CLASSICAL AND QUANTUM MECHANICS OF THE DAMPED HARMONIC OSCILLATOR H. DEKKER Physics Laboratory TNO, Den Haag. P.O. Box 96864, The Netherlands 50 ig) “RH [NORTH HOLLAND PUBLISHING COMPANY - AMSTERDAMPHYSICS REPORTS (Review Seton of Physics Letters) 80, No.1 (181) IoI12. Nom Holland Polish Company CLASSICAL AND QUANTUM MECHANICS OF THE DAMPED HARMONIC OSCILLATOR Comers: s 1 + Addess August 15, I8I-Agot 15.182; Dept. of Chembty, Manachusts Intute of Techaoogy, Cambridge, Mas M2138, USA Fetroction| Histories! survey Bateman’s dual Hamittonian 231, Cisieal mechanics 32. Quantum mechanic! ypectram 33. Wave packets 34. Summary Kana time-dependent Hamitonian 411, Clasieal mecha 42. Quantum mechanes 43. Quantum fucteation, 44, Summary The semiinfite S. Preliminaries 52. The mechani! moe 83) The elcricl mnt $4, The dasieal diped oxltor 5. Normal mode expansion 56. The quntal damped oxitor 5. Connection wth Callola-Kanai model 5K Senin’ treatment 5.9, Trammssion ine Rotations 5.10, Summary ‘The dual Hamitonian and complex caus f91, Compex symplectic model 62, Madied Boppthcory 63. Sommary Complex phase spc weastivation H. DEKKER Physics Laboratory TNO, Den Ha. P.O. Bor 956, The Netherlands” Received May 1981 3 7 b 4 % » » » 2 u cS » n » u 2 u » ” ” o 1, Preliinanes 112 Reduction ofthe dual phase space 13, Quantization 74. The mater equation 78. Quantum fuctuations 16, Summary The quintum optics osiator AL Intocton remarks 12. The mode Hamionin 1. Reduction ofthe Hiket space 14, The master equation AS, Summary asse's pure stat representation ‘91. The Sehrdinge equation 92. Wave packet solutions 93. Summary Elements of complex Hamiton-Jacbi theory 104. Mase clase dynamis 102. Complex phase space dbmamics 103. Summary 1, Kestin’s moatnear Sehringe equation 1H. lai! mechanics 13. Quantum mechanics 13. Summary 12, Summary ad Ra emahs Relerences {nde c 3 ” e zee “s n xs a ” * ww Single onder for this issue PHYSICS REPORTS (Review Section of Phys Leters) 1, No.1 (181) Copies ol his issue maybe shined a he price given below, All orders should he ent ircty othe Publisher, ere mat he accompanied by check Single ie pie DA. SLM), potage ncoded 0370-1573/81/0000-0000/$28.00 © 1981 North-Holland Publishing Company1H. Dekker, Clas and quantem mechanic of the damped harmon extn a Abert ‘The rltions between varies treatments of the classical inary damped harmonic oar and ts quantization ae investigated, In the course ofa Bisons! survey typical features of the problem ae dacused on the bass of Havas clavsal Hamann and the ybantam mechinial ‘Ssmann-Hasse-Albreeht mals coined bythe Munchen: Garching clear pice group, Is then show how by imposing a estistion om the fase eajctres inorder to coatect he Hamilonan withthe en. the tie dependent Bateman-Mrse-Feshnach-Bopp Harlin leads to the time-dependent Cadila-Kanai Hamiltonian. Canonical quntation of ether formulation entails a woaton of Henenber'spiacigle. By means ofa une treatment of th the electrical and mechanical semi-infinite transmison line. ths detect relsed tthe dnegard of abbtonal ‘quantum factatons that ae intrmucay connected withthe dination The dficuies ofthese models are discussed, Then its proved that the Bateman dual Hamiltonian is connected 1 a recently developed complee simplest formation by a simple canonical transformation, The {uadamestal commutators til ear. Therefore itis demonstrated how. either separating the dua oscillators according to & modified version of ‘opps orginal Weatment or reducing cawial comple phane space by 2PsAlgtaiin over the miror image subopace, a quantum cont sustion obtained that leads to Dekkers master equation foliwing the usual operator algbra. The dissipation induces ational Actuations “The same density operator equation js shiwn to are i quantum optics inthe weal coupling het, Nevt for weak (con, Hanes pute sate conuition is used to derive an equivalent nonnear bat nomconsering frictional Schrodinger equation. It moles 2 paricuat non-Hermitian Hamitonian sbich, i tewrien interns of Dekkers complex variables reveal an elegant extension ofthe cavacal Hamlin Final this formalism ywped to make contat with Kosin’s ud dynamical SchrBdigst-Langesn equation cob then 1, Introduction ‘The obviously irreversible, dissipative behaviour of the vast majority of physical phenomena in the every-day world, in particular when contrasted with the reversible nature of our basic models. has {dled science with an intriguing and apparently enduring problem. In fact, it was not until the times of Newton before it was clearly formulated that the fundamental laws of mechanics indeed were of the reversible type. Dissipation then arises in principle {rom interactions between the actually observed system and another one (or more) into which energy flows in an irreversible manner. As it usually turns out, the details of the structure of the other system (often called the reservoir or bath) are quite immaterial especially in the weak damping limit. However, the calculations from first principles are cumbersome and little transparent as a rule, and the ultimate transition from reversibility to irrever- lity remains & difficult problem the solutions of which can often barely stand a serious criti examination, Luckily, not so difficult to account for dissipative forces in classical mechanics ina Phenomenological manner, at least not in the Newtonian formulation {1.2}. Noteworthy examples are Stokes’ linear frictional force proportional to the velocity v, Coulomb's friction ~efe [3]. Dirac’s radiation damping ~4 [4], and the viscous force ~¥» [5,6]. The phenomenological approach applied to the quantum domain will be the main topic of this article. Unfortunately. as opposed to the Newtonian (vector) formulation, the situation is much less comfortable within the framework of the Hamiltonian (scalar) formalism by which theoretical physics has advanced so much. The deeper reason for this difference can be found in the derivation of the Lagrange-Hamilton variational formulations from D’Alembert’s principle [1.2.7]. Apart from the fact that this derivation can be carried through fruitfully only for systems with holonomic constraints, the latter principle (which in fact is the dynamical generalization of the static principle of virtual work) is au fond equivalent to Newtonian mechanics only if the virtual work of the forces of constraint is zero. And this is clearly not true for friction forces The troubles are immediately manifest for one dimensional systems with Hamiltonians that do not explicitly depend on time, at least if these Hamiltonians are required to represent the physical energy of the system [8-10]. However. the time-independence of the forces and the auvillary conditions is not required for the Hamiltonian theory to be valid. Evidently then, one could have resort to in principle4 1H Dekker, Casueal and quantum mechames ofthe damped harmon oscar three alternatives. First, one could proceed to multi-dimensional systems (see e.g. (11. 12]). Second. the Hamiltonian could be allowed to have an explicit time-dependence (e.g. (13. 14]). Third, one could attempt Hamiltonians that have no relation to the energy [8. 15]. so-called mathematical Hamiltonians or merely generators of the motion, Usually. these possibilities (all of which have indeed been discussed in the literature) are intertwined. In the present text several examples will be encountered in the course of an attempt to elucidate the relations between quite different historical approaches to the problem of the classical damped oscillator. The difficulties with the Hamiltonian description of frictional phenomena become the more prom- inent at the microscopic quantum level as quantum mechanics is an essentially Hamiltonian theory. Of course, and with an even deeper physical significance than in classical mechanics. the fundamental quantum dynamical laws are of the reversible type. A closed system. say the universe. is supposedly described by a Schrédinger equation. The system's dynamics is governed by the Hamiltonian that represents its total energy. and which is a constant of the motion. Principally. as in classical mechanics, dissipation is merely observed for open systems, being only a part of the universe. The dissipation arises from the subsystem’s interactions with the rest of the universe, again often referred to as the reservoir ‘or bath, whether it be thermal or not. Confining oneselves to the subsystem of interest. the explicit microscopic reservoir variables are eliminated from the description by means of projection operator techniques or tracing procedures (see e.g. [16-25]. Usually this extracts from the Liouville-von Neumann equation a (generalized. or Nakajima-Zwanzig) master equation for the reduced density operator. That is, in effect one reduces the considered phase space by means of an appropriate integration procedure over the irrelevant actual dynamical variables of the reservoir. AS a result the sservoir can only be recognized in the description through a few parameters, such ay the friction constant, These features are familiar, for example, from the theory of Nyquist’s formulae for the thermal noise in an electric resistor [6, 26-30]. Once more. the essential point is the observation that the Uetailed microscopic structure of the bath is mostly irrelevant (see also e.g. (31-33). The latter recognition has given impetus to the search for a so-called mesoscopic description of reduced systems [34]. One might approach this idea either philosophically or technically from two points of view. First, physics has often advanced by looking for the simplest possible description that can be considered as comprehensible from an intelligent anthropomorphic position, and that conforms to observation. Why should one carry along part of the way a tremendous amount of information that in fact will be eliminated in due course? Actually, what is the physical significance of such effectively hidden information? Of course, the prime significance could very well be the intellectually satisfying achievement of unification, but otherwise the answers to these questions are definitely nontrivial (see also [35-37}). Second, the ab-initio calculations are not only cumbersome and the approximations involved at the various stages often hard to control. but they are also afflicted with particular problems that have not yet been resolved entirely satisfactory. As a typical example serve the divergencies occurring in certain correlation functions (e.g. {38, 39] and section 5 of the present text), which may have either physical or mathematical origins (e.g. [40-42]). In any case, these deficiencies seem to be intimately connected with the infinite number of degrees of freedom which the reservoir must possess in order for the dissipated energy never to return to the considered subsystem, The existence of a regime where the concept of phenomenological friction applies is easily conceived in the realm of classical physics where the particle under investigation is much more massive than the constituents of the reservoir (see also [43}). Also in Brownian motion the concept is certainly legitimate, albeit that in general there exist at this scale in between macroscopic and purely microscopic phenomena so-called fluctuation-dissipation relations (or generalized Einstein relations: see eg,1H. Dekker, Catsicel and quantum mechanic ofthe damped harmon oxilave 5 (6.21, 24, 28, 30, 34-38, 44-68). The earlier mentioned Nyquist relation is a typical example. In classical thermal systems these relations are of the general type 2AkeT = S,(0) ay where 5,(0) is the de spectral density of the fluctuations and where A is the relaxation constant, By way of simple illustration. consider the Langevin equations for a particle of unit mass (21. 34, 37, 38. 56- 60, 65-67} x (12) B= map +E) a3) The noise term, which here is a real physical fluctuating Newtonian force, is assumed to represent Gaussian white noise, with EM)=0; E+ TEU) 2D8(r) (ay ‘This generates a time-homogencous Wiener process [29, 48. 56, 62, 68-76]; D is the diffusion coefficient, The spectral density of the fluctuations in this case reads S.(w)=2 f (6(2) £0) c08 wr dr = 20, as) so that 5,( 2D. Further. the formal solution for the velocity is PU) =e py + ‘cae ME) dL. (1.6) Rather than treating the adiabatic limit A+ (so that p(t) = £(1)/A, and where the problem is readily reduced to one of position only, leading to the common diffusion equation [37, 44. 68, 77-84). we calculate the velocity autocorrelation function directly from (1.6). By means of (1.4) one obtains in the long time limit 1-+, (lt +7) pO) = (DiAye*” (7) Thus one finds for the mean kinetic energy E=\p)= DPA. (1.8) On the other hand, we know from different reasoning (€.g. [28,30]) that in thermal equilibrium the equipartition law yields E= SkaT per particle per degree of freedom. Combining these results, one indeed arrives at the relation (1-1),* 1H. Dekker, Clarice and quantum mechani ofthe danped harmonic oxi In the classical regime. where kaT > ha. keT is the average energy per unit bandwidth. According to quantum theory (see €.g. (38, 41. 50, 57, 85-94]), when ka T= hw. this quantity should be replaced by Planck's function, P(w)= (Nu + Show, a9) Nu = (exp(htw/kyT)~ 1)", (1.10) which can also be cast in the form P(w)= Ye coth(helkeT) qn ‘These formulae present the energy available from the reservoir per unit bandwidth. In connection with the Brownian motion, (1.2) and (1.3), the precise significance of the frequencies w is not so obvious as there is no resonance phenomenon involved here at some characteristic frequency. in contrast with the harmonic oscillator case. These questions will be further touched upon in the sequel of this paper. For the moment being, it is important to observe that in the pure quantum regime, where kaT A, we have lel$. which stil includes the Hasse (c= +!) and the Albrecht (¢ = 0) Hamiltonians.” The expectation values of the position and momentum operators may be obtained directly from the Schrddinger equation (2.10), They obey the correct Newtonian equations following Ehrentest's theorem: OE 213) (Y= -24Q)- 2°) ey ‘Their general solution is given by (2.5), with w = (Q*- A2)', However, the Hamiltonian (2.7) equals the energy of the tor only if (W)=0. For nonzero A. this is the case only if either ou = xpx + xp) {pXx)=00 o if c= 0, which seems to be an argument in favour of Albrecht’s case.” Defining further ov = (x)~(x)* and op = (p")= H or a ast their chine counter) te inn appre Fut, te ‘that throophout the pecent teat represents the fice onl fede. while we wwe ola reqeeney. This convention ea ome 8 [Lb i the revere of Hse [18,8 one considers force tee mation fr the genetal Sewmane-Hlavc-Albeccht mel ue. =O), the range ¢< 11 immediately Fuld ost pPhysaly at ead toa exponential iereve ofthe Kinetic encry, Ths i cau sen 0 2.)Hi. Dekker, Caracal and quantum mechanic of he damped harmin oxi " Gin = Op VO. 1s) gg = ~ANCOpy 26 (2.16) Gu = 20 + Aeon 17) Considering Gaussian wave packets. the standard treatment of Heisenberg’s uncertainty relation provides us with the additional relation {136. 164] Gp Fux = + OF 2.18) which allows the construction of a second order equation for o,. separately. Introducing the width w, of the Gaussian wave packet by means of WA = [mw 3] "7 exp[= (r= (x) Fw 2.19) one obtains (see also [136)) Wotan = Awe which immediately reveals the existence of a stationary solution oy, = the associated uncertainty product reads e2n ding (2.18), the quantum mechanical fluctuations are not only preserved but even enlarged by the dissipation, in fine with earlier made more general remarks (sce e.g. the Introduction, below (1.2)). Unfortunately, the above stationary solutions for the Hasse-Allbrecht model are not (asymptotically) stable (34, 37, 136, 238, 259-269]. For that reason we have purposely refrained from adding them the argument = %, It is easy to see from (2.20), that this damped quantum oscillator will exhibit persistent, undamped oscillations in its width if the system is initially not exactly in the above found steady state. This feature is not observed in any of the other known models, and it is quit unlikely to be physically realistic (136]. In closing the story on the Miinchen-model, it may be noted that it was recently indicated by Stocker and Albrecht [43] how these nonfinear frictional potentials can be generated formally within the framework of the fluid dynamical interpretation of the classical real space Hamilton-Jacobi theory and Schrdinger’s wave mechanics. Each of the above mentioned theories is essentially concerned with a Schrédinger equation representation of the dissipative system and the associated effective Hamiltonian. However. in view of the experience with fully quantum mechanical treatments of open systems (see e.g. [16-25. 50, 55-58, 61, 87-89, 200, 270-275}), actually mainly concerned with oscillators and their interactions with oscillator-reservoirs, it seems more natural to seek a proper description in terms of mixed states rather than pure states. Anyway, a quantum mechanical phase space description by means of a dei operator is fundamentally more general than by means of a Schrdinger state vector. In the case of a closed system (the universe) it can always be cast into a pure state representation, but not so vice versa (Gee in this context also for instance (270, 276-287]). For that reason, already in 1975 [59] we have introduced in fact the density operator from the very beginning in our mesoscopic approach to quantum friction. The theory was further developed successively in (184, 24, 195, 196, 288, 258]. Trains of thought2 1H, Deter, Classical und quantum mechan ofthe damped harmon oscilator along similar lines were revealed independently by Burzlaff [289]. Huguenin [2%] and Balazs (291). In his mesoscopic approach, the present author has always emphasized the relation between Heisenberg- Langevin quantal noise sources (operators. not classical functions) and additional diflusive-like con- tributions in the master equation for the density operator.” This notion gives rise to typical quantum- mechanical fluctuation-dissipation relations. For the linearly damped harmonic oscillator the charac- teristic example is (see (7.36) and [24)) Duy - Dp =i, 2.22) where the D's are diflusion coefficients. Dekker’s theory employs complex dynamical variables that are directly connected with the usual quantum mechanical commutator algebra (see also (57. 58. 160- 164. 293-303], and more specifically: [114. 115, 137, 181, 197, 289, 290, 304-309). It ensures that in the limit of vanishing friction the known undamped oscillator results emerge. and it also preserves the uncertainty principle for nonzero dissipation. ‘At the end of this historical survey it is interesting to note that Hasse in 1979 (107,310, 311] succeeded in making contact between the master equation and nonlinear Schriidinger equations in the limit of weak damping (radiation damping limit). As it turns out, this relation implies a constraint on the possible diffusion coetlicients. As expected. and as will be shown in the sequel, this condition guarantees the survival of the uncertainty principle and the vacuum fluctuations for the oscillator under the action of dissipating forces. In agreement again with the earlier expressed general ideas about dissipation and Muctuations, the quantal vacuum fluctuations appear to be increased by the damping” (see also (2.21). Hasse’s nontinear frictional Hamiltonian looks somewhat similar to the heuristic Stissmann—Hasse- Albrecht species (2.9), but is definitely different. For example, Hasse’s model does not show the peculiarity of the undamped, persistent oscillations in the width of the wavepacket. It is further interesting to observe that Dekker’s theory of the linearly damped harmonic oscillator leads to a class of specific diffusion coellicents all satisfying Hasse’s pure state Schrddinger representation condition if considered in the weak damping and long time limit, which seems to add to the credibility of that phase space approach (288). As the outline of the present paper should be clear from the table of contents and from the abstract, we now start our comparison of the various approaches to the damped oscillator problem in more detail by considering the time-independent Bateman-Morse-Feshbach-Bopp Hamiltonian. 3. Bateman’s dual Hamiltonian 3.1, Classical mechanics ‘The classical deterministic equation of motion for the simple one-dimensional damped oscillator with unit mass in Newtonian mechanics reads The wigifcance of he nose operator the original paper [9] (xe abo in patclr [2 bas at always Deen propery recogsized [19,250] Bat ee forthe cmtary e122) To be pec, the increase mcasred with expect othe “abulute vacuum”. the lnlstedundamped oxi ese (which phyily teat The teas of the zero-point uctatons shows tha the exciton i not laced inthe “ablute vacuum”, ut frat wth a reserve {eg the electomapnctic or graitalional elds) which may be in it own ground sale, Sematically. the lem ground state i perhaps more appropiate than vacuum steHE. Deter, Clascal and quanto mechanics of the damped harmon oct B EF UE+ Mr =0 G1) Its general solution has already been given in fact in (2.5). The equation (3.1) can be obtained as the Euler-Lagrange equation (La) -La=0 G2) from the Lagrangian L= ik - Oxi + Ale — 2), G3) that is by a variation of the auxillary variable ¥ (1,2. 11, 113-116}. On the other hand, varying x leads to E-2S+ ME =0, Ga) which clearly represents the time reversed process of (3.1). The system (3.4) is called the mirror-image oscillator of (3.1). The canonical momenta for the above dual system of oscillators become peLse anak, Fela sites. @5) Obviously, these differ essentially from the oscillator’s mechanical momenta, However, this is in no sense an obstruction to the classical theory and the Hamiltonian is easily found to be H = ph-A(xp- 3p) + ok, G6) with w = (07 a7)! representing the reduced actual frequency. If not indicated otherwise. this reduced frequency will be assumed to be real throughout the present text, That is, we shall mostly be concerned with the underdamped case although occasionally results can be taken over to the overdamped case. Hamilton's equations, fH oh ; G7) f2Ha. of course lead back to (3.1) and (3.4). Almost trivially, the Hamiltonian (3.6) is a constant of the motion: the energy dissipated by the original oscillator is completely absorbed at the same pace by the mirror-image system. Nevertheless, in general this H is not equal to the energy of the total closed system, not even in the limit of vanishing friction A | 0. This presents a first hindrance to the canonical quantization of the physical system, as noted before, It can be remedied, however, by restricting the solutions to the Hamiltonian equations to those for which both oscillators start off from the same initial state. Yet there exists another difficulty. Namely, the only nonzero Poisson brackets of the above classical theory would lead in quantum mechanics to the commutators” or ypraphical reasons, andifthere cane no misunderstanding, we denote commutators by [ether tha |]. Anicommetators wil aways be witen sf1" 1H, Dekker, Casical and quantum mechanic of the damped harmon exter AL LRasin. G8) [nal Evidently, in view of (3.5) this does never imply a nonzero commutator between for instance the position x and the mechanical momentum p = i. not even in the limit A | 0 where we know it should exist. Thus, unless bold measures are taken, the quantum theory based on the Hamiltonian (3.6) can be expected to be endowed with difficulties. Nonetheless, it is rather interesting to survey the attempts at a quantum theory based on the Bateman dual Hamiltonian. To the author's knowledge, there are only two papers to date on the subject. Bopp's treatment [114] (see also [108] is closely related to a symplectic formulation (197] that will be discussed furtheron; his presentation is somewhat loose and confusing.” The canonical quantization procedure devised by Feshbach and Tikochinsky [113] is of a higher subtlety. It is this ‘more sophisticated approach that will be sketched here. 4.2. Quantum mechanical spectrum As usual, the dynamical variables x, ¢ and &, are considered as operators in a linear space. They obey the commutation relations (3.8); all other commutators are zero. The basic Hamiltonian is the apparently Hermitian form (3.6). Then introduce the following annihilation and creation operators: G9) where «” is the customary Hermitian conjugate of a, Note that the bar can be considered as a formal operation that will be called mirror conjugation.”* For the sake of completeness, we also list the inverse transformations: a an eee, i vz a= Re eH), G.10) The only nonzero commutators, after the above canonical transformation, are G.11) (3.12) For example, his @ and a are not the sual Hermitian adjoint, See section 62, = "Toe expliiti lew ofthe eto of motion (31) aad (8), mito conjuation i defined a (A> A: gee A1H, Dekker, Casical and quantum mechanic ofthe damped harmon oxilator 1s and their conjugates, so that in effect Filet A-iee+ 2. 6.13) 1 + =) Wigle-A-iwe- a1. G4) ‘These new operators obey the same algebra as in (3.11), that is (al. st']=[2,2"]=1, all other commutators being zero. The transformed Hamiltonian (3.6) reads H= Hot Hy G.16) Hy = haol'sl - 8'R), @.17) Hy=iha(a'a'- la). G.18) The cigenvalues of "sf and B"A are the common ra, nu = 0.1,2,.... 80 that the eigenvalues of Ho are fex(nq~ nu) with cigenstates |na, mu), belonging to Hilbert space. Notice that Hy represents the srence rather than the sum of two free oscillator Hamiltonians (see also [116]). In the limit A | 0, the known harmonic oscillator results emerge infact only if the @-oscillator is kept in its ground state, for which |na,0)= 0. In order to investigate the eigenspectrum of the full Hamiltonian (3.16), it is of advantage to define y= Moll - 8'9), G and to introduce b.= Ua B" + ft), 4, = Silot'H - fB), 21) 4b. = Moll + 93’) (3.22) Notice that Hy= hobo, Hy = 2hAd, 3) The cigenvalues of d, are natnut |), the eigenstates being the same as those of the free Hamiltonian Hy, ic. of do. Indeed, Jo commutes with ¢,, as well as with 4, and ¢,. The algebra of these spin-like operators is given by [ond] =i de. (3.24)6 1H, Detker, Clascaland quanton mechan of he damped hamoncesilior [$2] =id.. (3.25) [d.d.1=i4,. 8.26) Be aware of the difference in sign in (3.25) and (3.26) with respect to the usual spin algebra (see e.g, [164, 300, 312-314). One easily computes now that O5)= 5-5 (3.27) Of course. $5 commutes with a. dy and d.. It is the only Casimir operator for the present algebra (as, the rank is unity; see e.g. (163, 300, 315-318]. Further, let us conventionally introduce Mngt na). (3.28) = Mta-ne). 0 and label the eigenstates of Hs. do and g. as jn) rather than |r. ta). To relate the eigenproperties of ¢, (in which we are interested in view of (3.23)) to those of 6. (which are known), the Baker-Hausdorff relation 3] for the present algebra, J. cos tid. sing, G29) explads dy exp(-no. is particularly useful (see also [168] p. 167). Namely, with = +7/2 one obtains b= ti expl(n!2)b.} db; explt(7/2)b.b 3.30) Therefore, the eigenvalue equation may be written as Via = ile + WS, Gal) where Win! = expl#(/2)¢.} im) - 6.32) The eigenfunctions ¥%;? do not belong to ordinary Hilbert space because they can not be normalized in the usual manner. In fact, their conventional norm can be calculated, similar to the Wigner D-functions, by means of Racah's methods (related to the better known Clebsch-Gordan coefficients; see e.g. (163, 164, 300, 303, 317, 318, 324-331). The result is [115] Gil exp. him) = SEINE ("YC core. 6x) where the sum runs over all integers /. Evidently, the norm (3,33) becomes infinite if ¢ = +7/2. This may be remedied, however, by redefining the length of a vector as the inner product (WW) = SE Mien G34) where ¥7 represents the time reverse of ¥, and where S indicates summation or integration (with theDekker, Claseal and quantum mechanes ofthe dampel harmonic xan v appropriate measure) over vector components, In the same way matrix elements, and in particular expectation values. can be defined as (HEY) = SEF Ko G35) F being an arbitrary operator working to the right. The absence of a finite norm in the usual Hermitian sense invalidates the standard proof that eigenvalues of self-adjoint operators are real. Therefore. the eigenvalue equation (3.31) can be accepted as a proper result Since go and @, commute, they may have simultancous eigenstates. Indeed, the eigenstates ¥2! are also eigenstates of $v as dn commutes with @, too. So, the complete eigenspectrum of H will be known if we determine the allowed values of m for given j. Defining b.