arXiv:quant-ph/0412052v1 7 Dec 2004 i uhltrwe Nyquist ba- when firm later a much on between put sis was connection fluctuations intimate related This and dissipation friction. the to fusion ea elpoie rtln ewe h dissipative as known the fluctuations, between thermal link the impeding first the a and provided forces well as he h aso thermodynamics with of combined laws ingeniously the he which methods statistical it fqatmmcaisi h al 90sw can we 1920’s the early After the in context? mechanics quantum this of in birth play fluctuations quantum current-fluctuations. and voltage- of density spectral the udmna set fQatmBona Motion Brownian Quantum of Aspects Fundamental inmto nhis in motion nian INTRODUCTION I. the with phenomena. driven confronted noise when quantum of a of aware world a of be that out must shortcomings out point and one pitfalls decay we subtleties, of Furthermore, dissipative series the for state. situation of metastable re- the problem corresponding illustrate quan- We the the generalized for dynamics. a duced equation of master terms tum integral in general- path or a a with formulation, equation, nonlinear Langevin dealt quantum dissipative be ized can A harmonic dynamics damped quantum theory. and a response thermodynamics of linear combining case upon topic prominent the oscillator model discuss the we to for roadways Here, theoretical dissipation. of quantum variety There a the fluctuations. with exist quantum thermal friction classical of connects the strength evolu- of that generalization relation These time quantum Einstein systems. a the quantum to open lead on equilibrium of thermal properties constraints influ- tion of severe frictional symmetries nonlinear impose The exhibit and linear that ences. in systems motion quantum Brownian tum htrl oqatmmcaisadteassociated the and mechanics quantum do role What letEnti xlie h hnmnno Brow- of phenomenon the explained Einstein Albert quan- of description the with deals work This isenrelation Einstein nerlmtoooyi ple otedcyo eatbes metastable of decay the subt to noise. of applied series is a methodology to time-reve integral point the La furthermore, of quantum and, consequences dynamics generalized the quantum nonlinear discuss We the of approach. role is both, gral The of dissipation context oscillator. quantum quantum the of harmonic op in damped issue for a The hold of must problem equilibrium. that quan thermal consequences celebrated important in the some from discuss in out noise we Starting quantum of physics non-equilibrium. the on ary elaborate we work this With Augsbu 86135 Universit¨at Augsburg, f¨ur Physik, Institut Ingold Gert-Ludwig H¨anggi and Peter nu mirabilis annus hc eae h tegho dif- of strength the relates which 1 2 nti ineigwork pioneering this In . n Johnson and f10 yueof use by 1905 of 3 considered g Germany rg, slsmer o noe dissipative open an for symmetry rsal ae sitdb unu Brownian quantum by assisted tates u utaindsiaintheorem fluctuation-dissipation tum eisadpsil ifls h path The pitfalls. possible and leties e feetos us-atce,adaie sasse b assisted nature is quantum the alike, which trans- and for the noise quasi-particles, and electrons, biolog- tunnelling of the and fer example, nano-scale For many systems. in ical source noise prominent the equilibrium, theorem. in fluctuation-dissipation and quantum fluctuations function quantum response associated gener- the a the between forward put connection they valid work ally their include In to Johnson effects: and quantum Nyquist Einstein, by relations the e yNqitfrtefis ieteitouto of the introduction of the substitution time the pa- first 1928 via energy the the noise of mechanical for paragraph quantum Nyquist final by very per the in encounter eertdwr yCle n Welton and the of Callen precursor by a work oscilla- constitutes celebrated harmonic thus remark the Nyquist’s of tor. contribution) the energy out point leaving zero (but energy quantum averaged thermally r h ucinlitga ehdfrdsiaiequan- systems dissipative for systems These method se: tum integral quan- per functional motion describe the Brownian are several to quantum and on methods noise with elaborate tum equivalent of We consistent but principle be the alternative balance. and to detailed thermodynamics order of (quantum) law in latter second This fluctuation- necessary the the is Callen-Welton). of property version la (`a quantum at theorem must the dissipation noise obey thermal from times the motion particular, all In Brownian principles. quantum first modelling of schemes noise. Johnson-Nyquist func- a classical a to temperatures occur high as sufficiently does drastically At crossover change temperature: noise of this tion of features The tcatcschemes stochastic Schr¨odinger equations h unu ersinhptei n/rteMarkov the and/or forces, hypothesis Langevin regression quasi-classical quantum of the use ob- ap- wave the rotating be the proximation, must others, approximations. among which involve, looking pitfalls naive pitfalls Such even delicate making when of well served as series and, a situation classical identify the to differences distinct gvneuto n h ahinte- path the and equation ngevin hra qiiru n nstation- in and equilibrium thermal xmlfidwt h fundamental the with exemplified unu utain sdiscussed is fluctuations quantum ihu ob,qatmflcutoscnttt a constitute fluctuations quantum doubt, Without ihti okw hl rsn aiu ehd and methods various present shall we work this With n isptv unu systems quantum dissipative en, kT 8 h unu agvn(prtr approach (operator) Langevin quantum the , rmtecascleupriinlaw equipartition classical the from 6,7 n iedpnetdie quantum driven time-dependent and 10,11 12 rtecneto stochastic of concept the or , ndigs,w alatninto attention call we so, doing In . cannot 5 h generalized who eneglected. be 4 ythe by 9 y , , 2 approximation6,8,13. This result is the quantum version of the fluctuation- dissipation theorem as it relates the fluctuations de- d scribed by SBA(ω) to the dissipative partχ ˜BA(ω) of the II. THE QUANTUM FLUCTUATION-DISSIPATION response. THEOREM AND ITS IMPLICATIONS In the spirit of the work by Nyquist and Johnson we consider as an example the response of a current δI As already mentioned, in 1951 Callen and Welton through an electric circuit subject to a voltage change proved a pivotal relation between equilibrium fluctua- δV . This implies B = I and, because the voltage couples tions and dissipative transport coefficients. Note also to the charge Q, A = Q. The response of the circuit that this quantum fluctuation-dissipation relation holds is determined by δI(ω) = Y (ω)δV (ω) where the admit- true independent of particle statistics. The following tance Y (ω) is identical to the susceptibility χIQ(ω). As a cornerstone achievements can be found in this primary consequence of I = Q˙ , the symmetrized power spectrum work5: of the current fluctuations is given by SII (ω)= iωSIQ(ω) • The generalization of the classical Nyquist’s for- so that we obtain mula to the quantum case. ¯hω S (ω)=¯hω coth ReY (ω) (6) II 2kT • The quantum mechanical proof that susceptibilities are related to the spectral densities of symmetrized ¯hω ¯hω =2 + ReY (ω) . correlation functions. 