sσe′,=the,hybridizationmatri es −I (where for the ollinear magnetizations, as in the ase of non- ↑ ↓) are diagonaland angle independent. This al- T =0 25 magneti leads, the (cid:28)nite temperature ondu tan e an lowsthe useof the Meir Wingreen formula.GT(0h)e2 beputintermsonthedot'sspe traldensity.25Thisisnot ondu tan eisthussimplyproportionalto| d | . The T = 0 the aseforarbitraryangles. Theproblemisthatfornon ondu tan e has been omputed using NRG and L θ ollinear magnetizations, the two spin hannel of the is shown in Fig. 4 as a fun tion of for di(cid:27)erent values R δε and leadsaremixedandthestandardapproa husedto of the splitting . The ondu tan e rea hesthe unitary θ = π/2 eliminate the lesser propagator an not be applied. To limit for (antiparallel ase) and is minimum our knowledge, in these ir umstan es, the only dire t forparallelmagnetizations. Thoughthis resultmaylook wayofevaluatingthe ondu tan eat anytemperatureis unusual, this is expe ted sin e the splitting is zero for θ = π/2 through the Kubo formalism. This requires evaluating restoring the Kondo e(cid:27)e t and is maximum for θ = 0 p(0) = 0 two parti le propagators, a matter that requires sophis- . Remember that with the de(cid:28)nition of ti ated numeri al odes. There are however some limit- the DOS we used in NRG. ing ases that illustrate the general behavior as we will Thisresultshouldbe ontrastedtothesituationwhere 5 1 δε=0.1D anomalous denominator omes from the fa tGthre imagi- δε=0.25 D narypartofthe dotretardedGreenfun tion dσ issim- δε=0.5 D 1/Γσσ T =0 θ pply − Tat T (wp,hθi) h depends on the ang(lTe/Tan)d2 K K . At (cid:28)nite ≪ , deviations of order G u from are expe ted using the Fermi liquid theory. On 2e/h) the other hand, at high temperature T ≫ TK(p,θ), the 20.5 ( ondu tan e is mu h smaller and an be omputed by G/ renormalized perturbation theory: 2e2 3π2/16 G= (1+p2cos(2θ)) . h ln2(T/T ) (24) K 00 π/4 π/2 3π/4 π Noti e thattheexpressionforthe hightemperature on- θ du tan e has the same angular expression as the one for a tunnel jun tion between two ferromagneti leads on- versely to the low temperature ase. FIG. 4: Condu tan e as a fuδnǫ =tio0n.1Dof,t0h.2e5aDn,g0le.5fDor di(cid:27)er- ent values of the bands shift . Other parameters as in Fig. 1. V. CONCLUSION anexternalmagneti (cid:28)eldabletorestoretheKondozero In this paper, we have studied a quantum dot in the bias anomaly is added. We onsider onstant bulk DOS as in Se . III. Ele trons with spins ↑ have a di(cid:27)erent Kondo regime between two non ollinear ferromagneti DOS at the Fermi energy than ele trons with spins ↓, ele trodesby ombiningpoormans alingandNRGte h- p(0) = 0 T = 0 resulting in 6 (as in [13,14,15℄). At , the niques. We have shown using NRG that the Kondo zero bias anomaly is in generalsplitted for all angles between Kondo singlet is therefore formed and the Fermi liquid G the two lead magnetizations (ex ept for the antiparallel theory an be applied. The ondu tan e through the T = 0 G u ase). This splitting an be ompensated by an external devi e rea hes at its maximum ondu tan e magneti (cid:28)eld. Based on the spe tral densities, we have whi h reads: 2e2 1 p2 addressedtransportpropertiesforboth ompensatedand Gu = h 1 p2−cos2(θ) . (23) un ompensated ases. − Note added Some of the results presented in this pa- G θ = 0 per have been also independently obtained in a re ent Noti e that here reap hes the unitary limit foθr= π/2 preprint27. whatever the value of and is minimum for p G at (cid:28)xed . In order to determine to determine u, we A knowledgments. (cid:21) PS