ebook img

Statistics and noise in a quantum measurement process PDF

4 Pages·0.36 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 Statistics and noise in a quantum measurement process

Statistics and noise in a quantum measurement process Yuriy Makhlin1,2, Gerd Scho¨n1,3, and Alexander Shnirman1 1Institut fu¨r Theoretische Festk¨orperphysik, Universit¨at Karlsruhe, D-76128 Karlsruhe, Germany 2Landau Institute for Theoretical Physics, Kosygin st. 2, 117940 Moscow, Russia 3Forschungszentrum Karlsruhe, Institut fu¨r Nanotechnologie, D-76021 Karlsruhe, Germany 0 0 0 2 the initial state of the qubit. The origin of the mix- The quantum measurement process by a single-electron ing is that the charge operator (the measured quantity) n transistor or a quantum point contact coupled to a quantum andthe qubit’s Hamiltoniando notcommute. The third a bitisstudied. Wefindaunifieddescriptionofthestatisticsof J time scale appears in the dynamics of the current in the the monitored quantity, the current, in the regime of strong 8 measurement and expect this description toapply for a wide SET. Consider the probability distribution P(m,t) that 2 melectronshavetunneledthroughtheSETbytimet. It class of quantum measurements. We derive the probability distributions for the current and charge in different stages of wasshown[7]thatafter acertaintime tmeas information ] l the process. In the parameter regime of the strong measure- about the state of the qubit can be extractedby reading l a ment the current develops a telegraph-noise behavior which outm. Asexpectedfromthebasicprinciplesofquantum h can bedetected in thenoise spectrum. mechanicstheobservationofthequbitfirstofalldisturbs - Introduction. The long-standing interest in funda- its state. Hence t τ . The measurementprocessis s meas ϕ ≥ e mental questions of the quantum measurement received only effective if the mixing is slow, t t . In the mix meas m new impetus by the experimentalprogressin mesoscopic oppositelimitofstrongmixingtheinform≫ationaboutthe . physics and increasing activities in quantum state engi- qubit’s state is lost before a read-out is achieved. t a neering. The basic idea is to use as meter a device, able The distribution function P(m,t) describes the statis- m to carry a macroscopic current, which is coupled to the tics of the charge which has tunneled. The distribution - quantum system such that the conductance depends on ofthecurrentintheSETp(I,t)andcurrent-currentcor- d the quantum state. By monitoring the current one per- relations require, furthermore, the knowledge of corre- n o forms a quantum measurement, which, in turn, causes lations of the values of m at different times. In earlier c a dephasing of the quantum system [1–3]. The dephas- papers on the statistics in a SET [7] or a QPC [8,9] the [ ing has been demonstrated in an experiment of Buks et behavior of P(m,t) at times shorter than tmix was de- 1 al. [4] where a quantum dot is embedded in one arm of rived. Effects due to the additional knowledge acquired v a ‘which-path’ interferometer. The current through a by an observer [10], and the possible influence of the 3 quantum point contact (QPC) in close proximity to the wave-functioncollapseonthe monitoredcurrent[8]were 2 dot suppresses the