2 eβhω¯ − 1 For a single degree of freedom, linear response the- In the high temperature limit kT ≫ ¯hω, we re- ory yields for the change of the expectation value of an cover the results of Nyquist and Johnson, i.e. SII (ω) → operator-valued observable B due to the action of a (clas- 2kT ReY (ω). For the Markovian limit of an ohmic re- sical) force F (t) that couples to the conjugate dynamical sistor, where Y (ω)=1/R this result simplifies to read operator A SII (ω)=2kT/R. The quantum version was already an- t ticipated by Nyquist in the last paragraph of his 1928 2 <δB(t)> = dsχBA(t − s)F (s) . (1) paper . However, he made use of the original expres- −∞ Z sion of Planck which yields only the second contribution Here, δB(t)= B(t) − 0 denotes the difference with present in the lower line of (6). Nyquist thus missed the respect to the thermal equilibrium average 0 in the first term arising from the vacuum energy which already 14 absence of the force. The reaction of the system is con- appears in a paper by Planck published in 1911. tained in the response function χBA(t) with a so-called On the other hand, in the extreme quantum limit dissipative part kT ≪ ¯hω, we find that SII (ω) → ¯hωReY (ω). In par- ticular, this implies that at zero frequency the spectral d 1 weight of the current fluctuations vanishes in the generic χ (t)= [χBA(t) − χAB(−t)] . (2) BA 2i case where the admittance does not exhibit an infrared divergence. The Fourier transform of χd (t) will be denoted by BA We cannot emphasize enough that the quantum χ˜d (ω). It is worth noting here that only when A = B BA fluctuation-dissipation relation (5) and corresponding does this part in fact coincide with the imaginary part of ′′ implications hold true for any isolated, closed quantum the complex-valued susceptibilityχ ˜ (ω). BA system. Thus, upon contracting the dynamics in full The fluctuations are described by the equilibrium cor- phase space onto a reduced description of an open quan- relation function tum system exhibiting dissipation these relations hold
CBA(t)= <δB(t)δA(0)>β (3) true nevertheless. Therefore, care must be taken when invoking approximations in order to avoid any violation at inverse temperature β =1/kT . The correlation func- of these rigorous relations. We next consider the role tion is complex-valued because the operators B(t) and of quantum dissipation for an exactly solvable situation: A(0) in general do not commute. While the antisym- the damped quantum harmonic oscillator dynamics. metric part of CBA(t) is directly related to the response function by linear response theory, the power spectrum of the symmetrized correlation function III. QUANTUM DISSIPATION: THE DAMPED HARMONIC OSCILLATOR 1 SBA(t)= <δB(t)δA(0) + δA(0)δB(t)> (4) 2 A. Equilibrium correlation functions depends on the Fourier transform of the dissipative part of the response function via Let us next consider the most fundamental case of a simple open quantum system, namely the damped har- ¯hω monic oscillator. This problem could be tackled by set- S (ω)=¯h coth χ˜d (ω) . (5) BA 2kT BA ting up a microscopic model describing the coupling to 3 environmental degrees of freedom to which energy can hyperbolic cotangent at ω = ±iνn with the Matsubara be transferred irreversibly, thus giving rise to dissipation. frequencies νn =2πn/¯hβ become important as well. Af- Such an approach will be introduced in Sect. IV.A. On ter performing the contour integration in (5), one arrives the other hand, the linearity of the damped harmonic at6,15,16 oscillator allows us as alternative to proceed on a phe- ¯h γ nomenological level. This approach is closely related to Sqq(t)= exp − |t| the usual classical procedure where damping is frequently 2Mω¯ 2 sinh(¯hβ ω¯)cos(¯ ωt) + sin(¯hβγ/2)sin(¯ω|t|) introduced by adding in the equation of motion a force × proportional to the velocity. cosh(¯hβω¯) − cos(¯hβγ/2) ∞ Classically, the motion of a harmonic oscillator subject 2γ ν exp(−ν |t|) − n n . (10) to linear friction is determined by Mβ (ν2 + ω2)2 − γ2ν2 n=1 n 0 n t X 2 Mq¨ + M dsγ(t − s)q ˙(s)+ Mω0q =0 . (7) In the limit of high temperatures the second term van- −∞ Z ishes and the first term yields the classical correlation In the example of an electric circuit mentioned in the function. Quantum corrections to this term are relevant previous section, a damping kernel γ(t) with memory at temperatures of the orderhω ¯ 0/k or below, and these would correspond to a frequency-dependent admittance. corrections may be obtained from weak coupling theo- In the special case of ohmic friction corresponding to ries like the quantum master equation approach17,18,19,20. Y (ω)=1/R, the damping force is proportional to the However, there is another regime at temperatures below velocity of the harmonic oscillator, so that the equation ¯hγ/4πk. Here, the second term may initially be small, of motion reads but nevertheless it may dominate the long-time behav- 2 ior of the correlation function. This becomes particu- Mq¨+ Mγq˙ + Mω0q =0 . (8) larly apparent in the limit of zero temperature where In (7) and (8) the mass, frequency, and position of the the exponential functions in the second term in eq. (10) oscillator are denoted by M, ω0, and q, respectively. Due sum up to an algebraic long-time behavior, i.e. Sqq(t)= 4 −2 to the Ehrenfest theorem, the equation of motion (7) is −(¯hγ/πMω0)t . Its relevance for the dynamical evolu- still valid in the quantum regime if we replace q by its tion of the damped harmonic oscillator depends on the 21 expectation value. As a consequence, the quantum me- details of the initial preparation . Although the alge- chanical dynamic susceptibility agrees with the classical braic decay results from the zero temperature limit it can expression15,16 also be observed at low, but finite temperatures during intermediate times before an exponential decay with time 1 1 22 constant ν1 sets in . The occurrence of additional time χqq(ω)= 2 2 (9) M −ω − iωγ˜(ω)+ ω0 scales besides γ at low temperatures leads to shortcom- whereγ ˜(ω) denotes the Fourier transform of the damping ings with the quantum regression hypothesis and allows for the decay of correlations on time scales longer than kernel γ(t). 13 As mentioned before, the response function directly γ. yields the antisymmetric part of the position autocorre- lation function Cqq(t). It therefore suffices to discuss the B. The reduced density matrix and the partition function symmetrized part Sqq(t) defined according to (4). Fur- thermore, our linear system with linear damping repre- sents a stationary Gaussian process so that all higher In the previous section, we have seen that the dynamics order correlation functions may be expressed in terms of of a damped harmonic oscillator can be fully described in terms of the position autocorrelation function (10) and its second order correlation functions.15 In addition, equilib- rium correlation functions containing momentum opera- time derivatives as well as the (classical) response func- tion. If one is interested only in equilibrium expectation tors p can be reduced to position correlation functions by means of p = Mq˙. The dynamics of the damped har- values of arbitrary operators acting in the Hilbert space of the harmonic oscillator, it is sufficient to know the re- monic oscillator can therefore entirely be described by the response function, i.e. the Fourier transform of (9) duced density matrix. By means of arguments analogous 6,15,16 to the dynamic case presented in the previous section, the and Sqq (t). In the case of ohmic damping,γ ˜(ω) = γ, the po- reduced density matrix can only depend on second mo- ments of position and momentum, β ′ 2 tribute to the correlation function Sqq (t) at all tempera- × exp − 2 − 2 (q − q ) . 