You are on page 1of 7

Physics Letters A 375 (2011) 16301636

Contents lists available at ScienceDirect

Physics Letters A
www.elsevier.com/locate/pla

Quantization of the damped harmonic oscillator revisited


M.C. Baldiotti, R. Fresneda , D.M. Gitman
Instituto de Fsica, Universidade de So Paulo, Caixa Postal 66318-CEP, 05315-970 So Paulo, S.P., Brazil

a r t i c l e i n f o a b s t r a c t

Article history: We return to the description of the damped harmonic oscillator with an assessment of previous works,
Received 1 July 2010 in particular the BatemanCaldirolaKanai model and a new model proposed by one of the authors. We
Received in revised form 23 February 2011 argue the latter has better high energy behavior and is connected to existing open-systems approaches.
Accepted 3 March 2011
2011 Elsevier B.V. All rights reserved.
Available online 5 March 2011
Communicated by A.P. Fordy

Keywords:
Dissipative systems
Quantization with constraints
Coherent states

1. Introduction ally in time, and many of the pathologies usually appointed to the
quantum theory can be seen as an artifact of its innite time indef-
The problem of constructing a quantum theory for the damped inition. Nonetheless, regardless of these disputes, there has been
harmonic oscillator as well as for similar dissipative systems, e.g. fruitful applications of the BCK theory as a model of dissipation,
radiating point-like charge, has attracted attention for already at least in the case of the canonical description of the FabryProt
more than 50 years. In spite of the success of interaction-with- cavity [10].
reservoir approaches, we feel there is still room for some formal The underlying understanding of dissipative systems is that
developments in the direction of a closed theory approach. In this they are physically part of a larger system, and dissipation is a
Letter we analyze a novel Lagrangian model for the damped har- result of a non-elastic interaction between the reservoir and the
monic oscillator, which was recently proposed in [1] as a particular subsystem. Thus, quantization can conceivably follow two different
case of a general procedure for nding action functionals for non- approaches: the rst one takes the classical equations of motion of
Lagrangian equations of motion. Even though we nd it is locally the system and applies to them formal quantization methods, try-
equivalent to the renowned BatemanCaldirolaKanai (BCK) model ing to overcome the diculties related to the non-Lagrangian na-
[24] (see Section 2 for a revision), a complete global equivalence ture of the system. Without attempting to exhaust all the literature
is absent, and we believe it has some formal advantages over its on this matter, we cite, for instance, the use of canonical quanti-
predecessor, regarding the high energy behavior of solutions to the zation of classical actions [11,12], Fermi quantization [13], path-
Schrdinger equation. Notwithstanding these formal discrepancies integral quantization [14], doubling of the degrees of freedom [2,
between the two models, we show that they share a very close 15,16] (in this connection, we ought to mention the interesting re-
physical interpretation with regard to their asymptotic behavior in sult obtained in [17], where it is shown that a (2 + 1)-dimensional
time and to their physical functions. gauge theory with a ChernSimons term can be interpreted in the
Ever since the proposition of the BCK model, there have been infrared limit as the Bateman dual-system for the damped har-
divided opinions as to whether it really describes the damped har- monic oscillator see also [18] for a related approach), group
monic oscillator or dissipation. There are those who dispute it as a theory methods [19], complex classical coordinates [20], propaga-
possible dissipative model [57], and those, in addition to the orig- tor methods [21], non-linear Schrdinger equation [22], and nally
inal authors, who maintain it accounts for some form of dissipation a constrained dynamics approach [1]. The second approach aims
[8,9]. As we explain further on, the BCK theory is not dened glob- at constructing a quantum theory of the subsystem by averaging
over the reservoir, see e.g. [5,2329]. A more mathematically rig-
orous approach using the system-plus-reservoir idea is to replace
the usual dynamical group for closed systems by a completely pos-
* Corresponding author. Tel./fax: +55 11 3091 7177.
itive semigroup, which was proposed in [30,31] (for an extensive
E-mail addresses: baldiott@fma.if.usp.br (M.C. Baldiotti), fresneda@gmail.com
(R. Fresneda), gitman@dfn.if.usp.br (D.M. Gitman). study, see [32]). The Markovian semigroup approach of the previ-

