<<

arXiv:chao-dyn/9703012v1 21 Mar 1997 en fsvrldffrn hois o ntne h in- the instance, for theories, different several of means decoherence the of enumerating elements and functional. of histories necessity possible the the all of because be cannot done Wigner computation on efficiently numerical based that maps and high distributions the the out” quantum For “smears the that noise suggested difficult. also low- Brun the extremely limit, in temperature is calculations limit performing inte- noticed, temperature path he the As involves ver- gral. through which quantum oscillator approach the Duffing decoherence derive damped the and to environ- forced tried the the has of of [8] sion dimension Brun the the [7]. regarding ment of problem effects a always method the and has operators diffusion of simple by state approximation environment struc- complicated quantum an the demands any general, the often out In and smears chaotic so ture. the and between quantum behavior the periodic sin- that the a found “kicks” with and ensemble noise simulated the of They environment member gle one [6]. of form behavior Lindblad the assumed using of the by on equation based oscillator master model non-linear diffusion state driven stud- quantum and the [5] damped Ralph a and ied Spiller to Recently, difficult extremely numerically. is com- solve that the equation instance, to master for used inevitable via, monly seems processes it stochastic since difficult with system relatively deal quantum a is in Dissipation treat and interact to general dissipative. in These open and are noisy thus Hamiltonian. and environment nor their with isolated electro- neither of cavity however,are measurements), a quantum in NMR fields, atom magnetic SQUID an (e.g., junction, Josephson systems with mechanical has sys- dissipative quantum work of Many chaos little quantum tems. very at looking contrast, in By done been [1–4]. extensively ied 05.45.+b,03.65.Sq,05.40.+j,42.50.Lc numbers: PACS certa counterpart. is classical the chaos the that while in system show forbidden this studies in Numerical exists chaotic dynamics. only the in appears double- force term noise a driving quantum with A equations chosen. quan- estab- potential Duffing well we Through quantum theorem, the oscillators. explicitly Ehrenfest’s harmonic lish and of equations bath Langevin heat tum a with it h isptv unu ytmhsbe tde by studied been has system quantum dissipative The stud- been has systems Hamiltonian of chaos Quantum isptv unu ytmi rae eeb coupling by here treated is system quantum dissipative A eateto hsc n etrfrSaitclMechanics Statistical for Center and of Department unu hoi trco nadsiaiesystem dissipative a in attractor chaotic Quantum h nvriyo ea,Asi,Txs78712 Texas Austin, Texas, of University The .VnetLuadWlimC Schieve C. William and Liu Vincent W. inly 1 H of is bath Hamiltonian a complete with interacting the [9–11], atom Then an photons. of equilibrium this system of a example is familiar model of The assembly an [9–11]. assumed oscillators is harmonic here many which of freedom, bath” of “heat degrees external an with interacting linearly n naoedmninltm-needn one po- bounded time-independent tential one-dimensional a in ing force. impossible driving the external to is an due chaos of system Duffing absence whereas classical ro- model corresponding the this chaotic for in that show exist and now bustly are obtain formal- results to Langevin Numerical possible of theorem. system Ehrenfest’s means a and by ism for potential equations write double-well and Duffing of [9–11] quantum dissipative Refs. a of the that study explicitly to will we similar model paper, quantum this In double- found a potential. with well was system (Hamiltonian) chaos conservative the quantum for theorem (apparent) Ehrenfest’s ostensible via where motion of from equations pro- directly Heisenberg [3] dynamics semiclassical Schieve of system-bath derivation and a the posed Pattanayak of above Recently, of model Most generic interaction. one. a a classical considered in the given theories to explicitly similar be quite can quan- system fashion of the dynamics formal- the for since Langevin operators us tum quantum for interesting the more these, is [5–7,11], ism Among equations master [12]. the etc. [11], quantum equations the [10], [9] Langevin motion Vernon Brownian to and application Feynman its and of approach functional fluence nwihw aeasmdta l amncoscillators mass harmonic quency all same that the assumed possess have we which in oitouedsiain emytk h system the take may we dissipation, introduce To ie omtto eain mlcti 1 r [ are (1) in implicit the relations and commutation times system the tween i [ˆ approximation Langevin quantum distribution the g frequency derive continuous standard a to Assuming the for [11] equations. apply motion Ref. may we of of Then procedure equations system. Heisenberg complete the the down write to p ¯ h, ( l ω od hs e scnie atceo ntms mov- mass unit of particle a consider us let this, do To , = [ p fhroi siltr n sn h rtMarkov first the using and oscillators harmonic of ) ˆ Q, ˆ n P n [ˆ and 0 = ] ˆ 2 x ˆ V 2 n ω ( + [ = ] n Q ˆ V ihisHamiltonian, its with ) n that and ( Q Q, ˆ ˆ + ) x ope System, Complex & p ˆ l n , p ˆ [ = ] X n n = ] C ( n , P m ˆ i 2 p ˆ ¯ hδ m steculn osatbe- constant coupling the is n 2 x ˆ u a aedffrn fre- different have may but ln n n + [ = ] o ti straightforward is it Now . hoclao.Teequal The oscillator. th 1 2 mω , P ˆ H n 2 sys p ˆ (ˆ x n 0 = ] n = − 1 2 mω C P ˆ , n 2 [ˆ x Q ˆ n 2 + Q, l ˆ , ) V 2 x ˆ ) P H n ˆ ( = ] Q = ] sys ˆ , ) . (1) 2 ˙ g(ω)C (ω) 2γ hQˆi = hPˆi, (6a) 2 ≡ (2) mω π ˙ hPˆi = −hV ′(Qˆ)i− γhPˆi + η(t). (6b) where γ is assumed constant, we then have ˙ We now expand the equations around the centroid by Qˆ = P,ˆ (3a) ˆ (n) ˜n (n) using the identity F (Q)= n F Q /n!, where F = ˙ ′ n n Pˆ = −V (Qˆ) − γPˆ + ξˆ(t), (3b) ∂ F (Qˆ)/∂Qˆ | ˆ ˆ , and for a double-well potential of Q=hQi P V (Qˆ)= − 1 aQˆ2 + 1 bQˆ4 obtain the closed system of the where ξˆ(t) is the operator due to the 2 4 stochastic differential equations heat bath and γ defined by (2) clearly represents the constant damping coefficient. Suppose that the system Q˙ = P, (7a) and the bath are initially independent (at t → −∞) so P˙ = aQ − bQ3 − 3bQ¯hµ − γP + η(t), (7b) that the complete density operator can be written into ρ = ρsys ⊗ ρb, and that the bath is initially thermal, µ˙ = α, (7c) ρ ∼ exp(−H /k T ). Then, ξˆ(t) has the properties (see, 1+ α2 b b B α˙ = +2µ(a − 3bQ2) − 6b¯hµ2 − γα. (7d) for example, [11]) 2µ