8 The second moments are found to read 0 ∞ + −0.25 2 1 1 = , (13) −1 β β ω2 + ν2 + |ν |γˆ(|ν |) n=−∞ 0 n n n 0 0.5 1 1.5 2 X kT/hω¯ where we have introduced the Laplace transform of the 0 damping kernel 2 FIG. 1 The weak coupling correction ∆q to 2 The properties of this partition function become more 26,27 6 eralized Langevin equation in a nonlinear situation . A popular model for the dynamics of a dissipative quantum system subject to quantum Brownian noise is 4 ρ obtained by coupling the system of interest to a bath of 0 harmonic oscillators. Accordingly, we write for the total ¯ hω Hamiltonian 2 p2 H = + V (q,t) (20) 0 2M N 2 2 0 5 10 15 pi mi 2 2 2 ci + + ω x − qcixi + q (E − ǫ )/¯hω 2m 2 i i 2m ω2 0 0 i=1 i i i X FIG. 2 The density of states defined by inversion of the re- where the first two terms describe the system as a particle lation (19) for a weakly damped harmonic oscillator with of mass M moving in a generally time-dependent poten- γ = 0.1ω0 exhibits broadened peaks close to the energies tial V (q,t). The sum contains the Hamiltonian for a set ǫ0 + n¯hω0. A delta function at the ground state energy ǫ0 of N harmonic oscillators which are bi-linearly coupled is not shown explicitly. The dashed line represents the aver- with strength c to the system. Finally, the last term, age density of states. i which depends only on the system coordinate, represents a potential renormalization term which is needed to en- sure that V (q,t) remains the bare potential. This Hamil- IV. DISSIPATION IN NONLINEAR QUANTUM SYSTEMS: THE GENERALIZED QUANTUM LANGEVIN tonian has been studied since the early 60’s for systems EQUATION (QLE) which are weakly coupled to the environmental degrees of freedom9,18,19,20,28,29,30,31. Only after 1980, it was re- 32 A. Bath of oscillators alized by Caldeira and Leggett that this model is also applicable to strongly damped systems and may be em- For nonlinear systems the arguments given in the pre- ployed to describe, for example, dissipative tunnelling in 31 vious section no longer apply. In particular, second-order solid state physics and chemical physics . correlation functions are not sufficient anymore to com- One may convince oneself that the Hamiltonian (20) pletely describe the damped system. An alternative ap- indeed models dissipation. Making use of the solution of proach to quantum dissipative systems starting from a the Heisenberg equations of motion for the external de- 33 Hamiltonian at first sight does not seem feasible because grees of freedom one derives a reduced system opera- the absence of time-dependent forces implies energy con- tor equation of motion, the so-called generalized quantum 9 servation. However, as we will see below, once it is real- Langevin equation ized that dissipation arises from the coupling to other de- t dV (q,t) grees of freedom, it is straightforward to model a damped Mq¨(t)+ M dsγ(t − s)q ˙(s)+ = ξ(t) (21) dq quantum system in terms of a Hamiltonian. Zt0 A well known technique to describe a statistical dy- namics governed by fluctuations is given by the method with the damping kernel of generalized master equations and the methodology of N 1 c2 generalized Langevin equations. This strategy is by now γ(t)= γ(−t)= i cos(ω t) (22) M m ω2 i well developed for thermal equilibrium systems. Here i=1 i i the projector operator methodology17,18,19,20,25 yields a X clear-cut method to obtain the formal equations, either and the quantum Brownian force operator for the rate of change of the probability, i.e. the gen- eralized quantum master equation (QME) or the (gen- ξ(t)= −Mγ(t − t0)q(t0) erally nonlinear) generalized quantum Langevin equation N (QLE). + ci xi(t0)cos(ωi[t − t0]) (23) i=1 Already for the case of relaxation towards a unique X thermal equilibrium specified by a single temperature T , pi(t0) + sin(ωi[t − t0]) . the equivalence between the two approaches is not very miωi transparent26. A crucial role is played by the fluctua- tional force which explicitly enters the equivalence, such The generalized quantum Langevin equation (21) ap- as corresponding cumulant averages to an arbitrary high pears first in a paper by Magalinski˘ı9 who started from order. This fact is not appreciated generally, because one (20) in the absence of the potential renormalization term. often restricts the discussion to the first two cumulants The force operator (23) depends explicitly on the ini- only, namely the average and its auto-correlation. It is tial conditions at time t0 of the bath position operators a fact that little is known about the connection of the xi(t0) and bath momenta pi(t0). The initial preparation generalized master equation and the corresponding gen- of the total system, which fixes the statistical properties 6 of the bath operators and the system degrees of freee- In some physical situations a microscopic model for the dom, turns the force ξ(t) into a random operator. Note external degrees of freedom is available31,34. Examples that this operator depends not only on the bath proper- are the electromagnetic modes in a resonator acting as ties but as well on the initial system position q(t0). To a reservoir or the dissipation arising from quasi-particle fully specify the reduced dynamics it is thus of impor- tunnelling through Josephson junctions35. In the case of tance to specify the preparation procedure. This in turn an electrical circuit containing a resistor one may use the then also fixes the statistical properties of the quantum classical equation of motion to obtain the damping kernel Brownian noise. Clearly, in order to qualify as a stochas- and model the environment accordingly. This approach tic force the random force ξ(t) should not be biased; i.e. has been used e.g. to model Ohmic dissipation in Joseph- its average should be zero at all times. Moreover, this son junctions in order to study its influence on tunnelling Brownian quantum noise should constitute a stationary processes36, and to describe the influence of an external process with time homogeneous correlations. impedance in the charge dynamics of ultrasmall tunnel Let us also introduce next the auxiliary random force junctions37. η(t), defined by This scheme of the QLE can also be extended to the η(t)= ξ(t)+ Mγ(t − t0)q(t0) (24) nonequilibrium case with the system attached to two baths of different temperature38. Two most recent ap- which only involves bath operators. In terms of this new plications address