= dF be. (3.36) one easily derives from the commutation relations (3.24)-(3.26) that [d..6-]= tid... (3.37) so that $. resp. are the raising resp. lowering operators for the eigenstates of .. In view of (3.31). m can be taken to be nonnegative without loss of generality. Let my be the smallest value of m. Then bint 3.38) where the vector is nonzero, Multiplication of (3.38) from the left with 6. and using (3.36) leads to (OF =[de be]- 62) Hint = 0 G39) Invoking the commutator (3.26), the eigenvalue equation (3.31), the Casimir relation (3.27) and the known eigenvalues of go, one gets mi= j*, so that m = jl, 1+ ili] + L[j]+3.... In summary, the eigenstates of the Bateman-Feshbach~Tikochinsky Hamiltonian are WM) = exp{—2i wjt +A (2m + IF (r/2)b< Hine), (3.40) the corresponding eigenvalues being Uibil+s HD = uj ihAQm +1); m Gal) If, in view of the earlier noted limit A 0, we keep the ‘-oscillator in its ground state with 1m = 0, 50 that 2j = 2m = ny =n, the eigenvalues become HY = nhosifnt a; 0 20.1.2. 42) This is almost identical to the spectrum found by Bopp [114]. In fact, Bopp’s procedure is somewhat ambiguous in the separation of the dual oscillators. As will be seen more precisely furtheron, (3.42) is, however, one of the possibilities allowed for by Bopp’s theory,H. Dekker, Clascal and quantum mechanics of he damped harmonic oxilltor 3.3. Wave packets ‘The result (3.42) for the eigenvalues indicates that the dissipation also afficts the lowest state m = 0, as the imaginary part of the eigenvalues remains nonzero.” In this respect, it should be recalled that the Feshbach-Tikochinsky quantum mechanics is based on the Bateman Hamiltonian for the dual x, <- system, but that the mirror image oscillator in fact is nothing but a mathematical device invented in order {o remain within the framework of conservative Hamiltonian classical mechanics. It should be emphasized that the only physically observable system is the damped x-oscillator for which, however, the correct commutator (x,p]= ih is absent from the theory (see (3.5) and (3.8)). Therefore, it is particularly relevant to investigate in more detail the behaviour of the damped oscillator on its own. In principle the eigenfunctions can be used to calculate the propagator (see e.g. [72]) and more general wavepackets. But since the relation of the above obtained eigensolutions with the expectation values of the original dynamical variables for the physical oscillator is not that straightforward (note for instance that the <-oscillator in fact contains a mixture of both the damped and the adjoint oscillator; and e.g. 4. and x do not commute), it makes sense to turn to the time-dependent Schrédinger equation itself, ‘The solutions for the mean values and higher moments can be calculated directly from it, without knowing the wave function explicitly, because of the linearity of the present system (see also eg, (34,332,333). The Schrdinger equation may be written as iN, = [WAG ~ EA) + w*k]Y, G43) The time reverse reads” Hi RWT = [Apt ACxp~ Fp) + ore]. (3.44) The equation of motion of the mean value (3.35) is easily obtained as (FY (FH -(H"Y'F), (3.45) where the superscript a indicates the real adjoint operator (e.g. [334)). It is not difficult to see that for the Bateman Hamiltonian (H™)"= H, so that (3.45) reduces to its usual commutator form (FY = -G/h\(R HY). (3.46) By means of the commutation relations (3.8) one now easily computes the mean value equations (y= Ga), 47) GY = -AG) =~ wx), (3.48) which, in line with Ehrenfest’s theorem, agree with the classical results and lead back to iis pecnty this fatate that was eliminated in {1 in a ater adhoc mann. To be cer time reversal is efectated by means of T= (i= 1 rmx pane Ek Pana1M, Dekker, lsc and quanton mechan of e damped harmon cil ® (x) + 2A) + (x)= 0, (3.49) the quantal version of (3.1). The mirror image oscillator is obtained from (3.47) and (3.48) by interchanging x, p and & a and letting A-»—A. For the second moments one finds xy = 2A (pe) +P) wx"), (3.50) wy (8) = -2A07)+ 26a) (3.52) 2A (FF) — 200i), G51) which form a closed set of equations pertaining in effect precisely to the physical damped oscillator. Note that. in view of the algebra of this model, (px) = (xf). Introducing the mechanical momentum p= =~ Ax and the variances (see also the discussion of the Siissmann—Hasse— Albrecht theory, in the Introduction) One = Mpx + xp)— (p) (x). (3.53) me = (9")- PY. (3.54) ou = (8)- (2, (355) one obtains Ge = 2p + yp = PO. (3.56) Gey = ~My - 2 op. (3.57) Gun = 2 (3.58) Note that presently (px) = (xp). Contrary to the Siissmann-Hasse-Albrecht model (2.15}2.17), the present Bateman model apparently allows only for the steady state dps = Opp = Gus =0. Actually, the time-dependent solution for ou, with o..(0) = A/22 and d.-(0)=0 (see also (136, 156, 180]) is easily found to be oul By a [+A sin20e+245 sin? ot] 659) ‘The associated momentum spread is, om(ty = 82 [1A sin20r+245 si? wt] (6.60) Clearly, in the end the uncertainty product» H. Dekker, Casical and quantum mechanics of the damped harmon olan Gye [sy CY aera] os ing Heisenberg’s principle. Notice that this violation occurs with certainty once A¥0, no matter how small the friction coefficient may be. If A = 0, the model (3.56)-(3.58) does, seem to reproduce the free oscillator equations, and allows for a steady state op: =0. Ope = ou. However, as in the dual model p and x commute. even if A = 0, the relation (2.18) reduces in this case t0 oo. = 03, = 0, which is manifestly at variance with well-established quantum phy: vitably tends to zero, viol 34. Summary Although Bateman’s time-independent dual Hamiltonian is correct classically. it has its problems in quantum mechanics. It ean not readily be associated with the system's energy and it does not lead to the correct fundamental commutator. Therefore, it violates Heisenberg’s uncertainty principle for the physical oscillator, even in the limit of vanishing friction A | 0. This feature cannot be swept under the carpet by a sophisticated treatment of eigensolutions. 4. Kanai’s time-dependent Hamiltonian L. Classical mechanics ‘The explicit solutions to the classical Hamiltonian equations (3.7) for the Bateman are iltonian (3.6) x(0)= [xy cos wt + (Ie) sin at], a HO) = e"[acos wt - wy sin wf) ay and similar results for #(¢) and }(¢). Let us first transform the irreversible part of the motion away by means of the canonical generator”” FaxPeMt+iPeo™, 3) which contains both the old coordinates and the new canonical momenta , &. One finds X=Fy=xe™, Fo=te™, Ga) ‘The new Hamiltonian H' = H+ F, becomes H'= 93 + u°XX, 3) Nove that 1X cine derive within Ihe framework ce time reser Mental 19 complex conjugation the quantum thea. "Thi enceator Fle. 9 A) isthe Ff Goldstein (I 1 Fevtbach-Tikochinsky theory 4 = falling he consent ening a8 tha‘HH. Detker. Classical and quantom mechanics of he damped harmonic oir u which describes undamped oscillations at the reduced frequency w, as expected from comparing (4.1). (4.2) and (4.4). Note that the new dynamical variables fulfil the same commutation relations (3.8) as the old ones. Of course, H’ is a constant of the motion, equal to its il value H"(0)= PoPo+ w® XoXo, For this to coalesce with the energy (at least in the weak damping limit) is suffices to consider Py = Py and Xo= Xo. Because of the time-reversible nature of the new solutions, this entails P(f) = P(e) and X()= X(0). It is now convenient to extend the real coordinates and momenta into the complex plane, and to introduce a canonical transformation from X, 9, X, F to Q. P,Q, P by means of X=HQ+O)+(/2wKP-P), X= 4Q+O)-(i/2w\KP- P). P=¥P+P)+(wl2Q-O), F=HP+P)-(wl2(Q-O). 4.6) where the new variables can be taken to be real. In the end one only considers those solutions for which Go = Qy and P, = Po, which confines the original variables again to real phase space. Note once more that the bar can be formally considered as @ mirror conjugation. The inverse transformations may be written as PHiW(P+P)-iw(X-X)]. Q=-G/2w)(P-F)+iw(X+X)], the other two following by mirror conjugation. Note the differences with the Feshbach-Tikochi transformations (3.13) and (3.14). The only nonzero commutators (Poisson brackets) for the new variables, of course, are [QPl=in, (0? (8) Inserting (4.6) into (4.5) yields H’=\P?+w'Q?)+ (P+ 0") 9) ‘This Hamiltonian evidently represents the energy of two independent identical oscillators. For that reason we shall focus our attention on one oscillator only, say the Q, P-system, The canonical generator F=110+:AQ? (4.10) then transforms Q. P into Y, [1 as (sec also (119]) P=Fo=M+iQ, Y=Fu 11) Hence, the pertinent oscillator from (4.9) is described by (sce also (119, 120, 137, 181)) H’= P+ AY + PY? (4.12) One further transformation (from Y, II to y. 1), generated by FerYe™, (4.13)x 1H. Dekher,Casicl and quantum mechanics ofthe danped harman exilltor yaF,=Ye™, (44) while H" = H'+ F,, finally leads to Hee heats lemay? ai This is the well-known Caldirola-Kanai Hamiltonian (13. 14]. As noted before, historically, the corresponding Lagrangian was already presented in Bateman’s original paper {11}. See further e.g. (10. 12, 96, 116-156, 180, 181, 275, 335]. In view of (4.14). (4.11). (4.6) with the constraint Q = Q and P = P. and (4.4). one readily infers that x= X e™ = Qe = Ye™'= y. In the remainder of this section we replace therefore y again by x 4.2. Quantum mechanics The Hamiltonian (4.15) has in fact been constructed purposely so as to represent the energy of the system, at least in the weak friction limit. As noted earlier, this scems to be one of the prerequisites for the construction of a quantum theory. However, from the equation of motion for x, that is (3.1). itis known that the irreversible part of x(¢) behaves like “and so does £(¢). Further, according to the Hamilton equation (4.16) the canonical momentum behaves like e™. But then (4.15) would imply that there is no secular change in the energy of the oscillator and, hence, there is no genuine dissipation at all (10, 275]. To be explicit. if fs the solution for x(t) from (4.1) with j,=0 for convenience, makes use of (4.16), substitutes into (4.15) and time-averages the Hamiltonian over any of its oscillation periods w/w, one independent cesult H's, = $x}. This feature is most easily clarified by looking at the Lagrangian belonging to (4.15). It may be written as (see Bateman (11] and e.g. [10, 128, 136)) 5 (= 2" (4.17) with mass m(t) = my exp(At), our choice being mo= 1. This Lagrangian immediately allows for an interpretation in terms of kinetic and potential energies. Physically, the situation described by (4.17) might be realized for example by a frictionless pendulum consisting of a pail collecting rain [10]. See fig 4 Notwithstanding these considerations, the Caldirola-Kanai Hamiltonian (4.15) has been used to study dissipation in quantum mechanics (e.g. (12-14, 96, 116-138]). Turning the only nonzero classical Poisson bracket into a commutator gives [x ]=ik (418) Using again the Hamilton equation (4.16), this implies in terms of the damped oscillator’s mechanical1H. Dekker, Claeeal and quantum mechanics of he damped harmonic oxilitor 3 meme pai Fig. 4.1, Model ora simple harmonic omilatior with atime dependent mas (after (1. momentum p = ¥, that [xp]=ine™ (4.19) Hence, as is indeed found in known explicit quantum mechanical solutions, Heisenberg’s uncertainty principle will be violated in the course of time.” This forbidding feature was already noticed by Kanai [14], and most clearly formulated by Brittin [12]. It should be clear, however, from the preceding discussion, that this can be considered as an inappropriate physical interpretation. If one agrecs to really equate the model's Hamiltonian to the mechanical energy, the solution of the problem becomes simple indeed: the quantal features of the system become increasingly irrelevant in the course of time as the growing mass takes the particle from the initial microscopic domain to the ultimate macroscopic, classical regime. ‘Thus, as far as the damped oscillator is concerned, it does not seem that we have much gained beyond the time-independent Bateman-model (3.6). Nevertheless, before taking another view of the matter, let _us briefly survey the results from the quantum theory based on the time-dependent Bateman—Caldirola-Kanai Hamiltonian (4.15). The Schrddinger equation in the coordinate represen- tation reads thy = Wey + Lem MPY, (4.20) Although the Hamiltonian is nonstationary, probability is conserved as the continuity equation patizeO 21) A recent clam that this can be remedied by simply considering the quantization in phase space as done in [137] i untenable. See ako forteronu 1H. Debter, Casal and guanton mechanic ofthe damped harmonic ocr holds with the probability and current densities respectively given by prey. j=e MHI) Wd vd). (4.22) It is easily verified that (4.20) allows for the following set of so-called pseudo-stationary states (see e.g (96. 119, 129, 133):* y= Np expl-i(n + Yor + 'At— (w + iA) e™s72h] H,(e"xV Th) 4.23) where, =(w/rh)Q%nty", with n=0,1,,... For curiosity. compare this with (2.11) for the ‘Siissmann-Hasse-Albrecht model. And for the sake of completeness it could be mentioned that Dodonoy and Man’ko (e.g. [133]. see also [10]) denote (4.23) as loss-energystates.“* 4.3. Quantum fluctuations Since (4.20) is a linear partial differential equation, the pseudo-eigenstates (4.23) can be used to construct more general solutions by superposition, for example Gaussian wavepackets (see e.g. (116, 128, 129, 131, 135, 136, 180)). See also (2.19). The mean values, of course, follow the classical path obtained from (2.13) and (2.14): the widths of the packets require further investigation. From the Schridinger equation one obtains the usual equation of motion (sce e.g. (3.46)) for the expectation value of an arbitrary function of the canonical operators x and zm. For the second moments one re finds (an) = 6 ™(n) -e™ NU, a» (ny =~ O(nx + x), (4.25) (ey (wx + xm). (4.26) It may be noted en passant that the Gaussian Heisenberg relation (2.18) presently holds not for p and x, but in terms of a and x; that is Onn = +02, (427 where the variances are defined by (3.53)~(3.55) with p replaced by m, Rewriting (4.24)-(4.26) in terms of the variances for the mechanical variables p= ¢ and x, by means of (4.16). leads to 2G pe + Opp ~ VO (4.28) = Hay 20 op. 429 Gur = 2p (4.30) “Hie aware of we dilerences in ott, For instance. [133] #2 ad wate precy inkrchanged Fin [133 the thors cone fin analogy wth fet time peri amitonane) the penocy of the Caliols-Kanai Hamitonan over fopacements ih along the imaginary time Sts, ie, Hi wh) = HU). Thete consideration dct tad 19 a predo-igenspecium with ‘complex egenvalcs Hi = Rly +14¥in +!) The itiieanee of he spectra, however, remains in Yin,H Dekker, Case and quantum mechanics of the damped harmon exile s which are readily observed to be identical to the corresponding equations (3.56)-(3.58) for the Bateman-Feshbach—Tikochinsky model. The only improvement that has been obtained is the recover- ing of the correct quantum mechanics if A = 0. The solutions to (4.28)-(4.30) with o,,(0)= h/20 and Gu(0)=0 have already been explicited in (3.59)-(3.60), while the consequences for the uncertainty principle are shown in (3.61). Of course, other initial conditions are possible too. As an example, one might investigate the behaviour of the first and second moments for the pseudo-stationary states (4.23). Note that the pseudo-groundstate yy is Gaussian. It is not difficult to see that (x) is strictly zero for any rn. Hence (x)= 0, so that ()=0 and (p)=0, Further, using the standard properties of Hermite polynomials, one finds ohne @3)) Substituting this into (4.26) learns that (mx + x7) is time-independent for all states y. Therefore the L.h.s. of (4.24) is zero, wt leads to (rym AE (nt 4x Hence, the product Oo ug (ntay 4.33) is conserved in the course of time. This formula is the generalization of a result from 137] to arbitrary n, However, it must be emphasized that (4.33) applies to the canonical momentum in stead of the proper mechanical momentum.” In terms of the latter one finds wnt, 3) Cn which, like (3.61), inevitably tends to zero even for the lowest state n = 0, thus violating the uncertainty Principle whenever A>0. 44, Summary In the course of improving the energy representation, the time-independent Bateman Hamiltor has been related to the time-dependent Caldirola-Kanai Hamiltonian by a succession of canonical transformations. However, if this Hamiltonian is indeed taken to be equal to the mechanical energy, there can be no energy dissipation and the described system rather is a mass-accreting pendulum. The alternative consideration of the Hamiltonian as a mere generator of the motion, leads to a violation of the fundamental quantum mechanical commutator in the course of time. In fact, the uncertainty should aso be noted therfore that (239) ~epeially wth m * Ofor the seo rounisat eseovally diferstrom Deber'sresh [195 See abo frter om in he pretem arte e§,e9 (7.51,» 1H. Dekker, Cascal and quantum mechanic of Be damped harmonic oscil product is identical to the exponentially decaying form (3.61) connected with Bateman’s dual Hamil- tonian, 5, The semi-infinite transmission line Sil. Preliminaries In this section we will attempt to shed further light on the deficiencies of the above discussed theories, in particular that of the time-dependent Hamiltonian, by considering an explicit physical model. In fact we shall treat a mechanical model and an electrical model in a unified manner. The essence of these models is the coupling of a privileged. tagged oscillator to a reservoir consisting of an infinity of tiny other oscillators, and to let the latter system tend to have infinite size. In that case, simply classically speaking. signals emitted by the tagged oscillator, which is the one under actual investigation i.e. the subsystem of interest, can never (that means, within any finite time interval) return to it by being reflected from the boundaries of the reservoir. The electrical network model that will be discussed in the sequel is an adapted version of that of Stevens, who's explorations were originally based on a suggestion by M.H.L. Pryce [157]. The mechanical analogue of the electrical semi-infinite transmission line is a semi-infinite string. The present formulation is akin to recent work of B. Yurke and O. Yurke (39]. Within the model's framework the results are exact, More or less similar approaches have been presented for instance by Senitzky (50] and by Ford, Kac and Mazur [38]. More general reference can be made eg. to (16-23, 41, 42, 53-58, 63, 92, 93, 168-179, 200, 27 S]. Further, itis interesting to note that historically the infinite harmonic chain was already discussed in the early days by Hamilton [223] and in the beginning of this century for example by Schridinger [339]. Let us first consider the mechanical model, depicted in fig. 5.1. A finite mass m, has been attached to a chain of identical masses my = m (k= ), located at positions 2, and bound to each other by harmonic forces. The mass my is further attached to a special harmonic spring with Hooke constant by, while it is constrained by a frictionless guide to move up and down the x-axis. Clearly, if the tagged mass mip were uncoupled from the harmonic chain, it would itself just be a simple undamped harmonic oscillator with eigenfrequency £2 = (by/ma)'*. But if connected to the chain, it will excite the transverse modes of this reservoir. In that way the oscillator may dissipate some of its energy into the chain, at least for some time, Fe. Mode fora mechanical ransision tine1H. Dekker, Casscal and quantum mechanics ofthe damped harmonic xilar ” The latter proviso takes care of the following. If there were only a small number of elements in the chain, the signal travelling away from the tagged oscillator would return to it almost immediately. If the stumber of elements would tend to infinity while keeping the physical length A of the system finite, the chain becomes an elastic string as depicted in fig. 5.2. But even in that case, by reflection from the far end of the string, emitted signals return to the actual oscillator after a time period of the order of 2.4/c, ¢ being the average propagation velocity. The details of the dynamics would. of course. involve the particular nature of the boundary condition at the far end. Clearly, only if A becomes arbitrarily large there will be the possibility of dissipation for the tagged oscillator over an arbitrarily long time interval. ‘The infinite length is most easily incorporated by using periodic boundary conditions. which are analogous to tying the transmission line to a characteristic impedance in order to prevent reflections. ‘The characteristic impedance, attached to the finite line at z = A, is simply equivalent to “an infinity of more line” (72] Fig. $2. The mechanical transmission ine of 5. inthe limit 0, but ict andthe ength Kept ie, J she tension of he sng ater 0, It can hardly be overemphasized that there are thus two essential steps in reaching irreversible behaviour if starting from reversible mechanics. First, the reservoir must have an infinite number of degrees of freedom over which the energy drained from the low dimensional subsystem of interest can be distributed. This is effectuated by taking the continuum or field limit, In terms of the normal (Fourier) modes of the reservoir, however, this still means a discrete spectrum entailing a finite recursion time. That is, the system will exhibit periodicity, which is a typically reversible phenomenon. Therefore, second, the system must have infinite size. This is sometimes called the thermodynamic limit (see eg. (28, 34, 37, 52, 238, 332, 340, 342)), In terms of the normal modes this means a continuous spectrum. It implies an infinite recursion time. As will become evident, both steps must be taken for the following program to be carried through. 5.2. The mechanical model ‘The classical mechanical Lagrangian for the harmonic system shown in fig. 5.1 reads SF tun om) where J = b(d-du) is the bias force applied to the little springs, which posses a rest length fy and Hooke constant b, = b. The quantity J is usually called the tension of the string (e.g. [343}). The index» Hi, Dekker, Casucal and quanti mechanics of the damped hamoneoxloe K€ [0..N] measures the distance along the z-axis, which is along the chain, according to zx = ke The tagged oscillator is at : = 0. The canonical momenta following from (5.1) are Paz Las mi. 62) so that the Hamiltonian becomes, sacri trish 3 ott SS wn 6.3) ‘The first Hamilton equation. as usual, reproduces (5.2). The second equation leads to Pee -Hay= (Tiere tas KEL 6.4) H = ~bota + (Tid xr xa) (63) Be so that the (Newton-Lagrange) equations of motion become mig = (Teer at teas KEL 5.6) moka boku = (Fld) (X= x0) 67) Letting now the number N of chain oscillators tend to infinity. keeping the length of the string fixed at A= Ne, so that /| 0, and confining oneselves to differentiable physical solutions x(z, f) the above set of coupled differentiabdifference equations leads to the following two coupled partial differential equations: Kal N= Cras — 2>0, (8) (0,1) + 27x(0, 1) = 4.210458) 69) Equation (5.8) is just the usual wave equation (e.g. (83, 113]) for the transverse vibrations of the string. ‘The speed of propagation is given by c= (Tip)". with p being the mass density p = mid, kept finite while ¢ | 0. Further 2 = (bu/my)"?, while x = J/my can be considered as the coupling constant of oscillator and reservoir. In terms of the fields, the Hamiltonian may be written as H= pO.) F + O7[x0. OF + f Mp. OP + if. OF bdz (5.10) where the canonical momentum density is conveniently defined as p(s) = pile if k 2 1, while p(0) = pa, and where we have chosen p = mu= 1, so that pe. = x1) 6.11)1H. Dekker, Casal and quantum mechan ofthe damped harmon exile > for all z. With these choices” the only nonzero Poisson brackets lead to the commutators [x20 ple". 0] = 1h - rv'40, (3.12) {x00.0.p0. 0] =ih (6.13) 5.3. The electrical model The electrical analogue of the above mechanical model has been sketched in fig. 5 resonant circuit, with selfinductance fp and capacity eo. is coupled to a chain of identical tuned circuits with selfinductance l, = I and capacity e, = ¢(k ). As indicated in the figure. all transverse currents are taken positive into the positive x-direction, all longitudinal currents are similarly defined along the positive z-axis. Going around any circuit, Kirchhoff's voltage law (usually called his second law: see e.g. [37. 44.345) yields The tagged giles + leh ~ queslenes= 0. 6.14) where q represents the charge at the kth condensor. and where J, is the longitudinal current through the kth circuit located at 2. Differentiating (5.14) with respect to time and using the relation quer = Fuees for the transverse currents, one obtains Teale t hel Laalens 0, (6.15) with the understanding that J ..0= ‘a. The other Kirchhoff law, expressing current conservation at the vertices, relates the transverse currents to the longitudinal currents as, Fane Lee (6.16) pte ht oT anode aie ° @ zs zs 2 >? Fig 5. Mode! for an elec rancmision lite fate (157, An aerate, xedia |) evintoduce apa dependent macedensiyp(2)= p+ m+) Inthateaw.(8|2)and(SSjeanbe presented a singe forma, in fet (812,x H. Deer, Clas and quantum mechanic of he damped harman exile Inserting this into (5.15) one finds lig = (Velen 2h thea): KEL 6.17) (6.18) oho (Meo)lo= (Ve Yh which are immediately seen to be completely identical to their mechanical counterparts (5.6) and (5.7) In the limit of an infinite number of circuits in the chain, keeping the length of the transmission line fixed again at A=N¢/, one obtains the partial differential equations (5.8) and (5.9) with the appropriate definitions: the propagation velocity becomes ¢ = (Iil'e’)"3, where I" = Ud and e' = cl are the specific selfinductance and capacity respectively; the characteristic frequency of the tagged circuit becomes = ({Ilvco)', while the coupling constant is « = IMac’. The model, be it mechanical or electrical, will now be further explored on the basis of (5:8) and Sat, The classical damped oscillator Following d'Alembert the general solution of the wave equation (5.8) may be written as (sce e.g. [s3) (2.1) = xalt 2c) + xu(t+ zie). (5.19) The xq, respectively x» represent waves travelling to the right (outgoing. so to speak, from the point of view of the tagged oscillator) respectively to the left (incoming from the bath), Using (5.19) one readily verifies the identity ¥(2, 0) = —(Ie)x (2, )+ Beanz), (5.20) ‘so that (5.8) can be recast into the form: #(0,)+2A.2(0,1)+ 2? x(0,) = 4Ax0,(0,1), (521) where A= «/2c. It should be noted that d’Alembert’s solution typically applies to the wave equation. ‘That is, the exact and complete separation (5.19) into left- and right travelling waves would in principle not have been feasible if we had not already taken the limit to continuous fields (i.e. the first step mentioned earlier: N+ =, | 0, A = Ne finite), Hence, the equation of motion for the tagged oscillator could not have been given the appearance of (5.21) (compare especially with (157). Now in prequantal classical theory one may conceive of the situation that there are no incoming ‘waves at all present in the transmission line at the initial time, say ¢=0. Exciting at that instant the tagged system at z = 0 will, of course, produce waves travelling to the right. But, as noted earlier, if the transmission line were semi-infinite, these outgoing waves would never be reflected back, i.e. they ‘would never be transformed into incoming waves. In other words, if xa(z, 0)=0 for all z= 0 and A ==, cone has classically that xg(z, ¢) =0. Hence in that case (5.21) reduces to ¥0, 1) + 2A 20, 1)+ 0? x(0.1)=0, (8.22)H. Dekker Cusica and quantum mechs ofthe damped harmon oxo u which is exactly identical to the classical equation of motion ( oscillator. See also fig. 5.4 However. in general the very special initial condition of no incoming wave excitations being present in the reservoir field will not be encountered in real physical systems. Normally there are at least thermal excitations. and these do not discriminate between left- and right travelling waves. And in quantum mechanics. even at zero temperature there would still exist the zero-point excitations of the oscillators in the transmission line. The basic reason for this. of course. is that if the line is capable of transmitting signals. its microscopic elements must be emitters as well as absorbers. Hence. in quantum theory the inevitable excitation/de-excitation processes in the reservoir can not be circumvented. For that reason, in constructing the model's quantum mechanics we must carry along the incoming waves. i) for the linearly damped harmonic Fig. $4 The mechanical damped olin, 5.5, Normal mode expansion ‘The total field in the transmission line can be written in terms of its normal modes as x2,.9=235 Vag expl-iay (t~ 2/c)} + By expl-iay(¢+ zfe)}+conj.], 2) where “conj.” means complex (Hermitian) conjugation in classical (quantal) theory. However, for a finite transmission line the left- and right travelling modes are not independent of each other as a consequence of the boundary conditions. The most general boundary conditions pertaining to the wave equation are of the so-called mixed type (see e.g. [83, 113, 346, 347]), but it suffices to illustrate the point by considering for convenience the Dirichlet condition x(A, 1) = 0. This leads to Ax expli anAle) + By exp(-iaAle) = 0, Al exp(-i aAle)+ BE expliarAlc)=0, (6.24) so that ATA, = BIB. 625) Hence, we infer the relation Ac= By expl-2ies), (5.26)2 1H. Dekker, Casicl and quantum mechanic ofthe damped harmonic oxilltor ‘ex as yet being an arbitrary scalar. The boundary condition at the near end z = 0/of the transmission line is simply given by the equation of motion (5.9) for the tagged oscillator. It gives in principle two constraints involving the As. By and their conjugates. Inserting the relation (5.26), these constraints are found to be identical and one is left with tg 4 = (0?