hξˆ(t)i =0, (4a)Here we have written Q, P for hQˆi, hPˆi. We remind the ∞ ′ 2γ¯h ¯hω ′ reader that all physical must be obtained by h[ξˆ(t ), ξˆ(t)]+i = dωω coth( )cos ω(t − t), (4b) π 0 2kBT averaging over stochastic noise η(t). When damping con- Z stant γ is set to zero, the system is Hamiltonian with 2 2 in which [···]+ denotes an anticommutator. The average P π a 2 2 1 b 4 4 Hextended = + − (Q + ρ )+ 2 + (Q +3ρ + h· · ·i is over all bath variables. The operator nature of 2 2 2 8ρ 4 6¯hQ2ρ2), for which µ → ρ2 and α → 2ρπ in (7). The ξˆ(t) can be reduced by using the strategy of the adjoint above equations of motion can be reduced to those of commutative representation [13]. We can define a new Ref. [3], in which Hamiltonian semiquantal chaos was re- ′ 1 ˆ ′ operator η(t) by η(t)ρ(t ) ≡ 2 [ξ(t),ρ(t )]+ for all t and ′ ′ ported. t , which yields [η(t), η(t )] = 0. This means that η(t) is Also, in the ¯h → 0 the first two equations a c-number function of time. Let us replace the opera- ˆ of (7) decouple from the fluctuation variables (µ, α) and tor ξ(t) by the c-number η(t) in (3). Its 1- and 2-point the noise term vanishes as T → 0 [19]. The well-known correlation functions are given by (4). In the following classical Duffing equations without external driving are we will restrict ourselves in low temperature limit (not recovered (for T → 0) discussed in [8]), T → 0, in which quantum effects are most important. In this limit the noise η(t) is given by Q˙ = P, (8a) ˙ 3 hη(t)i =0, (5a) P = aQ − bQ − γP. (8b) γ¯h ∞ hη(t)η(t′)i = dωω cos ω(t′ − t) This shows that (7) are indeed the quantum analog of π Z0 Duffing equations with the presence of a quantum noise γ¯h 1 = − . (5b) term, which serves as a Langevin driving force. Now a π (t − t′)2 question perhaps rises upon how effective our equations (7) could be to give the description of exact quantum dy- To obtain this, an exponential cutoff (e−ǫω) was used, namics. The procedure of using Gaussian wavepacket to letting ǫ → 0 after integration [14]. describe the motion of a particle may be subject to large While the dependence of bath operators has been elim- errors and will even break down after a certain time scale inated in it, the quantum Langevin equations (3) are [20]. This has yet to be decided. It has been argued, still operator equations and are as difficult as that in for example, by Heller [17] who has extensively studied the deterministic quantum . To make fur- quantum chaos without noise that the generalized Gaus- ther progress, we make an additional assumption that sian approximation works extremely well. Also, recently the wave packet of the system can be described by Ashkenazy et al [4] have computed the time development the squeezed [15]. Then we have the of the in the presence of a potential bar- relations [3] hQ˜2mi = (2m)!(¯hµ)m/m!2m, hQ˜2m+1i = rier in a bounded well for a long time. They numerically 0, hP˜2i =h ¯(1 + α2)/4µ, hQ˜P˜ + P˜Q˜i =hα, ¯ where confirmed the appearance of Hamiltonian chaos due to O˜=Oˆ − hOˆi henceforth and h· · ·i denotes the expectation tunneling first suggested in Ref. [3]. value. It has been shown that these are exactly equiva- In order to proceed with simulations of the stochas- lent to those derived from the generalized Gaussian wave tic differential equations (7), we write [11] η(t) = functions [16–18]. The equations of motion for the cen- ∞ |ω|γh¯ −iωt troid of a wave packet representing the particle are given −∞ dω ν(ω) 2π e , where ν(ω) is a random func- from (3) by tion with the properties: hν(ω)i = 0, hν(ω)ν(ω′)i = R q