0375-9601/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.physleta.2011.03.009
M.C. Baldiotti et al. / Physics Letters A 375 (2011) 16301636 1631

ous references is directly related to the appearance of resonances, framework of quantum mechanics. Consequently, we write the
as discussed in [33] using the framework of Gelfand triples and subscript BCK (BatemanCaldirolaKanai) to label the Hamiltonian.
rigged Hilbert spaces (see also [34]). Formal canonical quantization of the Lagrangian action (3) is
It is immediately clear that two different outcomes must fol- straightforward,
low from these approaches. Indeed, in the rst case, we consider
the system as being closed, and thus naturally only pure states [q, p ] = i , [q, q] = [ p , p ] = 0, H BCK = H BCK (q, p ), (5)
can result, and they represent physical states without any further
restrictions. In the second case, the subsystems states will be nec- and coincides with Caldirola and Kanais quantum theory. Solu-
essarily described by a statistical operator, such that only mixed tions to the Schrdinger equation with Hamiltonian H BCK have
states are physically sensible. In spite of the general understanding been found in the form
that the second approach seems to be more physical and probably   
should produce a more adequate quantum theory for dissipative
 1/2 1/4 t
nBCK (q, t ) = 2n n! exp i E n t +
systems, in this Letter we wish once again to analyze the rst, for- 2

mal, approach. Its comparison to the second one will be given in
q 2  
a future work. Our analysis will be particular to the damped har- ( + i ) e 2 t H n qet , (6)
monic oscillator as an example of a dissipative system, and it will 2
be devoted to the comparison of two different canonical quantiza-
where E n = (n + 1/2), = 2 2 and H n are Hermite poly-
tions.
nomials. These are the familiar pseudostationary states [38,39,11]
In Section 2 we revise the BCK theory, emphasizing problems
or loss-energy states [8] which are also eigenstates of the Hamilto-
which were not considered before. In Section 3 we present the
nian H BCK + 2 (q p + pq) with eigenvalues E n . Even though |nBCK |2
rst-order theory, dene some useful physical quantities, and we 
depends on time, the total probability dq |nBCK |2 = 1 is time-
also construct the coherent and squeezed states of the rst-order
independent, as can be seen by the transformation of variables
theory in order to obtain the classical limit. In Section 4, we prove
q  q = qe t . Moreover, the mean value of the BCK Hamiltonian
the local equivalence of the theories presented in Sections 2 and 3,
in the pseudostationary states is constant,
and discuss possible divergences related to their global behavior.
In Section 5 we present some nal remarks and discuss our results BCK


as well as open problems. n
H BCK
nC K
 


2 1
2. BCK theory = nBCK
E n (q p + pq)
nBCK = n+ , (7)
2 2
One of the peculiarities of dissipative systems, which hindered which is a reection of the fact that in the classical theory de-
early quantization attempts, was the non-Lagrangian nature of the ned by (3) the average of the Hamiltonian over the period of one
classical equations of motion. In the particular case of the damped oscillation is constant. On the other hand, mean values of the me-
harmonic oscillator of constant frequency and friction coecient chanical energy E = 12 (q2 + 2 q2 ) decay exponentially with time,
> 0, the second-order equation of motion  E n = e 2t  H BCK n [38,11]. Coherent states for the BCK theory
are given in [9], for which the uncertainty relations are
q + 2 q + 2 q = 0 (1)
1
cannot be directly obtained as the EulerLagrange (EL) equation q p =  . (8)
of any Lagrangian, since it fails to satisfy the Helmholtz conditions 2 2
[35]. Nevertheless, there is an equivalent second-order equation for Now we draw attention to the high-energy behavior of pseu-
which a variational principle can be found, namely, dostationary states. In Appendix A we consider asymptotic (in n)
  pseudostationary functions, and for these one has the limiting
e 2t q + 2 q + 2 q = 0. (2) eigenvalue equation
The exponential factor is known as the integrating multiplier, and

it is enough to make the above equation satisfy the Helmholtz con- H BCK nBCK = E n (q p + pq) nBCK
ditions [36]. The fact that a Lagrangian can always be found for the 2
 