~ wi)/2Aor, (6.27) where we have used the relation «/c = 2A. Finally, inserting (5.26) into (5.24), the eigenvalues follow from wAle~g.=(k-Hm; K=1,2... (5.28) In fact, the above procedure is precisely the one used by Yurke and Yurke (39].” Since their ultimate results are identical to those obtained presently,“* we here face an explicit confirmation of the general expectation that the precise form of the boundary conditions at the far ends of a system should somehow become irrelevant in the thermodynamic limit of infinite system size A+ (see also (348-350)). Therefore, it is a useful common practice to introduce the most appropriate, though physically somewhat artificial, boundary conditions for large systems, namely periodic boundary conditions, from the outset. So we return to the normal mode decomposition (5.23) and require x(z, 1) to be periodic over A. This leads to the two conditions Ag By = Ax expli andle)* By exp(-iaaAle). 6.29) and their conjugates, Either adding or subtra ed in (5.29) yields in stead of (5.28) ig the two equations cont Dens (6.30) wAle = 2kz ; which fixes the eigenvalues without requiring any relation between the mode amplitudes. 5.6. The quantal damped oscillator The normal mode amplitudes are now determined by the initial conditions as (hax J dz exp(-i anz/c) [ax x(z, 0) +i plz, 0)]. 631) a = Chay? f dz exp(iwaz/e) fon x(2,0) +i ple, 0), (632) plus their conjugates. In these formulae we have used the canonical relation (5.11). By means of these FLet ks and gy = 2-8 for complete equivalece ** Hhowcver there anil problem invlsed a their eigenfunctions ar at sly orthogonal. This act reuites farther study. i may be related t probleme of divergences, See ee (8),Hi Dedker, Clasico and quanti mechanics of the damped haron oslo s transformations one easily verifies that the only nonzero commutator for the reservoir field variables, that is (5.12). leads to [Av Ail=(B.. BI] = bu. 533) any other conceivable commutator being zero. Evidently, the Ax. By and Ai. BY are the annihilation respectively creation operators for the normal mode excitations. We have now prepared the system adequately for the limit 4+ (.e. for the second step to be taken). essentially implying the (linear) independence of the incoming and outgoing waves. Now substitute the pertinent part from the normal mode expansion (5.23) for the incoming driving field on the r.h.s. of the equation of motion (5.21) for the actual oscillator. Then define y()= x0.) n0). (634) n= In. (5.35) such that n,(f) must be solved from i++ = 1d LM CBE expliont)= By expC-ien)« 636) with the initial conditions ¥(0) = m(0) = 0. Asa result one obtains tray + ay =0, 6.37) with y= ¥(0.0) means of elementary analysis [351. (0.0) and_y(0) = x(0,0), The required solution from (5.36) is easily found by 52] to be m(onia [M2 (B, a.(0-coni.1, (5.38) sepa eee ~ P+ 2h, ay et (Leppert) 639 {+ axlw)+ Si alo, (5.40) ¢ w = ((F— A)" is the reduced classical frequency as before. What has been achieved is in fact teresting. The result (5.37) shows that y(#) represents the position of a classical damped oscillator, even in the case that there do exist incoming signals from the semi-infinite transmission line In other words, the dynamics of a classical damped oscillator can be obtained in an exact manner if one subtracts the incoming reservoir signals (thermal noise, for instance) from the actual oscillator variables. ‘This represents a well-defined example of so-called smoothing (sce eg. [29.66.353]) by nois subtraction (157]. The explicit solution for the smoothed variable y(t), satisfying (5.37) and the required initial conditions, reads y(0) = e-{x(0.0)c0s at + (1/a)[(0.0) + Ax (0.0) sin ea} . Gal)a 1H. Dekker, Case and quantum mechanics of the damped harmone oxcilator 5.7. Connection with Caldirola-Kanai model Following Stevens. we now conjecture that a time-dependent canonical transformation exists from (0.1), p(0.1) to y(t), (2) with y() given by (5.34) and the new, noise-subtracted momentum defined as, pO. t= H(A (642) (0 In view of the initial conditions on n(t). one immediately concludes from (5.13) that Ly). rO)] =i, 6.43) while any other commutator at 1 = 0 will be zero. This gives reason to calculate the time derivative” LO. 70} Cad eb iP L(0.9. Bt e'™= Be}, 6.4) where we have used (5.11), (5.21) and (5.36). According to (5.32), the normal mode amplitudes are expressible in terms of the initial values x(2.0) and p(:.0). For the further evaluation of the commutators on the r.h.s. of (5.4). it is therefore useful to observe that (5.41) expresses the smoothed position y(@) for all times entirely in terms of the initial values x((,0) and (0.0) at z= 0, As these commute with any of the field variables on?* : €(0, A), it can be concluded that y(e) = (0, 1)— n(0) commutes with the driving force of the bath at any time, Therefore, (3.44) is equal to = [y(o. mio] = 21a e™ {1-5 Ve into. Bem — Bem} (5.45) Since n(t) has been expressed in terms of the BL, By by means of (5.35) and (5.38)-(S.40), the commutator can now be calculated explicitly, Using the algebra (5.33), one easily obtains [y(n r= (5.46) We now take the actual infinite system size limit .1 +=, so that in view of (5.30) Cri) Z(le f du’, 647) which yields by. #0 = (48) he much spl. ¢ rer. while the arse commutator S13) at: = 0 wise atthe preven meta arn is comaied ia (31) and (32) aly with meas One may of coure comput he 7" Note for the sae of til. hat ss pomingtarH Dekker Cased and quantum mechanics ofthe damped harmime exile s where we have further used the relation 2A/c = cplT = plmy= | (see (5.8)-(5.11)] can be evaluated analytically as ” The integral ‘ | ooo ae | (ema wrarrele af C Hence, br. 0) =0, ‘so that, in view of its initial value (5.43), [y@. 7o]=ih 631) at any time 120. Other commutators can be handled similarly to show that y(0) and x(¢) are indeed proper canonical variables. Note that in proving (5.51) we have used the nature of the solution for y(t) in going from (5.44) to (5.45). In fact, the precise explicit form (5.41) of that solution, together with the corresponding expression for #(), was used by Stevens [157] to arrive at the same conclusions. Thus. to be clear, whereas Stevens’ proof involves the explicit formulae for the smooth (noise-subtracted) oscillator variables, the present treatment rather makes use of the explicit expressions for the noise (0). ‘The results are the same. Further note that the transformation replaces x(0., 1) and p(0.1) by y(t) and (1). but that it does not affect the normal mode amplitudes of the reservoir z € (0, )."" From (5.34), (5.11), (5.42) and (5.37) it is clear that the canonical equations for the noise-subtracted tagged oscillator variables are (5.52) Hi,=-e"My 653) Evidently, the total Hamiltonian H’ will therefore consist of two commuting parts H" and H’, where H" generates the dynamics of the smoothed oscillator and where H'% merely pertains to the reservoir, That is. Hi = H+ HalAy AL: Ba BLD. (5.54) (5.55) The latter effective Hamiltonian for the smoothed oscillator is an exact isomorph of the original Foe the elcid mn he choice = my =| atte int = ly, 4 tha again 2We* le?» Py Ue Belo (S18, ‘Recall ht v(t an iiarty 0 commute withthe Bk fe (540-(84)}, The sme argument of core apis the“ H. Dekker, Castel and quantam mechanics ofthe damped harmon oscillator Bateman-Caldirola-Kanai Hamiltonian (4.15). However. the different physical significance of the canonical variables in (5.55) and (4.15) must not be forgotten. The present mode! does not violate the fundamental commutator (x. p]=if in the course of time. in contrast with the standard Caldirola— Kanai theory. As we have seen, the single reason for this is the absence of noise (jie. the usually thermal, incoming signals for the same subsystem, the reservoir. that accounts for the dissipation) in the latter phenomenological classical approach, These findings. originally most clearly formulated within the present context by Stevens in 1961 [157]. were in fact corroborated at about the same time by Senitzhy (50. 384]. However, the latter did not discuss the connection between the harmonic oscillation-reservoir model and the Bateman~Caldirola~ Kanai model. His treatment is limited to the weak damping approximation (frst order in A). whereas the present formulae are still exact within the model's framework. Actually, itis somewhat hard to see good reasons for restricting the treatment to the weak damping limit. It is very well conceivable that the single privileged oscillator. which is under investigation, is strongly damped while the vast majority of the oscillators in the essentially infinite reservoir still is in the same equilibrium state as in the weak damping case, The amount of energy dissipated by the tagged oscillator will always be negligible compared to the total energy content of the reservoir. See further section & SS, Svin'in’s reamment Consider the classical Hamiltonian (5.55). The canonical generator Femytmen()-ye™ al) (5.56) transforms y, 7 into x. 7, as follows: ae, ytn) (5.57) Compare these formulae with (5.34) and (5.42). The new Hamiltonian H™= H+ F, becomes Ht = sear +e Mx? — eB) x, (5.58) where an irrelevant term independent of m.,x has been disregarded, and where we have used (5.35) and (6.36), so that Bray Hee exp(i ext) — By exp(-iwst)). (6.59) ‘The Hamiltonian (5.58) has precisely the form of the Hamiltonian introduced by Svin‘in in 1976 [180]. It was recently also used by Brinati and Mizrahi {181}. investigating its quantum mechanics in complex phase space (see also [156, 195)). Svin‘in treated B(t) as a classical, Gaussian delta-correlated thermal Langevin noise source in order to remedie the observed violation of Heisenberg’s principle in the original Caldirola-Kanai quantum theory (see refs. quoted below (4.15). That is. Svin'in’s Ansatz is quite akin to the Langevin-Wiener assumptions (1.4) for the process &(1) as discussed in the Intro- duction. Let us presently calculate the correlation function of B(#), as given in (5.59). explicitly. Considering1H. Dekker, Casical and quantam mechanic ofthe damped harmon oxillaor a the noi in principle to be of a quantum mechanical origin. we determine the symmetrized Weyl Wigner (276, 277] correlation function in order to obtain a real outcome. Then. using the properties of a thermal bath (see e.g. (19. 23. 50, 37, 164)). (BLBi) = (exp(hax/kaT)~ 1) Sar. (8.60) one easily obtains in the infinite system limit: BC+ 2). BOL) = f Aw) c0s.'r da’, 561) 7 where 9 represents the modified Planck function (1.11). Evidently, the noise from the transmission line is quite different from being delta-correlated. Actually, although the thermal part converges, the integral in (5.61) does not even exist because of the diverging zero-point contribution. Only in the extreme classical limit ky, (5.61) can be reduced to (B(t+ 7) BE) = (E(t + 1) (0) = 4AkyT3(7) (6: From comparison with its definition (14), we sce that this would imply a diffusion coefficient D=2AkwT.” Of course, the integral in (5.61) does converge if we drop the zero-point fluctuations, but that would spoil the cure of the deficiencies of the damped oscillator as observed in the Caldirola—Kanai theory. On the other hand, treating the bath (semi-) classically and ascribing the vacuum fuctuations in fact merely to the tagged oscillator must be considered rather unsatisfactory. to say the least. Notwithstanding that the correlation function (5.61) can not be calculated in general, the spe density defined by (1.5) does exist. If w >0, one obtains q Sy(w)= 4A Pw). (5.3) while Sy(~w) = S(w). Defining a frequency dependent diffusion coefficient by means of Sulw) = 2D(w), one has D(w) = 2AP(w). 's Ansatz amounts to the assumption that only the actual value of the spectral density at the free oscillator frequency w = £2 will be relevant to dynamical quantities of the damped oscillator. Intuitively, this will at least require weak damping. But, as should be understood from the preceding, this is by no means sufficient For complete equivalence with Svin‘in’s treatment [180] we can simply replace in the Hamiltonian (5.58) B(0) by £(2) with the properties (I.4) and the diffusion coefficient given by D = 24P(M), Hence." (E(Q)=0, (E(t + 2) E(N) = 2D 4(r), D=AhQ cothnQlkyT) (6.64) This is tice as large a the difeson cei inthe Intrastion. See expec (1.1), (15) and (1). Both ere a there 24 hax been chosen asthe telaation constant ofthe energy. However. presetly not only the momestam Bul ab the potion carticn energy, Thetloe. in thermal equiriam now E = kyT instead of {ka7 (ue below (1) shih explains the factor two nthe desion cocticient o Note that Svivia's @ = 2, while iw the preset 1? andthe present w BN™ HE Dekker, Cac and quantum mechanics ofthe damped harmonic excl ‘As usual, one may compute the equations of motion for expectation values from (3.46) with H +H”, or equivalently directly from the Schrdinger equation Ghd = Hm. x. (OW (5.65) Note that the averaging procedure will consist in fact of two parts: first. the common quantal averaging with J. and second, the thermal averaging of &(/). In terms of the mechanical variables the result for the variances reads Gps = ~2ho pe + Op ~ Pon + Ux) £0). (5.66) gy = ~Sdohy ~ Poy, + 2p) EO) (5.67) Gu = 20 (6.68) The quantities (p)€) and ((x)£) are easily calculated following the lines of standard Markov-Langevin theory (sce e-g. [19, 21. 23, 24, 37, 57-59, 65, 66, 92] and the Introduction). That is, formally integrating the equation of motion of the as yet merely quantum mechanically averaged momentum, (p)" = -24(p)- 4) + 60), (5.69) and using (5.64) and ((x(1’)) €(0)) = 0/if ¢' <1, one finds ((p)g) = D. Similarly, by means of (x)‘ = (p), one obtains (x)) = 0. Hence, Gp. = Dt + Oe Dare, (5.70) in = App ~ Poy, + 2D, 571) Gag = hn (672) D given in (5.64). The only, but essential difference between (5.70)~(5.72) and the original Caldirola— Kanai model (4.28)-(4.30) is the present additional time-independent term in the equation for opm. which is due to the noise. It is a matter of straightforward calculation to obtain the solution with (0) = 0 and 044(0) = A/20. It reads [ison Ge) ][t+Bsinaore2ASstae]. 679 An h oult)= ano (star 7)* 3° Tyr) the associated result for the momentum spread being 12 coy (AL) 4 Me Phan (at ie Om(t) = “Scot at) tre [1-« coth (sf (5.74) The ultimate value of the uncertainty product becomes ein) 0.) = cont (575)H. Dekker, Casal and quantum mechanic of the damped hamonc oscillator » 0 that the uncertainty principle is preserved. albeit-as discussed -somewhat artificially. As they should, (5.73) and (5.74) reduce to (3.59) and (3.60) in absence of the noise.” Finally. it is worth mentioning that the above formulae can also be obtained entirely classically from the Fokker-Planck equation (see e.g. (21. 32, 34-37, 4447, 51, 56-58. 62. 66, 73, 76, 90. 92. 231, 242. 332, 333]) for the probability density“* W(p, x,t) associated with the Langevin equation (5.69). W, = pW, +P W, + 2A9W), + DWip (5.76) This connection will be made more precise in the sequel. See section 7. 5.9. Transmission line fluctuations Returning now to the genuine transmission line model, it should be noted that in principle it can be solved exactly, without enforcing the Markov hypothesis. Namely, the dynamics of the well-defined noise-subtracted variables is governed by the simple linearly damped harmonic oscillator equation (5.37), that can be solved easily. In fact, the evolution of the position is given explicitly by (5.41). Let us just consider the final state in some more detail. Clearly y() = 0, so that the actual oscillator position acquires the noisy value x(0,%)= n(). Hence, denoting (0, ) by x(t) for convenience. (x()) = 0. More generally y"(=)=0 with n= 1.2,.... For the second moment one thus gets, using the above result for the first moment, (x7())=(n7(#)). Invoking (5.35) and (5.38) for n(1). and using the properties (5.60) of the thermal bath, one finds“** ea oes a es 6m ), where the modified Planck distribution (1.11) (o=¥3)4 Det, Confining ourselves to the zero temperature case T = reduces to Mw") She, the integral in (5.77) is most easily evaluated setting a’ with 0’ dao’ = Aw dx and w = (f2=A2)'*. The upshot is aoa arccote ( “a ) (5.78) This result agrees with Yurke and Yurke’s [39] and is also consistent with the formula of Ford, Kac and “= In the weak friction limit (5.78) can be expanded in a Taylor series as Al-3+-@)) 6 where, for future comparison, one should note the occurrence of a term linear in A. 2 Recall hat (420-4, fo the Calista Kapa mate catia (354)-(8.88 forthe dust Bateman ve 7° Sic in (S458 ys the Weyl-Wigner syemitied expense, th Une maser equation (576) si fax aed by Wigner’ dst anton [277 See ao xpcily i view of (54) aT = 0, (2,256, Real the elation 2c 1, Sco (8-8) seer in| iy = 243040 = 1124 In|] reps the Planck function Py Remodel Pet T= Wamlcakula thir (2), thf» 4 aula = 2 se Se SILo H, Dekker Casal and quantum mechanics of the damped harmonic exile Finally it is not difficult to see that the formula pertinent to o79() will be (p°(=)) = Gi*(=). At one obtains Opel 2hA w dw’ 2 ote ea) As it stands.” this expression cannot be evaluated as the integral diverges logarithmically. Hence, the semi-infinite transmission line mode! does not allow the evaluation of the uncertainty relation. unless we would disregard the vacuum fluctuations from the oscillators in the reservoir. That, however. would imply the violation of the uncertainty principle for the tagged oscillator at low enough ierpeednee 5.10. Summary ‘The semi-infinite transmission line model shows that two steps must be taken in order to arrive at irreversible behaviour of a subsystem (the tagged oscillator at the near end of the line) if starting from reversible dynamics for the system as a whole. First, for finite system size, the number of degrees of freedom must become (spatial continuum or field limit). Second, the size of the system must become infinite (spectral continuum or thermodynamic limit). Then left and right travelling waves in the line can be completely separated. In quantum mechanics there will always be fluctuations present in the incoming waves at the subsystem. This noise source remedies the violation of the fundamental commutator for the tagged oscillator. The noise is non-white, and unfortunately the uncertainty product for the damped oscillator cannot be evaluated as it becomes infinite, save for the limit of vanishing frietion a 0. 6. The dual Hamiltonian and complex caleulus 6.1. Complex symplectic model 6.1.1. Classical mechanics Recently a complex symplectic formulation for the damped harmonic oscillator has been proposed by Dedene [197]. touching the problem from the point of view of geometric quantization (see eg. {355-369]). Apparently, Dedene was stimulated in his investigations by earlier attempts of the present author to describe the damped oscillator in terms of complex dynamical variables [59, 24, 195, 258). The relation between these complex variables theories will be studied in some more detail furtheron, Currently we will show that Dedene’s Hamiltonian is connected with the time-independent dual Bateman Hamiltonian (3.6) by a simple complex canonical transformation. In fact, it is precisely the transformation used previously by the present author [59] We do a comiler a feyactcy cute [6]. mor any monaco etic moicaion of the mrs, Nor ds we com the extreme liu 4 [spay tn tha ose (88) ea be evaluated 35 waa trading Dia’ sdsadibtion, Using 4) en hen he a = AID the Sundar ret forthe wndamped hare onclatr erovdnate The extension of auton theory ever the extreme ME A! tortemly under ivesigation. The have Hay ae closely elite to the “pace fine” of Hadamard [S18] andthe moka Ihory of ener functions devcopey Schwa {S10 Temple [53 ad Liphtl [S21] Sean (4,91, 522 29, The rus ihe pushed ehewhere, See ahothe fate below (5290H, Detter, Classical ond quantum mechanics of the damped hamonc oxo a In the canonical transformations 1) the dynamical variables z and i Z = m, (as well as their complex (or Hermitian) conjugates) play the role ‘of a canonically paired coordinate and momentum. The wiggle on z indicates a formal complex (Hermitian) mirror conjugation. That is, one must both interchange the real system variables by means of the earlier defined mirror conjugation {A++-A:x, p>, 7}, see (3.9)-(3.10), and take complex (Hermitian) conjugates. Hence, (or 2"). The inverse formulae of (6.1) are easily found to be sezee-s), oe fBE429. azee-m. ae Bere. (62) One should very clearly realize the difference between (6.1) respectively (6.2) and (3.9) respectively G.10) (aside, of course, from the factor VA). The complex variables introduced in the Feshbach— Tikochinsky theory each involve a mixture of both the original and the artificial mirror image oscillator in terms of their mechanical variables, whereas the above definitions instead involve mixtures in terms of the canonical variables. In fact, writing (6.1) in terms of the mechanical variables one obtains (consult (3.5)) [p+(-iw)x], [b-@-ie)s}, (P-@+iod}, (63) Vig Pt ation}, while (6.2) leads to (-z*), a O° Tre [w-ia)z + (w+ia): (64) (o-idet rine], Bea ‘The formulae in the first column of (6.3) and (6.4) are exactly identical” to those discussed earlier by the author in connection with the real physical damped oscillator (59, 24, 195]. The fundamental ‘commutators of the Bateman-Feshbach-Bopp-Tikochinsky theory, (3.8), are transformed into * Set or instance in [195 formalae (49) and (4.10), ¢= 20)! and w =i 20)".2 H. Dekker Classical and quantum mechanics ofthe damped harmonic xia [nz]=A, 5) ‘As the transformation does not involve time explicitly, straightforward substitution of (6.2) into the dual Hamiltonian (3.6) leads to HaX+H*, (66) where H=(w-ia)zi. 67) This is precisely Dedene’s Hamiltonian.” which proves our assertion, Recalling that ‘equations are conveniently written as”” the canonical i=He, fH ms =H, =-H es. Kins 1-H =H (68) For instance, with , =i (ie. i 2*) and the algebra (6.5) one readily gets BH iwz-Az, 69) i= iwt Az (6.10) These equations must be supplemented with their complex conjugates, which stem from the other part of H, ie. %*. Clearly, (6.9) describes the damped oscillator, while (6.10) pertains to the mirror image system. Using the transformations (6.4) one obtains [59, 195] dep, (6.11) B= -2p-0'x, 12) and =p, (613) p=rp-2, (6.14) which are equivalent to (3.1) and (3.4) respectively, ie. describe the damped oscillator and its mirror image, as it should. * There it a mine difernce in sgn, not nA (3 it might sem 3 rt lance) bat nw: Sc for instance Dedene's equation of motion for Note furter, that i [10] the Bar denotes complex conjagation insead of real mor conjugation, "inthe symplectic formolation| 107] the egustionsof motion are from the Hamiloiansringivenby the vector Fel x = ~ilw =i AN2 ~ £0) operating on a function F= esp a8 F = Gr(F) = br2dP* and simian for 21H, Detter, Claicel and quantum mechanic of the damped harmon oxi 2 6.1.2. Quantum mechanical spectrum Let us briefly consider the quantum mechanics based on the Hamiltonian (6.6)-(6.7) with the algebra (6.5). Obviously, (6.5) is very similar (apart from a trivial scale factor Vf) to the commutation relations for the usual oscillator creation and annihilation operators. However, it should be observed that is not the usual Hermitian conjugate of z, but requires an additional (real) mirror conjugation. Following Dedene [197]. we shall here call this formal conjugation generalized Hermitian, for short G-Hermitian.” If we now define a G-Hilbert space with the G-inner product (compare e.g. with (3.34) wiv’) Vier (6.15) teins = 8. the norm conventionally chosen as (yj) = I, then the evaluation of the eigenvalues of the operators N_= iz/h and N. = =""/h can be carried through in the standard manner (sce e.g. [57, 16164, 300, 313, 314)). In brief, using (6.5) for instance for z and Z, one readily shows that these operators indeed are the annihilation respectively creation operator on the elements of the G-set {In.)}, for which Nfn.)= nln). 6.16) ‘That is, one finds (6.17) (6.18) where we have used the normalization of the eigenvectors and é,- = és. = ¢7. because #. = n.. Hence, the lowest eigenvalue of N. must be zero, and the spectrum is given by n.=0, 1,2,.... Further, as usual ¢,. =n! In the same way one finds ¢.=(n_+1)"?. A similar analysis for the G-set {n.)} with N,|n,)= m.|n,) shows that n, = 0, 1, 2,.... It should be noted that these sets of eigensolutions of H are completely independent of each other. Clearly, the Hamiltonians % and ", defined in (6.6), are defined on separate G-Hilbert spaces, say G.. Notice, however, that itis not the physical oscillator and its mirror image that have been separated. ‘The separation of H into ¥ and " deserves a more careful consideration. Let us first allow forthe usual Weyl symmetrization (see also [370]) for both the N_ and the N, oscillator. Hence, H = (iz + 2'8'+ he)-iA(éz - 28"), (6.19) where €=1 for Weyl ordering, and ¢ = 0 for so-called normal ordering (57, 299].** Evidently, the ordering (if identical for both subsystems) does not show up in the damping part of the Hamiltonian. ne 4 the pinot GoHeriin, but that swtal p=)" = a as is easly vee by expt evaluation ‘Any conceivable ordering i possibe in principle, Se ep [57 714 H. Dekker, Classical and quantum mechanics ofthe damped hamon oxi This latter part, on the other hand, obviously allows for a nonzero separation constant, say 6, as follows: H=H.+H., (6.20) = w(Ez + ihe) iA(Ez +388), (6.21) H. = w(2"E" + dhe) +i A(z"F +485). (6.22) ‘The eigenspectra may be given conveniently in a single formula, setting n. nas HP = (nt helo tin + $S\WA; 0 =0,1,2.... (6.23) Leaving aside here the precise observable physical significance of these spectra, it may be noted that the choice € = 0, § = 2, makes (6.23) an isomorphism of the Feshbach-Tikochinsky formula (3.42). On the other hand, with the choice «= 6=0 one obtains the Bohr spectrum as discussed by Dedene (197). Further, the choice reproduces the so-called Bohr-Sommerfeld~Maslow spectrum (see also {360)), another possiblity discussed in [197]. Finally (but see furtheron), the choice € = 1, 8 = 0 leads to Bopp’s spectrum (114). 