2 δ(ω − ω′), ν(ω) = ν(−ω). One can easily check that in their studies for classical stochastic chaos. η(t) generated in this way satisfies (5). It follows that To conclude we note that the quantum Duffing equa- (7) can be numerically solved for each realization of the tions in low temperature limit have been explicitly es- random process η(t) by the Runge-Kutta method for the tablished from the quantum Langevin equations. These ordinary differential equations once the random sequence equations manifestly display great advantage of numer- of ν(ω) has been generated. Quantum chaotic attractors ical computation in comparison with others [5,8]. Nu- are then found for weak damping after some evolution merical results show that these equations exhibit the sta- time (typically around 1000 in our simulations). For one ble chaotic attractor for weak damping while as already realization of stochastic process η(t), Fig. 1(a) shows the known the classical counterparts certainly forbid chaos structure of the quantum chaotic attractor in the phase due to the absence of an external driving force. To our space. The fact that the attractor diffuses out due to knowledge, this is the first study of the dynamical behav- the quantum noise agrees with the previous results [5,8]. ior of the dissipative quantum system without external In particular, one may compare this attractor with that driving showing the quantum chaotic attractor for such of Ref. [5] for the case of strong noise. Lyapunov expo- a system. More detailed studies shall be presented else- nents λs and the fractal dimension Df are calculated for where [19]. each realization by using the standard method as given We should bear in mind that some assumptions have by Ref. [21] and, as seen in Fig. 1(b), are saturated af- been made. First, all results are based on a simplified ter time of the order t = 10000 with a small residual theoretical system-plus-environment model for which we oscillation less than 1% (λlargest ≃ 0.124500). For this assumed that: i) the heat bath consists of an assembly case, we furthur checked the largest of harmonic oscillators; ii) there is a continuous distri- by using an alternative computation algorithm [22] and bution of oscillator frequencies; iii) the coupling of the found that the Lyapunov exponents λlargest ≃ 0.124470 system to the bath operators is linear and the coupling agreeing with the above number up to an accuracy of 1%. constant is a smooth function of oscillator frequency; We have computed 1000 samples (realizations) syn- and iv) the stochastic process is Markovian. All these chronously and then have obtained the distribution of are quite well-known and usual in the study of a dissi- Lyapunov exponents λ and a probability map (see Fig. 2). pative quantum system (see, for example, Refs. [9–11]). As seen in Fig. 2(a), the largest Lyapunov exponents for Second, the squeezed coherent state (or equivalently the all realizations are conclusively positive. Note that the generalized Gaussian wave-packet) has been used to ap- fractal dimension Df is very close to four due to the proximate the true wave function of the system. The full weak damping (γ =0.002) since Df → 4 as γ → 0 where quantum is thus restricted into a truncated the system becomes Hamiltonian. Now let P(Q,t) de- “semiquantal” phase space [3,20]. It has been shown by note the probability of the system at position Q at time Ashkenazy et al. [4] via computer simulation that this ap- t. We make use of a probability map [23] defined by proximation does not break down for a long time. Fully P(Q,t) → P(Q,t + τ) for constant Q and τ to further understanding its validity is still an open task. characterize the behavior of the noisy quantum system. The authors would like to thank Arjendu K. Pat- It has been shown [23] that for regular behavior the prob- tanayak for many stimulating discussion during his re- ability function P(Q,t) does not depend on time once cent visit. the motion of the system is stable and therefore the map P(Q,t) → P(Q,t + τ) should only consist of a single point. Hence, Fig. 2(b) justifies the chaotic behavior of this noisy quantum system as well [24]. By contrast, it is well known that for the corresponding classical Duffing equations (8) only point attractors, describing regular [1] Chaos and quantum physics: 1989 Les Houches Summer , can be allowed due to the absence of the ex- School, ed. by M. -J. Giannoni, A. Voros, and J. Zinn- ternal driving force. While it has been suggested [3,4] Justin, (North-Holland, Amsterdam, 1991). that the tunneling effect induces chaotic behavior in a [2] L. E. Reichl, The Transition to Chaos in Conserva- Hamiltonian quantum system, the dynamical effect of the tive Systems: Quantum Manifestations (Springer-Verlag, quantum noise in (7) should have played a crucial role to New York, 1992). form the stable chaotic attractor in our dissipative quan- [3] A. K. Pattanayak and W. C. Schieve, Phys. Rev. Lett. tum system. Otherwise, the system has to be attracted to 72, 2855 (1994). the bottom of either well with zero momentum. In other [4] Y. Ashkenazy, L. P. Horwitz, J. Levitan, M. Lewkowicz, 75 words, we report here that it is the quantum noise that and Y. Rothschild, Phys. Rev. Lett. , 1070 (1995). [5] T. P. Spiller and J. F. Ralph, Phys. Lett. A 194, 235 leads to the chaotic attractor. Indeed, this is somewhat (1994) and reference therein. reminiscent of the classical fact that either multiplicative [6] G. Lindblad, Commun. Math. Phys. 48, 119 (1976). or additive noise may induce homoclinic crossing and so [7] R. Schack, T. A. Brun, and I. C. Percival, J. Phys. A: chaos, as suggested by Schieve, Bulsara and Jacobs [25]