one-dimensional problem such that its EL equation is equivalent to i  
a given second-order equation was established by Darboux [37]. = En + nBCK + O n1/4 . (9)
2
As was already mentioned, Eq. (1) is traditionally considered
to be non-Lagrangian, albeit the existence of a questionable ac- The appearance of imaginary eigenvalues is actually explained by
tion functional [6,7] which reproduces the equivalent equations of taking into account the domain of the operator q p + pq, which,
motion (2). In this respect, we have to mention that an action prin- as is shown in Appendix A, does not include the asymptotic part
ciple for the equation of motion (2) was rst proposed by Bateman of nBCK . In this connection, it should be noted that the above is
[2] in terms of the Lagrangian in contrast to [8,9], where it is claimed that the pseudostationary
states are eigenstates of H BCK with eigenvalues ( + i )(n + 1/2).
1 2 
LB = q 2 q2 e 2t . (3) One overlooked aspect of the BCK theory is that EL equation (2)
2 obtained from the Lagrangian (3) is only equivalent to the equation
If Bateman had constructed the corresponding Hamiltonian formu- of motion of the damped harmonic oscillator (1) for nite times.
lation, he would have discovered that the corresponding Hamilto- Eq. (2) is indicative that the theory described by (3) is not globally
nian theory is canonical without constraints and with Hamiltonian dened, i.e., it is not dened for innite times. In effect, the man-
ifold diculties which appear in connection to the t limit,
1  2 t 2 
H BCK (q, p ) = e p + 2 e 2 t q 2 . (4) such as the violation of the Heisenberg uncertainty principle and
2 the vanishing of the ground state energy for innite times, can be
This Hamiltonian was proposed independently by Caldirola and seen as consequence of inadvertently assuming Eqs. (1) and (2) are
Kanai [3,4] to describe the damped harmonic oscillator in the equivalent for all values of the time parameter.
1632 M.C. Baldiotti et al. / Physics Letters A 375 (2011) 16301636

3. First-order action y (0) p, so that the above quantum brackets have the familiar
form
3.1. Action, Hamiltonization, and quantization
[q, p ] = i , [q, q] = [ p , p ] = 0,
Next we consider the canonical quantization of the damped
harmonic oscillator based on the alternative action proposed in [1]. x(0) q, y (0) p . (15)
The idea was to reduce the second-order equations (1) to the rst- The differential equations for x and y can be easily integrated to
order system
x(t ) = e t q, x(0) q,
x = y , y = 2 x 2 y , (10)
t
y (t ) = e p , y (0) p . (16)
for which, according to the general theory [40,41], the action func-
tional has the form The quantum Hamiltonian is obtained from the classical Hamilto-

1    nian (13) as a function of the operators x(t ) and y (t ) (16): it does
S= dt y x x y y + 2 xy + x e 2t .
2 2 2
(11) not depend on time at all,
2
The EL equations of motion derived from (11) are locally equivalent   1 
H = H x(t ), y (t ), t = p 2 + (q p + pq) + 2 q2 , (17)
to (10), 2
S   S where we have used Weyl (symmetric) ordering for the mixed
= y + 2 y + 2 x e 2t , = (x y )e 2t . product 2 xy. The Hamiltonian H governs the time-evolution of
x y
the state vector in the Schrdinger picture, and it has appeared
Note that, as in the case of the BCK theory, the theory fails to in a number of different contexts: in [43,10] with regard to the
describe the damped harmonic oscillator as the time approaches electromagnetic eld in a resonant cavity; in the quantization of
innity, and thus one should expect problems in the quantum the- the complex symplectic theory [44]; and in [45] in connection to
ory such as violation of the uncertainty principle for innite times, the Lindblad theory of open quantum systems for the damped har-
for instance. monic oscillator.
This action describes a singular system with second-class con- Since the Hamiltonian is time-independent, the evolution op-
straints, and furthermore, these constraints are time-dependent
erator is given simply by U (t ) = e i Ht , with H given by (17). The
(we follow the terminology of the book [42]). Even though the
Heisenberg operators x and y corresponding to the classical vari-
constraints are explicitly time-dependent, it is still possible to
ables x and y are
write the Hamiltonian formalism with the help of Dirac brackets
and perform the canonical quantization, as is explained in [42].  
1
In order to do this, one must extend the initial phase space of x = U 1 x(t )U = e t cos t + sin t q + e t sin(t ) p ,
canonical variables = (x, y , p x , p y ) by the inclusion of the time t

 
and its associated momentum . As a result, the Poisson brackets 1 t 2 t
between functions dened on the extended phase space is
y = U y (t )U = e cos t sin t p e sin(t )q,