6.13. Quantum fluctuations ‘The dynamics of the total system is described by the Schrédinger equation, that we presently write formally as ial =H). (6.24) If the two subsystems, the dynamics of which is generated by %, are independent (i.e. uncorrelated) at the initial time, they remain uncorrelated for ever in this complex-calculus-Bateman model. This is readily shown by setting | )= |+)|-) at any time, leading to ial-)" l-), ial+) = +). (6.25) To be explicit, the G-Hermitian adjoint equations read Cle, AGH = IRL (6.26) From (6.21) and (6.22) itis readily verified that the %. are G-self adjoint, i.e. 3. = %-. Let us consider ‘now the equation of motion for an arbitrary function F_ = F(z, 2). Since usually F-|~)is again a vector in G_-space, using the G-inner product definition (6.15) together with (6.25) and (6.26) one obtains the equation of motion in the conventional commutator form. As the analogous conclusion, of course, applies to an F, = F,(2", 2"), we have (Fy =~ Fa HD. (62) More general functions F(z, £2", #") can almost always be written as sums of factorizing terms F.F-.H Dekker. Cassicl and quanaun mechames ofthe danped humonic oxilltor s ‘As an immediate applicati of (6.27) one obtains the correct equations ay w(z)-Alz), w(Z)+Az), (6.28) on G., and their complex conjugates independently on G... As to be expected in view of (6.27), notice that the equations of motion are entirely independent of a particular choice of e and 6. Considering the model's quantum mecitanical fluctuations for the physical oscillator, however, the consequences of the above performed separation of the total system H into two noninteracting subsystems %. are more or less fatal. Namely. since F = z'z factorizes as F = F.F.. one obtains (2"z)™(z2")=(s"Xz) at any time, oF o.'. = o<'=0 (which implicitly defines the variances in the usual manner). By means of the transformation formulae (6.3) one easily infers the consequences for the variances in terms of the proper mechanical variables of the real physical damped oscillator. One finds: Orolt) + 2Agpe(t) + Po(t) = 0. (6.29) Evidently, we are in fact not even able to let the system start off at ¢= 0 from the free oscillator ground state for which oj, = 0 and dy = Mou, = AM/2. Of course, the fundamental reason for the failure of the formalism is the fact that neither of the separated Hilbert spaces G. contains the physical oscillator as ‘And the separation became feasible in view of the fact that both z and 2 commute with both as the only nonzero commutators are given in (6.5). Since this algebra is connected with the Bateman-Feshbach-Tikochinsky commutators (3.8) only by means of the simple complex canonical transformation (6.1). we are back again at the origin of the problems that have already be discussed in section 3. Presently it is even more clear that the dual Hamiltonian on its own ean never describe the correct oscillator quantum mechanics, as it can not even do so if A = ‘A possible obvious improvement of the situation could be obtained assuming that, before turning on the dissipation at ¢=0, the system has been endowed on 1<0 with the appropriate quantum correlations by some other means (most likely the correct free oscillator Hamiltonian). In that case the factorization | ) = |+)|-) is not valid at t= 0, and, switching on the dual Hamiltonian at that time, the subsystems |) remain correlated as they were, since from then on they are noninteracting again. With these initial assumptions the separation of the dynamics with #. and X, is in fact not allowed and one should consider the dynamics of the system as a whole, generated according to (6.24). Actually, this must be the situation that has been considered intuitively in the discussion of the Feshbach~Tikochinsky ‘model in section 3. From (6.24). implying in fact (3.46), one again obtains (6.28) for the first moments, while the equations of motion for the variances now become” jaar 2dose (6.0) = Liao 2a, 31) Gee Baer 6.32) so that “Weshould be cle tht in the prevent forma rs #* H Dekker, Casal and quantum mechanes ofthe damped harmonic oxi ye, (6.33) ull) = on(0)e™, 634) er)=erQe™. (635) Using the transformation formulae (6.4) one readily obtains for the position and momentum spread of the damped oscillator the expressions Ou = ~(U20) Ove stove. (6.36) Opp = (I2w)[(w -iAFore +2 (6.37) Latting the system start at ¢ = 0 from the proper free oscillator ground state, where op(0)= 27ou(0 AD and om =O. amounts to o-.(0)= AACA i e202. 00) = WA(A +i @V20f and a0) ‘ANM2a. Inserting (6.33)-(6.35) with these initial values into (6.36) and (6.37) finally yields outt)= fer [144 sin 200 +245 sint wt. (638) ant em [1 Asin 201+ 24ssin* ot] 639) which are easily seen to be identical to (3.59) and (3.60). Hence, the thus obtained solutions from the complex calculus for the dual Bateman model are the same as those from the real calculus of section 3. Clearly, a prerequisite to obtain (3.59)-(3.60) or (6.38)-(6.39) is the preparation of the actual oscillator, by an expedient other than the dual Hamiltonian, initially in a proper quantum state: the Bateman Hamiltonian does not in any way have the power to do so. Therefore, we again face the troubles of the Killing of the vacuum fluctuations. 6.2. Modified Bopp-theory 62.1. Separation of the dual oscillators ‘The above analysis has oncemore shown that the roots of the troubles are at the incorrect fundamental commutator (3.8). That is, the physical oscillator and its artificial, purely mathematical mirror adjoint do not commute, whereas the dynamical variables of the physical oscillator on its own do all commute with each other, even in the limit of vanishing friction. No doubt, rather radical measures must be taken in order to improve the case. An interesting hint in the proper direction may be found in Bopp’s largely overlooked 1973-paper [114]. In the sequel we shall present a modified version of Bopp's treatment, Bopp’s Hamiltonian, closely related indeed to Bateman’s dual model.” is quite akin to the complex 10 curious to note that Bopp [13] apparently was unaware ofthe work of Bateman [I] and Morse and Festhach (11) in this mater Incidentally probably due othe ited aceety of [13 Feshbach and Tikochindy [15,6] in thee arn were ansmae of Bopp pape. The present author's atention wax drawn to [1] and [1S] daring awit to Gorchig/Mnchen in 199 [372] See sho [1,109,297H. Dobler Clascal and quantum mechan ofthe damped harmonic exile symplectic form (6.6)-(6.7), or alternatively (6.20)-(6.22).” As shown above. however, a separation of the total Hamiltonian into two commuting parts one of which solely pertains to the physical damped oscillator will be a highly nontrivial matter. Now, note that in the complex formulation, as in the real cease in fact, the equations of motion for the damped oscillator and the mirror adjoint system are actually fully separated. See for instance (6.28). In the complex calculus the dynamics of the physical system on it own may be profitably written as fa-iKe, 2 be (6.40) for which the term improper (incomplete**) Hamilton pair has been coined previously [196]. The only remnants of the artificial mirror adjoint system in these expressions are Z and its complex (Hermitian) conjugate, although they do not appear in the resulting equations. The essential notion oncemore is that the physical oscillator and its mirror image are fundamentally independent in Newtonian mechanics. Evidently, the physical content of (6.40) remains unaltered if one simply disregards the bars. Ad- mittedly. this may be considered as a bold leap."** Indeed. its substantiation must arise from the resulting physics. in particular the ensuing quantum mechanics, to be discussed hereafter, ‘The fundamental commutator (6.5) from the complex dual calculus now leads to laa’]=1, (41) where for convenience we have introduced the conventional variables a = :/VWi and a’ = ='/VI. The pertinent transformations back to the real coordinate and momentum for the damped oscillator are immediately obtained from the first columns of (6.3) and (6.4) as (59, 24, 195} 1 1 . a= Pt O-iok], a= Feelp+A tio). (6.42) and a * - Fale] i\Ee-c ‘ p= VE tw -ina to tian’) (6.43) By means of these formulae it is readily verified that (6.41) transforms into**** [xp]=it (6.44) * Complete formal equivalence maybe obtained by repaig NV respectively 2'VE, "VB by Boe’: pert. a esgectely 1," Note, however ha hese operators ae ot nny guaranteed 10 behave a ordinary Heri conjugates. Thi ac is ny ase touched upon in [114 ee frthern in the present text, Parental, Bop wes Wey-ordering ize» 1, at at cn be ea verfed thatthe parameter ¢ will never appear in observable dynamics eeause the pertinent pat ofthe Hanilonan wit alays only ofer in commutator xpress his posit ean be deta rom he ott for convenience. The neignbence of the separation ontant 8 fe (621)(622) lew obvious: but wil be ited ot in due coun Sethe fit owing (867, The pair of equations (6) may he amndred incomplete i the tense that. though merely deerbing the dyaamice of the damped itr vile # and 2" at tans (til makes reference to the mio image sytem tro he ourence of #2 ic wa peel his ap tat wasn aly uncovered sch in ef [13 rr igdetlly,(642}-(643) ate fat only special numbers of 2 mich roar gro of tandormations conecting (hl) and (8 See e9“ Hi Dekker, Cascal and quantum mechenics ofthe damped harmonic ole Within the present context, this is an important upshot. Namely, it means that henceforth it must be possible to construct a formulation for the damped oscillator which intrinsically includes the usual quantum mechanics in the limit of vanishing friction. The price that has to be paid for the correct, operator algebra is the Hermiticity of the Hamiltonian, together with a somewhat unusual form of the equations of motion. From (6.40) itis seen at once that the latter become [59] aa (6.45) Ghar. with the generator for the damped oscillator now obviously given by X= hw-idjata. (6.46) Clearly, 3€* # . The complex (non-Hermitian) generator (6.46) isin fact precisely Bopp’s Hamiltonian H’ (apart from the unobservable Weyl ordering constant). For more on non-Hermitian Hamiltonians, often in connection with the so-called optical model potential, one should consult e.g. [10, 24, 59, 102, 164, 218, 300, 303, 373-386]. It should be noticed that (6.46) indeed reduces to the well-known harmonic oscillator Hamiltonian if A | 0 [160-164, 300, 303] (see also [190, 192-194]). Finally, for the sake of completeness, let us explicitly state that for the mirror image oscillator one now separately has a= -GUA)E 5. (6.47) H = hw +iada’ (6.48) which indeed can be considered as the formal (real) mirror conjugates of (6.45)-(6.46). Note that in effect we have really separated now the physical system from its artificial adjoint. 622. Quantization In order to investigate the quantum mechanics of the Bopp model for the damped oscillator, one constructs an ordinary Hilbert space (in lieu of a G-space), wherein the quantum dynamics is given by the Schrédinger equation il y= A(w-iaya'al ), (6.49) its usual Hermitian adjoint given by HiA( |= C[a(w +i d)ata (650) Evidently, the eigenvalues of the Hamiltonian become H,=nhw-id); n=0,1, (651) Compare this with (6.23). Note, however, the differing physical meanings here and there, connected with the differing Hilbert spaces, A first to observe from (6.49}-(6.50) is that, in consequence of the non-Hermiticity of the Hamiltonian, the initial norm of the quantum states will not be conserved in the course of time. Hence, in order to remain within the framework of the conventionalH Dekker, Clusal and wantin mechanics ofthe damped harmonic oxilar ” probabilistic interpretation. expectation values will be defined in the general sense (Y= CIFIMK I). (652) F being an arbitrary operator. Let us now introduce the pseudo-density operator w as the projection operator w =| ){ |, the state vectors obeying (6.49)-(6.50). The equation of motion for w becomes (see esp. [270]) i hv = How — WH (653) It is interesting, not in the least for comparison with Bopp’s original treatment (114). to write (6.53) in the number of representation” where (see e.g. (57, 58. 270. 298, 299, 302, 387]) w= [n) Wa (ra, (6.54) in|wlm) (6.55) One easily obtains Wm = 1 (= M)Wam = AC + IM) Harm (6.56) Notice that, as expected, the trace of the pseudo-density matrix {Wan} decays and, secondly, that diagonal and off-diagonal elements do not mix as time proceeds. Even stronger. in their evolution the elements Wan do not interact at all, For the latter reason, (6.56) is almost trivially solved by Won) Pam) CEMEMHAA IH (65 where we have choosen to set Wnn(0)= pan(0), which for the moment being just defines pam (0). At arbitrary times ¢2 0, the appropriate density matrix {pan} is defined as Pam (t) = Wan (t/trace w(t), (6.58) so that trace p(t) = 1, while (F)= trace pF =) panFaun (6.59) as usual. Note, however, that the present theory provides us with w rather than p. In order to obtain p(t) one must specify particular intial conditions, i.e. Pan(0)= Wam(0). Therefore, in principle, no ‘equation of motion for the proper density operator of the damped oscillator exists within the present context that does not involve the system's initial state.* “te aware of the diference betwen the prea and the ust (ee oxo) number states. In pape the later area Kncar combination of the former, nd vice ver, 7" powily more pec relation Between thi facet andthe cheered non Markovian behaviour ofthe quantum mechani anambon le (ccctiom 8) remain 38 ye unepined» H. Dekker, Cascal and quanta mechan of the damped harmon oxlaor It is illustrative (and rather unexpected) that the above description allows for the paradoxical stationary states, quite similar to those observed in the Sissmann-Hasse-Albrecht model [129, 253, 254} (see (2.7)-(2.12)) and in Kostin’s theory (198, 199] (see section 11).” Namely, if one sets Wan (0)= Pam (0) = BamBnny ONE obtains by (6.57) trace w(t)= exp(~2Anot), so that (6.58) leads to Pam(t) = BnimB nn (6.60) Evidently, an(t)=0 represents an equation of motion that essentially involves the special pure state initial condition. Moreover, as such it is not unambiguous as, for instance, the equation i fim (n= m)pan. with completely arbitrary time scale, would do as well 6.23. The density operator Having started in fact from a classically damped oscillator (see e.g. (6.11}-(6.12) and (2.13}-(2.14)) fone should, in view of Ehrenfest’s theorem, pay special attention to classical initial conditions. Stationary quantum states as such do not behave very classical. For instance. (p)= 0 for any n since (nlaln) = (nla'|n) = 0. The latter formulae immediately indicate the way out, In order to obtain states for which (p) #0, which in regard of (2.13)-(2.14) is a prerequisite for observing classical dissipation {108, 109]. one should consider the most classical states of quantum theory (184, 302]. namely the coherent” states a). By definition, the states |a) have the eigenproperty ala)= ala), 661) a being a complex continuous valued number. These eigenstates of the annihilation operator have been extensively studied (see esp. (57, 298, 299, 301, 302, 371)), notably in relation to coherent optical phenomena in laser theory (see e.g. (20-22, 35-37, 58, 92, 184, 388)). They form a (overcomplete set ‘of not strictly orthogonal states, properties expressed by (itm) flay aja = 1 (662) Karla)| = exp(ala - BP). (6.63) where d (6.54), the density operator p can be expressed in the coherent state representation, for example, ay Re ard Ima and where the integration covers the entire complex a-plane. In analogy with p= | Palayalda, (664) where the diagonal, P- or Glauber distribution P(a) is a so-called quasi-probability density.“*** The *trshoaldhe noted hat au fond the seminfnite ransminsonline (cto som imi estate a one might in princi prepare the exe's incoming mes syvem in an aPiary pote number sate that (BLA,)=m (ee ef. (él) Of coure, the notin of 8 temperature becomes moot in that ce, These minimum uncertainty sats [293,29] ate ako fequenly denoted ax Glauber sates ee €g (32), +" the fansuon Pith ntinurlly rdcred sat ety of Lol [7], who abo Unease este creat pdsting (ce ao 71), Assach, Pla) isonestoune rtd pom ote anc, ror the reer hl at get cane wth he mies“ poetlensty” (wen conection wh a: ee ($3 al “yuna orP) {he ler meant Ho sy tat my Hake on aeatbe wae, UMA wth 4 peer eas deeH. Dekker, Classical and quantum mechanics ofthe damped hamonc exile st coherent states are a specific linear combination of the number states, namely ns moat la) = expt ') exp(aa”j0) = exp(-f (6.65) Let us now assume that the damped oscillator is at 1 = 0 prepared in a pure coherent state, say |av). Hence, corresponding to P(a) = (a ~ ao) 5(a* ~ a3) in (6.64), one has p() = la) (ao). so that using (6.65) pion(0) = exp(—larol*) aSas™/Vnim! . (6.66) Substituting this expression for pam(0) = Wam (0) into (6.57), one easily computes the sum of diagonal elements of {Wan (1). with the result trace w(t) = expl-lad(1-e™)] (667) Inserting (6.67) into (6.58) one obtains” expl-laul? =i a(n — m)r— An + mt] (6.68) It should, however, be pointed out that (6.68) as it stands is not the most useful density matrix one might wish. Namely, by means of the transformations (6.43) it is not diflicult to verify that a pure damped oscillator coherent state implies a4, = R/2w and oz» = AN?/2w. Since both of these variances are greater than their free oscillator ground state values, we face the difficulty (which is quite akin to one earlier noted in the symplectic treatment along with (6.29): but be aware of the differences) that the system can not even start off with the usual minimum uncertainty. 6.2.4, The master equation Of course, more general initial conditions than the damped pure coherent state are possible, but they would spoil the derivation of (6.68). Even worse, they would spoil Ehrenfest’s quantum-classical correspondence theorem. For instance, it is a matter of somewhat detailed but in principle straightfor- ward calculations to show that with initial Gaussian conditions, the distribution P(a..) remains a Gaussian (see also [62)), but leads to mean values, e.g. for (x(0)), that deviate from the correct classical ones by contributions that do in general not vanish in the limit fr | 0. To be more explicit."* the analysis leads to the conclusion that these unwanted terms are zero only in three cases. First, if A=0 (for arbitrary initial conditions). Second, if x(0) = p(0)= 0 (for any A and arbitrary initial variances). Third, i the system is initially in a pure damped oscillator coherent state (for all A and arbitrary p(0) and x(0). Hence, on the one hand we see that more general initial states than the coherent ones, if incorporated in Bopp’s model from the outset, generally violate Ehrenfest’s principle and prohibit the derivation of "This seem the appropriate pce Gee abi the fasts pressing (6. 1 pint ut ta the imate FEA For poh Le HL 6 independent of any ational separation” constant (8) inthe an: Hermitan part of the Hamiltonian (48) H woul! merely mali Both the ‘numerator andthe denominator in (6. with the same eta fal exponential deaying (or growing fone wishes) factor, shih thes drop xt **Genealy speaking the abtionl tems ate peopoetional 9 ALe "flO p= eg, wheter chars ep represents the damp eile coeent te sal, Tutt as the ration Ce) = (pale hy teem amir rte= H. Dekker, Casscal and quantum mechanics of te damped harmonic onl (6.68). On the other hand, considering more general conditions by averaging in (6.68) over ag preserves the required quantum-classical correspondence, but in effect questions the validity of (6.49) as a proper Schrdinger equation for the damped oscillator. The latter point of view will be given more emphasis farther on (section 7). Let us then introduce the rather general density matrix = [ Plao% Fa, (669) 1%, given by (6.68). which is easily shown to satisfy the master equation [16-24. 34-37, 55-58, 62, 82, 114, 195, 196, 200, 270, 271, 274, 289, 332, 333, 340, 389-393] Bom = =1.0(0 = MY = ACHE Mom + AW GE DVT Dp stmt (6.1) Remarkably, the result (6.70) does not explicitly depend on the initial conditions. Nevertheless, it does involve the particular initial states used in its derivation in an implicit way. For instance. this facet is rather well illustrated noticing that (6.70) does not allow for the pure number steady states (6.60), contrary to (6.56) on which it is based.” The ground state m= 0 (which is a so-called random phase coherent state [S7]) is a noteworthy exception. It is also worth noting that Bopp [114] in fact, gave a somewhat heuristic derivation of the diagonal part of (6.70).7* Ban =~ Anon + 2A + partners and solved it subject to the initial condition par(0) = 8,2. Notice that, as in (6.56), off-diagonal elements do not mix in (6.70) and (6.71). Further, (6.71) is the well-known classical master equation for radioactive nuclear decay (see e.g. (34] p. 272) . Quantum fluctuations From (6.69) itis easy to obtain explicit formulae for the moments The first moments immediately lead to the correct classical expectation values (x(t)) and (p(1)). For the variances one finds (use (6.59) eo) reer, (672) Cas(t) = au, (6.73) Fal) = Tore + (1 a (6.78) aed = ou e™ (6.75) where for instance o4(0) = (ard) ~ (a)? , ete. Using (6.43) one obtains the relations ag = ~ (R120) O00 ~ (0'4 + Fan’) + 0 (6.76) This in tne wit the Lown pneitence (within the tae eis) of he P-dntabution fe pate number sates 211,212. 371). °° Bop's me has the same signicance 38 the pete fun1H Detker, Cleland guna mechanics ofthe damped hammic lene 8 Ope = (hie) i AP Fue + Plors'a + ua!) + (w+ iAP OW’ (6.77) pe = (Ww [(A +i w)o00— MO a's + Fas) (Aw) ae] (6.78) where ops is again the Weyl form (see (3.53). Notice the essential difference between (6.76)}-(6.77) and (6.36}-(6.37). which principally stems from the fact that presently a” and a do not commute. By means of (6.42) it is further observed that the usual free oscillator ground state initial uncertainty presently amounts to oa4(0) = A(A ~ i o)2a0, 0 4140) = A(A +i w)/2002 or a's(0) = (2220) ~ $7440) = (A/20)+ 5 Inserting (6.72)-(6.75) with these initial values into (6.76)-(6.78) finally yields” oll) = Fer™ [144 sin 200 + 2A sint + (Le) (6) am()= ee wef Asin der + 2ASsin? or] ME Le (680) ant) = ~e-™ sint a 2 (1 -e™), (681) Comparison of these results, in particular of (6.79}-(6.80), with (6.38)-(6.39) for the (complex) dual model, shows the important upshot that has been achieved by imposing the correct quantum mechanical algebra (6.41)-(6.44): the quantal vacuum obviously has not been turned off by the frictional forces [24]. i.e. Heisenberg’s principle is not violated here. 63. Summary ‘The time-independent Bateman Hamiltonian has been related to the symplectic damped oscillator Hamiltonian by means of a complex canonical transformation. The symplectic Hamiltonian can be separated, but the two uncorrelated parts are mixtures of the physical oscillator and its mirror image. Heisenberg’s principle is obviously violated. Then, following Bopp. the separation of the physical system and its artificial adjoint has been enforced. The resulting quantum mechanics has the correct algebra. Unfortunately, the derivation of the density matrix equation (6.70) from the postulated Schrddinger equation (6.49) involves somewhat unsatisfactory subtleties. As it stands, (6.70) guarantees the validity of both Ehrenfest’s and Heisenberg’s principle. 7. Complex phase space quantization 7.1. Preliminaries As noted already in the Historical survey (section 2), each of the theories discussed so far intrinsically involves the postulate of the existence of a Schridinger equation for the dissipative system. In essence, four types of dissipative Schrédinger equations have been encountered. First, a nonlinear Schrodinger Nove that presently af) #0. See (3.59 and [136,16s 1H, Dekker, Casical and quantum mechanic ofthe damped harmon exile equation (of the Siissmann—Hasse-Allbrecht type:” see section 2). Second, a multidimensional equation (two dimensional for the Bateman-Feshbach-Tikochinsky dual model, as well as for Dedene’s complex symplectic formulation: see sections 3 resp. 6; infinite dimensional for the Stevens-Yurke semi-infinite transmission line model: see section 5). Third, an explicitly time-dependent Hamiltonian model (namely that of Bateman, Caldirola and Kanai~*: see section 4). And fourth. a Schrddinger equation with a noteworthy non-Hermitian Hamiltonian (a modified presentation of Bopp’s model: see section 6). However, itis in fact well-known that the general state of an interacting subsystem will not be a pure quantum state,"*" but rather a statistical mixture to be described by a density operator [16-25, 50. 55-58, 61, 87-89, 200, 270-287, 298-303, 310-314, 375, 387-393]. For that reason. the significance of the assumed dissipative Schrédinger equations has been questioned. in particular by Burzlaff°*** [289] and the author (184. 24. 195]. In the preceding section it has been shown that Bopp’s Schrédinger equation (6.49) has certain undesirable implications (notably the possible violation of the correspondence principle), whereas the density matrix equation (6.70) as such is much more satisfying. Of course, in questioning (6.49) one should not overlook the evident importance of the correct fundamental commutator in Bopp’s (modified) theory. As should be clear, the latter (ie. (6.44)) is intimately connected with the separation of the damped oscillator and its mirror image. which are (quantum mechanically uncomfortably) intertwined in the complex (symplectic) dual model. On the other hand, the symplectic formulation is intentionally on good terms with the notions of classical and quantal phase spaces (symplectic manifolds and Hilbert spaces), employing the methods of geometrical quantization (197, 355-369}. Let us therefore return to Dedene’s complex symplectic calculus for the damped oscillator and introduce the phase space description from the very beginning (ie. already in the classical formulation) in an explicit manner. Then, imposing the correct algebra (as in Bopp’s theory) and eliminating the mirror image system from the description, yields a continuity (or: master) equation in the reduced phase space of the physical oscillator. Only after these (classical) preparations, the system will be quantized. That is, rather than seeking a pure state Schrddinger representation (which need not necessarily exist), one allows from the outset for arbitrary mixed states represented by a density operator. Of course, the latter representation contains the former in special cases (e.g. if A = 0). 7.2. Reduction of the dual phase space Consider the Liouville equation for the closed complex dual system of oscillator and mirror image, in the phase space spanned by 2, Z, #, =i2* and m,=iz*. It may be written as (c.g. [1])°°""" R,=-ilH.R}, a) where R is the total phase space density, the Hamiltonian H = 2 + X° is given in (6.6)-(6.7) and where =i ,] can be identified as classical Poisson brackets.“*°7* More explicitly, (7.1), reads in view of (6.5) wo other types of nonlinear Schrager equation wil be dncwued farther on: Hase's non-Hermitian made (ction 9) and Kowt's Schringer-Langvin or Maid drama! equation ction 1). "One could aha catalogue ere the Stevens orem ine mol ater the Hime dependent anonial warsfmation thin the preset context we hall defines pure quinom sate 43 sate obeying any sto Schrier equation iis among to note that Bopp promoted Burra thes [2 In anther mation [7,366-M9] i eeads By = ~FuR, with y= Ex * Exe: othe vector fl specified in the fomote a 8) Recall that H withthe agcbra (8) 8 quantum mechanically wrung. mo classe. Tetfore, ome vould notice tha the pectimiary manipulations here are in ela phe space1H. Dekker, Clasical and quantum mechan ofthe damped harmins ave s {Hy R,-i RH, +c0nj., 72) where “conj.” denotes complex conjugation. Note that H* = H. As it stands. (7.2) is not convenient for the intended reduction of phase space. Therefore, we rather cast it into the form of a continuity equation, Eq. (7.2) is evidently equivalent to” Ry = (HR), -i(RH,).4+ con}. 3) Recalling e.g. (6.8). noticing that for the classical Dedene Hamiltonian %* = %, and trivially inter- ‘changing the second term on the r-h.s. of (7.3) with its complex conjugate, one readily casts (7.3) into the form Ry = i(H eR). +0 -R).: +00). 4) which clearly has the structure R, = (R‘+ mirror conj.)+ complex conj.=(R‘?+ complex conj.)