3 Math. Gen. 28, 5401 (1995). [8] T. A. Brun, Phys. lett. A 206, 167 (1995). [9] R. P. Feynman and F. L. Vernon Jr., Ann. Phys. (N.Y.) 24, 118 (1963). [10] A. O. Caldeira and A. J. Leggett, Physica 121A, 587 (1983). [11] C. W. Gardiner, Quantum Noise (Springer, Berlin, 1991), Chap. 3. [12] Other procedures to deal with dissipation can be found in, for example, G. Parisi, Statistical Field Theory (Addison-Wesley, Redwood, 1988), Ch. 18, and reference therein, for the stochastic formulation of quantum me- chanics. [13] This technique with its applications was discussed in Sects. 3.5, 3.7 and 9.3 of Ref. [11]. [14] This was given in Ref. [11]. Since the high frequency be- havior of the spectral function is proportional to |ω| and so is very badly behaved in noise theory, we have to im- − pose a high frequency cutoff by introducing e ǫω(ǫ ≪ 1). The physical justification for this comes from the form (2), where the factor g(ω)C2(ω) comes into play. This factor must drop off sufficiently rapidly at high frequen- cies because for wavelengths shorter than λcutoff = 2πcǫ a wave of displacements of atoms becomes meaningless. [15] A. M. Perelmov, Generalized Coherent States and their Applications (Springer-Verlag, Berlin, 1986) and refer- ence therein. [16] R. Jackiw and A. Kerman, Phys. Lett. A 71, 158 (1979). [17] E. J. Heller, in Ref. [1]; see also, E. J. Heller and S. Tomsovic, Phys. Today 46 (7), 38 (1993). [18] Y. Tsue, Prog. Theor. Phys. 88, 911 (1992). [19] For details, see, W. V. Liu and W. C. Schieve, (in prepa- ration). [20] A. K. Pattanayak and W. C. Schieve, Phys. Rev. E 50, 3601 (1994). [21] A. Wolf, J. B. Swift, H. L. Swinney, and J. A. Vastano, Physica 16D, 285 (1985). [22] H. G. Schuster, Deterministic Chaos (VCH, Weinheim, 1995), sec. 5.2. [23] T. Kapitaniak, Chaos in System with Noise (World Sci- entific, Singapore, 1988). [24] Because we have only run 1000 realizations (so the ac- curacy of P(Q,t) cannot be better than 0.001), the map seems to contain only finite elements. We argue that for an infinite number of realizations (i.e., in the thermody- namical limit) the map shall consist of infinite number of points. [25] W. C. Schieve and A. R. Bulsara, Phys. Rev. A 41, 1172 (1990); A. R. Bulsara, E. W. Jacobs, and W. C. Schieve, Phys. Rev. A 42, 4614 (1990).