the problem of the thermal conduc- random force the QLE (21) no longer assumes the form of tance through molecular wires that are coupled to leads an ordinary generalized Langevin equation: it now con- of different temperature. Then the heat current assumes tains an inhomogeneous term γ(t−t0)q(t0), the initial slip 24,27 a form similar to the Landauer formula for electronic term . This term is often neglected in the so-called transport: The heat current is given in terms of a trans- “Markovian limit” when the friction kernel assumes the mission factor times the difference of corresponding Bose ohmic form γ(t) → 2γδ(t). For a correlation-free prepa- functions39. ration, the initial total density matrix is given by the Furthermore, the QLE concept can also be extended product ρ = ρ (t )ρ , where ρ (t ) is the initial sys- T S 0 bath S 0 to fermionic systems coupled to electron reservoirs and tem density matrix. The density matrix of the bath alone which, in addition, may be exposed to time-dependent assumes canonical equilibrium, i.e. driving40. The corresponding Gaussian quantum noise is N 2 1 pi mi 2 2 now composed of fermion annihilation operators. ρbath = exp −β + ω x , (25) N 2m 2 i i i=1 i ! X with N denoting a normalization constant. The statistical properties of the random force η(t) then B. Consequences of time-reversal symmetry follow immediately: η(t) is a stationary Gaussian opera- tor noise obeying Let us now discuss some further properties of this QLE. not <η(t)>ρbath = 0 (26) If the potential V (q,t) in (20) does explicitly de- 1 pend on time t, the dynamics of the full Hamiltonian S (t − s)= <η(t)η(s)+ η(s)η(t)> bath (27) (20) obeys time reversal symmetry. It is thus an imme- ηη 2 ρ diate consequence that the reduced dynamics must be N 2 ¯h c ¯hωi invariant under time reversal as well. This must hold = i cos ω (t − s) coth . 2 m ω i 2kT true despite the fact that the QLE has been constructed i=1 i i X to allow for a description of quantum dissipation. It is Being an operator-valued noise, its commutator does not thus instructive to see how the validity under time rever- vanish sal emerges from the contracted description in terms of N 2 the QLE in (21). ci [η(t), η(s)] = −i¯h sin ωi(t − s) . (28) m ω Given the time of preparation t0, reversing the time i=1 i i X amounts to substituting time t by t0 − (t − t0)=2t0 − t. Setting for the initial position operator q(t0)= q0, the Using again the random force η(t) we can recast the QLE last expression in (27) is also valid for the noise correla- dynamics after the time reversal into the form tion Sξξ(t) of the noise force ξ(t) provided the average is − now taken with respect to a bath density matrix which 2t0 t dV (q) contains shifted oscillators. The initial preparation of the Mq¨(2t0 − t)+ M ds γ(2t0 − t − s)q ˙(s)+ dq bath is then given by the new density matrixρ ˆbath; Zt0 2 = ξ(2t0 − t) 1 pi ρˆbath = exp − β (29) N 2m = η(2t0 − t) − γ(2t0 − t − t0)q(t0) . (30) i i X 2 2 miωi ci + x − q0 . Setting next x(t)= q(2t0−t) and observing thatx ˙(t)= 2 i m ω2 i i −q˙(2t0−t), x¨(t)=¨q(2t0−t), we find after the substitution 7 of the integration time u =2t0 − s from (30) the result The literature is full of various such attempts wherein one approximates the quantum features by correspond- t dV (q) ing colored classical noise sources, e.g. see Refs. 43,44, Mx¨(t)+ M du γ(u − t)x ˙(u)+ 45. Such schemes work at best near a quasi-classical t0 dq Z limit44,46, but even then care must be exercised. For = ξ(2t − t) 0 example, for problems that exhibit an exponential sen- = η(2t0 − t) − γ(t0 − t)x(t0) . (31) sitivity, such as the dissipative decay of a metastable state discussed in the next section, such an approach Noting that the damping kernel is an even function gives no exact agreement with the quantum dissipative of its argument, γ(u − t) = γ(t − u), and that x(t0) = theory31,34. It is only in the classical high temperature q(t0), we find upon changing all signs of the initial mo- limit, where the commutator structure of quantum me- menta pi(t0) →−pi(t0) for the noise forces the relations chanics no longer influences the result. Perfect agreement η(2t0 − t)= η(t) and ξ(2t0 − t)= ξ(t). We conclude that is only achieved in the classical limit. the time reversed motion x(t) = q(2t0 − t) indeed obeys The study of quantum friction in a nonlinear quan- again a QLE of the form (21). This even holds true in tum system by means of the QLE (21) is plagued by the Markovian limit where γ(t − s)=2γδ(t − s) as one the fact that the nonlinearity forbids an explicit solution. can convince oneself by smearing out the delta function This solution, however, is needed to obtain the statistical symmetrically. The QLE then reads for all times t properties such as mean values and correlation functions. This (unknown) nonlinear response function also deter- dV (q) Mq¨(t) + sgn(t − t0)Mγq˙(t)+ = ξ(t) , (32) mines the derivation of the rate of change of the reduced dq density operator, i.e. the QME, and its solution of the open quantum system. where sgn(x) denotes the sign of x. The very fact that the QLE acts in full Hilbert space of The dissipation is reflected by the fact that for times system and environment also needs to be distinguished t>t0 the reduced dynamics for q(t) exhibits a damped from the classical case of a generalized Langevin equa- (quantum)-behavior on a time scale given by the Poincar´e tion. There, the stochastic dynamics acts solely on the 41,42 recurrence time ; the latter reaches essentially infinity state space of the system dynamics with the (classical) for all practical purposes if only the bath consists of a noise properties specified a priori47. sizable number of bath oscillator degrees of freedom31,41. • The quantum noise correlations can, despite the explicit microscopic expression given in (27), be C. Subtleties and pitfalls expressed solely by the macroscopic friction kernel γ(t). The use of the generally nonlinear QLE (21) is limited This result follows upon noting that the Laplace trans- in practice for several reasons. Moreover, the applica- formγ ˆ(z) of the macroscopic friction assumes with Rez > tion of the QLE bears some subtleties and pitfalls which 0 the form must be observed when making approximations. Some important features are: 1 N c2 1 1 γˆ(z)= i + . (33) 2M m ω2 z − iω z + iω • The QLE (21) is an operator equation that acts in i=1 i i i i the full Hilbert space of system and bath. The cou- X + pling between system and environment also implies With help of the well known relation 1/(x + i0 ) = an entanglement upon time evolution even for the P (1/x) − iπδ(x) we find that case of an initially factorizing full density matrix. + Together with the commutator property of quan- Reˆγ(z = −iω +0 ) tum Brownian motion, see eq. (28), we find that N π c2 the reduced, dissipative dynamics of the position = i [δ(ω − ω )+ δ(ω + ω )] . (34) 2M m ω2 i i operator q(t) and momentum operator p(t) obey i=1 i i the Heisenberg uncertainty relation for all times. X By means of (27) we then find the useful relation This latter feature is crucial. For example, the non- Markovian (colored) Gaussian quantum noise with real- Sξξ(t)= Sηη(t) (35) ∞ valued correlation Sξξ(t) = Sξξ(−t) cannot simply be M + ¯hω substituted by a classical non-Markovian Gaussian noise = dωReˆγ(−iω +0 )¯hω coth cos(ωt) . π 0 2kT force which identically obeys the correlation properties Z of (Gaussian) quantum noise ξ(t). An approximation of In the classical limit this relation reduces, indepen- this type clearly would not satisfy the commutator prop- dent of the preparation of the bath with ρ orρ ˆ, to the erty for position and conjugate momentum of the system non-Markovian Einstein relation Sξξ(t)= MkTγ(t). The degrees of freedom. relation (35) is by no means obvious: It implies that a 8 modelling of quantum dissipation is possible in terms of equation dynamics of the corresponding reduced density macroscopic quantities such as the friction kernel γ(t) matrix.52,53 and the temperature T . For other coupling schemes be- The solution of the QLE involves the explicit time- tween system and bath we generally can no longer express dependence of both the friction and the potential forces. the correlation of quantum noise exclusively in terms of These in turn determine the statistical properties of the macroscopic transport coefficients. As an example we density matrix. As a consequence, the friction force en- mention the coupling of the system to a bath of two-level ters the QME in a rather complex manner. This can systems (spin bath) rather than to a bath of harmonic already be verified explicitly for a parametric dissipative oscillators48. oscillator dynamics, where the time-dependent driving Note also the following differences to the classical sit- enters the diffusive kinetic evolution law of the reduced uation of a generalized Langevin equation. density operator or its equivalent Wigner transform52,53. For the bilinear system-bath interaction with the bath • The quantum noise ξ(t) is correlated with the ini- composed of harmonic oscillators it was possible to inte- tial position operator q(t ).49 This feature that 0 grate out the degrees of freedom of the bath explicitly. For the description of quantum dissipative systems it V is important to realize the analogy between the propa- gator and the equilibrium density matrix. The latter is ωb obtained by replacing t by −i¯hβ. We thus obtain from V (36) the path integral representation of the equilibrium b density matrix q(¯hβ)= q q0 ′ 1 1 q ρ (q, q )= Dq exp − SE[q] , (37) β Z ′ ¯h β Zq(0) = q ω0 where Zβ is the partition function. This integral is called imaginary-time path integral in contrast to the real-time path integral (36). Note that in (37) the action S[q] has been replaced by the so-called Euclidean action SE[q] FIG. 3 Cubic potential as defined in Eq. (40). which is obtained by changing the sign of the poten- tial term as a consequence of the transition to imaginary B. Application: The dissipative decay of a metastable state times. In imaginary time we therefore have to consider the motion in the inverted potential. A local potential minimum may be metastable due to The connection between classical and quantum me- the environmental coupling and quantum effects. Cor- chanics becomes particularly apparent in the path inte- respondingly, there are two escape mechanisms: ther- gral formulation. The dominant contribution to the inte- mal activation which dominates at high temperatures grals in (36) and (37) arise from the stationary points of and quantum tunnelling which becomes important at low the action, i.e. the classical paths. Quantum effects have temperatures. To be definite, we consider the cubic po- their origin in fluctuations around the classical paths. tential Therefore, it is useful to decompose a general path into M q the classical path and a fluctuation around it. Expanding V (q)= ω2q2 1 − (40) 2 0 q the action in powers of the fluctuations the second order 0 term yields the leading quantum corrections. Higher or- which is depicted in Fig. 3. The barrier height is given by der terms are often neglected within a semiclassical ap- V = (2/27)Mω2q2 and, in this special case, the barrier proximation which becomes exact for linear systems. b 0 0 angular frequency ωb equals the well angular frequency In the previous section we have derived an effective ω0. equation of motion for the system variable by eliminating In Fig. 4, the decay rate is shown in an Arrhenius plot. the external degrees of freedom. The same procedure At the so-called crossover temperature T0, see Eq. (42) may of course also be carried out within the path integral 6,31,34,58 below, there is a rather distinct transition between the formalism . The influence of the environment is thermal regime on the left side and the quantum regime then contained in an effective action which has to be on the right side59. Furthermore, we observe that the added to the action of the system and which in imaginary 6,31,34 thermal regime is larger for stronger damping, i.e. the time is given by system becomes more classical. While a real time approach to dissipative decay is hβ¯ hβ¯ 1 feasible61,62,63,64, a simpler alternative is provided by an S [q]= − dτ dσk(τ −σ)[q(τ)−q(σ)]2 (38) eff 4 imaginary time calculation where the partition function Z0 Z0 Zβ is considered. Since the potential (40) is not bounded where from below, it is no surprise that strictly speaking Zβ does not exist. From the path integral point of view +∞ M there exists an unstable fluctuation mode around the bar- k(τ)= |νn|γˆ(|νn|) exp(iνnτ) (39) rier which leads to a saddle point in function space. One ¯hβ n=−∞ can circumvent this difficulty by performing the integra- X tion in the direction of steepest descent. The partition andγ ˆ(z) denotes the Laplace transform of the damping function and as a consequence also the free energy then kernel γ(t). The effective action (38) is clearly nonlocal acquire an imaginary part which may be related to the and can thus not be expressed in terms of a potential. decay rate31,32. For details of this relation we refer the If the potential renormalization term in the Hamiltonian reader to the discussion in Ref. 65. (20) would be absent, there would have been a local con- The transition between thermal and quantum regime tribution in (38). The selfinteraction of the paths in- can be well understood within the path integral picture duced by (38) via the kernel (39) decays for ohmic damp- by considering the possible classical paths of durationhβ ¯ ing only algebraically as τ −2 and therefore represents a in the inverted cubic potential. For high temperatures or long range interaction. short imaginary timeshβ ¯ the only classical solutions are 10 0 This method seemingly is superior to any perturba- tive scheme that treats the system-bath coupling to low orders only, such as the weak coupling master equation methodology17,18,19,20. There are recent developments in −40 the strong friction regime, where an alternative descrip- ) 0 tion in terms of a quantum Smoluchowski equation is /ω promising68, see also the contribution by Grabert, Anker- hold and Pechukas in this special issue. ln(Γ −80 A consequent use of the so-called rotating-wave