+ mirror conj. Recalling (6.40), one again notes that the first term (plus its complex conj.) on the r.h.s. of (7.4) completely describes the phase space flow for the damped oscillator on its own, whereas the remaining terms just pertain to the mirror image. Therefore, the elimination of the artificial mirror system from (7.4) is most easily accomplished by an integration with respect to 2 and 2°, with the appropriate measure. The latter need not be specified (but see [195}) as it suffices to indicate this operation by a classical projector S,, so that R=SR+(1-S)R, (75) SR=p, (76) where p represents the reduced phase space density, also called the relevant part of R (see e.g. (16-25]). However, since the integration is in fact not over a genuine canonical pair, one must redefine the remaining subspace algebra. As in Bopp’s theory (section 6), one naturally imposes the conventi algebra on the physical oscillator’s subspace. That is, en passant rescaling the dynamical variables again as 2/V = a and z*/Vi = a*, one is concerned henceforth with the fundamental commutator (6.41). oF (6.44), and the transformation formulae (6.42}-(6.43). Eq, (7.4) becomes jor. on with = hw -id)ara, (7.8) the latter in fact given earlier in (6.46). It should be noted that as yet (7.7) still is a classical, ¢-number equation so that the ordering of quantities is actually immaterial. However, it will turn out that considering the quantal analogue of (7.7), one should specify the ordering of %,p and X° even for Previous ostote% 1H Dekker. Casal an quantum mechamses of he damped harmon oscar linear systems if #”* X. As is well known. in principle such ambiguities cannot be circumvented in quantizing classical systems (ie. in adding quantum fluctuations”), although in many cases the ap- propriate choice can be inferred from intuitive arguments or imposing additional constraints (such as general covariance; see e.g, [228, 231, 394, 395}). In [195] we have given a somewhat intuitive reasoning concerning the detailed structure of the quantal version of (7.7). Presently. we shall rather leave the ordering undecided until we have extracted some more definite consequences from the theory.“* The final conclusion will be the same as in {195}. The so-called complex Hamiltonian ¥ featuring in (7.7) generates the classical, deterministic dynamics according to the equations of motion (6.45), which are obviously akin to the usual canonical equations, But it should be noted that, as a consequence of (6.45), this complex variable theory differs in a number of respects essentially from the usual canonical theory. although it naturally reduces to the latter if #* = H (Le. for the damped oscillator if A | 0). For instance, although # does not explicitly depend on time it need not be a constant of the motion. It also need not represent the proper mechanical energy of the system. Further, % is in general not invariant under a complex canonical transformation.”** Therefore, the theory should rather be called quasi-Hamiltonian. Nevertheless, does play the role of a kind of complex potential determining the trajectories in complex phase space. Moreover, it should be noted that the continuity equation in phase space is always more general than the standard density equation for the incompressible phase space fluid, which expresses Liouvilles theorem (compare e.g. (7.3) and (7.2)). The latter is contained in the former in the case of closed, nondissipative Hamiltonian systems only (see esp. (24. 37]). Hence, if one is not starting from a complete, fundamentally reversible Hamiltonian description of, so to speak, the entire universe (and ‘one is even almost never interested in anything of that sort [35-37, 402)). then the Liouville-continuity equation’** is a sound starting point. In particular this seems relevant if one is aiming at a pro- bahilistic description of the system. And, to be clear, we shall here entirely conform to the conventional, essentially probabilistic interpretation of quantum mechanics. Incidentally, other approaches to the (quantum) dynamics of open (non-Hamiltonian) systems which are closely related to the above basic ideas, may be found for instance in [61, 271, 289-291, 403-416). In many cases a (Markovian) semi-group property (ie. in fact a master equation; see also eg. (73,417]) is introduced as the fundamental dynamical postulate (see esp. (271], and references contained therein). This is quite analogs to ang Langevin ie woursest deterministic egetins nthe cai theo of dockatic proces fe. [2 537,56, 59, the observable resale depend for instance onthe choice of eoorinate, See ep. (3. and ah [2.37] Ingen. the abigitis ‘cin be lated up one hs a welded micsncoic phyaal model for the pres *" Ar the rato srertrning et wy se that fon 77) icon esa ha miter an ting sant e) Wo a ep conNan se, fr hth «and 5.423) ad the fone preseing (0) wl nlc the eas aang fs the presse Fat 77 Sch a rasbrmation bs Sonsentinaly dine preserving the bave form af the ution atom 4) See a eg Ua the one dimensional comple ae his mt evi, Ad hone tesstormation hat leave the Pon Brackets (4) vacant rather oem & seu ‘he standard real variable theory ths invariance is deste), Note that (641) i Tat i nol a peaper comple canonical anaormation i the elasieal theory hat ce [24 HH, HA) The comple character of generally breaks the invariance of the elsal Psu Reaches under the rari time-dependent (anol) ensformation tht wakes the dynamical problem, OF caune, there it mo fundamental need to forme Ctaieal mechani terms of Sacks rather han in Siete notation, mo even ned quatre the stem (ee (Tes. p38; of our, the imponace of he bracket fepteestation i he historical devebpment of comeeative quantum mechinis shu ole dpreiaed 1s) JOO) For some move deta on the above apes of the clues! compleyvatable they. see the seton om the Hint The momenta i the Mrture ut una, Although wasenhat opp if thre cin he mo mnderaning dhe coats eu sit mini ae beck comcntonaly Lave equation, mile the dhaanis we itrincls atric oe eg [A The mst ‘uation wl ease of the mos general vty toe Hn nan tec oa hee demote haem Noutaan eto (ce ange for open te futh mined sts) 8 Naka Zanze exam fe [IN=18 28)1H. Dekker, Casal and quantum mechani ofthe damped harmon exile s 7.3. Quantization Let us then consider the quantization of the Liouville-continuity equation (7.7), as it stands.” To this ‘end we let the classical dynamical variables become operators in a conventional Hilbert space, The only nonzero fundamental commutator is given in (6.41). ie. [a.a°]= 1. which fixes the algebra (see for a particularly clear treatment Louisell’s book [$7]). Since the ordering of . p and 3” is not irrelevant in quantum mechanics, we introduce a real continuous ordering parameter 0ScS1 and write the quantum version of (7.7) as FAC apt Hae P. +(1-c)pH hat 7.9) a By means of the general operator relations (e.g. (57)) laFI= Far. [a.F] (7.10) F being an arbitrary well-behaved operator function, one readily casts (7.9) into commutator form, with the result** fia Fla’. 1-c)[a, Xp + cpla, X}]+ con)... OAD) First of all, notice that by using the fundamental commutator (6.41) in going from (7.9) to (7.11) one simply endows the mechanics with the usual algebra. This is a conditio sine qua non for obtaining the well-known nondissipative results. Further, note that (7.11) by construction guarantees the conservation of probability, ic. (trace p)' = 0. In (59] it has been proved that the classical # can always be taken real whenever there is no dissipation, that is if the process is time-reversible (see also (300, 331, 413), Consider then (7.11) with Hermitian 2, i.e, "= % = H. It is not difficult to show that for Hamilton- ians of the form (sce e.g. [57, 164, 195, 302]) H=ha(ta'ata’s(ytasty, (7.12) a(t) and s(t) being arbitrary scalar functions of time, (7.11) reduces to the usual commutator equation b= —WANH, p). (7.13) independent of a specific choice for the ordering parameter c. The Hamiltonian (7.12) generates the ‘most general dynamics of a driven linear oscillator. For more general Hamiltonians the above reduction *Reauiing the ual oxiltorqasotom mechanic inthe Fini A=, the continuity equation a terms of and p lads ta violtion of the uncertainty principle whenever A# 0. Moreover ia terms ofthe cal variables the ordering arguments of [195] cannot be wed a there is 0 Hamitonian whatever, “We have choten to mite rather p, in expressions where the expt diferenialreqreentaton ofthe operators snot shown, See aos 1H. Dekker, Clasical and quantum mechanics of he damped harmon olan usually depends on c or ceases to be valid (in principle for nonlinear systems).” It is quite interesting, however, to observe that for the rather basic Hamiltonian H= hag(t)a'a + hag(t)b'b + g(tha'b + g*(t)ab". (7.4) which describes two interacting oscillators, the reduction to (7.13) is still valid, even if one generalizes, 7.14) straightforwardly to an arbitrary number of oscillators. See also the section on the quantum optics model Let us now return to (7.11) with the Hamiltonian (7.8). The Hermitian part of this Hamiltonian clearly falls into the class (7.12)-(7.14). Inserting (7.8) into (7.11) one obtains = ~iela'a.p]~A(l— cia’. ap] + {pa'. al) + Act(pa a" + (a. a'p)). sy Although, of course, any averaged quantity can be calculated directly from (7.15). it is convenient and instructive to represent it in terms of a quasi-probability density. This can be done in different ways (see e.g. [20-22, 57, 182-184, 298-302]). Here we choose the diagonal P-distribution, defined earlier in (6.64), which implies c-number averaging of normally ordered products (a')ta', Following Louisell [57]. one then assumes p to be in antinormally ordered form and by means of (7.10) arranges all terms on the rhs. of (7.15) into that order, whereupon one may simply replace the operators by their coherent state eigenvalues (see e.g. (6.61)). This leads to = (Ati w)aP)« + (A -i@)(a*P) a2 — 2ACP sane s (7.16) which is a Fokker-Planck equation with a nonpositive diffusion coefficient. The mean values (a(t)) and (a*(t)) immediately lead back to the correct results in terms of the real dynamical variables x and p, satisfying Ehrenfest's theorem. The equations of motion for the second moments are easily found from (7.16) to be (a?)" = -2(1 +i ea"), 17) (a"?)" = -2(A ~iw Xa"), 18) (a*a) = -2A(a*a)—2Ac. (7.19) Hence, in the long time limit (a)+0, (a*?)+0, while (a*a)——c. Since further (a)->0 and (a*) 0, ‘one obtains for the variances (compare with (6.72)-(6.75)): (7.20) Faw) = aul’ (721) ou) “Quite generally, Hamiltonians with moce complicated contribution in tems of he cretion and anniistion operatrs can Be considered 3¢ effective Hamiltonians, which ace actualy inferred ftom perturbationl considerations Based on a mote Tandamcaal, clement interaction. Fat umn, the Later athe asi rctere a ad a" Se fo nance the quntam tice mone [923 SIT AL SRS A983 TS, [M0, 384, 306-92 and, mutatis matands (lection are fermions). the then of suprcondtvity [¥7, 419-421] Se alo [15] the present section ‘on the quantum optics model and e@ (718)1H. Dekker, Casical and quantum mechs of he damped hamome wile » By means of (6.76)-(6.77) the variances in terms of the real space variables are then easily calculated, with the result (1-2c)h0* Rw. 0.22 pe (*) On Evidently, in order for these results to make sense at least ¢ $3, that is0Sc uncertainty product one obtains }, Moreover, forming the pw) ou (#) = (I= 2c)?" 4e*, 023) If A | 0, this uncertainty product approaches the value (1~2c)"A?/4 2 A7/4. The latter inequality only hold: or ¢ 50. Hence. in order to fulfil the well-established principles of conservative quantum mechanics in the limit of vanishing friction, one must conclude that ¢ = 0. With this value for c, the basic master equation (7.11) reads fa. #lp|+ pela’ al. (12a) which precisely is the master equation invented previously in [195]. In that paper. in fact with an eye at the pseudo-density operator equation (6.53), it was argued that structurally one expects 3 respectively 2H” to stand on the left respectively right hand side of the density operator. This argument immediately leads to (7.24). See also the footnote preceding (7.9). 74. The master equation Setting ¢ = 0 in (7.15) gives i wla’a,p]~A(la’, ap] + [pa", al) (7.25) This upshot is in fact well-known, notably in quantum optics (the laser well below threshold; see esp. [19-22, 57, 92)). It is further in line with Lindblad’s general form for the generator of the quantum dynamical semi-group in the Markovian approximation” [271]. And it also agrees with the results of Burzlaff (289] and of Huguenin [290].°* Consulting (6.54}-(6.55), with w replaced by the proper density operator p, one readily writes (7.25) in the number representation, It yields Bom = ~i 2( = Mm )Pam — AC + 2 )Pam + 2AV HF INF Dn etm et (7.26) Comparison with (6.70) immediately shows that (7.26) is identical to the result from the modified Bopp-theory (see also (290]). Hence, the formulae (6.72)-(6.81) for the dynamical variances can be carried over without any modification. "Notice that presenti we have not separately posted his approrimation, Compare eg a0 wth [19,24], Remember that there is evidence that he Markov approximation i not valid in the quantum domain (ef kaT Ru} See section 5. See ae. [3 27, 2) * athe tranormationorealcoorinates. note thatthe p > £ + A dflerssomemhat rom our mechanical p~« Further tisworth observing that Buraa expitly invokes Lind there,” HE. Dekker, Clascal and quantum mechanics ofthe damped harmonic oxillator Let us now transform back from the complex variables to the physical real ones. Inserting the transformation (6.42) for a and a” into (7.25) one obtains UH ol-if (x 001-0 ~ str br oll 32 Ux. ecoll- sh ep. ol +L. oD» (727) where (see also [24, 190, 195, 398]) A= ip? + ialp. x]. + 7x? (7.28) “The master equation (7.27) is comfortably east into the more general form {196] i (Hap) if poled y(n Delp. spl] px Dal B.0ll~ js Das. (729) where Ho= Sp? + 30x? (7.30) represents the free oscillator Hamiltonian, and where the dif pp Bo 3!) Pam ay* Om Ia can be defined by means of the so-called generalized Einstein relations (see eg. (57, 58, 184]). If we define as usual the drift operators according to Gy =(D.), (PY =(D,). (7.32) we infer from (7.29) that De=p, Dp = -2Ap- "x, (7.33) which once more confirms the validity of Ehrenfest's theorem. The generalized Einstein relation defining for instance D,, can now be explicited as (57, 58, 184] 2Dyx = ~(Djx) ~ {pD,) + (px), Os) which using (7.29) and (7.33) yields Dpyx = (Dpr + Dap) 3iRA, (7.35) while Diy = Dj.. Hence, Duy ~ Dye =ifh. (7.36)Hi, Dekker, Casal and quanti mechanics ofthe damped harmon ol oo This formula, already mentioned in the Historical survey as (2.22). was reported originally in [24]. where it played a crucial role in the derivation of the master equation. It represents a typical example of a quantum mechanical fluctuation-dissipation relation (see €.g. (21, 24, 57-59. 92] and the Introduction). connecting the (operator) noise (expressed in the diffusion constants) with both the fundamental ‘commutator (expressed in Planck's constant) and the damping (expressed in the friction constant).” ‘The density operator equation (7.29) will considerably gain in clarity if it is represented in terms of the Wigner-distribution function (24, $7, 195, 277, 278, 299]. In the coordinate representation this quasi-probability density may be given as Wo.x)= shy [emma iylole+ yy ay. 3 In two previous papers (24, 195] we have transformed (7.29) into its Weyl-ordered c-number equivalent directly by means of (7.37), invoking the fundamental eigenvalue relations in the coordinate represen- tation. Let us presently rather do the calculations in the coherent state representation. Actually, we can then start from (7.16), with c = 0, for the P-distribution. In this representation. the latter is related to the ‘Weyl-Wigner function by the Gaussian convolution (see e.g. (21, 56, $7, 92, 182-184, 298, 299, 371]) Wea)=2 f exo(-2le~ al) P(A)eB 8) From this integral relation one easily deduces the transformation P, = D(dlda, a)P + W, = B(d/aa, a + sa/da*)W, (739) where @ represents an arbitrary well-behaved differential operator function. By means of (7.39) one transforms (7.16) into W, = +iwaW),+(A-iw)a* W) art AW ane (7.40) Reintroducing now the mechanical variables according to (6.42}-(6.43) by means of a= Feel -ieks. e Taga tion, at) it is an elementary exercise to show that (7.40) becomes [195] fA ha? - pW, + 07xW, + AW) 5 +3 Wag (7.42) ‘This is a linear Fokker-Planck equation with diffusion coefficients precisely given by (7.31), while the Ht, fotowing for istance the Introduction, we felate the dilusion coeficiens to Gauuian white noise operators, (7.36) leads 1 (0+ ‘.Glei)=21AA AC) tn tat, in sch a form the futuation~ 1, while on the other hand A Ar<1. The assumption is meant to imply that, in the interaction picture, p varies only slowly over At. Furthermore. the rapid relaxation of its variables and its additionally assumed infinite size, are taken to imply that the reservoir, described by pn, preserves its statistical properties in the course of time. Indeed. the reservoir will be considered throughout to be in its own thermal equilibrium state at temperature T. ‘The coarse-grained time rate of change for p(t) reads = try RUF AD= RO 629) at ‘The general solution for AR(é) = R(t+A1)- R(t) from the Liouville-von Neumann equation (8.26) may be written iteratively as (e.g. (20,22, 24]) Reray= Riot S(-Z) f ai fans Pantie (ei) (Hi (4). ROL “I (830) Having worked one’s way back in time from ¢+ Af to & with Af> re, one sets R= pp on the right hand side of (8.30), with pu Ton. «3p Pa = Ela aulail (632) af exp Folk). (833) where j = | or 2 for the two level atom, and where Z, is the canonical partition function for the species 1, From (8.30) one has through second order: Ree a)=RO-ES f dhlgua'b, "+ con). pln! -prS if dh, i dgua'b, e+ conj.,[g,a'by 2" "+ con). pllpull ey a Terms in (8.30) with n 23 necessarily introduce nontinear terms into the equations of motion and, hence, should not be relevant to the lincarly damped oscillator. Actually, the higher order. nonlinear terms are difficult to estimate but finally (ie. after tracing over the reservoir etc.) turn out to be roughly of relative order A At(a"a). Since in the Markovian approximation A A1-< I. the mean number of photons should not be too high, ic. (a"a)«(1).* Inserting now (8.31)-(8.33) for px into (8.34), and performing in tn the patty clawieal cet the perfect Mackonan kit is pow in princi ee seton $9), 0 that 312(h')L0, See eg (SA. Esen though (UM) that cae the near approximation ean be mace ex then Se aso [M31] Further, atpating the al formule seston wl be popertoal ta e(wK te PE and to Nq/ 1 In summary. the molest tem can he seared poste (0) he Mark _ssumption bokls god. (2) the excitation ofthe taged mecitor i ot to high (the interaction mate element pe evant om il, mb (2) he atomic density sony moderate1H. Dekker, Casical and quantur mechanic ofthe damped harmonic xilaer " view of (8.29) the trace over the reservoir. the first order term on the r.h.s. vanishes because it contains only single fermion operators. The second order term yields ptt 8=p0- pS fat f dtotreQ.0). 635) with Qu (ts. fo) = (Uy + Vy digul? expli(w — a )fi ~ i(w - w, )f3} + conj. , (8.36) Us (ab, ab pp) 63) Vu = [pprab.a'b,| (8.38) and where we have already used the standard properties of thermal fermion systems, namely that tra(pabsb,) is always zero, while tta(peb|.b,) is nonzero only if «= v. Let us perform the trace of one term in (8.35) in some detail. For instance, try Un = (a°ap ~ apa’) tralpnb. bs) 639) By means of (8.31)-(8.33) one easily evaluates the trace: ttalond.b,) = troub.b,) TT trp. =D pantuilbsby lai) = pu22= Zz' expl-F.SkuT). (640) T Tiles) pauluilbbe laid so that, evidently, only those atoms that are initially in their lower state contribute to this interaction term. Similarly one obtains tte V, = (paa" = a'pa)Z," exp(-E, sku). 41) Having thus in effect calculated tra Q,, we next concentrate on the time integrals occurring in (8.35). It will be useful to define the spectral function se-a)= af drerplio-o,)ntf de exp(— ile ~ on), 6.42) which is easily seen to be independent of the global time 1, Interchanging the orders of integration one is left with sho~ 0.) = 3 [ats (Ar 1) explo ~ 0, ) 43)” 1H. Dekker, Casicl and auant mechanic of the damped harmonic oxilur 84. The master equation With the simple, provisional result (8.43) at hand the reduced secular master equation (8.29) is written as 6 = ila’. pal - xa", ap]+conj.. (8.44) where x; = xj(w). with j= 1,2 and ssw) =o Zs! exp(-B w/koT)tel? (we). 643) The atomic filter function + (w~w,) clearly picks out the relevant portion of the total reservoir spectrum (see also [299, 338, 469]}. If in that relevant range the reservoir spectrum is densely dis- tributed, which requires the number of atoms to be very large, the sum over jin (8.45) can be replaced by an integral over w, such that (compare with (5.47) of the trans line model) E> favre J nao’) do" = N Jotorde, (8.46) which defines p(') and where N represents the total number of atoms in the system's volume A? which interact with the radiation oscillator. The coefficients x,, (8.45), can now be written as n(w)= | (0 - @)o(w {1 + exp(tha' kT)" das", (6.47) where the plus and minus signs belong to j=1 and j= 2 respectively, where we have invoked the t result for the partition function exp(-E,/keT), (8.48) with ftw’ = 8-8: >0, and where we have further introduced the function” (0) = N [g(wVAP po’) 6.49) Notice, in view of (8.23) for g(w") = guy that ow’) will be independent of Planck's constant and that it is proportional to the spatial density N/A? of the atoms (sce also the footnote below (8.34)). ‘The atomic filter function (8.43) has both a real and an imaginary part, which can be easily calculated exactly, Defining s(o-0)= w-w) tis"w-w) (8.50) *Efectvety, combining (846) and (8). one bas Ey I= eof.1H, Dekker, Clascal and quantum mechanics ofthe damped harmonic olan 2 cone obtains Moo) = Gall coe —e1 Ar}, (81) Wo-w)= caf Se ghar aye . (852) Here we come across well-known subtleties. Although At is quite short compared to the characteristic time constant 7,= /A of the tagged oscillator, it must be considered as quite long compared to the decorrelation time of the reservoir in order for the Markov assumption to make any sense. That is, as noted before, 1/At< Awe, Awe being the effective bandwidth of the bath as now expressed in terms of ‘o(w') in (8.49). See fig. 8.2. Evidently, .'(w'- w') picks out only a small portion of the broadband reservoir spectrum, namely near the actual oscillator frequency w. Indeed, if we formally let At tend to infinity, (8.51) is recognized as one of the standard representations of the Dirac delta function (see ¢.g [163). Analogously, in that limit (8.52) leads to Cauchy's principal value of the pertinent integrals over w'. Hence, in the Markovian limit one has” 2(o-e')= r8(o-w')+B (S45). (8.53) Also separating the x, (8.47), into their real and imaginary parts according to x,(w)= «/(w)+ix(w) ‘one now obtains Ki(w)= ro(w) [1+ exp(tholksT)|", (B54) bn 26 Tat Fig. 42. Real (a) and imaginary () par ofthe reserve spectrum fle function of), according to (KSU-4$2). San the reser Banh “Ths upshot is consent with introducing a complex frequency w +i and letingSt-+™ in (3, and in the end considering € 10. See e, (7.143.164,1 Det, Chacon aon mca of dagen vette xe) Crm) {S22 a0" (855) The result (8.55) is a clear-cut example of a Kramers-Kronig dispersion relation (see eg [166, 303, 314, 465, 470-480] and also [113] p. 371 and (481] chapter 5). In the extreme classical regime Aw kyT, one sees that x{=0 while «i= o(w). Recalling that to a very good approximation the broad reservoir spectrum o(w) can be assumed to be constant over the entire relevant frequency range, (8.55) is usually taken to show that the frequency shifts x7 are effectively nullified in both limits. Only in the transition regime fw = keT some subtle dispersive phenomena are expected. Anticipating our particular interest in the case T= 0, we indeed take «, real, ie. x= 0 in the sequel A in section 7, it will be useful to represent the master equation (8.44) in terms of coherent states, Following the procedure outlined below (7.15), one arrives at the Fokker-Planck equation P= (A+iw)(@P)a + (A-i@)(@*P) a+ 2ANP oe (8.56) for Glauber's P-function.** In going from (8.44) to (8.56) the opportunity has been taken to transform back from the interaction picture to the Schrdinger picture, Furthermore, we have identified AeKinni, (857) = KIA (8.58) By means of (8.54) for xj and comparison with (1.10). itis readily verified that N, = (exp(hayknT)~ 1 (859) does represent the standard thermal Planck function. Evidently, at T = 0 (8.56) is identical to (7.16) with ¢=0. Indeed, in the Schrodinger picture the presently derived Markovian master equation (8.44) for the reduced density operator reduces at T = 0 to ela'a, p]~ Ada’. ap] + [pa". al), (8.60) or with A = o(w), which is exactly identical to the basic result (7.25) of the mesoscopic complex phase space quantization. ‘The doppiness of thete considerations is acknowledge. However, since they depend on the det ofthe actally unkaown reservoir spectrum ou hey ae cl to improve and, as sul, wl be aken for ranted "tn overt uplate and more o es complete the lit of references: ce alo [486 ‘Hence, the present upshot ako agrees withthe result (670) from the modified Bopp-heory. See ako a recent preprint by Devlene [457] stich however ont contains the dagoea! dynamics (a ia Bopp orignal paper (Ud): see further the remarks ia Section 6. Fall i shoul be soticed that in both sections 7 and 8 he ealulation have beet performed wing the feetlonal ereaion aed anebation operator rather than the freeones. fact, itt not wo ey to compare the complex phae space quantization ain the free operator inthe elective complex Hamitonian (sce £€. [236 2) wth ab-eio model elelatons, Namely in that ese, scoring to Agarwal [S, db, the aniewnan! teractions ithe amittonian (16) must be cared along in ode forthe clasial reduced lequency wtoberepoduceia he revalts However 2.nocinvcetion 7.08 ‘the ful Harioian (14) the maser equation (7.34) doesnot eral reduce othe standard Schrnnger-Liowille equation 50 tha the omparison not ntermally consent. Rather. one shoul! ako do the made! eakulaions fom (24)1H Detker. Cast and quater mechanics ofthe dumped harmon exile 85. Summary Starting from the well-known microscopic Hamiltonian for an atomic reservoir interacting with a single quantized radiation oscillator, the reduced master equation for the latter subsystem has been derived. It describes linear damping if the usual sequence of rather subtle steps is taken. The rotating ‘wave approximation is made and the Hamiltonian is written in terms of the creation and annihilation operators of section 7 in order to guarantee the radiation field to oscillate at the proper classical frequency w. For the model Hamiltonian (8.21) the master equation (7.24) is equivalent to the conventional Schrddinger-Liouville-von Neumann equation, from which the further calculations are made along standard lines. The interaction between atoms and rac is assumed to be weak, and neither the density of atoms nor the initial excitation of the oscillator should be too high, otherwise the effective equations of motion would become nonlinear. A Markovian assumption is needed for the thermal reservoir, that is taken to possess a dense and broadly distributed frequency spectrum, which implies an infinite number of atoms and, hence, an infinite size of the system. The resulting coarse- grained master equation reduces in the zero temperature limit precisely to the essential upshot (7.25) of the mesoscopic complex phase space quantization of the previous section. 