would like to thank interest- have extended the s attering approa h developed by Ng ing ommuni ations with J. Martinek. Part of this re- and Lee26 to polarized ele trodes following [15℄. This sear hwassupportedbythe ontra tPNANO`QuSpins' of the Fren h Agen e Nationale de la Re her he. 1 6 D. Goldhaber-Gordon, H. Shtrikman, D. Mahalu, D. J. Park, A. N. Pasupathy, J. I. Goldsmith, C. Chang, Abus h-Magder,U.Meirav,andM.A.Kaster,Nature391, Y. Yaish, J. R. Petta, M. Rinkoski, J. P. Sethna, H. D. 156 (1998). Abruna,P.L.M Euen,andD.C.Ralph,Nature417,722 2 S. M. Cronenwett, T.H. Oosterkamp, and L.P. Kouwen- (2002); W. Liang,M. P. Shores, M. Bo krath, J. R. Long, hoven, S ien e 281 , 540 (1998); F. Simmel, R. H. Bli k, and H. Park, Nature 417, 725 (2002). 7 U.P. Kotthaus, W. Wegsheider, and M. Bli hler, Phys. M. R. Buitelaar, T. Nussbaumer, and C. S hönenberger, Rev. Lett. 83, 804 (1999); J. S hmid, J. Weis, K. Eberl, Phys. Rev. Lett. 89, 256801 (2002); M. R. Buitelaar, W. and K. v. Klitzing, Phys. Rev. Lett. 84, 5824 (2000); Belzig, T.Nussbaumer,B.Babi ,C.Bruder,andC.S hö- W. G. van der Wiel, S. De Fran es hi, T. Fujisawa, J.M. nenberger, Phys. Rev. Lett. 91, 057005 (2003); P. Jarillo- Elzerman, S. Taru ha, and L. P. Kouwenhoven, S ien e Herrero, J. A. van Dam, and L. P. Kouwenhoven, Na- 289, 2105 (2000). ture 439, 953 (2006); J.-P. Cleuziou, W. Wernsdorfer, V. 3 J. Nygård, H. Cobden, and P. E. Lindelof, Nature 408, Bou hiat, Th. Ondar uhu, M. Monthioux, Nature Nan- 342 (2000). ote hnology 1, 52 (2006). 4 8 M. R. Buitelaar, A. Ba htold, T. Nussbaumer, M. Iqbal, A.N.Pasupathy,R.C.Bial zak,J.Martinek,J.E.Grose, andC.S hönenberger,Phys.Rev.Lett.88,156801(2002). L. A. K. Donev, P. L. M Euen, and D. C. Ralph, S ien e 5 P. Jarillo-Herrero, J. Kong, H. S. J. van der Zant, C. 306, 86 (2004). 9 Dekker,L.P.Kouwenhoven,andS.DeFran es hi,Nature Semi ondu tors Spintroni s and Quantum Computation 434, 484 (2005). edited by D. D. Aws halom, D. Loss and N. Samarth 6 19 (Springer, Berlin, 2002). R. Swirkowi z, M. Wil zynski, and J. Barnas, 10 N.Sergueev,Q.-F.Sun,H.Guo,B.G.Wang,andJ.Wang, ond-mat/0511506. 20 Phys. Rev. B 65, 165303 (2002). J.König,J.Martinek,J.Barna±,andG.S hön,CFNLe - 11 P.Zhang,Q.-K.Xue,Y.Wang,andX.C.Xie,Phys.Rev. turesonFun tionalNanostru tures",Eds.K.Bus hetal., Lett. 89, 286803 (2002). Le ture Notesin Physi s 658, Springer, 145-164 (2005). 12 21 B. R. Bulka and S. Lipinski, Phys. Rev. B 67, 024404 A. C. Hewson, The Kondo Problem to Heavy Fermions, (2003). Cambridge Univ. Press (1993). 13 22 R. Lopez and D. San hez, Phys. Rev. Lett. 90, 116602 W.Hofstetter, Phys.Rev.Lett.85,1508(2000); R.Bulla, (2003). A.C.Hewson, andThPrus hke,J.Phys.: Condens.Mat- 14 J. Martinek Y. Utsumi, H. Imamura, J. Barnas, S. ter 10, 8365 (1998). 23 Maekawa, J. König, and G. S hön, Phys. Rev. Lett. 91, T. A. Costi, Phys. Rev. Lett. 85, 1504 (2000); Phys. Rev. 127203 (2003). B 64, 241310 (2001); P. S. Cornaglia and D. R. Grempel, 15 J. Martinek, M. Sindel, L. Borda, J. Barnas, J. König, Phys.Rev. B 71, 245326 (2005). 24 G. S hön, and J. von Delft, Phys. Rev. Lett. 91, 247202 C. Gonzalez-Buxton and K. Ingersent, Phys. Rev. B 57, (2003). 14254 (1998) 16 25 M.S. Choi, D.San hez,andR.Lopez, 92,056601 (2004). Y. Meir and N. S. Wingreen, Phys. Rev. Lett. 68, 2512 17 Y. Utsumi, J. Martinek, G. S hön, H. Imamura, and S. (1992). 26 Maekawa, Phys. Rev. B 71, 245116 (2005). T.K. Ng and P.A. Lee, Phys. Rev. Lett.61, 176 (1988). 18 27 C. J. Gazza, M. E. Torio, and J. A. Riera, Phys. Rev. B D. Matsubayashi and M. Eto, ond-mat/0607548. 73, 193108 (2006).