ap- proximations also may entail some danger. It safely can be applied only in the weak coupling regime for resonant situations. We remark that the use of the rotating-wave −120 approximation implies a violation of the Ehrenfest the- 0 20 40 60 orem in the order of γ2,13,69 which is clearly small only −1 ¯hω0/kT in the weak coupling regime, i.e. for γ ≪ ω0 , with ω0 denoting some typical time scale of the system dynamics. FIG. 4 Arrhenius plot for the decay rate of a metastable state. The same remarks apply to the failure of the quantum re- The damping strength varies from the upper to the lower gression theorem13,70,71: Again, the effect might be small curve as γ/2ω0 = 0, 0.5, and 1 (data taken from Ref. 60). for (i) very weak damping, (ii) not too low temperatures obeying kT ≫ ¯hγ and (iii) not too short evolution times. The generalized quantum Langevin equation discussed the constant solutions q = 0 in the well and qb = 2q0/3 in Sec. IV is formally exact for nonlinear quantum sys- at the barrier. Below a temperature given by the positive tems. Its practical use is typically restricted, however, to solution of linear systems for which the response can be evaluated in closed form. This holds true even for time-dependent ν2 + |ν |γˆ(|ν |) − ω2 = 0 (41) 1 1 1 b linear systems for which the response is still linear al- though the evaluation involves the use of numerical Flo- a second fluctuation mode becomes unstable, thereby in- 52 dicating a new classical solution which performs an os- quet theory . The lack of knowledge of this generally cillation around the barrier66. This new solution is asso- nonlinear response function also plagues the evaluation ciated with quantum tunnelling. Therefore, (41) defines of the corresponding generalized master equations. the crossover temperature which for ohmic damping is This problem of obtaining the generalized master equa- given by59 tion from the nonlinear generalized Langevin equation is not solved either for the classical problem with col- 2 1/2 ored noise27. It is also this very problem that limits the ¯h γ 2 γ T0 = + ωb − . (42) practical use of the various variants of recently derived 2πk 4 2 " # stochastic Schr¨odinger approaches12. As discussed above, stronger damping leads to a lower Likewise, the use of nonlinear, but non-stochastic crossover temperature and smaller quantum regime. It Schr¨odinger equations of the type discussed and surveyed thus makes the system more classical. A distinct feature in Refs. 72,73,74 can clearly not describe the time evo- of the dissipative quantum decay in the low temperature lution of a quantum mechanical mixture, nor do these nonlinear deterministic approaches obey, in general, the regime is its algebraic enhancement of the decay rate with 73,74 temperature31. For the case of an ohmic environment Heisenberg uncertainty relation . with a constant friction behavior at low frequencies one There have been repeated attempts since the early days finds a universal T 2-enhancement of both, the prefactor of quantum mechanics to explain quantum phenomena in and the effective action, with the latter dominating the terms of Einstein’s theory of classical diffusion. Early ef- 75 76 exponential rate enhancement31,67. forts in this direction were those of F¨urth , F´enyes , Weizel77 and Favella78. This credo has been popularized later by Nelson10,79 under the label of “Stochastic Me- VI. SUNDRY REMARKS AND CONCLUSIONS chanics”. It can convincingly be demonstrated, however, that a quantum dynamics is quite distinct from a clas- With this work we elucidated the topic of quantum sical Markovian – or even non-Markovian – stochastic Brownian noise which drives the dynamics of open dissi- dynamics11,80. This holds even more so, if one attempts pative quantum systems. We have emphasized the strong to incorporate the quantum dissipation for an open sys- implications that thermal equilibrium and time-reversal tem. symmetry (leading to detailed balance symmetry) im- These sundry remarks thus give clear evidence that the poses on the reduced system dynamics. We also pointed topic of quantum Brownian motion – although 100 years out the advantageous use of the path integral scheme for have passed since Einstein’s cornerstone contribution1– the case of nonlinearity and strong friction. cannot be considered as “solved”. For example, little is 11 presently known also for the description and the role of [11] H. Grabert, P. H¨anggi, and P. Talkner, “Is quantum quantum noise acting in steady state, far from equilib- mechanics equivalent to a classical stochastic process?,” rium situations, i.e. when several baths of different nature Phys. Rev. A 19, 2440–2445 (1979). and/or different temperature are coupled to the nonlin- [12] M. Roncadelli, “Connection between Langevin quantiza- 28 ear system of interest. tion and classical mechanics,” Europhys. Lett. , 379- 384 (1994); H. Kleinert and S. V. Shabanov, “Quan- The latter case is also of salient importance for the tum Langevin equation from forward-backward path in- description of the quantum dynamics of so-called quan- 200 81 tegral,” Phys. Lett. A , 224–232 (1995); L. Di´osi tum Brownian motors . In those applications quan- and W. T. Strunz, “The non-Markovian stochastic tum Brownian noise is utilized in combination with non- Schr¨odinger equation for open systems,” Phys. Lett. equilibrium (classical or quantum) fluctuations to per- A 235, 569–573 (1997); W. T. Strunz, L. Di´osi, and form exploitable work against external bias forces. In N. Gisin, “Open system dynamics with non-Markovian summary, the field of quantum Brownian motion is very quantum trajectories,” Phys. Rev. Lett. 82, 1801–1805 much alive and lots of challenges still need to be ad- (1999); H. P. Breuer, B. Kappler, and F. Petruccione, dressed and mastered. “Stochastic wave function method for non-Markovian quantum master equations,” Phys. Rev. A 59, 1633– 1643 (1999); J. T. Stockburger and H. Grabert, “Non- Markovian quantum state diffusion,” Chem. Phys. 268, Acknowledgments 249–256 (2001); J. Shao, “Decoupling quantum dissipa- tion interaction via stochastic fields,” J. Chem. Phys. PH gratefully acknowledges financial support by the 120, 5053–5056 (2004). DAAD-KBN (German-Polish project Stochastic Com- [13] P. Talkner, “The failure of the quantum regression hy- plexity), the Foundation for Polish Science [Fundacja na pothesis,” Ann. Phys. (N.Y.) 167, 390–436 (1986); P. Rzecz Nauki Polskiej], the Deutsche Forschungsgemein- Talkner, “Untersuchungen irreversibler Prozesse in quan- schaft via grant HA 1517/13-4 and the collaborative re- tenmechanischen Systemen,” Ph.D. thesis (Universit¨at search grants SFB 486 and SFB 631. Stuttgart, 1979). [14] M. Planck, “Eine neue Strahlungshypothese,” Verh. Dt. Phys. Gesell. 