9. Hasse’s pure state representation 9.1. The Schridinger equation Let us return to the complex phase space quantization of section 7. The basic outcome (7.25) of that procedure has been represented in terms of the original Hermitian coordinate and momentum in (7.27), For convenience, we repeat it here in the somewhat more general form (7.29). That is, b= fH. 0)-i2 Lalp.plel+ ps On + Da) (ol ol=zs DulPlp. ell Dobe ol. (0.0) where Hy is the free oscillator Hamiltonian, and where the diffusion coefficients have been specified in (731). The master equation (9.1) cannot be written in pure commutator form. In other words, there no ordinary Schrddinger wave equation corresponding to it, save for the case A =0 (ie. also all D's ‘equal to zero). Nevertheless, it turns out to be possible to find a special, stochastically equivalent Schradinger equation in the limit of weak damping A 10. However. as will be shown in the sequel. this is possible only if the diffusion coefficients in (9.1) obey Hasse’s pure state representation condition. See [107, 310, 311] for Hasse's original derivation, and [196] for a slightly generalized formulation.” ‘One may trivially write the density operator p in (9.1) as (0) = arte) + pilt), 02) where pit) is of the order A and represents a small correction to p(t). As p(t) is taken to describe a nondissipative system that can be represented by a pure Schrodinger state vector, say | }, one has at any time See ako 292] for related considerations* Hi Dekier, Classical and quantion mechani ofthe damped harmonic ola ils) = polt)=| «|. 03) ‘An arbitrary density operator possesses the property tr p*(t) 1. In order that it represents a pure state. it is both necessary and sufficient (see e.g. [163] p. 336) that (9.4) po If the system starts off from an initial pure state, then tr p°(0) ind (9.4) is guaranteed if the time derivative of its ILh.s. is identically zero. Therefore, since (p’)= pp +p, we multiply (9.1) respectively from the left and from the right with p. and add up the resulting expressions. In a first order (Born [55, 163, 164, 303, 314, 464, 488-490]) approximation one then replaces p(t), in terms that are already of order A. by po(t). This leads to A iA = fs puro +14 Un pox] + 2p + fa Dn + Dear + spor [or + ap), pile = [& puPosl-~ spol) = jz Da 2ouP"on~2Ap.2npos + [0 03} ga Dor 20x70 ~ 2 pox. + Lp.) @: Notice the anticommutators in the dissipative terms. Now using the idempotency property 9’ (9.3) and noticing that for an arbitrary operator function F po from oF pa = (F)po 9.) one easily evaluates (9.5) further as: 0) =~ fla IER LW pal + Dont fs Dye + Deg ox + 39)+ pet 3p)~ (oye ~ Ce. le = fa Dul(?)~ 2pI0 + P* pale ps Dml)~ Bade + pale 7 where (px + xp)~ Mpx + xp) + (p)x ~ (x)p. 9.8) Note that the convenient choice (W)=0 only involves a nonobservable time dependent change in the phase of the wave function. As discussed below (9.4), the propagation of a pure state is guaranteed if the trace of (9.7) is equal to zero. This requirement immediately leads to Hasse’s condition” (196, 258] ob7a, (9.9) 9, Dy*D, 20d Dye + Dip Dasdyp + Dope — (Dys + Dap Op “tn Have’ origina notation (31: 4 by. Dy1H. Deer, Clasica and quanten mechanic of he damped hamoni siltr n where the usual definitions (3.53)-(3.55) for the variances have been used. Since under the condition (9.9) for al times p%(t)= p(t), and since within the Born-approximation we may again replace py(t) by p(t) in (9.7), the latter formula can be written as (see also eg. (59}) B= ~(UAY[Ho+ AW, p]+ (WAND. ple. 9.10) where we have introduced the operator [196] D= AA + (Uh)(Dye + Dap) (px + xp)+ Xx + xp)~ (px ~ (Ip) ~ (UR )Decl(p") = 2p)p + p*}~ (UD ypl(2°) ~ 2x) + 27}. oy Now defining the nonlinear, non-Hermitian Hamiltonian = Hot AWHid, e.12) the result (9.10) can also be presented as (compare with (6.53)) b= -WAYGp - po"). 9.13) which manifestly shows that the wave function obeys the Schrédinger equation ihd,= Ob. (0.14) It is amusing to compare (9.14), with the Hamiltonian (9.12), with the Séissmann—Hasse-Albrecht Schridinger equation (2.10). The presently derived frictional potential (9.8) appears to be identical” to the heuristically proposed W in (2.9) with c=, being one of Hasse’s choices [129]. However, the present model has quite different features in view of the additional operator 2. 9.2, Wave packet solutions ‘The equations of motion for the moments are easily calculated from the Schradinger equation (9.14), or alternatively from the equivalent density operator equations (9.10) or (9.13). For the zeroth moment, the norm, one has 1)" = QIAKD) = 0 @15) by virtue of (9.11) and the pure state condition (9.9). In fact, (9.15) just re-expresses the latter. For the first moment (x) one finds (x)= (A) (x, Ho + AW) + (RY Lx. D].) 9.16) ‘The contribution from the potential W vanishes and one obtains * Apart fom an unimportant phase feo.* 1H Detter, Casical and quanti mechanic ofthe damped harmonic meter (x)" = (p) + 2A(x) + 2/A7\(Dye + Dap Mxpx) — (x7) (p)} QUA Das (pep) — (px + xp)(p) + (x) (p")} = (ZI )Dyp (2°) ~ (x) (2°)} (9.17) Defining the third deviation-moments (ic. from the mean) Same = (Ue — (Hd) (¥ ~ (WN) (= (x). 0.18) where 1, v and x each stand for x oF p, (9.17) is rewritten as (2)°= (p)+ QUA? W(Dpe + Dep ape ~ Derg ~ Deer} + (21h Xx) (2). @.19) Since (D)= 0 in view of (9.15). one is left with” (x) = (D)+ 2M (Dpe + Dep) ape ~ Das Opep ~ Depexeh 9.20) which reduces to the correct result (x)’ = (p) as obtained from the original master equation (9.1) for wave packets with zero skewness. As in view of the linearity of the oscillator problem the propagator of (9.14) will be a Gaussian, zero skewness is guaranteed if the system is initially in @ Gaussian state, This is, indeed, the usual assumption. Under the same conditions (p) = ~2A(p) ~ DX x) + (2RM(Dye + Dep) pup ~ Decpre ~ DyelTene} (9.21) also reduces to the correct equation for (p). Confining ourselves henceforth to Gaussian wave packets, it is a matter of straightforward application of the Gaussian theorem to obtain for the second moments, for instance, Gag = ope + Dax + (ARYA?) (D) ~ (lA)? (D) + (/R)Dex Open ~ ih? ~ oF) - (9.22) Once more, (D)=00 in view of (9.15). Further, the expression between brackets is exactly zero on account of the uncertainty relation (2.18) for Gaussian wave packets. Similarly calculating the equations for the other variances, one obtains altogether Gigg = Wap + Opp ~ Perey + Dye + Dep, 023) Gp = Aa — 2? 0p, + 2Dyp, (0.24) Ger = 2p: + 2Dax- 9.25) Comparison with (7.43)-(7.45) readily proves that, albeit only for Gaussian wave functions, the frictional Schrddinger equation (9.14) indeed represents the same dynamical state as the original master equation (9.1). ‘The essential proviso on the above results is that the pure state condition (9.9) must be satisfied. In Novice diernce in ign with respect to (31)1H. Dekker, Classical and quinaam mechani of he damped humane oxi ” general, this will not be the case in the dynamical state, But, even in the final steady state. if it exists, the fulfilment of Hasse’s condition is nontrivial. Nevertheless, at least we can show that the above theory is selfconsistent in the long time limit. Namely. since the stationary solution of the Schrédinger equation (9.14) with the Hamiltonian (9.12) will be a Gaussian, the special form (2.18) of the uncertainty relation holds in that case. It then remains to show that the steady state solutions of (9.23)=(9.25) do imply (D)=0. The pertinent steady state fluctuations have already been given in (7.46)-(7.48), from which we first of all learn that (2.18) can be specialized to Spo) Gue(®)= Ih? = Dene) (9.26) Inserting (7.47) for o»(2). one has Decl O00 (2) + 22 pe(%)] + Dyp Gus) = 7A (9.27) Most easily using (7.43), with dp. = 0, for the quantity within the square brackets in (9.27), one finds Das Opp) + Dyp us (#)+ (Dp + Dip Dex = SPX (9.28) Finally, once more using D.. =~op(®) in the last term on the Lh.s. of the equality (9.28), it is immediately seen to represent the steady state version of the pure state condition (9.9), which is what we set out to prove. Vice versa, if the steady state fluctuations follow from (9.23}(9.25). and moreover,” if (and only if) they satisfy Hasse’s condition (9.9), then the Gaussian wave packet uncertainty relation (2.18) holds. It is quite interesting to see, as is most easily verified by mere inspection of (7.31), (7.49) and (9.9), that the specific fluctuations emerging from the complex phase space quantization of section 7 precisely fall into this category (this was originally noted in (196]). In other words, the dissipative ground state of the damped oscillator of section 7 has the special property that it fulfils Hasse’s pure state condition and that, hence. it can be described in terms of a Schridinger wave function, ”* 9.3, Summary It has been discussed how the quantum mechanical master equation for the linearly damped oscillator can be approximated by a nonlinear frictional Schrddinger wave equation with an essentially non-Hermitian Hamiltonian. Two requirements must be met: first, only Gaussian wave packets should be considered, and second, Hasse’s pure state representation condition must be fulfilled. Unfortunately, the latter is generally violated. It has been shown, however, that the specific steady state fluctuations predicted by the complex phase space quantization of section 7 do satisfy Hasse's condition, such that the dissipative ground state can indeed be described exactly by means of a Gaussian wave function. This realy is a nontrivial ational condition. Counter examples are easily constructed ‘Tee remarks scem in pce Fit the wavepacket ie geen, of coune, i a0 common, fee exile minimum uncertainty packet ‘Second in trasent phen dynamical state wil defi be dillerent (rom He's parametrized by a eal scalar degree of freedom, that (ut Hasse condition exaty in the lng time mit. Se [238] for.i lane “0 1H Debier Clase and quantum mechanics ofthe damp ha 10. Elements of complex Hamilton-Jacobi theory 10.1. Hasse’s classical dynamics 10.1.1. Complex calculus As noted in the previous section, below the Schrédinger equation (9.14). there is a close similarity between Hasse’s potential (9.8) and the Siissmann-Hasse-Albrecht-potential (2.9) with ¢ = !. although the additionally occurring non-Hermiticity of Hasse’s Hamiltonian (9.12) should not be forgotten. In a recent paper [43] Stocker and Albrecht indicated how the nonlinear frictional potentials (2.9) of Siissmann, Hasse and Albrecht might be considered as formally generated within the framework of classical real space Hamilton-Jacobi theory (see e.g. (I. 2. 7. 9. 223, 224]) and the closely related fluid dynamical interpretation of Schrédinger's wave mechanics (see e.g. (190, 201-204, 216, 225)). Let us therefore consider the classical Hamiltonian analogue of Hasse’s theory In the classical formalism it is convenient to set (F(x. p)) = F(x(0).p(0)). where F(x. p) isan arbitrary function of the intrinsically prescribed functions of time x(t) and p(t). The latter are the solutions of the dynamical problem. The complex Hamiltonian (9.12), where Hy is the free oscillator Hamiltonian (7.30), W is given in (9.8) and 2 follows from (9.11), is now comfortably written as” = Hy + AD + pO) (x x(9) + é (Dp. + Depp - p(t) (x - x(0) — j Des(P ~ PO) ~ 5 Duel — XC)? (10.1) H,(0), ie, the classical mechanic Evidently, by choice in faet, along the classical trajectories S40) fonian equations of motion with energy.” It is particularly interesting to investigate the usual Ham (10.1), One obtains =H, + ACK ~ x(0)+ WUA)MDpe + Dep M(x ~ x(0) ~ 2ih)Dex(P - PCO). (10.2) B= Qe = - Ax - AD + PCO) Wh Dpx + Dap (P ~ pl) + 2(i/A)Dpo x ~ x0) - (10.3) Since by definition we are looking for the solutions x = x(t) and p= p(t), these equations reduce to 6.16.12), dsp, p=-2p-x, (10.4) which correctly describe the classical linearly damped harmonic oscillator. Notice that this result is obtained independent of a particular value of the diffusion coefficients. Of course, quite generally it would be rather problematic for imaginary terms to enter the classical real variables equations of motion, Although as noted, the classical dynamical theory obviously holds for arbitrary diffusion coefficients, Note thatthe difusion coetcens (731) ate ofthe order A 0 thatthe coficiens in (10.1 do have proper claims =" Remember alin the quam they me ave (9) =H),1H. Dekker, Casical and quantum mechanics ofthe damped hamonc exile a we will confine ourselves in the sequel to the specific case (7.31), =f "20" Da Dy BH, DD =, (10.5) which is quantum mechanically closely related to the canonical transformation (6.42)}+(6.43), or equivalently to the first columns of (6.3) and (6.4). In the classical formalism it is preferred to eliminate Planck's constant and, hence, to use the complex variables z, 2* rather than @ = 2/VA, a* = 2*/VA. Inserting P= gle -iaye + otiae'] (10.6) and the diffusion coefficients (10.5) into (10.1), one obtains Da -iA)etz +Baz"z+ oO, (10.7) where g(t) is an in principle dynamically arbitrary function of time that can be used, for instance, to make {9(0) real (and even represent the true mechanical energy) along the classical trajectory. It is well-known that the Hamiltonian (10.7) can be obtained from the Lagrangian (see e.g. (24, 182, 195, 300, 331, 418) Qaizti Oe.2.0). (10.8) so that the momentum conjugate to the (chosen) coordinate z reads 7, = iz*, which is line with the modified Bopp-theory (section 6) and the complex phase space quantization of section 7. The canonical equations of motion become wes ciwz- Az, (10.9) wz* Az" +2A(2"-2°(), (10.10) which along the proper trajectories z = z(t) indeed reduce to the correct results. Compare e.g. with (6.9); see further (24, 182, 195, 196, 258, 288] and [197]. Notice for example, following Bopp [114]. that E42AE +2 = 4 DE-(A-iw)E = (Et Ar tiwz) =O. (0.11) In the preceding we have used the time independent canonical transformation (10.6) which preserves the basic Poisson-brackets of coordinate and conjugate momentum. That is” {xp} = {2 Medes (10.12) “although the quantum mechanical erature (153,161, MO i reasonably coment i denoting committe brackets hy [and Pisin cts by [Jin the real of clasical mechanics varios notations forthe ater ate Aoating around (se eg [1.2 7.9) For defiiteness (0013) razed to higher dimensional yes amounts t U. ths = Ss Uatan~ Ran) For convenience, we further et €8 Luly = U.P2 H. Dekker. Classical and quantum mechamerof he damped harmon exter Of particular interest, however. are time dependent transformations. Of course, especially that transformation is of value that solves the dynamics or, within the present context, at least removes the irreversible part. It must then immediately be pointed out, that such transformations need not be canonical in the usual sense as contained in (10.12). Consider in view of the linearly damped oscillator problem, for example, the dynamical transformation from z to z’ according to zeze™, mame (10.13) Remember that , =i z*, so that 7,-= i 2'*. Substitution of (10.13) into (10.12) yields (2), Meda =O, (10.14) so that (10.13) is certainly not canonical in the sense of leaving the fundamental bracket invariant, However, on the other hand, it is obvious that the remaining reversible dynamics can be obtained from the Hamiltonian zz", (10.15) (10.16) in licu of (10.14). The other way around, if we had started from the well-known conservative dynamics, (10.15)-(10.16), the transformation (10.13) would lead to an exponentially decaying bracket in (10.12), i.e. for instance, {Phew (10.17) which is reminescent of the unwanted feature encountered earlier in (4.9) within the Caldirola-Kanai theory, and which has been seen to be closely related to the violation of Heisenberg’s principle in quantum mechanics. Of course, in classical theory (10.17) has no basic physical, but only mathematical implications. It is instructive to briefly investigate the time dependent transformation (10.13) in slightly more detail. Introducing, for convenience, the notation z’= #%(z, 1), one determines the equation of motion Fe AGTE- BB Wet Bay (10.18) making use of the general form of the original Hamiltonian equations (10.9)-(10.10). The expression in brackets on the r.h.s. of (10.18) is easily identified with the Poisson bracket (10.14). Noticing further that, i, does not depend on 7,-= iz'*, one obtains the transformed Hamiltonian = eM H+ mA + 2 KO, (10.19) where x(t) is an as yet arbitrary function of time, The latter is fixed by the “complex dynamical consistency requirement”, demanding that1H. Dekker, Cased and quantum mechanics ofthe amped harmonic exile ® (10.20) In words, (10.20) requires the operations of complex conjugation and time differentiation to commute with each other, at least along the true classical trajectory. This condition is both necessary and sufficient in order for the complex dynamical formalism to be selfconsistent. In the present case, using (10.19) and recalling that 73(z,¢) = z', (10.20) readily leads to (=i RT — eT dle-ew = 2A 2) (10.21) Inserting this result for (1) into (10.19) finally shows that the original Hamiltonian (10.7) indeed transforms into (10.15).” Transformations like (10.19) are not obtained within the standard treatment of the theory of canonical transformations”” (see e.g. [1. 2.7. 9). At least. this is consistent with our earlier observations concerning the conservation of Poisson brackets under canonical transformations. Indeed, theit in- variance can be derived once the usual form of the canonical generators is given (see especially Goldstein's [1] treatment. proceeding via the Poincaré integral invariants and the Lagrange brackets). However, the basic definition of a canonical transformation is that it preserves the form of the canonical equations of motion. This definition, almost trivially, implies that the most interesting time dependent transformation, namely the one that solves the dynamics, is always canonical. No doubt, transforming to the initial state, the new variables are time independent by choice and, hence, can always be derived from the zero Hamiltonian.’** In other words, the dynamical Hamilton-Jacobi transformation is canonical per se, whether it preserves Poisson brackets oF not 10.1.2. The Hamilton-Jacobi equation Let us now recall the complex canoni omentum rel ion ind extend for a while the formulation to an N-dimensional problem.**7* Suppose then that we were given a function Z(2, 1), with the property izt=Sai (10.22) Note that 2 = {2} -s2w}, and that the ths. of (10.22) solely depends on 2, not on z* Separating (10.22) into J imaginary parts, one obtains 2N relations connecting 2N unknowns. obvi dynamically equivalent to (10.15). The tte jas the inplnt member of grup of dynamically indwtinguhable generar. Thi ambiguity intimately connected with the arbtarines ofthe difuson cieficens in the orginal el variable Hamitoian (1) wih respect the eae dames (102410) ‘Awan vet mcerincinesmas heath tini the shana aia Gog wl be t fancy consi the problem othe waa "even pane spe”. Soe eee acon eae [7] Of cure, ma he he 3 conan, nema! of he ew “aac” Gable "Ta hese, Ns aiteary a he pine spe il a 77 Nate ere lilerenns with wal hey of anon ‘of 40299 = Souhih ick aelation Peep a Ova this alse fox, For tha purr ne au the seve al of th tans fem foe cme 4” 8. where the patel arabs ste Te MO IRE the a cn Inte comp shh hey he sccm tn fast, the open comple cugaon el aitoan and rar snes rasormatinsin tems real pce sncchanal ara. The sand ase stain the luton casi eben nicas 1H. Deller Clascal and quantum mechanic ofthe dump! harmon oxilot to determine that particular generator Z(z, t) that makes this function z(t) identical to the solution of the dynamical problem. To achieve this, consider the total differential of 22. On one hand. by (10.22). it can be considered as a function of z and ¢, so that ,dz,+ zi dt (10.23) Using the general form of the canonical equation (10.9) for z in the sum, and (10.22) for z% in the last term on the r.h.s. of (10.23). one obtains Gb.e7dt-i(Sa).n dr (10.24) Dz. (On the other hand, from the second canonical equation (10.10) one has (10.25) Equating (10.24) and (10.25) leads to Gat Dh Oat (10.26) Now notice that 2 isa function only of z and , but that in (10.26) the Hamiltonian © still depends on both z and z* (and possibly on t, of course). Substituting (10.22) for z* into s, we define” (2*(2,£). 5.) 84'(2. 1). and consider at LOZ. (10.27) Substituting (10.27) for ©... into (10.26), the result may be written as +E) =TDEt,- 2 We (10.28) In view of (10.22), the rh.s. of (10.28) vanishes termwise identically. Hence, §'(z. t)+ S.(z,¢) is a function of time only. The latter dependence reflects the well-known possibility to choose a different zero reference point for the Hamiltonian at each instant ¢, without affecting the classical dynamics. This has been expressed earlier, for example, in (10.7) by the arbitrary function g(t) in the Hamiltonian ©. In conclusion, we may wet E+ S.2,N=0, (10.29) One should of coun ot conse this (2) with he in W0.as}-1030,1H. Dekker, Cac and quantum mechanics ofthe damped harmonic sie s or, in extenso, in an obvious notation: HCI S..2.0+S,=0. (10.30) This is the complex Hamilton-Jacobi equation relevant to Hasse’s classical dynamics. As usual, it has the structure of a first-order partial differential equation in N +1 variables and, consequently, a complete solution must involve N + 1 independent constants of integration. In the particular case of the damped oscillator Hamiltonian (10.7), the Hamilton-Jacobi equation (10.30) specializes to* 2G, ()-Atiw)zS, +S, =0 (10.31) Although (10.31) can, of course, be solved by general techniques (e.g. the method of characteristics [78, 83, 113, 346, 491, 492]), mere inspection of this special case immediately reveals that it is satisfied by See az eter, (1932) which does contain two constants of integration ay, a. As usual, the additive constant a arises because (10.31) does not involve the generator & itself, but only its partial derivatives with respect to z or f. But, by the same argument, the additive constant is irrelevant for the solution of the dynamical problem as contained in the transformation formula (10.22). Applying the latter to (10.32). one readily concludes that a, =i 2*(0). It is finally easily verified that the currently obtained solutions for z(t) and 2*(t) indeed satisfy the original canonical equations (10.9)-(10.10). That is, in view of the time independent complex canonical transformation (10.6), they properly represent the classical linearly damped harmonic oscillator, (3.1) or (6.11}-(6.12), in real space. 10.2. Complex phase space dynamics 10.2.1. The HamiltonJacobi equation So far certain elements have been given of the complex Hamilton-Jacobi formalism associated with Hasse’s classical dynamics. Let us now recall that Hasse’s Hamiltonian, (10.7), (10.1) or (9.12), has been derived in section 9 in the weak friction limit from the complex phase space quantum mechanical master equation (9.1), or (7.29). It will be interesting to investigate the underlying classical mechanics of the latter ina little more detail. Thereto we return to the beginning of section 7, or to Bopp's modified theory in the second part of section 6, From either (7.7)-(7.8) or (6.45)-(6.46) we infer the so-called improper or quasi-Hamilton equations [59, 195, 196, 258, 288], which are repeated here in terms of the rescaled classical variable z = aVA: (10,33) (10.34) As before, 7, 2° can formally be considered as a conjugate momentum. Of particular importance, “Note the iference Between By = Z(t. and B,(0) & (2.0 although it tums ou tobe of no importance inthe present eae« 1H Dekker Cusical and quanto mechanics ofthe dumped harmon eave ‘once again. will be to find the generator of the canonical transformation that is equivalent to solving (10,33). Suppose therefore, as in (10.22). that we are given a function /(2.). with the property” izt=F,. (40.35) According to this relation one then has available a complex function =*= z°(:.t). which in principle suffices to obtain functions 2(¢) and 2*(t). The total differential of z*, considered first as a function of z and and using the first equation from (10.33) and also (10.35), leads to 4: 21H dt~iF,). de (10.36) This is the immediate analogue of (10.24). However, unlike (10.25) one presently has. on the other hand, from the second equation of (10.33). that dz*=itt dr (10.37) Equating (10.36) and (10.37), and defining (24, 9. 195] H=HiT, (10338) H and T being real, one obtains HAH t (Pade =e. (10.39) in lieu of (10.26). Introducing the function '(z. 1) = H(z 2,0), the equation (101.39) reduces to (+P) = ATs (10.40) One must be aware of the fact, that on the Lh.s. of (10.40) the partial derivative with respect to z concerns a function of f and z only, but that on the r-h.s. it applies to a function of (possibly) 1, and both z and 2*. Keeping this in mind, (10.40) leads to the quasi-Hamilton-Jacobi equation H(2.+ Fz. = HZ), (0.41) where the dissipative “inhomogencity” J reads"” di (10.42) _genfrus We confine ouneles fr the momen! heing 103 one-dimemiona complex valued procen. The genealiation of he prevent Hamiton-eoh {ecm to higher mensional yteme is ponfvl, bth claweally and within 4 quantum methaneal content, a wil be further cai bythe simate rsa nd in ection Cleat, thn integration prscdre may gst in diiculy in higher dimensonat sem. See abo section TI concerning Kontn’s moan Ietional Shebinger equation [12,136 198, 199.23, 27,254)Hi. Dekker, Caccal and quantum mechanic of the damped harminc nave ” Since in the absence of damping #* = % =H is real (59]. so that I” vanishes identically in that case and (10.41) takes on its common structure. |. the “inhomogeneity” # 102.2. The effective Hamiltonian In the special case of the linearly damped oscillator, comparison of (10.34) and (10.38) shows that [= -Az"z, Hence, [., =-Az* and by (10.35) the integrand in (10.42) becomes iAY,. Therefore, $=-2AA(z.1), apart from an irrelevant arbitrary function of time. and the quasi-Hamilton-Jacobi equation (10.41) yields: DAS-~(AFiw)S, +S, (10.43) Its interesting to compare this upshot with (10.31). According to the theory of characteristics (sce e.g (78, 83, 113, 346, 491. 492)). the general solution of (10.43) is given by f= oF) (oa) where F(y) is an arbitrary function of y = z exp(i wt + At). Applying the transformation relation (100.35) to (10.44) at the initial time ¢ = 0 gives F,, =i z°(0), so that Fy) = i yz"(0)." Hence, nen, (10.45) which is identical to the generator E(z, 2) found in (10.32) for Hasse’s model, Finally, it is quite interesting to note that (10.43) suggests the possibility of defining a prop. effective Hamiltonian, say ¥, looking at it as if it were a genuine, “common” Hamilton-Jacobi equation, Then H=(w~id)z*2 + AP), (10.46) which generates the following proper canonical equations: B= 71K = -iwz- Az, (10.47) (10.48) H, sias* + Ast t2AS, Since, by (10.35), Z.=iz*(t) along the true trajectory. (10.48) is in fact identical to (10.10). So, (10.47)-(10.48) do correctly describe the damped oscillator and X can indeed be considered as a genuine classical Hamiltonian. It will be further exploited in the next section. 10.3. Summary In the first part of this section the classical analogue of Hasse's pure state Hamiltonian dynamics (ection 9) has been investigated. The classical dynamics is independent of the diffusion coefficients, * The intention constant can best elt 20 fo dest aflst the tranMormation, a6 encountre eae. For the same reson an ve constant doesnot appear in (10-4). Remember that (103) nique, sae for an arity ative fanstion of ime= 1H. Dekker, Casica and quantum mechenic ofthe damped harmonic oxcilaee Hasse’s Hamiltonian has been rewritten in terms of Dekker's complex canonical variables. The generator of that particular time-dependent canonical transformation that solves the dynamics (and which need not leave the Poisson brackets invariant), has then been shown to obey a complex Hamilton-Jacobi equation. In the second part of this section similar techniques have been applied to the classical dynamics underlying both Bopp's modified theory (section 6) and the complex phase space quantization of section 7. A complex quasi-Hamilton-Jacobi equation has been derived. from which a new proper Hamiltonian could be identified. This Hamiltonian will be discussed further in section 11. LL, Kestin’s nontinear Schrédinger equation ILI, Classical mechanics In the previous section the ideas of a complex Hamilton-Jacobi formalism have been applied to both the Hasse and the modified-Bopp-Dekker dynamics for the damped oscillator. From the latter a novel proper classical Hamiltonian X of a peculiar nature emerged. In principle, the availability of a true Hamiltonian offers new vista's on the application of conventional quantization procedures. In view of the unconventional nature” of X, however, it seems wise to at least transform back to the real classical coordinate and momentum, x and p. Consult (10.6), (6.42}-(6.43). oF the first columns of (6.3)-(6.4). 