 
F G F G F G x(0) q, y (0) p , [q, p ] = i , [q, q] = [ p , p ] = 0.
{F , G} = + + F G.
x px y p y t From the above expressions, one also nds
With the above denition, the rst-order equations of motion (10)
are equivalently written in terms of Dirac brackets, d x d y
= y , = 2 x 2 y ,
  dt dt
= , H (x, y , t ) + D ()
, x = y = 0, (12)
which coincide in form with the classical equations,
where the Hamiltonian H (x, y , t ) and the constraints are

d x d y
1  = {x, H + } D ()
= , = { y , H + } D ()
= . (18)
dt dt
H (x, y , t ) = y 2 + 2 xy + 2 x2 e 2t ,
2 Thus, Heisenberg equations (18) reproduce the classical equations
1 1 of motion, and therefore mean values of x and y follow classical
x = p x ye 2t , y = p y + xe 2t . (13)
2 2 trajectories.
The nonzero commutation relations between the independent vari- Moreover, the only nonzero commutator becomes
ables are
[x, y ] = ie 2t ,
2 t
{x, y } D () = e . (14)
which matches the classical Dirac bracket (14). Thus, the resulting
Quantization of this system follows the general method described quantum theory at least obeys the correspondence principle.
in [42] for theories with time-dependent constraints. One intro- One can easily nd solutions of the Schrdinger equation by
duces time-dependent operators (t ) which satisfy the differential making a (time-independent) unitary transformation H = S 1 H S,
equations d/dt = i {, } D () |= with initial conditions subject to  
an analog of the Dirac quantization, i 2
S = exp q , S 1 p S = p q, (19)
 
2
(0), (0) = i {, } D (0 )
=(0) ,
one obtains
1 1
p x (0) y (0) = p y (0) + x(0) = 0. 1  
2 2 H = p 2 + 2 q2 , = 2 2 . (20)
The above operatorial constraints allows us to work only in terms 2
of the independent operators x and y. Let us dene x(0) q and The Hamiltonian (20) has the familiar stationary states
M.C. Baldiotti et al. / Physics Letters A 375 (2011) 16301636 1633

n (t , q) = exp(it E n )n (q), Thus, by (7),


     
 1/2 1/4 BCK

BCK 2 1
n (q) = 2n n! exp q2 H n ( q), n
E
n = e 2t n+ = n | E M |n .
2 2
 
1
H n = E n n , E n = n + . 3.2. Semiclassical description
2
Therefore, the wave functions 3.2.1. Coherent states
Finally, we obtain semiclassical states for the damped harmonic
n (q) = S n (q) oscillator from the coherent states of the simple harmonic oscil-
    lator using the unitary transformation S (19). To this end, we
 1/2 1/4 q2
= 2 n! n
exp ( + i ) H n ( q) (21) introduce rst creation and annihilation operators a+ and a and
2 the corresponding coherent states | z,
are eigenfunctions of the Hamiltonian H , and the solutions to the 1 1  
corresponding Schrdinger equation are a = (q + i p ), a+ = (q i p ), a, a+ = 1,
2 2
 
n (t , q) = S n (t , q) = exp(it E n )n (q), | z = D ( z)|0, D ( z) = exp za+ za , a| z = z| z. (26)
H n (q) = E n (q). In terms of these creation and annihilation operators, the Hamilto-
nian (20) is
One can also check directly that (21) are eigenfunctions of H with
 
eigenvalue E n by using properties of Hermite functions. + 1
We now dene some useful quantities to be used in Sec- H = a a + .
2
tion 4.2 for the purpose of establishing the physical equivalence
between the approaches presented here. Let us write the classical Thus, the coherent states for the Hamiltonian H are S | z and the
Lagrangian energy as mean values of x and y in these coherent states are
L L 1 2  1  
EL x + y L = y + 2 xy + 2 x2 e 2t . x  z| S 1 x S | z = e t ze i t + ze i t ,
x y 2 2

The corresponding Weyl-ordered Schrdinger operator for the La- t  it 
grangian energy is  y = i e ze ze i t x.
2

1 2  One can now easily verify that the mean values of the coordinates
E L = E L ()
= = p + (q p + pq) + 2 q2 = H , (22)
2 x and y follow the classical trajectories,
which coincides with the Hamiltonian and thus with its Heisen- d d
berg representation, and is therefore conserved. Likewise, we de- x =  y ,  y  = 2 x 2  y .
dt dt
ne the following conserved energy in the BCK approach,
The pathological behavior of the rst-order theory with regard to
1  the limit t can be seen here in the computation of the un-
E = q2 + 2qq + 2 q2 e 2t = H BCK + qp ,
2 certainty relation
d S 1
E = (q + q)e 2t , x y = e 2t . (27)
dt q 2
which is constant on-shell. Its image as a Weyl-ordered Schrdin- The unphysical result of the above violation of the uncertainty
ger operator is principle is an indication that the rst-order theory is not dened
for all values of the time parameter, as was pointed out in connec-
E = H BCK + (q p + pq), E nBCK = E n nBCK . (23) tion to the classical equations of motion and to the commutation
2
relation between x and y. Another indication of the failure of the
Now consider the mechanical energy in the theory with La-
theory at innite times is in the observation that the radius of the
grangian (11) at = 0:
trajectory of the mean values vanishes:
1    
EM = y 2 + 2 x2 . z = re i = x2 +  y 2
2