interference. However, since passing also discussed. Here we develop a systematic approach, 4 electronsinteractwiththecurrentonlyforashortdwell- based on the time evolution (Schr¨odinger equation) of 1 0 time in the dot, the meter fails to distinguish between thedensitymatrixofthe coupledsystem. Thisapproach 0 two possible paths of individual electron; only a reduc- allowsustostudyaveragesandcorrelatorsofthecurrent 0 tion of interference has been observed. This situation is and charge. Since, due to shot noise, instantaneous val- t/ referred to as a weak measurement. uesofthecurrentfluctuatestrongly,westudythecurrent a For a strong quantum measurement a different setting I¯, averaged over a finite time interval ∆t. We calculate m isneeded,whereaclosedquantumsystemisobservedby themixingtimeinaSETandderiveanalyticexpressions d- ameter. Thenasufficientlylongobservationmayprovide forp(I¯,∆t,t),aswellasP(m,t),validonbothshortand n information about the quantum state. This situation is long time scales. We study the noise spectrum of the o realized when a single-electron transistor (SET) is cou- currentand find thatin the limit of strongmeasurement c pledtoaJosephsonjunctionsingle-chargebox,whichfor (t <t )thelong-timedynamicsischaracterizedby : meas mix v suitableparametersservesasaquantumbit(qubit)[5,6]. telegraph noise,withjumpsbetweentwopossiblecurrent Xi The analysis of the time evolution of the density matrix values, corresponding to two qubit’s eigenstates. ofthecoupledsystemdemonstratesthemutualinfluence The results are of immediate experimental interest. A r a between qubit and meter, i.e. measurement and dephas- recent experiment demonstrated the quantum coherence ing [7]. in a macroscopic superconducting electron box [11], but The measurement process is characterized by three the coherence time was limited by the measuring device. time scales. On the shortest, the dephasing time τϕ, The SET-based measurement should extend the coher- thephasecoherencebetweentwoeigenstatesofthequbit ence time, which combined with experimental progress is lost, while their occupation probabilities remain un- in fast measurement techniques [12] should increase the changed. Latermeasurement-inducedtransitionsmixthe maximum number of coherent quantum manipulations. eigenstates, changing their occupation probabilities on Master equation for the measurement by a SET. The a time scale tmix > τϕ and erasing information about system of a qubit coupled to a SET is shown in Fig. 1. 1 ThequbitisaJosephsonjunctioninthe Coulombblock- current in the SET. aderegime. Its dynamicsis limitedto atwo-dimensional At low temperatures and transport voltages only two Hilbert space spanned by two charge states, with n = 0 charge states of the middle island of the SET, with N or1extraCooperpaironasuperconductingisland. The and N +1 electrons,contribute to the dynamics. Trans- island is coupled capacitively to the middle island of the lationalinvarianceinm-spacesuggeststheFouriertrans- SET,influencing thetransportcurrent. TheSETis kept formation ρij(k) e−ikmρij(m). Expanding in the N ≡ m N in the off-state during manipulations on the qubit [7], tunneling term to lowest order, we obtain the following P with no dissipative current and no additional