1 1 : Fyre ios, Veg b tation ay into (10,46), readity yields H = (Li Aeo)(p? + ape +222) + 2A H(e(p.2).0) ) Anticipating the real space Hamilton-Jacobi theory, a function $(x, ¢) can be introduced, with the usual property PaSa. (13) which is the real mechanical analogue of (10.22) or (10.35). Using (11.3) we define the function (x. 1) according to H2p.2).1)= AS. + (Ato) = L(x), (ita) and consider 71, = S24 (11.3) one then calculates iz*z,., In the latter we have invoked (10.35). With the aid of (11.1) and gy. -[s oe +35 oe 2ape + a)| : (ts) Foe inane, it ie by no means eae what the quantum analogue ofthe complet generator tel should be1H, Detter, Clasicl and quant mechanics ofthe damped harmon oiay ” Hence, Help. 2).1)= Sx.) Spx + 35 (p?+ apr + Wx), (6) plus an arbitrary function of time that may be set equal to zero by the same token as usual. i.e. dynamical invariance. Inserting (11.6) into (11.2). one obtains the upshot X= pt eras. ay If, considering again p and x as independent conjugate variables, this result is used as the Hamiltonian in the canonical equations of motion, it leads to (U8) Mx - DS, (tL.9) Because, in view of (11.3). S.z = p(#) along the true trajectory. (I1.8)(11.9) are identical to (10.4), and hence. indeed properly represent the original Newtonian damped oscillator (3.1), The real classical Hamilton-Jacobi equation Hx S.)+ S.=0, (tio) associated with (11.7), becomes AS +P +YS.F+ 5,20 (uy In this explicit form, (11.11) has been proposed earlier by Razavy [190, 226]. It is also mentioned in [43] within the framework of extended fluid dynamics, 11.2. Quantum mechanics 11.2.1. The variational principle ‘The Hamilton-Jacobi equation (11.11) appears to be amenable to quantization following a method originally due to Schrddinger {225]. In fact, the method is still usually cited in the textbooks on quantum mechanics as giving the exact solution to the (Rayleigh-Ritz [493-495]) variational problem (sec eg. (162-164, 300, 303, 314, 331, 496)). Its application to dissipative quantum physics stems from Razavy’s work [190, 226]. The method, however, is not free of subtleties.” Although we have not much to say concerning improvements at present, at least an attempt will be made at a viable formulation. In the 11s amusing to note Seheinge's footnote [225]: “Ex entgeht mi nicht da diese Formuterung nicht gad eindeatig it” Notice further. comparing [25] and (19,226. that Sebrdinger introduces a rea valed wave function @ (wbih cern allowed inthe ttonay tates oridert) according 0 $ In. However i his way, the wavefunction i connected withthe clascal rincpal fenton Sin a rather unnuat say (conc eg (72 13, 10,13, 16, 201-231, SD, HR, 31 314,417 97-80). On the other hand, Raraylalowe the common tehniqu of imeoucing 2 complex valed wavefunction o (which certainly i equred inthe dynamical cae) according to $= (hilind. $= AYN 6 or 'S= (HDI In(ld") The ambiguity that each of these expeecone mute iaeoduced a its “apyeoriate™ place the caial oul. Un fe. the two diferent competion of the wavefunction with thecal picipal function hay seem compatible, Final. ote that hese problems are baal elated tothe undarped tem, ether han 10 the dsipaion, See futher the present tet*” HE Dekker, Clarucal and quantum mechanics of the dumped harmon exile first place it will be shown that the dynamical variational principle basically is a Lagrangian recipe. It reduces to the usual variation of the Hamiltonian (the Rayleigh-Ritz method for obtaining the eigenvalues and eigenfunctions) only at stationary states. Secondly. itis noted that in the classical theory there exists apparently only one rea! field S(x, ¢). The subsequent introduction of a complex field yx.) leading to a two dimensional variational problem (ie. with 64 and 64° as independent alterations). will be given a little more attention than usual (sce e.g. (190, 226]). Actually starting with the latter aspect, let us return to the Hamiltonian (x. S..) that can be read from (11.10) and (11-11). It is H = YS. F + 02x? + 2AaS+ g(t). (U1) where we have once more explicitly accounted for the invariance of the classical dynamics under the addition of an arbitrary time dependent function g(t) to the Hamiltonian.” Compare e.g. with (10.7) Next, a complex function is introduced according to OC, thx0) (i exo f S(x th]. (uty where both $ and the amplitude A are taken to be real. In the definition (11.13) it has been carefully ‘explicited that the classical princip.l function, besides depending on x and ¢, also contains an integration constant. For convenience, this parameter has been identified with the initial value x(0)= x», For a classical, deterministic process the actual value of x isa sure quantity. ie. itis specified with probability one. However, in view of the probabilistic nature of quantum mechanics. it will be important to recognize Xs in the wave function (11.13) as a basically free parameter. Actually, (+. tts) is not yet a true wave function. For instance, it docs not conserve total probability in the usual sense. Rather, it representsa quantum meckanical propagator (Green's function, or transition amplitude).“” See ¢..{70,72, 116, 228, 231, 387, 417, 500, S01, 54]. In the form (11.13) itis known to be exact for the undamped (possibly driven) harmonic oscillator”** [72]. From (11.13) one has S= (hI2i) In(yid*) (uta) ‘Schrédinger’s quantum mechani Hamiltonian as the functional I postulate now amounts to defining the expectation value of the k= [orxudr, (1s) so that |y/? attaches the meaning of a (relative [72]) probability density, and considering K(p*, @, ) as the This notion takes he place hereof the autonay imposed noematation condtion i [180,225] Actual, tht subarycoeition would lead to nonsensical revue the cve ofthe emia wave function (1113, which, on the other hand, must be ineodued nore to obtain seliconsstnt variational procedure leading tothe andar nese Schrager equation i 4, See als [02-504] andthe previous footnote "Foe the nonnative linear (in #)Schringer equation the important sperption principle ols, a that general wavefunction can be obtained a a linear weighted combination of propagators with diferent xa i. Wt) fost) (49) dro, Usortuntely, the superposition principle is arated fr the nonlinear fictional Sehangr equations. including Kosi’ See futher on oe# The exprenson (113) 8 aho the fint step inthe sandaed WKU-anajss. Recall thatthe WKBLeigenales are exact (in the fs ode analysis for the harmonic oir. Se eg. 162-164, 30, 0, 08, 313, 314, $81,496, 98, 99, 06-516], Further. (1.13) sao exact within the Caldola-Kanai moet ofthe damped oxo (ection 4).1H Dekker, Clase and quanta mechanics of the damped harmon exile o canonical generator of quantum dynamics. Notice. that (11.15) involves both S and A, Because there are no additional constraints in the present formulation, these functions can indeed be altered dependently. Hence, alternatively. we may consider 8 and yv* as independent variations in the following. Inserting (11.12) with (11.14) into (11.15) straightforwardly leads to Him? [ (U1) (ia? 205d. + OIE dx +80? [Pure de wits f wv incaia*) de + eto [ uted (i.t6) The combination of the first and last term within the square (kinetic) brackets is easily further evaluated using the identity (WINGY + (WMO AY = -2 Td + (WD.aLl(Y DLs quay) Evidently, in view of (11.13), the second term on the rhs. of (11.17) is definitely zero, Therefore. (11.16) reduces to Ke i [Pu de + WERE i ha In(yid*)+ @)] de (118) With A =0 this Hamiltonian is well-known in the standard theory of “second quantization” (s 22, 300, 331). By means of one partial integration in the kinetic term, and putting A= convenience, (11,18) can be written as Ko= [ urGp'+ Viva, (iy) with p=-ifd, and V={0%x°, The Hamiltonian Ky is immediately recognized as the harmonic potential c-number analogue of the atomic reservoir Hamiltonian (8.3). that has been employed in the quantum optics model of section 8. Following the usual procedure, the Hamiltonian (11.18) can be obtained from the Lagrangian L ». (11.20) Wf atvedr- Ker, It is instructive to compare the functional (11.20) with the function (10.8). Presently, the canonical ‘momentum (x,t) conjugate to the field g(x.) is 7=ihY*. The Lagrange variational principle 8 J Ldt=0, taken over from classical mechanics, now becomes. 8 [itura,~Wus.- vtud%e?~ i Aa in(ole*)+ o)]ax dr =0, ay which yields the Euler-Lagrange equation (see e.g. (1, 113, 300, 331]) This procedure is not unig. It ams, however as nondisptive Sebridinger equation aca in and contiiag only Jo 1H. Delier Casal and quantum mechan of he amped harmon ole Bhd = MP iae + PG MAG In (HIG) + (UHA + GW ai in consequence of a variation 64° #0, and the complex conjugate equation by the independent variation 89 0. 11.2.2. The Schrédinger equation Clearly, in absence of dissipation (11.22) reduces to the usual Schrddinger equation, which is linear in . Hence. in that case more general solutions than (11.13), which forms an essential element in the derivation of (11.22), can be readily constructed by superposition of propagators with different initial states. See also again the footnote following (11.13). Unfortunately, as will be obvious from (11.22 superposition principle ceases to be valid as soon as A# 0. The bold leap of the present theory is the postulate that, nevertheless, (11.22) does have physical significance for solutions other than (11.13). even if A¥ 0." Of course, of particular interest are those solutions that can be normalized in the conventional way, thus providing a proper probability density p(x, ) = |¥[*. From (11.22) and its complex conjugate ‘one finds pitie=2(a4zIme)p, (11.23) where j(x, 1) represents the usual current density J ARNO. ~ O50). (2a) Compare this e.g. with (4.21)-(4.22). Clearly, total probability will be conserved in the course of time if Img = ~hA, The real part of (0) produces an irrelevant scalar phase shift in the wavefunction, as usual, and can be set equal to zero. The Schriidinger equation (11.22) can now finally be written as ids = (Hot AW, (11.25) Hy again representing the free oscillator (2.8), and where We Ein(ehie*) ~ (In(oie*))) (11.26) As before (see e.g. (2.9) for the Sissmann-Hasse~Albrecht models, and (9.8) for Hasse’s pure state model), we have added in (11.26) a non-observable time dependent function in order to let (Wx) The result (11.25)-(11.26) represents Kostin’s Schridinger-Langevin”* equation [198, 199]. It has subsequently been discussed within the frameworks of the fluid dynamical interpretation of the “ove hs may rete he val f (1.2) 1th mak tion fi. otf {1.28)U0 29) repent the T= ee, A ated keer oe cou! ran sha pote Va = 0 which due 1 4 eis thermal reser. Abhough thi procedure ite cn the Srinin-Lagcin eaten (1A, and seston 8) othe Cakota-Kana\ model (econ 2) the preset version smote saying. Name the pure asc Kosn-Langevin nae source) need et feveae the erp Hutton lhe gpd quantum mechanical xan Tat here (> FEU) = 2Dx 6) th De = 2ANAD in ew ot D=24(Np + See eg (19+-(.1) and (5). Keni’ onal papers i inde V's impiatons hae been urher Ssused by Meser taH. Debder, Casicl and quantum mechanics ofthe damped harmonic exter 2 Schrédinger equation [201-204], stochastic quantization [205-213] and Hamilton-Jacobi. theory [43, 190, 226). See further (129. 136, 156, 247-250] and the Historical survey. ‘The Kostin model (11.25)-(11.26) has the remarkable property that every stationary solution of the free oscillator also solves the damped problem.” In general there is a regression towards these stationary states (203, 207, S17]. according to*” d= dntydenp [FO S09], 27) where y = x ~ x(t), and where x(1) represents the classical trajectory of the damped oscillator. i.e. (2.5) or (5.41).°** Further, the ¢, represent the undamped stationary state eigenfunctions: aly) = He exp[-ilr + M0 y7/2h] Hay OIA), (11.28) with 4. = (O/rh)“@'n!)", Finally, So = $.(x(0), tt) in (11.27) stands for the free oscillator action calculated along the dissipative trajectory, ie. Si : Lolx(). 4€¢) de, a2») 20-70, (1.30) 11.2.3. Wave packet solutions ‘At the ground state n=0, the wave function (11.27) is a Gaussian. Compare with section 2, especially (2.11) and (2.19), In that ease one obtains, with (x) = x(0) lwP = ro) exp [- “ (3h) which compares with the wave packet (2.19) of the Siissmann—Hasse~Albrecht models at the stationary width 20,, = w7 = A/A, ie. for ¢=0 (Albrecht’s choice; see (2.12)). However, as has been shown by Remaud and Hernandez [136]. the stability properties of the Kostin fluctuations are essentially different from the Miinchen-models. Computing the first moments directly from the nonlinear Schrddinger equation (11.25)-(11.26), one finds (x)’ = (p) and, using p = ~ifd, (py = ~AP(x) +i AAU G")).«) = ~2%(x) ~ 2A(p) , (11.32) invoking the usual definition of expectation values, as contained for example in (11.15). Simitarly calculating the equations of motion for the variances, the results for dp. and dux emerge q * Thisprperty nese the andom pct Vain pesca at over temperatute ee peso Gt, Comat etn 2a, 109 (oe tuner omen ‘The won (1127 erst rom hat sen in [286,207], ut arse (ean 3 n= yh pest inf, Ln] Te iersnceie pss face on et in (SA Bero Hi. Dekker, Carica! and quanto mechonic ofthe damped harmon oxcllaur straightforwardly and appear to be identical to (3.56) and (3.58) of the Bateman-Feshbach~Tikochinsky theory, (4.28) and (4.30) of the Caldirola-Kanai model. (5.70) and (5.72) of the Kanai-Svin‘in treatment; and to (7.43) and (7.43), or to (9.23) and (9.25), if in the latter two cases one sets the diffusion coefficients Dj. + Diy = Dex = 0. The equation for dp» is more involved and. as will be shown, leads to a somewhat unusual result. Obviously, (7) = A px + xp)— Alp? In(wid*))). (1133) Writing the dissipative term on the r.h.s. explicitly in the coordinate representation, one ob s (py = Apr + xp)+ an fore == wed) dr 2a foo ar +an? Jlorronu+ worvery dr. (13a) By partial integration the first dissipative integral in (11.34) is noted to vanish, while the second leads to ~2A(p?), Moreover, applying (1.17) to the last integral on the r.h.s, of (11.34) leads to 9" = -A%px + xp)~ Haig) + aN? f (UP) (USIO" + bald) dx. (1135) Substituting now the general form (11.31) for the det integral in (11.35), gives FF of a Gaussian wave packet into the OY = =O + 3p) ~ AAUP) = OM on) f = EDM DP «de (11,36) Either performing a partial integration, or once more using (11.31) and the very definition of o., yields the final result. In conclusion, the Gaussian variances for the Kostin mode! obey the following set of equations: Gp * 2A ps + Op ~ Poe, (1137) = Aggy ~ 20 ap + HM oan (11.38) (11.39) This result agrees with (136). The stationary solution, contained in fact in the state y= Yo of (11.27). is easily found: op.(#)=0, opp(2)= Qou(#) and onp() ou() = h7/4. That is, there is minimum un- certainty with o..(%)= A/2 and opp(%) = hAI2, representing indeed the free oscillator ground state. Notice that, due to (11.38), the dynamics of the quantum mechanical fluctuations for the Gaussian Kostin-oscillator can not be brought into the form of any of the other models treated in the previous sections, Actually, the general dynamical solution of (11.37)-(11.39) is not known in closed form [136] However, if ae 7A the Kostin equations (11.37)-(11.39) reduce to (3.56)-(.58), so that for largeHE Dekker, Gasca and quantum mechanics ofthe damped harmon oxo 6 coordinate spread the solutions are given by functions of the type (3.59) and (3.60).* What will happen if the variances come close to the oscillator ground state values. is clarified by means of a linear stability analysis of (11.37)-(11.39). Setting a(t) = oe (*)[1 + €(0)]. (11.38) can be linearized to = dag ~ 20 ope — AAD oe, + SHAD. (140) Note in passing. that (11.40) cannot be obtained from a diffusion equation with constant (i.e. x-independent) dif_usion coefficient Djp. The eigenvalues of (11.37), (11.39) and (11.40) are 4A and ~A=2i(Q?— A7/4)"2, Hence, the Kostin-oscllator widths (ie. for instance, the energy) are asymptotic- ally stable about the ground state. Compare this with the remarks below (2.21). concerning the Miinchen-modebs Let us finally note, more or less for the sake of completeness, that because (j) the Kostin-oscillator is described by means of a Schrodinger wave function, because (i) the canonical momentum p = ~i hd, is identical to the mechanical momentum, and because (ii) a Gaussian wave packet solution exists, one may conclude that the “normal” uncertainty relation (2.18) can be applied to (11.37)-(11.39). Hence, as for the Sissmann-Hasse-Albrecht oscillators, one can find a second order equation for a. separately”” Introducing once more the width w, according to w? = 2a... one obtains We + 2Aw, + Dw, = wt (at) Compare this with (2.20), and notice the presently occurring damping term on the Lh.s. of (11.41). which explains the gross differences between the Kostin and the original Munchen-models. Analysis of (11.41) completely confirms the earlier conclusions concerning the stability of the Kostin-width about its ground state value w.(#) = (R/M2)", as it should [136] 113, Summary The classical Hamiltonian, obtained in section 10) on the basis of the complex quasi-Hamilton—Jacobi formalism associated with the modified Bopp-Dekker dynamics (sections 6 and 7), has been written in terms of the real space canonical variables. It has subsequently been quantized using an adapted dynamical version of the original Schridinger-Razavy theory. The result is Kostin’s nonlinear frictional Schrddinger equation. It violates the superposition principle and is difficult to generalize to higher dimensional systems, Its one-dimensional stationary state and Gaussian wave packet solutions have been discussed in some detail, The latter always deeays into the free oscillator ground state. The Kostin damped oscillator respects Heisenberg’s uncertainty principle. 12, Summary and final remarks jon in classical and quantum mechanics has been discussed from different point of views. both microscopically and phenomenologically. In fig. 12.1 the general interrelations between the various kana rewumean "| sense cron | Kostas wna] theory i quantum oohks oszilator i Hasse’: Munchen Messer's Sun'in's theary rogels treatment treatment Fig. 123, Block diagram showing specie relations Between varius theres ofthe damped harmonic oxilar.H. Dekker, Clawsiea and quantum mechan ofthe damped harmonic ose ” dynamics. From the quantal point of view this is to say that they all satisfy Ehrenfest's principle, However, the actual quantum mechanics of the various models reveals a considerable variety in fluctuation behaviour. Closer inspection of the models further shows that none of them. neither the microscopic nor the mesoscopic ones, are completely satisfactory in all respects. For instance, the Bateman-Feshbach~ Tikochinsky theory is at variance with the uncertainty principle. even in the limit of vanishing dissipation: the transmission line model involves infinit in's classical reservoir requires quantum mechanical properties: Bopp’s modelling is. in fact. in line with Ehrenfest’s principle only for certain special. nearly classical initial conditions: for arbitrary nondissipative Hamiltonians. Dekker's theory is ‘not always exactly equivalent to the usual Schrddinger description: the evaluation of the quantum optics oscillator involves quite a number of delicate steps: Hasse’s pure state representation condition is usually not satisfied in the dynamical state: and Kostin’s Schrédinger equation can hardly be generalized to more than one single oscillator. This list should suffice here. For further comments the reader is referred to the pertinent sections. Let us in the end again remember that dissipation basically arises from microscopic time-reversible interactions, and that it is observed essentially looking at a subsystem of the universe. Such problems are intrinsically difficult, although the observed subsystem dynamics is often relatively simple. In conclusion, there can be no doubt that dissipative phenomena ~ being in their modelling often at the borderline between miicroscopy and macroscopy-will continue to be a challenging subject (431. $30-S63}.* Acknowledgements I would like to express my gratitude to Prof. D. ter Haar (Oxford) for his suggestion to extend the original draft of this article to its present scope. It is also a pleasure to thank Dr. G. Denene (Leuven), Dr. R.W. Hasse (Minchen/Garching), Dr. B. Yurke (Cornell) and Prof. K.W.H. Stevens (Nottingham) for a critical examination of the handwritten text, and for their stimulating response. Prof, N.G. van Kampen (Utrecht) is gratefully acknowledged for his alert comments on the typewritten version. 1 am further indebted to Dr. E.S. Hernandez (Buenos Aires). Dr. M. Ichiyanagi (Osaka). Dr. M. Razavy (Edmonton), Dr. B. Remaud (Nantes), Dr. BS. Skagerstam (Gateborg) and Prof. Zhang Yung-Lun (Shanxi) for useful and encouraging correspondence. In addition, I wish to acknowledge the skill of Monique van Stuyvenberg in typing the manuscript. References** FIRST CITED IN SECTION 1 Ln] H, Gotdcin, Casual Mechanics (Adon. Wevey, Reading, 190) [2] LD. Landau and EM. Lishis, Mechanics (Pergamon, Oxford, 140) [3] RA. Coulomb, Mem. de 'Acad. Roy. Si, 10 (188) 16 *Is] PAM. Dita, Poe: Roya Soe, London AUST (1938) 188 Ref (530) seems tobe the east eeference tothe problem of quantization of a damped sytem. Ref. [3] prese quantum mechanical operator equation for radiating eccron [S32]. See a (S33. ** Relerenceslbeid with an aera expecilytelevan the damped ciao models cued inthe tou Key ppershave heen aed 2 second ars, The remaining elerences ate ofa more general malar fit derivation ofthe100 Hi. Dekker, Classical and quanto mechanics ofthe damped hamonicescilator [5] LD. Landau and EM. Lithit, Pid Mechancs(erramon, Oxford, 1959) [6] BG. Levieh, Theoretical Physics Vol. 4 (North-Holland, Amsterdam 1973), (7) C. Lanczos, The Variational Principles of Mechasis Univ, Press, Toronto 18970) ia] P Hava, Nuovo Cim. Suppl. (1957) 8, *19] LA. Pas, Analytical Dynamics Heinemana, London, 195) {10} Ray, Arm, J. Phys 47 (1979) 66, [U1] H. Bateman, Phys. Rev 38 (1931) 81, 112] W. Brin, Phy. Rev. 77 (1950) 396, 3] P.Caldrola, Naowo Cim. 18 (1981) 393. [13] E Kana, Prog. Theor. Pays. 3 (198) 0, 15] M. Raray, Can. J. Phys. $0 (1972) 2037 6S. Nakajima Proge. Theor. Phys. 201958 988. 1] RW, Zang. J. Chem. Phys. 3 (1960) 138, tus) RW. Zanzi. Physica 30 (194) 110, *{09} F. Hane, Spriager Tacs in Modern Physics 6 (1973 98. SLR] He, Dekker, Ope, Comm. 10 (1974) 14 “{2t] H Haken, Revs. Mod. Phys 47 (1973) 67 “[22] H. Dekker, Physica BC (1975) 13,193. “[2] E.B. Daves. Quantum Theory of Open Systems (Academic, New York, 1976) *[2s] H. Deter, Phys Rev. AUB (1977) 2136. “1S 0. Penrose, Reps. Proge. Phys. 42 (1979) 1937, [26] HL Nyauit, Phys. Rev 32 (128) 10 [21] A. van der Ziel Sod State Physieal Electronics (MacMian, Landon, 1958) fox] LD. Landau and EM. Lisi Stata! Phys (Pergamon, London, 1989) [29] A. Papoatis, Probability, Random Variable ad Stochastic Process (McGrail, New York, 1988) [30] BG. Levick, Theoretical Phyvies Vol. 2 (North-Holland, Amsterdam, 1971), [51] Der Haar, Revs. Mod Phy. 27 (1988) 2) [h2] NG. van Kamen in: Fundamental Proms in Sttsial Mechanic, ef, £.G.D. Cohen (Noh Hind, Amsterdam, 1 [33] M. Razavy, Physica HA (1995) 39, [Bs] NG. van Karpen, Adv. Chem. Phys 4 (176) 2 [D8] Haken (ed Syeeracics: Conperatve Phenimena in Mohicomponent Systems (Teubner, Statgat, 173). [De] H. Haken (ed), Cooperative Elects: Progress in Synergtcs (North-Holland, Amsterdam, 173) [07] Haken, Syargeizs (Springer, Bed 177). [38] G.W. Ford, M Kae and P. Mazur, J. Math, Phy. 6 (15) 08, [9] B. Yarke and O. Yurke, Quantization of a Damped Harmonic Oxia, preprint Corel Univ no. 424) (Apri 19K) ‘lal VF. Weiskop and EP. Wigner, 2. Phys 68 (19%) 54 ‘[st] 13. Loder, Dissipative Motion, Thess, Utect (Rijabuzen Report 95,1976) *[32] K.Atbrecht, Zit Berindung vom dissipative nicht near Schrdinget-Gleichungen(Vortrag Garchng/Minchen, Aug. 197. ‘lss] W. Stocker and K, Albrecht. Ane. Phys 17 (109) 6 [sd] A. Einstein, Ann Phys 17 (195) 509 [45] GE. Unienbeck and LS, Orsein, Phys. Rew. 36 (1930) 823, [ss] S. Chandrsetnar, Revs. Mod. Phys. 15 (18) 1 [a0] MC. Wang and G.E, Unlenbeck, Rev. Mod. Phys. 17 (1948) 323 [ai] L. Onsager and S. Maclup, Phys. Rev. 911953) 1515 S12 [49] R. Kubo, in: Lectures in Theoretical Phys tthe Unversity of Colorado 1988 (Gordon and Breach, New York, 1950). “50] LR. Senitty Phys, Rev, 119 (1960) 670 [si] NG. Van Kampen, in: Faction Phenomena in Solids, ed. RE. Burgess (Academic, New York, 14), [52] L Tia, Generated Thermodynamics (MIT, Cambri. 1956) “{SS} Ferma, Physica 32 (1966) 27. ‘[St] M. Lax in: 1946 Brandeis Universiy Summer Insttate (Gordon and Beach, New York, 168), {5 G.S. Agarwal Pays Rew, Ad (197) 79. [se] Graham. Springer Tats in Modern Physis 6 (1973) “{57] W.H. Louse, Quantum Statistical Properties of Raation (Wiley. New York, 1973. [Ss] M, Sargent Il, M.0. Seuty and W.E, Lamb Jt Laser Physics (Ndsion Wesley, Reading, 1973). [9] H. Deter, Z. Phys. B21 (91S) 2. [eo] R Graham, in: Factaations, fnuabitis and Phase Tranion, eT, Rie (Plenum. New York, 1975) “{6l| G. Linabd. Rept. Math. Pays. 10 (1976) 39. [eo] H. Detter, Physica RSA (1976) 38, “1S E. Braun and S-V. Govoy, Physica A (1977) 397,H. Dekker, Casucal and quantum mechanics of the damped harmon oscilaor 101 {61} 191 Rabi and D. Ja, J Pos. ALB (4 LIS, {65|P Langevin, CR. Acad Sei 146 (918) 5. {06 RL Staton, Theory of Random None Vl (Garon and Beach, New York. 195) (67 R. Graham, 2 Pys. B26 (97) 97 (|S. Wiener: Acta Mat, $5 (3) 117. {| S. Wiener, Nonaca Problems in Random Theory (MIT. Cambri, 1958) {70| LM Gelfand and AM. Yap, J. Math. Phys. 1 (0 {711 AT. Bharcha Red Element ofthe Theory of Marko Pocees and thie Appication (MeGrawHil. New York. 10). {72} RP. Feynman and AR. Hibs, Quantam Mechanics and Pah Interas (McGraw-Hill New York 155), [73] W. eles, Ae Inteducton to Pobabiliy Thor and ts Applicaton (Wiley, New Yor, 19807) {P| RA. Sratonovch, Sel. Tras. Mat, Statist. Prob, 10 (1971) 2. 115] FW, Wiel Paya 37 (1987 10, {Tl H. Dekker Phyaca 103 (6) SH, {7 Lon Rayleigh, Pi Mog 3211) 424 {P| Frank and Rv Mies, Di Diferetil und Intgrsichvogen det Mechanic und Physik (Viewe. Braschi, 10835 Rsenber New York. 180 (79) HA. Kramers. Physica 7 (104 284 {11 Gitman and A. Stroh, Stchotic DiferentlEqsations Sprig. Bin, 7 [si] L Amol. Stochasiche Diferestiscangen(Oleshoug. Miachen 197) [ke] NG. van Kampen, Phys. Reports 2C (1078) 7. [Ri] LN. Sneddon Elements of Paral Diferental Equations (McGraw Hil Toyk, 195) [sd] DV. Wier, The Heat Equation (Academe, New Yor. 175) “(85 HB, Caen id 13 Welion, Pays. Rev 831951) sf] R Kubo J Phys. Soe. pm. 12 (1957) 57 SU] UR Senihy Pye Rew 111 (1989) % sf] UR Senioky. hye Rev. 249) 16] 3 Sewing. Math Phys. 2 (617: ak 29) *f] Ran, C.Schmal and W. Welch, 2. Phy. 1 (ee) 37 sj] Kobo ns Proe It Cont. Sst Mech Kyu 198) Phe. Si Jp. Sup 26 (1% 1). “2] Haken Light and Mate, Enylopes of Phys, 8 Flag Senger erin 70) si] T. Arimio, ¥. Takako and F, Shit, Phys 100A (et) sis] Avan dee Zc. J Gong, WA. Kel Meta, RR Sch a MLW. Top, ysis 14 (7 [Ps] Beker, Tie der Elkin Vl 2 (Teshne, Rea, 1931) “[e] opp 2. Angew: Pay 1 (2) {91 Le fa Aacrbach and AM Cet, Phe. Let 474 (178) 18 [pil 1 Sickel, Fob d Phys 24 (94) 17. [DL Hi and JA. Whose, Phys Res. (1983) 12. [Wo] A. Boke and BR Motchon, in Nckes Stature (Renumin New York, 19) {00 M. Gide and G Teles, Phys, Rew 78 (198) Hi [W2] DHE Gros, Nac, Phys, A2H (175) 72 [Wa] We Nornberg, dour, Phys. 37 (976) CS [104] 1. Hotmann aod P.Semens, Nl Phys. A275 (1977) [Ws] HA. Wieden, 3 Phys SxS. Sap 8 (197) 7 [106] DHE. Grove and H. Katinonth, Phys Reports 4 (1974) 17, “{W"] RW. Hawe, Proc. Tops! Cot, on Large Ampltide Collie Naskar Motion, Keely Hangiy. dh A. Kis, J Nemeth an ZimSny (EPS, Badapest, 19M) Vol p22 *ta] RW. Hane, Nola Frition, Habitation (Uni, Minchen 1977) “{W] RW. Have Rpts. Pog. Phy. (197) 127 “[to] B. Yorke (Come Une). prt communication (40); se 3h [9). [U1 TJ. Rieger. DI. Selina JE. Merserent, Phys, Reve 86 (1972) 17, [012] K. Kawai MC. Vals and .D.Guntin, Phys Rev AT (1974385, secrION Maint releant w secon 3: {113} PM More and H. Fxthsch, Methods of Thcorticl Physics Vol 1 (MeGraw Hil, New York, 1983) Ss{11 F Bopp, Stier Bayer, AKad. Wiss, Math nature KL. (1W7) 67, 15] H. Feathach ad Y, Tikochinny, i: A Feuscht fr U1. Rabi, Trans New York Ac Se See. 2, Val. 36 (1977) 44 *ftte] Y. Tikochiesky. J. Math. Phys. 19 (1978 aKwm 1H Dekker Casical and quantum mechani of he damped harmonic oxi Mainly secton 4 *[117] P. Havas, Bull Am. Phys. So, (1950) 37. “{U18] EH Kerner. Bul Am. Phys. Sc. 21957) 383, [9] EH. Kerner Can. J. Phys 36 (1988) 371 *[0] K.9EHL Stevens, Proc Phys. Soe. London 721984) 1027 *SU21] KW HL Stevens and B. Josephson Je, Pr. Phys. Se. Londoa 741980) $6 [123] VW. Myers. Am J. Phy. 27 (1959 507 23] G. Valent. Rend. Lomb, Se ASS (941 286. 