The corresponding operator is equal to 2 t 
= e r 1 + 2 cos(t ) + sin 2(t ).
1 
E M = p 2 + 2 q2 e 2t = e 2t H e 2t (q p + pq). (24)
2 2
3.2.2. Squeezed-state
The mean value of E M in the energy eigenstates is One can also consider the family of conserved creation and an-
  nihilation operators
2 1
n | E M |n  = e 2t n+ .
2 b(t ) = cosh e i t a + sinh e i t a+ ,
Finally, we consider the observable dened by the mechanical en- b+ (t ) = cosh e i t a+ + sinh e i t a,
ergy in the BCK description,   d d
b(t ), b+ (t ) = 1, b = b+ = 0,
1 2  1  2 t 2  dt dt
E= q + 2 q2 = e p + 2 e 2 t q 2 e 2 t
2 2 and construct squeezed coherent states | z,  = exp( zb+ zb)|0
= H BCK e 2t . (25) [46]. For = 0 one arrives at the previous coherent states,
1634 M.C. Baldiotti et al. / Physics Letters A 375 (2011) 16301636

 z, 0|x| z, 0 =  z| S 1 x S | z, z, 0| y | z, 0 =  z| S 1 y S | z. The quantum theory can be readily obtained by a quantum
time-dependent canonical transformation2
For arbitrary values of and with x and y given by (16), one has
for the mean values of coordinates i t
D = exp (q p + pq) , (29)
2
e t  i t 
 z, |x| z,  = e z + e i t z cosh which is suggested from the classical generating function and the
2 relationship between old and new variables. The effect of D on the
  
e i t z + e i t z sinh , canonical variables is to make the dilation

t  it  D 1 q D = e t q, D 1 p D = e t p .
 z, | y | z,  = i e e z e i t z cosh
2 The dilation operator D has been discussed in a more general
  
e i t z e i t z sinh  z, |x| z, . setting in [49] in relation to conformal invariance of quantum sys-
tems, in [50] in relation to time-reversal, and in [51] in connection
The main interest in squeezed states is that they allow one to with the description of open systems by time-dependent Hamil-
change the uncertainty in either direction x or y by adjusting the tonians. In the latter, as is the case here, the dilation operator
parameter . For example, the uncertainty in x can be written for simplies the calculation of the evolution operator.
arbitrary values of as The dilation operator transforms the BCK Hamiltonian into the
rst-order Hamiltonian H ,
1 
(x)2 = e 2t (cosh + sinh )2
2 D 1 
 H = D 1 H BCK D i D 1 = p 2 + (q p + pq) + 2 q2 .
2
4 cosh sinh cos t  0, t 2
Since n (q, t ) satisfy the Schrdinger equation with H , it fol-
and it reduces to the preceding coherent states calculations for
lows that D n (q, t ) satisfy the Schrdinger equation with H BCK .
= 0.
The wave functions D n (q, t ) are indeed the pseudostationary
states, as can be seen by direct application of D. Thus, we
4. Comparison of the BCK system and the rst-order system
write nBCK (q, t ) D n (q, t ). Similarly, it also follows that U =
D exp(it H ) satises the Schrdinger equation with Hamiltonian
4.1. BCK model as a transformation of the rst-order system
H BCK ,