13 138–148 (1911). [15] H. Grabert, U. Weiss, and P. Talkner, “Quantum theory References of the damped harmonic oscillator,” Z. Phys. B 55, 87–94 (1984). [1] A. Einstein, “Uber¨ die von der molekularkinetischen [16] P. S. Riseborough, P. H¨anggi, and U. Weiss, “Exact re- Theorie der W¨arme geforderte Bewegung von in ruhen- sults for a damped quantum-mechanical harmonic oscil- den Fl¨ussigkeiten suspendierten Teilchen,” Ann. Phys. lator,” Phys. Rev. A 31, 471–478 (1985). (Leipzig) 17, 549–560 (1905). [17] H. Grabert, Projection operator techniques in nonequilib- [2] H. Nyquist, “Thermal agitation of electric charge in con- rium statistical mechanics, Springer Tracts in Mod. Phys. ductors,” Phys. Rev. 32, 110–113 (1928). 95, 1–164 (1982). [3] J. B. Johnson, “Thermal agitation of electricity in con- [18] F. Haake, Statistical treatment of open systems by gen- ductors,” Phys. Rev. 32, 97–109 (1928). eralized master equations, Springer Tracts in Mod. Phys. [4] It is worthwhile to recall that the validity of the equiparti- 66, 98–168 (1973). tion theorem is restricted to classical statistical mechan- [19] H. Spohn, “Kinetic equations from Hamiltonian dynam- ics. ics: Markovian limits,” Rev. Mod. Phys. 52, 569–615 [5] H. B. Callen and T. A. Welton, “Irreversibility and gen- (1980). eralized noise,” Phys. Rev. 83, 34–40 (1951). [20] R. Alicki, “General theory and applications to unstable [6] H. Grabert, P. Schramm, and G.-L. Ingold, “Quantum particles,” in: Quantum dynamical semigroups and ap- Brownian motion: The functional integral approach,” plications Lect. Notes Phys. Vol. 286, Chaps. II and III Phys. Rep. 168, 115–207 (1988). (Springer, Berlin, 1987). [7] T. Dittrich, P. H¨anggi. G.-L. Ingold, B. Kramer, G. [21] G.-L. Ingold and H. Grabert, “Sluggish decay of prepara- Sch¨on, and W. Zwerger, Quantum Transport and Dis- tion effects in low temperature quantum systems,” Lect. sipation (Wiley-VCH, Weinheim, 1998). Notes Math. 1442, 219–230 (1990). [8] M. Grifoni and P. H¨anggi, “Driven quantum tunneling,” [22] R. Jung, G.-L. Ingold, and H. Grabert, “Long-time tails Phys. Rep. 304, 229–358 (1998). in quantum Brownian motion,” Phys. Rev. A 32, 2510– [9] V. B. Magalinski˘ı, “Dynamical model in the theory 2512 (1985). of the Brownian motion,” Sov. Phys. JETP 9, 1381– [23] A. Hanke and W. Zwerger, “Density of states of a 1382 (1959) [J. Exptl. Theoret. Phys. (U.S.S.R.) 36, damped quantum oscillator,” Phys. Rev. E 52, 6875– 1942–1944 (1959)]; R. Benguria and M. Kac, “Quantum 6878 (1995). Langevin equation,” Phys. Rev. Lett. 46, 1–4 (1981); [24] G.-L. Ingold, “Path integrals and their application to dis- G. W. Ford and M. Kac, “On the quantum Langevin sipative quantum systems,” Lect. Notes Phys. 611, 1–53 equation,” J. Stat. Phys. 46, 803–810 (1987); G. W. Ford, (2002). J. T. Lewis, and R. F. O’Connell, “Quantum Langevin [25] H. Mori, “Transport, collective motion and Brownian equation,” Phys. Rev. A 37, 4419–4428 (1988). motion,” Progr. Theor. Phys. 33, 423–455 (1965); K. [10] E. Nelson, “Derivation of the Schr¨odinger equation Kawasaki, “Simple derivations of generalized linear and from Newtonian mechanics,” Phys. Rev. 150, 1079–1085 nonlinear Langevin equations,” J. Phys. A 6, 1289–1295 (1966). (1973); S. Nordholm and R. Zwanzig, “Systematic deriva- 12 tion of exact generalized Brownian-motion theory,” J. A. Schmid, “The quasiclassical Langevin equation and Stat. Phys. 13, 347–371 (1975). its application to the decay of a metastable state and [26] H. Grabert, P. H¨anggi, and P. Talkner, “Microdynamics to quantum fluctuations,” J. Stat. Phys. 59, 885–934 and nonlinear stochastic processes of gross variables,” J. (1990). Stat. Phys. 22, 537–552 (1980). [47] P. H¨anggi and H. Thomas, “Stochastic processes: Time [27] P. H¨anggi, “Generalized Langevin equations: A useful evolution, symmetries and linear response,” Phys. Rep. tool for the perplexed modeller of nonequilibrium fluctu- 88, 207–319 (1982). ations?,” Lect. Notes Phys. 484, 15-22 (1997). [48] J. Shao and P. H¨anggi, “Decoherent dynamics of a two- [28] I. R. Senitzky, “Dissipation in quantum mechanics. The level system coupled to a sea of spins,” Phys. Rev. Lett. harmonic oscillator,” Phys. Rev. 119, 670–679 (1960). 81, 5710–5713 (1998). [29] P. Ullersma, “An exactly solvable model for Brownian [49] P. Schramm, R. Jung, and H. Grabert, “A closer look motion,” Physica 32, 27–55, 56–73, 74–89, 90–96 (1966). at the quantum Langevin equation: Fokker-Planck equa- [30] R. Zwanzig, “Nonlinear generalized Langevin equations,” tion and quasiprobabilities,” Phys. Lett. 107A, 385–389 J. Stat. Phys. 9, 215–220 (1973). (1985) [31] P. H¨anggi, P. Talkner, and M. Borkovec, “Reaction-rate [50] H. Grabert, P. Talkner, and P. H¨anggi, “Microdynam- theory: Fifty years after Kramers,” Rev. Mod. Phys. 62, ics and time-evolution of macroscopic non-Markovian 251–341 (1990). systems,” Z. Phys. B 26, 389–395 (1977); H. Grabert, [32] A. O. Caldeira and A. J. Leggett, “Quantum tunnelling P. Talkner, P. H¨anggi, and H. Thomas, “Microdynamics in a dissipative system,” Ann. Phys. (N.Y.) 149, 374–456 and time-evolution of macroscopic non-Markovian sys- (1983); Ann. Phys. (N.Y.) 153, 445 (1984). tems II,” Z. Phys. B 29, 273–280 (1978). [33] For the explicit details of this calculation see Refs. 24 and [51] K. M. F. Romero, P. Talkner, and P. H¨anggi, “ Is the dy- 27. namics of open quantum systems always linear?,” Phys. [34] U. Weiss, Quantum Dissipative Systems, second edition Rev. A 69, 052109 (2004). (World Scientific, Singapore, 1999). [52] C. Zerbe and P. H¨anggi, “Brownian parametric quantum [35] U. Eckern, G. Sch¨on, and V. Ambegaokar, “Quantum oscillator with dissipation,” Phys. Rev. E 52, 1533–1543 dynamics of a superconducting tunnel junction,” Phys. (1995). Rev. B 30, 6419–6431 (1984). [53] S. Kohler, T. Dittrich, and P. H¨anggi, “Floquet- [36] G. Sch¨on and A. D. Zaikin, “Quantum coherent effects, Markovian description of parametrically driven, dissi- phase transitions, and the dissipative dynamics of ultra pative harmonic oscillator,” Phys. Rev. E 55, 300–313 small tunnel junctions,” Phys. Rep. 198, 237–413 (1990). (1997). [37] G.-L. Ingold and Yu. V. Nazarov, “Charge Tunneling [54] J.Luczka, “Spin in contact with thermostat: Exact re- Rates in Ultrasmall Junctions,” in: Single Charge Tun- duced dynamics,” Physica A 167, 919–934 (1990). neling, ed. by H. Grabert and M. H. Devoret, NATO ASI [55] N. G. van Kampen, “A soluble model for quantum me- Series B, Vol. 294, pp. 21–107 (Plenum, New York, 1992). chanical dissipation,” J. Stat. Phys. 78, 299–310 (1995). [38] U. Z¨urcher and P. Talkner, “Quantum-mechanical har- [56] L.-D. Chang and S. Chakravarty, “Dissipative dynamics monic chain attached to heat baths II. Nonequilibrium of a two-state system coupled to a heat bath,” Phys. Rev. properties,” Phys. Rev. A 42, 3278–3290 (1990). B 31, 154–164 (1985); F. Sols and F. Guinea, “Bulk and [39] D. Dvira, A. Nitzan, and P. H¨anggi, “Thermal conduc- surface diffusion of heavy particles in metals: A path- tance through molecular wires,” J. Chem. Phys. 119, integral approach,” Phys. Rev. B 36, 7775–7785 (1987); 6840–6855 (2003). P. Hedeg˚ard and A. O. Caldeira, “Quantum dynamics [40] S. Camalet, J. Lehmann, S. Kohler, and P. H¨anggi, “Cur- of a particle in a fermionic environment,” Phys. Scr. 35, rent noise in ac-driven nanoscale conductors,” Phys. Rev. 609–622 (1987). Lett. 90, 210602 (2003); S. Camalet, S. Kohler, and P. [57] R. P. Feynman, “Space-time approach to non-relativistic H¨anggi, “Shot-noise control in ac-driven nanoscale con- quantum mechanics,” Rev. Mod. Phys. 20, 367–387 ductors,” Phys. Rev. B 70, 155326 (2004). (1948); R. P. Feynman and A. R. Hibbs, Quantum Me- [41] P. C. Hemmer, L. C. Maximon, and H. Wergeland, “Re- chanics and Path Integrals (McGraw-Hill, New York, currence time of a dynamical system,” Phys. Rev. 111, 1965). 