024) HLH. Denman, Am. J. Phys 84 (196) 117 +128] LA. Malka, VI. Manto and DLA. Trfocor, Phy. Rex, D2 (1970) 137 1125] G4 Papadopoulos. J. Phys: As (1973) 179 1127] LH. Buch ard HH. Denman, Ar. Phys, 42 (1978) Mi 1128] G4. Papadopoios. 3. Pye. A7 (107) 38, 129] RM. Hase, 1. Math, Phys. 16 (175) 205 *{130] G1. Papadopoulos. Phys, Rew. DIL (197) 240, *{131] ER. Sein, Teoe Mat Fi, 2 (1975) 107 *{133] VV Dado, LA. Malka and VI. Mano, J. Stat Pye 16 (177) 37 [33] VV, Dodonow and V.E Manco, Noowo Ci, BSS (V7) 25, [4] J Meer, Lett. Math, Phys 2 (197) 2 [8] AD Junausis. GN. Beodimas and S. Sued, Phys Lett 7A (197) 6 004] 8 Remaod and ES, Hemander J. Phys A (19D) 2013 [137] A. annus P.Fiipakis and Th, Flips, Physi HA (190) St 198] A. anne. V. Papatheou and K. Vahos, Phys. Left 77A (58D) 12, 109] P.Havas, Phys, Rev. #8 C1951) 234, 13} K-Husim, Prog Theor, Pays. 9 (1989) 38 sftsi] Cabos, Rend. ft. Lomb. Se. ABS (989) 4. *{122| 6. Vaan, Naovo Cin, 19 (961) 12, 1B. Caias, Nuovo Cm. (98) 172. *{044| PG. Soma, Encria Noel, {3 (188) 316, SIMS] HR, Lewis and W.G, Risefed. J. Math, Phys 10 (I 15K sftse] A. Hoss, ete, Noowo Can (1 181. [87] HLH, Dene and UH. Much J. Math, Phys. 14 (1923) 326, [89] LA. Malkin, VIL Man’ko and DA. Talon, J Math Phys 14 (173) 5%, {189} LH. Buch an HLH. Denman, Phys. Rev. DIL (28) 28, {USI} VV. Domo. LA, Makin and VIL Man‘, Int J. Theor. Phys. 14 (1975) 3. S{ISI] VV. Dosowow, LA. Makin and VIL Mae'to, Teor, Mat Fit 28 (1978) tot 1882] PL, Leseh J. Math, Phy. 19 (1974 6 *{1S3] LR. Buran, MR. Feit, EFiskor and A. Monier, Phys att, TSA (IFN) 1 *T1SH]F, Negro and A. Tartaglia, Phys Lew. 77A (180) | {RSS} HE, Marina, Phys, A (9 L288, 186} J Messer, Acta Phys. Aust, 50 (1979) 8, Mainly section 8: 1ST] KWH, Stevens, Proc. Phy, So. London 77 (61) 5, [05H] KWH. Stevens, Phys Let. 75 (10) 403, [159] A. Binsin, Phys. Ze. (1917) 12 [160] PAM. Dirac, The Piciles of Quantum Mechanics (Univ, Press, Oxford, 18). [U6] W. Hei, The Quantum Theory of Radiation (Univ. Press, Oxford, 16/58) [162] HLA. Kramers, Quantum Mechanics Dover, New York, 195761. [163] A. Messiah, Qvantom Mechanics (North-Holland, Amsterdam. 16170). [16s] E, Merbacher, Quantum Mechanics (Wie, (96170). [065] D. ter Haar, The OM Quantum Theory (Psegamon, Oxford, 1967) [164] 14, Sakari, Advanced Quantum Mechanics (Addie. Wesley, Reading. 9477). [N67] A. Pie, Rees. Mod. Phys. $1 (1979) AAS, [168] F.C, Hemmer, Dynamic and Stochastic Types of Mition in the Lineat Chai, Thess (Trondheim, 195. [148] M. Toda, J. Phys. So. Japan 13 (1954) 126, {H70] M. Toda, 3 Phys. Se. Japan 1 (1989) 72, [IM] RA. Rabin, Mah. Phys. 11960) 09H Detter, Claseul and quantum mechani of the damped harmonic exile 3 (17) P Maur and E, Montrol J Math. Phys. 1 (1960) 70, [13] RE. Torn, Phrsica 26 (190) 298, {14} ¥.Kogue. 3. Phys. Soe Japan 16 (1961) 1. [105] ¥.Kopore. 3. Phys Soe Japan 17 (1982) 36 {19 M. Toda and ¥.Kogur, Pog. Theor. Phys. Suppl. 29 (1912) 17 [17] P Matar ad E, Braun, Physic 3 (195) 1973, [178] J Meianer. ia: Poe Int. Syn, Stat. Mech. Aachen 1964 orth Holand, Amsterdam, 1955, U9] S Carwsrto, Phys. Rew. AIL (175) {37,1407 [180] UR. Senn, Teor. Math, Fi, 27 (1976) 270, *{tsl] IR. Brat and 5. Maa, J Math, Phys. 2 (1980) 2184, sft] H. Dekker. 2 Phys. B24 (1976 21 *tvi] H Dekker, 2. Phys. B2S (1976) 88, *fted] H. Dekker. Phys B26 (1977) 7. [145] R. Kubo, Lciare Notes in Physics 3 (Spinger, Belin, 1973) p74 [196] P-HsngeZ. Phys, BM (17) 7. {157] A. Kitahara, W. Horstheme aed R,Lelever, Phys, Let, 0A (1979) 37 [188] NG. von Kampen, Phys. Let 6A (190) 1 199] H.Hembole J. reine u anges. Math 10 (180) 137, [0] M. Razay, Z Phys, B26 (77) 21 [191] RM Santi, ade. 11 497) 23 *{192] R Cowley, Phye Ree, AD (1979) 290 *{103] R Cawley, J Math, Phys 21 (194) 239. 199] R. Cawley, Pye, Rev, D2 (10) A, {198} H. Dekker. Physi 9A 107) 31. 94] H_ Deke, Phys Let. 784 197) 15, S197] G Dedene, Physics INNA (IH 37. Mainty section 1 [19H] MD. Koni, Chem, Phys $7 (973) 360, =TW] M.D. Konia, 1 Stat Phy 12 (1979) HS, [200] M.O. Sealy 49d KG, Witney, in: Prores in Opis X, ed E, Wolf (NorthHolland, Amnctda, 1972 0 E, Madelung, Z. Phys (936) 12 2} D. Babin, Phys. Rev AS (19) 1, [203] KK. Kan and J. Grin, Phys. Lest, S08 (1974) 23, 204] JD. Immele, KK. Ka and J. Grif, Nol, Phys, A24 (1975) 7 25) B.Skagestam, Phys. Let S48 (1975) 21 2] K.Yawe. Phys. Len, 88 (1974) 29 [207] B Skagerstzm, J. Math. Phys. 14 (1977) 4, ‘fpis] K Yoswe, J Stat Phys (1977) 11. R05] K Yasue, J Math Ps. 19 (10) STEW] K Yaw, Ann. Phys. 118 (1978) 39 [It] NMashimot, Let Nuovo Ci, 26 (1979) SS, [13] NA. Lemos, Phys. Let. 1A (1) 2, [21S] nes. 2. Phys, 2 (1952 a, £215] E. Neon, Pays Rev, 15 (6) 10%, [216] E. Neon, Dynamical Theories of Brown Motion (Uni. Press, Princeton, 147), [20] Lge la Peta Averbach 1. Mah, Phys 1 (94) 12. [IN] J. Fronteav and &. Teter: Arenas, Nuovo Cim, 368 (197) [219] J. Fronteay and A. Teles Atenas Nuowo Cim 4B (74) ¥. [220] GC Girard, C. Omero, A. Rimini and T. Weber, Riv. Noowo Cim, (197) 1 [221] St Davida, Let, Math: Pye. ¥ (19%) 27, [222] M. Davidon, Let. Math Phys 3 (197) 367 [223] WR, Hamiton, On a General Method in Dynamics ld ewsey) IRM: Coste Papers Vol. 2 (Cambridge. 140) pr 103-20 [228] G4. Jacob. Crete’ Journal 27 (17) 9: Volevngen der Dynami (Bertin, 1), | E_Schdnger, Ann. Phys. 79 (1924) 3 [P38] M_Rasany, Can. Phys. $6 (190%) 3 1M, van Viok, Proc, Nat. Asad Se US 4 (924) 17%, [e2M] BS. Dewi, Reve Mo. Pye. 29 (1957) 37us 1H Dekker. Cusicl and quanta mechanic of the damped harmonic silator [25]. Swinger. Ovtotum Kinematics and Dynan (Benjamin NewYork. 170) [350] WHC Mier Ad Chem, Phys. 2 (1974) 6 [31] Deter in Functions Inegrtin: Theory ad Applications. ch J, Asti aed E Tiapegi enum. New Yok, 1) 9,3 {232] L Bokumann,Vorewnge Sher Gabor Leng, HOA. [20] §Kotbck. formation Theory and Sass Mey, Ne {23 EX. Jaynes, Py: Re. 161957) 6. [35] EX Saves Phys: Re. 108 (957) 7h {236] NG. van Kampen, Physica 25 (199) 13 [237] 1 Pigoine Pr. In. Symp. Teste 1968 VAEC. Vienna. 19) p31. [28] P. Gnsdort and L Pippin: Themodyeanie Theor of Score. Sait {239 M. Eigen, Die Natarwsercaten 5 (1971 8 [20] D siepian, The Development of Isormaton Theory (IEEE, New York, 17) {2 G. Linda, JS. Phy (72121. notion to Ouatatve Apes of Sat Phenomena (Gordon and Breach, New Yok, 174) [Ri] 8, Weber, Revs, Mod. Pays. $0 (1978) 22, [245] R. Landsver, Phys, Rev. AIB (1978) 55, [246] Rate J Phys. AD 1679) Lins. 247] 14. Grif and K.K. Kan, Boll Am. Phys, So, 19978) S27 *[248| 1H, Weiner aed RE, Forman, Phys. Rev. BIO (197s) 128. 209} 14. Griffin apd K.K. Kan, Revs, Mod, Phys 8 (197) 267 250) L Bialik Birla and J. Mycity. Ann. Phys 10 (1976) 6, *f28t] ®.Ehvenest Z. Pays. 45 (1927) 85 “252 H. Mod Prog. Theor. Phys. $3 (1978) 117 ‘Mainly conceming the SHA modes *[253] G. Siwmana (Garching Machen, Los Alamos Seminar (vapubished 173) "1254) K: Aree, Phys, Let $68 (1975) 127 *f285] RW. Hase, Comp, Phys. Comm. 11 (1976) 38 *f256] RW. Howse, J. Phys ATL (ITA) 235, “f2ST| RW. Haws, in: Nuclear Stvtue Stuy by meats of Nuclear Reactions, Proc. Summer School Miki, Poland (197); Nukleoika 25 (as 133. 289] H.Debler, Phys Le DA (19M 3, {259] 1. Mati, Theorie der Stbilit einer Bewegung (Oldenburg. Minchen, 1950) [240] J, LaSalle and, Letcher, Sait by Lyspunows Dirt Method with Applications (Academie, New York, 1961) [B61] Vi. Zubor, Methods of AM. Lyapunov and thei Applications (US. AEC, Te. 4139 196). [202] L Cesar, Asymptotic Behaviour and Stability Problems in Ordinary Diteentil Equations (Springer, Bet [263] N. Mics, Nonineae Onions (Vae Nostan, Pigeon, 1962) [260] TR. Fowler J. Math. Phys. 4 (1988) 589. [26s] NIN. Krasort, Stability of Moon Uni. Press, Stanford, 1983, [260] T Yoshizawa Siabity Theory by Lyapunen’s Second Method (Math So, Tokyo, 1944) [267] 11. Willems, Stability Theory of Dynamical Systems (Nelson, Landon, 1970, [248] G,Nicts Ads. Chem. Phys. 19 (1971) 29. [269] W. Mas, Ann. & Phys 32 (1975) 107 1962) Mainly econ 7 {270 D. ter Har, Rept, Prog. Phys. 24 (1961) 34, *f271] 6. Linabind, Comm. Mat, Phy. 48 (1976) 19 {272} GL Soe, Fors. Phys. 24 (176) *[a73] ST. Als and E, Pragoneth, J. Math, Pos. 1K (1977) 219. [204] V. Gorin, A Kouabovsi aed E.C.G. Sudarhon, J. Math Phys. 17 (1906) 82, [275] D. Marcuse, Engineering Quantum Electrodyeamies (Harcourt, Brace and Word, New York, 197), [276] H. Weyl, Grappentheore und Quantenmechanik(Hirel, Leirig. 192) tans: The Theory of Groups and Quartum Mechanics (Roberts, New Yor, 193} {277 E. Wigner, Phys.Rev 0 (193) 739, [27H]. Moyai Pro. Cambridge Pil Soe. 45 (188). {2 F. Suppes, Pi ei, 2 (1061) 37 {280] NAL Bales and GG. Zipfel Je. Ann. Phys. 77 (973) 1. [PBI] E.Chen, J Math. Phys 6 (178) 257Hi. Deter. Classical and quantum mechanics of the damped harmon oxi ms [20] 16. Kriger and A. Posty, Physica HSA (L07) [280] R. Haberkoen, Mo. Phys 32 0978) 191, [283] 16. Kruger and A. Poy, Pysca BTA (197) 132. {285] MCV. Bery, Pil. Trans, Royal Soe 247 (1977) 237 (286) HJ. Kors, J. Pys. AL (197) BIL [287] H.W. Lee and MO. Selly. J. Chem Phys. 73 (980) 2258 *fPa8] HL Dekker, Phys. Let. 6A (1980) 342 289] J. Bur, Thess and peeprit, Minchen (197, “fv P. Hoguesin, Hele. Ps. Acta St (978) 3h, ‘p91] NAL Balas, Physica 4A (1078) 18 P92] M ehiyanag. 2. Phys. BA (1980) 13, (293) E. Schrodinger, Natrwis. 14 (1925) 6 [299] PAM Die, Proc, Royal Soc. AUIS (1927) 243, [295] EH Keanad, Z Phys 44 (127) 2s, “296 E.Sebringer. Abhandlungen zur Welenmechanik (Bath Leip. 1928), “RIT LR Seiaky. Phys, Rev. 95 (1958) 94 *fP98| RA. Gauber Phys.Rev. 13 (1963) 2766, *[299| RJ. Glauber. i: Quantum Optics ad Electonics, Les Houches. ef C, DeW ‘0 Li. Seki, Quantum Mechasics (McGraw Hil, New York, 1888). ‘fB0I] JR. Klauder and E-CG. Sudarshan, Fundamental of Quantum Opis (Benjamin New York. 1958) ‘foma] RJ. Gauber. in: Laser Handbook, ed. FT. Arecthi and E.O, Ssbule Dubois (North-Holland, Amerdam, 1972) pb. [303] A'S, Davydow. Qsantim Mechanics (Pergamon, Oxford, 19457) *fs04] V. Casivll nd P- Sept Phys.Rev. Dis (1977) 1080 ‘305 V. Cavell and P Seplar, Phys.Rev. DN (1978) 1082, *[06] BS. Skapertam, Phys. Le, A (1978) 7, *f307] AD. annus, G.N. Brodimas and LC. Papaocas, Phys Lett T1A (1979) 3 sf] V.V. Dodonor and VI Mano, Phys Rew. AN (197) $3, “fun V.V. Dodo. EV. Kurmyshev and Vi. Mako, Phys. Let. 1A (IMD) 10 (Gordon and Breach, New York, 1965) p43 Maint section 8: S10] RW, Hase, Nuc Phys. AMIR 197) 4, ‘efstt] RW. Hae, Phys. Let AS (1579) 197 SECTION 3 [12] W. Pau, Genera Principles of Wave Mechanic, ia: Encylopedia of Phys, eS. lige (Springer. Berlin, 198. [S13] LD. Landaa apd EM. Lithitz, Quantum Mechanics (Pergamon, London, 1950), [B14] 8G. Levies, Theoretical Physics Vo. 3 (North-Holland, Amsterdam, 1973) [BIS] HBG. Casini, Poe. Royal Soc. Anwterdam 3 (931) 8 [B16] G.Racab, Rend. Linc: 8 (190) 108 [B17] BL van Wacrdea, Die Grupentheoretiche Methose in det Quartenmechanit(Edwarus, Ann Arbor, I) [B18] G, Raab, Group Theory aed Spectroscopy (Lecture Noes, Paceton, ISL/CERN prepriet 6:8 1) [BIB] LE Compbell, Proc. London Math, Soc. 2 (1H) 18 [520] HF Baker, Pro. London Math, So. 34192) 37. [B21] F Hausdort, Ber. Verh Socks. AL. Wis, Mah-Nature., Lipig 58 (194) 19 [B22] GLH. Weise and AA, Maradudin. J. Mah. Pays. 3 (982) 71 [823] ET. Arcee, E.Courtens R. Gilmore and H. Thomas, Phys. Rey. A6(I973) 211 [329] G. Racah, Phys. Rev. 61 (192) 16, [325] G. Racab, Phys. Rev, 62 (1982) 38 [526] G. Racah, Phys. Rev. 68 (183) 307 [B21] ARE, Edmonds, Angular Momentum in Quantum Mechanes (Univ. Pres, Princeton, 1957), [323] EP. Wigner, Group Theory (Academic, New York. 1959), [329] G.Lyubarski, The Application of Group Theory in Physics (Pergamon, Onord. 1. [330] DM Brink and GR. Satcher, Angular Momentum (Univ. Press, Oxford, 12) “{33I| HA. Bete, ltermedate Quantum Mechanics (Benjamin, New York, 1961) [B32] NG. van Kampen, Can. J. Phy. 39 (1961 51 [333] H.Dekler, Pays 103 (168) 38, [B34] AN. Kolmogorov and SV. Fomin, Introductory Real Analysis (Premize-Hall. London, 1970) p221% H. Dekker, Clscal and quantum mechanic ofthe damped amon exilatar secTion 4 195] 1D. Logan, Ivanant Variational Principles (Academe, New Yor. 1779. SECTION s [236] RR. Pari and SV. Lavande. Phys. Le. 62A (197) 13 [B97] RR Par and SV. Lavande. Ps Lett. A (178 [535] H Dekker and JPM de Vreede, Opin Com, 30 (1979) 13 [309] E,Schinger Ann d Phys 8 (948) 916 sts] V Doh, Sod Sate Comm. 11 (973) 1273, [Bat] R Kobo. K. Matsuo and Kaba, 5. Sot. Pye 9 (973 51, [2] R.Kebe. in (35) [5] LD. Landau and EM. Libhie, Theory of Elsciy(Ptgamon, London, 195) [B44] G.Kicho. Porzndorts Ana, Phys 72 (84) 98 [his] G. Kirt, Ano Phys. Chem, 72 (7 32 {bi6| PR Carabedian Parl Diferetal Equations (Wily, New York, 16) [7] H, Detker Opies Comm. 16 (1976) 12, [8] CV. Raman eta, Pro Touian Acad. Sc. AS (181 317,489 {599] W. Lederman, Pro, Royal So. AR? (194) 3. [8] 31 san Hemmer, Pays Reports 6 (9 38 [BSI] EL tac, Integration of Ordinary Diletta! Equation (Over and Boyd Edinburgh, 1982) [B82] GM. Morphy Ocdnary Dieremial Equations and Tee Soto (Van Nowe. Prieta, 1 [ssi] C Lancron. Dicoure on Fourier Series (Over and Boy, Edinburgh 18) *T88] UR. Sentry, Phys Rev 118 (109) 27 SECTION 04 [B85] R. Abram, Foundations of Mechanics Menamin, Reaing 7), [69] A. Weinstein, Ae Math 6 (971) 19 [357] W Konstant, Leste Notes Mth. 17) (Springer, Het, 170) [64] J. Soariay, Stuctte des Stems Dytnigues (Dunc, Bar, 197), [391 Vain, Col mology and Dillerentil Forms (Dekker, New York, 173) [ws] DJ Sms, Pave Cant, Pi, Si. 78 (1979) 4 [Bot] DL. Siw, Col. In. CN RS. (Pars 237 (1975) tn. {269] J, Sates, Reps Math, Phys. 7 (1075) MI [sa3] £4, Flaherty, Lecare Notes Phys. Springer, ein, 1976), [39] DL Sm and NIM. Wontbose, Lecture Nees Phys, 83 (Spgs, Beri, 197) [343] Vi Arnold, Mathematical Method of Clasical Mechanics (Speinger, New York, 1978) [4] 144 Rawal, Let Notes Mth. 67 (Springs, New York, 178) p11 [347] DL. Simms, Lecture Notes Mah, 67 (Springer, New York, 1978) p81, [B44] DJ Simm, Pro. Cont, Feyaman Integral (CNRS, Marci, 978), [349] J Sitges, Geet Quantvation and Quantum Mecham (Springs, New York, 1 [300] Hs. Groenewol, Physica 1 (1046) 8, [B70] KE Cah and RA. Glauber, Phy. Rev. £77 (10) 17,18. SECTION 92 [372] K Altech and RW. Have, private communication (197), [373] H.Fesbach, CE Porter and VF, Weis, Phys. Rev. 9 (1944) [374 PE. Hog, The Optical Model of Ele Seutening (Oxtor, New Yor, 1980, [375] WE, Lamb Je, Phys Rew 14 (1968) ALES, [375] AS. Davy, Naclar Tory (Prone, New York, 1988) [377] AL. etter and KM, Watson, Advances in Thvretcl Physics (Academic, New Yor, 1988, [379] TA. Get, Am. J Phys. 3 (1967 207 [379] HL Haken, Aw Sli Sule Phys, (Pergamon Vieweg. 17 p38 [bal] CF. Tsang, Phys. Serta 10A (1974 9, [p81] §D, Doyle, 15. Ext, WJ. Thompson and OL. Wenser,Am J. Phys 43 (1918) 677 {x2} BL Sia, Phys Repents 20C (1928)1H. Deter, Clasica and quanaum mechanic ofthe damped harmonic osclaor uw [580] K Sivek Wiest and J. Wieryski,NucL Phys. AZM (1976) 115, (S84) 15, Eek and WJ. Thompron, Am J. Phys. 45 (977 I [585] B Saha, Ps. Le, 71B (1977) 283, {386] T bamoto, §. Krewald an A. Fear. Phys. Ltt 95B (1960) 16 [381] RP. Feyoman,Satisical Mechanics (Benjamin. Reading. 1972) “298] H. Haken i: Laser Handbook, ed, FTA. Ateceh and E.O, Sehul-Dubois (North-Holland, Amsterdam, 197) *f] M.O, Scully and W-E. Lamb Je Phys. Rev, 159 (1967) 308 *f390] M.O, Sealy and W-E. Lamb J Phys. Rev. 165 (1968) 286 *fp91] MO. Seay and W-E. Lamb Je Phys Rev. 179 (196) 368 *f92] M. Suen Il and M.O. Seu in Laser Handbook, eds. FT. Arechi and EO. Schule Dubos (North-Holland, Amsterdam. 197) p48. [E91] 1. Oppenheim, KE, Schuler and G.H. Weis, Stochatie Process it Chemial Phyct: The Maser Equation (MIT-Press, Cambri a SECTION? {398] B. Rodos Phys, Rev, 3 (192) 812. [295] S. Weinberg. Gravitation and Cosmology (Wiley, New York, 1972, 96] R. Gram, Phys Let DA (198) 38, [97] H. Deter. Physica 4A (1974) ¥. {[D0R] H. Dekker and H. Horner, private communication (197) {099 P. Dita, The Development of Ousntum Theory JJ. Oppenheimer Memorial Prive Acceptance Speech (Gordon and Breach, New York, wn) {400] 5, Wieder, Foundation of Quantum Theory (Academic, New York. 173). [401] 1. Meta, Historical Development of Quantum They: the Rie of Matnx Mechanics (Wile. New York, *[w] H. Dekker, Ned. Tj, Nath. 21974) 20. sats} 8, Daves, Comm: Math. Phys. 22 (971) St ‘[utt} E10, Daves, Z, Wabreh, theoie vere, Geb, 23 (1972) 2 sus] A. Kowakorki, Bull Acad Pol Sei Set. Math, Ase. Phys. 20 (1072) 103 *i06| A: Keak. Rept. Math Phys. 3 (1072247 suo} J Mebra and ECG Surshan, Nu Cin UM (1972) 28, fut] E., Davies, Camm Math Phys. 33 (197) 17 a} A. Kovahork, Hail Asa. Ba Sei Sex Math, Ae Pye. 21 (1979 6, *fs10] LY. Pale Comm. Math. Phys. 38 (1974) 28 [sit] EA, Dies, Cm, Math Phys, 39 (1978) a *fs12] RS. garden and A. Konsskoski, Ann. Phys. 4 (1978) SL *[913] J. Messer and B. Baumgatine, Z Phys. 832 (978) 1h, sTsta} B. Booweaw and C. Barabe, J. Math Phys. 19 (1) 10 s{s1s] MV. Berry and NL. Balas J Phys. AL2 (1979) a3, *[slo] LL. McCauley JJ. Pays 12 (1979) 1 {s17] G. Rowen, Formulations of Clases and Quantom Dynamical Theor (Acai, New Yor, 1 ‘[sls] EM. Henley aed W. Then, Elementary Quantum Feld Theory (McGraw Hil, New York, 82) [319] J. Barden, LN Cooper and JR. Sevier, Phys. Rev, 1 (1987) 175, [220] PG, DeGennes, Soperconductvity of Metals and Alloys (Benjamin, New York, 194). {{21] NHC Match, WHC Young and. Samponthar, The Many Bay Pablem in Quantum Mecharies(Combvie, Unie. Pres, Lemon, 197). STE22] M. Lat, Reve, Mod, hye 3 (166) 39, SH *f223] M. Lax i: Pays of Quantam Electronica, ous, PL Kelly, B, Lax and PE, Tatmenwald (McCtaw-Hil, New York 1h. *[224] HL Dekher. Physics Laboratory TNO Seminar (published, June 1977 SECTION [225] RAK. Wangsaew and F Bbc, Phys. Res (1982) 72%, [126] W, Pau in: Fesecbsit zum 6, Geburtstag. Sommerfeld (iret, Legg, 1928p. 30 [s07] L van Hove, Physics 25 (1957) 44, [s2X] EW. Monto i: Fundamental Problems in Saiscal Mechanised. E.G.D. Caen (Nuth-Hlland, Amsterdam, 12) [k09] 1 Pigone and P_ Rebs, Physica 27 (196) 62, ‘fst. Welch and F Haake, Z, Phys IAS (1985) 3 ‘[s3l] R Bengura and M. Kae Pye. Rev Lett. 4 (8) 1 [282] TH. Mamas, Bat. Comm, Elect, 7 (4) 673 [s31] A.G. For and TL, el Sy, Techn 3,49 (1961) 48vow 1H. Deer, Cascal and quantum mechan ofthe damped harmon xiaor [834] GD. Boyd and JP. Gordon, Bel Set. Teche. J) (961) 9 {455] GD. Boyd and H. Kogelit. Bell Syst Techn J $1 (1962) 167 [456] H. Risk, Z. Pays. I (196 190 [157] L. Rowen: Later Handbook, ed. FT. Arecci and EO. Schule-Dubois (North-Holland, Amserdam, 19°) p. 11 [28] WE. Lamb Je, in: Proc. ft School of Physics “Eario Ferm, Course 31. ed. PA. Miles (Academic, New York. 1h p78. {489} FS. Bares (ed). Laser Theory QEEE Press New York, 1972), {440} M. GoeppertMayer. Ana. Pay 9 (1831) 273 [1] JM. Worbek, in: Liter Handbook, ed F.T, Areceh ad EO, Schule Dubois (Noeth-alland, Amsterdam, 172) p37. [2] CW. Lamberts and H. Dekber, Physics Laboratory TNO Report PHL 49 (1978) 7-16 [3] EA. Power and S.Zieau, Phil Trans, Royal Soe. London 251 (199) £27, [sts] RG. Wooley. Mot. Phys. 22 (1971 1013 [S45] 2 ried. Phys. Rew. AR (12) 2838 {se} KHL Yang. Ann Phy. 101 (1976) [847] F Basan. Forney and A. Quitropani, Phys.Rev, Lett 39 (972) 107, [S48] DH, Kobe, Phys. Ree Let (1979) 83, [49] DH. Kobe, Phys Rev. Len. AID (1979) 298 [420] DH. Kobe and AL. Sm, Am. J. Phy. (197) 634 [Usi] G. Gryberg and E. Giscobino, J Phys BI (1979) LOS, [N52] WE. Lambe, Phys.Rev 5 (1952) 29, [455] F Schada, CR. Sond Jr and M, Heres J. Phy. B7 (1079) 1% [54] FY Wo, RE. Grove and S. Ezehic. Phys. Rev, Let. 38 (1975) 12 [45] HJ. Kime, M. Dagens and L, Mandel Phys Re. et, 39 (1977) 6, [4] M. Dagenai and L, Mande. Phys. Rev. AIS (1978) 2210. [M57] L. Mandel, Phys. Rev. AX) (19%) 15m, 4S] WEP. Healy, Phy, Rev, A22 (OMT) 291 [390] EA Power and T, Thunamachandean. Pye. Rey. A2 (1980) 2, Sea] Graham and M Haken, Z. Phys. 287 (9%) 3. *Hot] R. Lang. MO Selly and W-E, Lamb J Phys. Rev. AT (173) 17 *f4s| R. Lang and MO, Sly, Opti Coa. 91973) 3 {46} P dodan and E, Wigner, Z. Phys #7 (1928) 6, *foea] GS, Age, Phys Rev. 198 (199) 025, *f465] R, Lown, The Ovantom Theory of Light (Clarendon, Oxfod, 19%) “fe HLM, Noweneweis,Itrodetion to Quantum Optics Germ and Breach, New York, 1979) *[s] , Stenhoim, Phy. Reports 6C (1973) ‘lil AP. Karantcr and GI, Surdstovch Pog, Ou. Elect. 3 (1974) 231 [st] 1. Detc,Physes Laboratory TNO Report PHT 4} (197) 124 [390] RL. Kronig, J. Ope Soe Am. 12 (1926) 80 [e71] HLA, Kramers, A. Congr. Int. Fs Com 2 (1927) 8, [12] H.A, Kramers, Collected Seientibe Papers (Noh Hollnd, Amsterdam, 156), [493] 15. Tol, Phys. Rev. 108 (954) 18. [494] LD. Landoo and EM. Licht, Elettodynamics of Continuous Media (Person, Oxford, [495] WALI. Panofsy and M. Philp, Clie! Ekectricty and Magnetiom (Adiwon Wesley, Reasig, 142) [s74] A. Yas, Quantum Electonics (Wiley. New Yor, 1967) [eT1] MY. Klein, Optics (Wiley New York, 1970, [s7H] A. Maitland and MH, Dana, Laser Physics (NorthHolland, Amsterdam, 1). [s79] M. Born and E. Wot, Principles of Optics (Pergamon, Oxford, 1989979. [st 1 Hilgevoor, Dispersion Relations and Casal Desenpion (North-Holland, Amsterdam, 190), [ski] ECC. Tuehmarsh, latraduction tothe Theory of Fourie Interas(Clarendon, Ould, 14H) [sto] 5. Chaturved, PD. Drummond and DF. Wal, J Phys, AIO (197) LIN [sK3] PD. Drummond and H J. Carmichael, Optics Com. 27 (1978) [3%] PD. Drummond and DF. Walls, J Phys. ALS (1980) 735 [45] P.D. Drummond and CW. Gardiner J. Phys. 3 (0) 2383 [sth] PM. Nieto and LM, Simons J. Phys. Rev, Let. (1078) 207 “[s47| G. Devens, Phys Lew WA (Io) 21 section 9 {sta} Lord Rayleigh, Pi. Mag. 12 (81 [st] R. Gans, Ana . Phys. (1925) 2. [350] M. Bon, 2, Phys, 3 (126) 31H. Dekker, Casical and quansum mechan ofthe damped harmonic exile ww SECTION 1 [491] L Kuipers and R. Tinman (eds), Handbook of Mathematics (Pergamon. Oxford. 18). [492] CE. Peanoa (ed), Handbook of Applied Mathematics van Nosirand Reinhold, New Yor, 1972 SECTION 11 {:93] Lord Rayleigh, Theory of Sound Vol. (MacMillan/Dover, London New York, 197) section RX the oviginal work dates back to 173 [494] W. Ritz J. Reine Angew. Math. 135 (168) 1 [495] W. Ritz Gesammete Werke (Gautier-Villas. Pas. 1911) p. 192 281 [496] PAM, Mone and H. Fesbbach, Methods of Theoretical Physics Vo. (McGraw-Hill New Yor. 1953, [497] N. Born. Atommechanit (Springer. Lege. 1928) [04] HLA, Kramers, Z. Phys 39 (19) 808, [299] C. Estar. Revs. Mod. Phys. 29 (1988) 39. [S00] KK. Thoraber. io: Path Imterals snd their Applications in Quantum, tatiticl and Solid State Phys, eh. GL. Papadopoulos and LT Dewees (Plenum. New York, 197) p. *|s0i] G4. Papadopoaon. in: Path Interas. es, C.J Papadopoulos and 1T, Devteee (Plenum, New York. IK) p48 [s02] § Srebrenie and RF_W. Bade, J. Chem. Phys. 61 (1978) 338 [S03] 5. Srevenik, RE-W. Bader and TT. Nguyen-Dang.J. Chem. Phy. 6 (IWR) 47, 48, {S04} H.Graket P. Hingg and P Tlkne, Phys, Rew. 19 (1070) 20, [S05] RP. Feynman, Revs. Mod. Phys. 20 (038) 7 [S04] J. Linge J. de Math 2 (17) 16, 318 [S07] Lord Rayleigh, Roe. Roy. Sc. London ARS (191) 317, [SOK] Hdetey, Pre, London Math, Soe. 23 (121) 2 [S09] G, Wentel, Z. Pays. 4 (192) SK {Sto} 1. Bouin, Compt. Rend. 1K (1925) 28 [sli] ECC Kemble, The Fundamental Priainles of Quantum Mechanics (McGraw-Hl, New York, 1937) section 2 [Siz] RE Langer, Phy, Rev. St (957) [sia] WH, Fury. Phe, Rev. 7 (147) Sa, [sta] S.C, Mile Je and RL Goo Je, Phys. Ren 91 (1983) 178 [sis] EC. Titchmarsh, Egenfonction Expansions, prt I (Onfrd UP, Oxford, 196, [S16] U. Deisiaphave and R. Graham, Z. Phys. BY (97) 211 [St7] L dela PehaeAverbach and AM. Cet, Phy. Let 9A (1972) 65, SECTION 12 [sth] J, Hadamard, Lectures on Cauchy's Problem in Liner Partial Dierentil Eqtions (¥ale/Oxford Unis Pres. New Haven/London, 1923. [sto] L, Sehwar, Théorie dex dntnbatons Vos Ian Il Hermann et Cie, Pais 19881) [S20] G. Temple, Poe. Roy. Soe. AZ (195) 19, Fs21] MJ. Lig nvedaction to Fourier Anais and Generalised Function (Unie. Pes, Cambrge, 1984) [522] M. Ries, Comptes Rendus du Congr International des Mothimaiciens, Osi 1936, Tome Il (1937) exp... [523] L,Sehwars, Matematiche Methoden der Phyuk (Hermane/Bibiogaphsches Institut AG, Pats Ziich, 148/79) *{524] NG. van Kampen, Physica 20 (1984 ¢03 *[525]L. van Hove, Physica 21 (1985) $17. *[526]F Bloch, Phy, Rev. 102 (1956 16 “1527 F Bloch, Pays. Rev. 10 (1957) 136, S28] CL George, Paya 26 (19) 453, [529] AG. Redheld, Advances in Magnetic Resonance Vo I (Academic. New York, 1965) [30] RY, Seer Proc. Nat Acad, Se 1 (1932) 303. “TS31] A. Longer, Novo Gi, 2951) 5 {532] M. Razay, private communication (198). *1533] M. Raza, Lett, Nuovo Cin, 24 (1979) 293 [534] B, Mara | Prigogine and M, Coubage, Physica 9A (1979) 1 {535] H. Spohn, Res. Mod. Phys. 53 (980) 5. 536] J. Hares, Phys. Rey, AZ2 (1980) 1205. [S37] 1 Prgopine, From Being 1 Becoming: Tine and Compleity in the Pascal Sciences (Freemae, Oxford, 1M) [S34] H, Kagermann Physien 0S A (1981 48, 3 +[500] K_ Abiecht and RW, Hase, Factonsng Coupled Quantum Systems and Damped Nonlinear Schdinger Equations (Pepin, MnchealGarchiag 1981) to be published ee aso (2)uo 1H. Dekker, Clasical and quantum mechanic of the damped hormone oslo “{SW] G. Ghosh and RW. Hasse, Phys. Re. AZS (SKI) 16 “{StI] B. Yorke, Remarks concerning the qvantzaton of lamped elecronc crits by means of transmisivn Hines Private Communication, Conse Unverity, 181) se a0 (9,110, *[S10] DM. Mazen (Private Conversation, Univ, Groningen, Netherlands, 1981}; Mathemateat Aspects ofthe FKMmoel: we [131] and ao [3 {S01 JT. Lewis and LC Thomas, in; Proc. lat. Cont. Combertand-Lodge: Functional lategraton aed its Application. ed. AML Arthurs (Clarendon, Ord. 195) p97 *[s44] Zhang Labshun. Modiied Perturbation Method ofthe Damped Harmonic Oxilatr in Quantum Laser Theory (Shani Unis. China 98) submited to Physica A “{s49] K.Jesita and BS. Skagerstam, Reps Math Phys. 5 (199) 423, *[sus] Mt, Razany, Can J. Phys. $6 (19) 1019, *{SI7] ES Hermandes and 8. Remavd, Phys Let 754 (198 248, s[sut] B. Remind and ES Hernande, Physica 103A (IRD) 36 *[Si] M. Razaey Coheret States andthe Imeversble Motion of 2 Parle Coupled to 2 System of Oxilators (Preprint, Edmonton, 18/8). “{850] E'S, Herander and B, Remasd, Physics HSA (941) [8 *[sSi] 8 Remaud and ES. Hermandee. Physica 107 A (1981) S83, *{SS2]P. Exner and G..Koleron, Phys. Le. ABA (IRI) 203 *I5S3] M lehivanag, On the Probabilistic Descrition of May: Bay Systems inthe Schrager Petar (Preprint, Oak, 1981): see alo [3 “[S8I] RKC Coeprave and M,Sebawte Abdalla, Optica Acta 2 (141) 4S “I5S5] A Jaonasis, A. Lear. VPapathve, N. Patargan and K.Viatos, Phys OA (UL) *[sse) RK. Colgrave and M.Sehawe Aba, Phys. ALS (NI) 20 *[S] §, Okubo, Pye Rev. 2 (Ut) 27, “[S8H] N. Gin J Pays ALE (sD 289 [S99] F Neqoo and A Tata, Phys Res, A23 (UND) [81 [S40] EA Spiegel, Physica 1D (1980) 2% [sel] A'S. Piowsky and MLL Rabin, Physica 2D (KI) [s62] GP. herman and G M. Zaslvsy, Physics 20 (NT) 2, [se3] 8 Seca Jy, Ae (lost 2877 Almarens 85 TE ‘Asti 91%, principal fenton Aut pert: I Aa onciator aerormage Anublation opera ae esto ‘Antcomnatatr: 10, 13, 7, 8 67,757 ‘Na he funstio 71-13 Iaher-Htanor elton: 16 awh 670,78 Bathe tse Bolum act 0, THT? Binoy Hoarty cio: 27,32 roan moti 6.49 (Canonical momenta 8 1B. HBS 28M. AT 2. SH, ATER, AS (Camel tamdoraaton: 14,39, 20 ML 7, $6 8TN and eonlee Hamman, NI 2, Casimir operator: 16 Cauchy pmol ela: 78 Cay: rewnate Charset method of 867 CCharscert tie #9, 78 Coil foguensy ae rte fregaeney (Ciclo 7, 510,74, ML ‘wanearaning: 69, 3,78, se scat Comer ten: 81 8664, 18 random phase 82 Comat, LS1S, 2-24 29, BIS, 42,7, $7,470 dierent pero 57 CComscratit of peobahinty 23,57, 77.9% Contin eaton: 28, S857, 92 Coatineum pais 27. 8 postal 27, 34a 48.72 Conta, cize a 8, (Correa function: $. 37,92 oslo ftion: Coat gauge #8 Coupling constant: 2, 1, 67-49 Creston operators BY, 8, nds DAlembert’ principe: 3 DAtembert’s otiog: XP Damping underonereucat W103 Dense spectra: se oti Denutye clic S86 Dent} matiioeratn: 82, 9-40, (472, 177