Here we show how one can obtain the classical and quantum U
U (t ) = D exp(i Ht ), i = H BCK U , U (0) = I .
description of the BCK damped harmonic oscillator as a canoni- t
cal transformation of the rst-order approach (Section 3). At the One should not be to eager to jump to the conclusion that the
classical level, both systems are transformed one into the other by two theories here presented, the BCK theory and the rst-order
means of the following time-dependent canonical transformation1 theory, are physically equivalent solely on the grounds of the time-
dependent canonical transformation D. The existence of this trans-
1  1
q= x 2e 2t p y , e 2 t y ,
p = px + formation per se is insucient to prove physical equivalence, since
2 2 in principle, and at least locally, one can always construct a time-
1 1 dependent canonical transformation between any two given theo-
1 = y p x e 2 t , 2 = p y + xe 2t ,
2 2 ries, classical or quantum. For instance, given two quantum theo-
ries T 1 and T 2 , and corresponding evolution operators U 1 and U 2 ,
where (q, p ) and ( 1 , 2 ) are new pairs of canonical variables. In one can always write a general solution 2 (t ) of the Schrdinger
these new variables, the equation of motion become Hamiltonian: equation of T 2 in terms of the general solution 1 (t ) of T 1 by
means of the transformation 2 (t ) = U 2 U 11 1 (t ). Besides the nec-
q = {q, H BCK }, p = { p , H BCK }, = 0, essary ingredient of the unitary transformation relating the two
1  theories, it is imperative to show that the physical observables per-
H BCK = e 2 t p 2 + 2 e 2 t q 2 , taining both theories are also unitarily equivalent in order to prove
2
physical equivalence. A proof which we postpone to the next sec-
where the Hamiltonian H BCK (q, p , t ) is the canonically transformed tion.
Hamiltonian H (x, y , t ) (13) on the equivalent constraint surface
= 0. 4.2. Physical equivalence
It is useful to write the following relation between old coordi-
nates x and y and the new variables q and p: In this section we show that the two approaches presented in
this Letter are physically equivalent. In order to sum up the pre-
x = q + O (), vious results in a coherent whole, let us present both quantum
theories anew.
y = e 2t p + O (). (28) Starting with the BCK theory, it is dened by the Hamiltonian
This relation shows that x is physically equivalent to q, while y is H BCK (5) written in terms of the canonically conjugated operators q
physically equivalent to e 2t p, since they coincide on the con- and p (5) with the usual realization in terms of multiplication and
straint surface = 0. derivation operators in the Hilbert space HBCK of square-integrable
states BCK (q) with measure

1
The generating function for these canonical transformations depending on the
new and old momenta is
2
The operator q p + pq is well known in the literature in connection with the
Riemann hypothesis [47,48], where proof of its self-adjointness is given in the case
q R+ . We therefore provide a simple proof for our particular case q R in Ap-
F ( p x , p y , p , 2 , t ) = 2e 2t (2 p y ) p x 2e 2t pp y + e 2t p 2 . pendix A.
M.C. Baldiotti et al. / Physics Letters A 375 (2011) 16301636 1635