689–694 (1958). [58] R. P. Feynman and F. L. Vernon, “The theory of a gen- [42] P. Mazur and E. Montroll, “Poincar´ecycles, ergodicity, eral quantum system interacting with a linear dissipative and irreversibility in assemblies of coupled harmonic os- system,” Ann. Phys. (N.Y.) 24, 118–173 (1963). cillators,” J. Math. Phys. 1, 70–84 (1960). [59] P. Hanggi, H. Grabert, G.-L. Ingold, and U. Weiss, [43] S. A. Adelman, “Quantum generalized Langevin equa- “Quantum theory of activated events in presence of long- tion approach to gas/solid collisions,” Chem. Phys. Lett. time memory,” Phys. Rev. Lett. 55, 761–764 (1985). 40, 495–499 (1976). [60] H. Grabert, P. Olschowski, U. Weiss, “Quantum de- [44] A. Schmid, “On a quasiclassical Langevin equation,” J. cay rates for dissipative systems at finite temperatures,” Low Temp. Phys. 49, 609–626 (1982). Phys. Rev. B 36, 1931–1951 (1987). [45] S. K. Banik, B. C. Bag and D. S. Ray, “Generalized quan- [61] G.-L. Ingold, “Anwendung von Funktionalintegralen auf tum Fokker-Planck, diffusion and Smoluchowski equa- Transport- und Relaxationsph¨anomene in dissipativen tions with true probability distribution functions,” Phys. Quantensystemen,” Ph.D. thesis (Universit¨at Stuttgart, Rev. E 65, 051106 (2002); D. Banerjee, B. C. Bag, S. K. 1988). Banik, and D. S. Ray, “Solution of quantum Langevin [62] H. Hofmann und G.-L. Ingold, “Dissipative transport equation: Approximations, theoretical and numerical as- across a parabolic barrier,” Phys. Lett. B 264, 253–258 pects,” J. Chem. Phys. 120, 8960–8972 (2004). (1991). [46] U. Eckern, W. Lehr, A. Menzel-Dorwarth, F. Pelzer, and [63] J. Ankerhold, H. Grabert, and G.-L. Ingold, “Dissipa- 13 tive Quantum Systems with Potential Barrier. General [71] G. W. Ford and R. F. O’Connell, “There is no quan- Theory and Parabolic Barrier,” Phys. Rev. E 51, 4267– tum regression theorem,” Phys. Rev. Lett. 77, 798–801 4281 (1995); J. Ankerhold and H. Grabert, “Dissipative (1996). quantum systems with a potential barrier. II. Dynam- [72] M. D. Kostin, “Friction and Dissipative Phenomena in ics near the barrier top,” Phys. Rev. E 52, 4704–4723 Quantum Mechanics,” J. Stat. Phys. 12, 145 (1975). (1995); J. Ankerhold and H. Grabert, “Dissipative quan- [73] J. Messer, “Friction in quantum mechanics,” Acta Phys. tum systems with a potential barrier. III. Steady State Austriaca 50, 75–91 (1979) and references therein. Nonequilibrium Flux and Reaction Rate,” Phys. Rev. E [74] H. Dekker, “Classical and quantum mechanics of the 55, 1355–1374 (1997). damped harmonic oscillator,” Phys. Rep. 80, 1–112 [64] J. Ankerhold and H. Grabert, “Quantum tunneling and (1981). the semiclassical real time dynamics of the density ma- [75] R. F¨urth, “Uber¨ einige Beziehungen zwischen klassischer trix,” Europhys. Lett. 47, 285–291 (1999). Statistik und Quantenmechanik”, Z. Phys. 81, 143–162 [65] P. H¨anggi and W. Hontscha, “Periodic orbit approach (1933). to the quantum-Kramers-rate,” Ber. Bunsenges. Phys. [76] I. F´enyes, “Eine wahrscheinlichkeitstheoretische Begr¨un- Chem. 95, 379–385 (1991); P. H¨anggi and W. Hontscha, dung und Interpretation der Quantenmechanik,” Z. “Unified approach to the quantum-Kramers reaction Phys. 132, 81–106 (1952). rate,” J. Chem. Phys. 88, 4094–4095 (1988). [77] W. Weizel, “Ableitung der Quantentheorie aus einem [66] P. S. Riseborough, P. H¨anggi, and E. Freidkin, “Quantum klassischen, kausal determinierten Modell,” Z. Phys. 134, tunneling in dissipative media: Intermediate-coupling- 264–285 (1953); W. Weizel, “Ableitung der Quantenthe- strength results,” Phys. Rev. A 32, 489–499 (1985). orie aus einem klassischen Modell. II.,” Z. Phys. 135, [67] H. Grabert, U. Weiss, and P. Hanggi, “Quantum tunnel- 270–273 (1953). ing in dissipative systems at finite temperatures,” Phys. [78] L. F. Favella, “Brownian motions and quantum mechan- Rev. Lett. 52, 2193–2196 (1984). ics,” Ann. Inst. Henri Poincar´e Sect. A, VII, 77–94 [68] J. Ankerhold, P. Pechukas, and H. Grabert, “Strong fric- (1967). tion limit in quantum mechanics: The quantum Smolu- [79] G. C. Ghirardi, C. Omero, and A. Rimini, “Stochastic chowski equation,” Phys. Rev. Lett. 87, 086802 (2001); Interpretation of Qauntum Mechanics - Critical Review,” J. Ankerhold, “Quantum decay rates for driven barrier Riv. Nuovo Cim. 1, 1–34 (1978). potentials in the strong friction limit,” Phys. Rev. E [80] M. S. Wang and W. K. Liang, “Comment on “Repeated 64, 060102 (2001); L. Machura, M. Kostur, P. H¨anggi, measurements in stochastic mechanics”,” Phys. Rev. D , P. Talkner, and J.Luczka, “Consistent description of 1875–1877 (1993). quantum Brownian motors operating at strong friction,” [81] R. D. Astumian and P. H¨anggi, “Brownian motors,” Phys. Rev. E 70, 031107 (2004). Phys. Today 55 (11), 33–39 (2002); P. Reimann and [69] G. W. Ford and R. F. O’Connell, “Inconsistency of the P. H¨anggi, “Quantum features of Brownian motors and rotating wave approximation with the Ehrenfest theo- stochastic resonance,” Chaos 8, 629–642 (1998); P. rem,” Phys. Lett. A 215, 245–246 (1996). Reimann, M. Grifoni, and P. H¨anggi, “Quantum ratch- [70] H. Grabert, “Non-linear relaxation and fluctuations of ets,” Phys. Rev. Lett. 79, 10–13 (1997). damped quantum systems,” Z. Physik B 42, 161–172 (1982).β ¯h tures. At sufficiently low temperatures, the poles of the 4
β = 2 2 (12) Mβ ω + ν + |νn|γˆ(|νn|) 0 q n=−∞ n −0.5 X ∆ and −0.75 +∞ 2 2 M ω0 + |νn|γˆ(|νn|)
β according ∞ to (16) is depicted as a function of the temperature T . For γˆ(z)= dt exp(−zt)γ(t) . (14) kT ≫ ¯hω0, the correction becomes negligible. Z0 We note that for strictly ohmic damping the second moment of the momentum (13) exhibits a logarithmic model for the environment, the partition function Z for divergence which can be removed by introducing a fi- the damped harmonic oscillator can be obtained by the nite memory to the damping mechanism. For finite cou- requirement that it generates the second moment of po- 15 pling to the environment, i.e. for finite damping strength sition according to γ, the reduced density matrix (11) obviously does not agree with the canonical density matrix exp(−βH ) at 2 1 d S
β = − ln(Z) . (17) the same temperature, where HS denotes the Hamilto- Mβω0 dω0 nian of the undamped harmonic oscillator. In order to get an idea of the deviation of the true This leads to the product representation of the partition reduced density matrix from the canonical one, we con- function, i.e. sider the leading corrections to the second moment of ∞ 2 the position due to the finite coupling to the environ- 1 νn Z = 2 2 . (18) ment. Expanding in orders of the damping strength γ, ¯hβω0 ν + ν γˆ(ν )+ ω n=1 n n n 0 we obtain for ohmic damping Y
β(γ) γ 2 =1+ ∆q + O(γ ) (15) transparent if one relates it to a density of states ρ(E)