decoher- master equation (cf. Refs. [7,13]): ence. To perform the measurement, the transport volt- age is switched to a sufficiently high value, so that the ¯h d ρˆN + i[Hqb,ρˆN] current starts to flow in the SET. dt(cid:18)ρˆN+1 (cid:19) i[Hqb+Hint,ρˆN+1]! Γˇ eikΓˇ ρˆ n (cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1) = −ΓˇL ΓˇR ρˆN . (2) (cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1) N m(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1) (cid:18) L − R (cid:19)(cid:18) N+1 (cid:19) (cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1)(cid:0)(cid:1) I V (cid:0)(cid:1) (cid:0)(cid:1) x Here the operators V V g tr Γˇ ρˆ Γ ρˆ+πα [ ,ρˆ] , (3) FIG.1. The circuit of a qubit and a SET used as a meter. L ≡ L L Hint + Γˇ ρˆ Γ ρˆ πα [ ,ρˆ] . (4) R R R int + ≡ − H The Hamiltonian of the system is given by [13] represent the tunneling rates in the left and right junc- H=HSET+HqNb+HT+Hψ . (1) tions,with αL/R ≡RK/(8π3RLT/R) being the tunnel con- ductanceofthe junctionsinunitsofthe resistancequan- The first term is the charging energy of the transis- tum R = h/e2. The rates are fixed by the potentials K tor, quadratic in the charge Ne on the middle island, µ , µ = µ +V of the leads: Γ = 2πα [µ (1 = E (N N )2. The induced charge N e is L R L tr L L L − − HSET SET − 0 0 2N0)ESET] and ΓR =2παR[(1 2N0)ESET µR]. They determined by the gate voltage V and other voltages − − g define the tunneling rate Γ = Γ Γ /(Γ +Γ ) through in the circuit. The Hamiltonian of the qubit, N = L R L R Hqb the SET. The last terms in Eqs. (3,4) make these rates +N , includesthe Hamiltonianoftheuncoupled Hqb Hint sensitive to the qubit’s state. qubit and the Coulomb interaction with the SET Hqb The initial condition at the beginning of the measure- N (splitintotwotermsforlaterconvenience.) Inthe Hint ment, written down in the logical basis, basisofthequbit’schargestatestheyaregivenby = qb H e21n(eErqgbyσˆszc−aleEsJEσˆSxE)Ta,nEdqNb HanindtE=inEt ianrteNdσˆezt.erTmhineecdhabrygicnag- ρiNj(m,t=0)= |aa∗|b2 abb∗2 wN δm0 , (5) pacitances in the circuit, while E is the Josephson cou- (cid:18) | | (cid:19) J pling. We considerthe eigenstatesof , 0 and 1 ,as describesthequbitinapurestatea0 +b1 andtheSET qb H | i | i | i | i the logicalstates of the qubit. In this basis, is diag- in the zero-voltage equilibrium state. One can assume qb H onal, with the level spacing ∆E (E2+E2 )1/2, while that w =1 and w =0. ≡ J qb N N+1 Ek E⊥ Reduction of the master equation. In general the dy- Hint = Eii⊥nntt −Einikntt !, where Eiknt ≡ EintEqb/∆E and nmaamstiecrseoqfutahteioqnub(2it)’,sidsecnosmitpylimcaatterdixsi̺ˆn,cdeedscerpihbaesdinbgy(tdhee- E⊥ E E /∆E. Theterm describestheFermions int ≡ int J Hψ cay of the off-diagonal entries) and mixing (relaxation in the island and electrodes of the SET, while gov- T of the diagonal to their stationary values) may occur on H erns the tunneling in the SET. Here we assume weak similar time scales, τ t . However, under suitable ϕ mix couplingtotheenvironment,withrelaxationslowerthan ≈ conditions the mixing is slow, which is the prerequisite the SET-induced mixing. The opposite limit is dicussed for a measurement process. This is the case, if the qubit in Ref. [13]. operates in the regime with dominant charging