4 (b)

λ 0.8

0.6

0.4

0.2 Lyapunov Exponent

0.0 0.0 20000.0 40000.0 60000.0 80000.0 100000.0 time

FIG. 1. Quantum chaotic attractor for one realization with parameters:h ¯ ≡ 1, a = 10, b = 4, γ = 0.002. (a) The Poincar´esection Q-P taken by α = 0 ± 0.0015 andα ˙ ≥ 0; (b) The of the largest Lyapunov exponent λ. For this realization, the sum of all Lyapunov exponents in −3 phase space i λi ≃−4.011×10 and the fractal dimension Df ≃ 3.968. P

40.0 0.100 (a) (b) 0.080 30.0

) 0.060 τ ) λ 20.0 D(

P(Q,t+ 0.040

10.0 0.020

0.0 0.000 0.00 0.04 0.08 0.12 0.16 0.20 0.000 0.020 0.040 0.060 0.080 0.100 λ P(Q,t)

FIG. 2. The simulation for 1000 realizations with the same parameters as in Fig. 1. (a) The distribution D(λ) of the largest Lyapunov exponents with dλD(λ) = 1. The mean −3 λlargest ≃ 0.127, Df ≃ 3.969 and i λi ≃−3.957×10 . Lya- punov exponents are calculated byR means of Wolf et al [21]. (b) The probability map, takingP constant Q = 1.581 ± 0.05 (near the bottom of one potential well) and τ = 5. This map starts after a relaxation time of 10000.

5