+ As has been pointed out (see Dekker [44] for a comprehensive


BCK |BCK HBCK = dq BCK (q)BCK (q). review), the BCK oscillator has a pathological behavior for in-
nite times, since the (mechanical) energy mean values and the
Heisenberg uncertainty relation between the coordinate and the
The rst-order theory is dened by the Hamiltonian H (17) and physical momentum go to zero as time approaches innity, so
canonically conjugated operators q and p (15) realized as the usual that even the ground states energy eventually vanishes. Despite
multiplication and derivation operators in the Hilbert space H of these shortcomings, the quantum theory has a well-dened clas-
square-integrable states (q) with measure sical limit, and for time values less than 21 ln
the Heisenberg
uncertainty principle is not violated. It is our understanding that
+ this unphysical behavior is a result of extending the proposed the-
|H = dq (q)(q). ories beyond their validity. The fact that both theories are not
globally dened in time has its roots already at the classical level,
as a consequence of the non-Lagrangian nature of the equations of
We have seen that the time-dependent unitary transforma- motion of the damped harmonic oscillator.
tion D (29) maps the two Hilbert spaces, Finally, we recall the intriguing behavior of the asymptotic
pseudostationary states, which so far has escaped notice from all
D : H HBCK ,  BCK = D , works dedicated to the BCK oscillator. At rst glance Eq. (9) implies
that the BCK Hamiltonian loses self-adjointness as n approaches
and is a canonical transformation,
innity. Fortunately, these asymptotic states are not in the domain
D 1 of the BCK Hamiltonian and thus pose no threat whatsoever. On
H BCK = D H D 1 + i D . the other hand, there is no such constraint on the domain of the
t
rst-order Hamiltonian H , where there is no upper bound to the
Therefore, for every (t ) solution of the Schrdinger equation of energy eigenstates. This inconsistency can spoil the physical equiv-
the rst-order theory, D (t ) is a solution of the BCK Schrdinger alence of the two theories at high energies, but it is not altogether
equation. Furthermore, one has BCK |BCK HBCK = |H , as can unexpected. As the energy grows, that is to say, as n grows, the
be seen by noting that D (q) = e t /2 (qe t ) and wave functions spread farther out in space and become highly
non-local. Any canonical transformation, on the contrary, is only
BCK |BCK HBCK locally valid, and thus our time-dependent dilation transformation
D (29) cannot guarantee there are no global issues in establishing
+ +
    the physical equivalence of the two theories (or any two theories,
= dq e t qe t qe t = dq (q)(q) = |H . for that matter). This is not unexpected, since in general the non-
local properties of the wave functions can spoil any equivalence
that the two theories might have on the level of the Heisenberg
Finally, to complete the proof of the physical equivalence, it re-
equations.
mains to show that any two physical observables OBCK and O are
Darbouxs method for solving the inverse problem of the cal-
D-equivalent, that is, OBCK = D O D 1 . One can check that this is
culus of variations for second-order equations of motion of one-
indeed the case for the physical observables previously considered,
dimensional systems produces theories which, despite being clas-
such as the mechanical energies (24), (25)
sically equivalent, do not possess the expected quantum proper-
ties. Thus, whereas every classical one-dimensional system can be
E = D E M D 1 ,
Lagrangianized, in quantum mechanics one still encounters one-
and the conserved energies (22), (23), dimensional systems which can still be called non-Lagrangian.
For such systems, quantum anomalies resulting from limiting pro-
E = D E L D 1 . cesses (in our case, the limit t ) can be identied already
at the classical level by an asymptotic inequivalence between the
As a result of the equivalence, one can easily obtain the BCK original second-order equation and the EL equation resulting from
coherent states presented in [9] by merely transforming the coher- the Lagrangianization process.
ent states (26) given in the context of the rst-order theory. Taking We note that the quantum Hamiltonian (17) naturally appears
into account the relation y = e 2t p + {} between the physi- in the master equation for open quantum systems in the descrip-
cal variables of the two theories (28), one can see why the un- tion of the damped harmonic oscillator in the Lindblad theory [45]
certainty relations we obtain decay exponentially with time (27), and Dekkers complex symplectic approach [52]. This relation has
while those (8) calculated in [9] are constant. This has a simple yet to be claried, and it should prove useful in the comparison
explanation in terms of the physical equivalence of y and e 2t p: of the BCK theory and the rst-order theory along the lines of a
y is physically equivalent to the physical momentum of the BCK subsystem plus reservoir approach.
oscillator, and it is in terms of the physical momentum that the
Heisenberg uncertainty relations are violated (see Dekker [44] for Acknowledgements
a comprehensive review).
M.C. Baldiotti, R. Fresneda and D.M. Gitman thank FAPESP for
5. Conclusion
nancial support.

In this Letter we have proved that the classical and quantum


Appendix A
description of the damped harmonic oscillator by the BCK time-
dependent Hamiltonian [24] is locally equivalent to the rst order
approach given in terms of a constrained system [1]. This equiva- Large n limit of A nBCK
lence allowed us to easily obtain the evolution operator for the
BCK oscillator and many other results in a simpler manner, due Let us compute the action of A = q p + pq on the functions
to the time-independence of the quantum rst-order Hamiltonian. nBCK (6),
1636 M.C. Baldiotti et al. / Physics Letters A 375 (2011) 16301636

 
which are not square-integrable in the interval (, ). There-
(q p + pq)nBCK = i 2q + 1 nBCK
q fore, the corresponding deciency indices are (0, 0) and A is es-
  sentially self-adjoint.
= i (n + 2)(n + 1)nC+K2 i n(n 1)nBCK
2