energy: ThefulldensitymatrixcanbereducedtoρiN m (t)by jN′m′ tracing over microscopic degrees of freedom and keeping E , E +2E E . (6) qb qb int J | | | |≫ track only of N and m, the number of electrons which have tunneled through the SET. Here i,j =0,1 refer to Weak (E ∆E) or strong (E ∆E) coupling to int int ≪ ∼ a qubit’s basis. A closed set of equations can be derived the SET can be considered. In the latter case a faster for ρij(m) ρiNm, the diagonal entries of the density measurement is achieved (see Eq. (11)). N ≡ jNm matrix in N and m [14]. Solving these equations, we We first analyze the dynamics without mixing, i.e. we analyze the evolution of the reduced density matrix of put E⊥ = 0. In this case the time evolutions of ρij for int N the qubit, ̺ij(t) ρij(m,t), as well as P(m,t) the four different pairs of indices ij are decoupled from ≡ N,m N ≡ ρjj(m,t)andotherstatisticalcharacteristicsofthe eachother, each being characterizedby two eigenmodes. N,j N P P 2 The absence of mixing, further, implies the conservation equal-weightdistribution: ̺00(t) ̺11(t) exp( Γ t). mix − ∝ − ofoccupationsofthelogicalstates̺ii,fori=0,1,andwe Thereforethe two-peak structure appears only in the in- find two ‘Goldstone’ modes for k 1, with eigenvalues terval between t t Γ−1 . The measurement can ≪ be performed onmlyeaisf≤t ≤ mΓi−x1 1 meas ≪ mix λii(k) iΓik fiΓik2 . (7) The measurement takes longer than the dephasing, + ≈ − 2 t τ . Such measurement can be called non- meas ϕ ≫ Here Γi Γi Γi /(Γi + Γi ) are the tunneling rates efficient [10]: the information about the qubit is con- ≡ L R L R throughtheSET,Γ0/1 =Γ πα Ek andΓ0/1 =Γ tained in the SET already after τϕ, but can be read out πα Ek are the tunLnelingLra±tes inLthinetjunctioRns for Rtw∓o from the current only later. R int The quantum measurement with a QPC can be de- logicalstates(cf.Eqs.(3,4)),andfi 1 2Γi/(Γi +Γi ) ≡ − L R scribed in a similar way. The Coulomb interaction of are the Fano factors responsible for the shot noise re- the qubit with the current results in two tunneling rates duction. The other two eigenmodes decay fast, with the Γ0/1 = Γ¯ δΓ/2 for two qubit’s states. Tracing out mi- rates λii (Γi +Γi ). ± − ≈− L R croscopic degrees of freedom one arrives at the master The analysis of the eigenvalues, λ01(k) = [λ10( k)]∗, ± ± − equation[3] for the Fourier transformof the density ma- ofthefouroff-diagonalmodesinij,revealsthedephasing trix ρij(m), which can be rewritten as time τ of the qubit by the measurement, i.e., the decay ϕ time of ̺01 = ρ01(k = 0)+ρ01 (k = 0). It is given τby−1τ=ϕ−1w=Γ4Γ+EikwnNt2/(ΓLΓ+iΓnRt)h2eNiof+pE1pioknstit≪e caΓsLe.+ΓR, and ¯hddtρˆ+i[Hqb,ρˆ]=(cid:20)Γ¯ρˆ+ 21δΓ[σˆz,ρˆ]+(cid:21)(eik−1) ϕ N L N+1 R (4τ )−1eik[σˆ ,[σˆ ,ρˆ]]. (12) The picture is modified by the mixing at finite E⊥ . − ϕ z z int Wefindthatthemixingmaybetreatedperturbativelyif One can show that τ−1 1(√Γ0 √Γ1)2 is the decay λ01 E⊥ , which turns to be the case if the condition ϕ ≡ 2 − | ±|≫ int rateofρ01. Forρii theeigenvaluesaregivenby(7),with- (6) holds. Then, in the secondorder,the degeneracybe- out Fano factors. The measurement time (11) and the tweenthelong-livingmodes(7)isliftedandthelong-time dephasing time coincide, implying the 100% efficiency. evolutionofthe occupationsρii(k)=ρii(k)+ρii (k) is N N+1 Under the condition E max(∆E,τ−1) the per- given by a reduced master equation, J ≪ ϕ turbative treatment produces the reduced master equa- d ρ00(k) ρ00(k) tion (8,9), with the mixing rate =M , (8) dt ρ11(k) red ρ11(k) τ (cid:18) (cid:19) (cid:18) (cid:19) Γ =E2 ϕ . (13) mix J1+∆E2τ2 λ00(k) 0 1 1 1 ϕ Mred = +0 λ11(k) + 2Γmix −1 1 . (9) (cid:18) + (cid:19) (cid:18) − (cid:19) A phenomenon, termed the Zeno (or watchdog) effect, can be seen [3,8] in the limit τ−1 ∆E: the stronger For the mixing rate, Γmix, we obtain ϕ ≫ is the measurement,quantified by τ−1, the weakeris the ϕ Γ rate of jumps between the eigenstates, Γ E2τ . mix ≈ mix ≈ J ϕ 4ΓE2E2 The analysis of the SET mixing rate (10) in terms of ∆E2(∆E+2Ek )2+[ΓJ∆iEnt+Γ (∆E+2Ek )]2 . (10) the Zeno effect is more complicated. The rates τϕ−1 and int R L int Γ depend in this case on several parameters and no mix To understand the role of the mixing we assume first simple relation between Γ and τ is found. However, mix ϕ Γ = 0 in Eqs. (8,9). Then, for the initial condition in the regime E Γ Γ ∆E , these two rates mix int ≪ L ∼ R ≪ | | (5)weobtainρ00(k)= a2eλ0+0(k)t, ρ11(k)= b2eλ1+1(k)t, change in opposite directions as functions of ΓL ΓR, andP(k,t) P(m|,t|)e−ikm =ρ00(k)+ρ1|1|(k). From which is reminiscent of the Zeno physics. ∼ this we obt≡ain tmhe distribution P(m,t), which evolves Statistics of charge and current. The results of this from a peak δP(m) at t = 0 into two peaks with weights section apply to the SET and QPC alike. The statisti- a2 and b2, moving in m-space with velocities Γ0 and cal quantities studied depend on the initial density ma- Γ| 1|, and |w|ith widths growing as 2fiΓit. The peaks trix (5): e.g., P(m,t) = P(m,t ρ0). In the two-mode | approximation (8,9) this reduces to a dependence on separate after a time p γ ̺00 ̺11 = a2 b2. We solve Eq. (8) to ob- 0 2f0Γ0+ 2f1Γ1 2 tain≡P(m,−t γ0)=T|r|qb−[U|(m|,t)ρ0], where U(m,t) is the | tmeas = p Γ0−Γp1 ! . (11) Γin0v/e1r=seΓ¯FourδiΓer/2traarnesfcolromse,otfhUe(rke,stu)lt≡inegxdpis[Mtrirbedu(tkio)nt]i.sIf ± Thus measuring the charge m after t constitutes a meas e−δm2/2fΓ¯t strong quantum measurement [7]. However, at longer P(m,t γ )= P˜(m δm,t γ ) . (14) times t > Γ−1 the mixing spoils this picture. In par- | 0 − | 0 2πfΓ¯t mix δm X ticular, the occupations of the logical states relax to the p 3 Thefirsttermintheconvolution(14)containstwodelta- asthesumoftheshot-andtelegraph-noisecontributions. peaks, corresponding to two qubit’s logic states: At low frequencies ω Γ the latter becomes visible mix ≪ ontopoftheshotnoiseasweapproachtheregimeofthe P˜(m,t γ0)=Ppl m−Γ¯t , 1Γmixt γ0 strong measurement: Stelegraph/Sshot ≈4tmix/tmeas. | δΓt/2 2 (cid:18) (cid:12) (cid:19) +e−Γmixt/2 a2δ m Γ0t + b2δ(cid:12)m Γ1t . (15) | | − | | (cid:12) − a P(m=t;t)(cid:1)t b p(I;(cid:1)t;t) On the time s(cid:2)cale t(cid:0) Γ−(cid:1)1 the pe(cid:0)aks disap(cid:1)(cid:3)pear; in- mix ≡ mix stead a plateau arises. It is described by Γ tmix tmix Ppl(x,τ|γ0)=e−τ 2mδΓix I0 τ 1−x2 (cid:0)0 tmeas (cid:0)0 +(1+γ x)I τ 1n x(cid:16)2p/ 1 x(cid:17)2 , (16) (cid:0)1 tmeas<(cid:1)t<tmi(cid:0)x1 0 1 − − c p(I;(cid:1)t;t) d p(I;(cid:1)t;t) (cid:16) p (cid:17) p o at x < 1 and P = 0 for x > 1. Here I , I are the pl 0 1 | | | | modified Bessel functions. At longer times the plateau transforms into a narrow peak centered around m=Γ¯t. tmix The Gaussian in Eq. (14) arises due to shot noise. Its tmix effect is to smear out the distribution (see Fig. 2a). (cid:0)0 (cid:0)0 We also calculate P2(m,t;m′,t′|ρ0), the joint proba- tmeas<(cid:1)t(cid:24)tm(cid:0)1ix (cid:1)t(cid:29)tmix (cid:0)1 bility to have m electrons at t and m′ electrons at t′. This allows us to obtain the probability distribution FIG.2. Theprobabilitydistributionsofthecharge(a)and current(b–d). P(m,t)in(a)isrescaled,sothatthepeaksdo p(I¯,∆t,t ρ ) P (m+I¯∆t,t+∆t;m,t ρ ) (17) not move. Thetime-axis scale is logarithmic. 0 2 0 | ≡ | m X To conclude, we have developed the master equation of the current I¯ t+∆tI(t′)dt′ averaged over the time approach to study the statistics of currents in a SET or ≡ t interval ∆t. The evolution is Markovian,and we obtain: aQPCasaquantummeter. Weevaluatetheprobability R P (m,t;m+∆m,t+∆t)=Tr [U(∆m,∆t)U(m,t)ρ ] distributions and the noise spectrum of the current. 2 qb 0 for ∆t >0. The derivation of p(I¯,∆t,t) thus reduces to We acknowledge discussions with Y. Blanter, the calculation of the charge distribution (14) for differ- L. Dreher, S. Gurvitz, D. Ivanov, and A. Korotkov. ent initial conditions: p(I¯,∆t,t γ )=P(m=I¯∆t,∆t e−Γmixtγ ). (18) | 0 | 0 [1] I.L. Aleiner, N.S. Wingreen, Y. Meir, Phys. Rev. Lett. 79, 3740 (1997). The behavior of p(I¯,∆t,t) is shown in Fig. 2. A strong [2] Y. Levinson, Europhys.Lett. 39, 299 (1997). quantum measurement is achieved if tmeas < ∆t < tmix, [3] S.A. Gurvitz, Phys.Rev.B 56, 15215 (1997). attimest<t (seeFig.2b). Inthiscasethemeasured mix [4] E. Buks, R. Schuster, M. Heiblum, D. Mahalu, and currentis close to either Γ0 or Γ1, with probabilities a2 V. Umansky,Nature, 391, 871 (1998). | | and b2, respectively. At longer times a typical current [5] A.Shnirman,G.Sch¨on,andZ.Hermon,Phys.Rev.Lett. patte|rn| is a telegraphsignaljumping betweenΓ0 andΓ1 79, 2371 (1997). on a time scale t . If ∆t t the meter does not [6] Yu. Makhlin, G. Sch¨on, and A. Shnirman, Nature 386, mix meas ≪ have enough time to extract the signal from the shot- 305 (1999). noise background. At larger ∆t the meter-induced mix- [7] A. Shnirman and G. Sch¨on, Phys. Rev. B 57, 15400 (1998). ing erases the information, partially (∆t t , Fig. 2c) mix ∼ [8] S.A. Gurvitz, preprint,quant-ph/9808058. or completely (∆t t , Fig. 2d), before it is readout. ≫ mix [9] L. Stodolsky, Phys.Lett. B 459, 193 (1999). Thetelegraphnoisebehaviorisalsoseeninthecurrent [10] A.N. Korotkov,Phys. Rev.B 60, 5737 (1999). noise. Fourier transformation of the correlator [11] Y.Nakamura,Yu.A.Pashkin,andJ.S.Tsai,Nature398, 786 (1999). hI(t)I(t′)iρ0 = mm′∂t∂t′P2(m,t;m′,t′|ρ0) (19) [12] R.J. Schoelkopf, P. Wahlgren, A. A. Kozhevnikov, mX,m′ P. Delsing, and D. E. Prober, Science 280, 1238 (1998). [13] Y. Makhlin, G. Sch¨on, and A. Shnirman, to be pub- gives in the stationary case the noise spectrum, lished in ”Exploring the Quantum-Classical Frontier.” Eds. J.R.Friedman and S.Han, cond-mat/9811029. e2δΓ2Γ S (ω)=2e2fΓ¯+ mix , (20) [14] H. Schoeller and G. Sch¨on, Phys. Rev. B 50, 18436 I ω2+Γ2 mix (1994). 4

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.