(n + 2)(n + 1)n+2 + (2n + 1)nC K
CK References

 
+ n(n 1)nBCK
2 . (30) [1] D.M. Gitman, V.G. Kupriyanov, European Physical Journal C 50 (3) (2007) 691.
[2] H. Bateman, Physical Review 38 (4) (1931) 815.
The asymptotic form of the pseudostationary states for large values [3] P. Caldirola, Nuovo Cimento 18 (9) (1941) 393.
[4] E. Kanai, Progress of Theoretical Physics 3 (4) (1950) 440.
of n is [53] [5] I. Senitzky, Physical Review 119 (2) (1960) 670.
 1/4    [6] J.R. Ray, American Journal of Physics 47 (7) (1979) 626.
BCK n (2n 1)!!
q 2 e 2 t [7] D.M. Greenberger, Journal of Mathematical Physics 20 (5) (1979) 762.
2n (q, t ) = exp i (1)
2 2n!! [8] V.V. Dodonov, V.I. Manko, Nuovo Cimento B 44 (2) (1978) 265.
 t
  1/4  [9] V.V. Dodonov, V.I. Manko, Physical Review A 20 (2) (1979) 550.
cos (4n + 1)e q + O n [10] R.K. Colegrave, M.S. Abdalla, Journal of Modern Optics 28 (4) (1981) 495.
  [11] R.W. Hasse, Journal of Mathematical Physics 16 (10) (1975) 2005.
= 2n + O n1/4 ,
asym
[12] S. Okubo, Physical Review A 23 (6) (1981) 2776.
 1/4    [13] D.C. Latimer, Journal of Physics A: Mathematical and General 38 (9) (2005)
BCK q 2 e 2 t n (2n 1)!! 2021.
2n+1 (q, t ) = exp i (1) [14] D. Kochan, Physical Review A 81 (2) (2010) 14.
2 2n!!
    1/4  [15] M. Blasone, P. Jizba, Annals of Physics 312 (2) (2004) 354.

sin (4n + 3)e q + O nt [16] E. Celeghini, M. Rasetti, G. Vitiello, Annals of Physics 215 (1) (1992) 156.
  [17] M. Blasone, Annals of Physics 252 (1) (1996) 115.
= 2n+1 + O n1/4 .
asym [18] R. Banerjee, P. Mukherjee, Journal of Physics A: Mathematical and Gen-
(31)
eral 35 (27) (2002) 5591.

Taking into account that cos 4n + x cos 4n + 1x = O (n1/4 ) [19] J.M. Cervero, J. Villarroel, Journal of Physics A: Mathematical and Gen-
eral 17 (15) (1984) 2963.
for any real , and substituting the above in the right-hand side of [20] H. Dekker, Physical Review A 16 (5) (1977) 2126.
(30), one has [21] Y. Tikochinsky, Journal of Mathematical Physics 19 (4) (1978) 888.
  [22] M.D. Kostin, Journal of Chemical Physics 57 (9) (1972) 3589.
+ O n1/4 .
asym
(q p + pq)nBCK = i n [23] R. Feynman, F. Vernon, Annals of Physics 24 (12) (1963) 118.
[24] H. Haken, Reviews of Modern Physics 47 (1) (1975) 67.
+ O (n1/4 ), so
asym
On the other hand (q p + pq)nBCK = (q p + pq)n [25] A.O. Caldeira, A.J. Legget, Physica A: Statistical and Theoretical Physics 121 (3)
(1983) 587.
  [26] D. Walls, G. Milburn, Physical Review A 31 (4) (1985) 2403.
+ O n1/4 ,
asym asym
(q p + pq)n = i n [27] H. Grabert, Physics Reports 168 (3) (1988) 115.
[28] L.H. Yu, Sun Chang-Pu, Physical Review A 49 (1) (1994) 592.
or to the same approximation, [29] I. Joichi, S. Matsumoto, M. Yoshimura, Progress of Theoretical Physics 98 (1)
  (1997) 9.
(q p + pq)nBCK = i nBCK + O n1/4 . [30] G. Lindblad, Communications in Mathematical Physics 65 (3) (1979) 281.
[31] A. Kossakowski, Reports on Mathematical Physics 3 (4) (1972) 247.
asym [32] R. Alicki, K. Lendi, Quantum Dynamical Semigroups and Applications, Springer,
Proposition. The asymptotic pseudostationary functions n are not 2007.
in the domain of A = q p + pq. [33] D. Chruscinski, Annals of Physics 321 (4) (2006) 840.
[34] A. Bohm, H. Doebner, P. Kielanowski, Irreversibility and Causality: Semigroups
and Rigged Hilbert Spaces, Springer, 1998.
We rst nd the domain of A for which it is symmetric. Con- [35] Hermann von Helmholtz, Journal fr die Reine Angewandte Mathematik 100
sider, for , D ( A ): (1887) 137.
[36] W. Starlet, Journal of Physics A: Mathematical and General 15 (5) (1982) 1503.
, A   A ,  [37] G. Darboux, Leons sur la Thorie Gnrale des Surfaces, vol. 3, Gauthier
Villars, Paris, 1894.

d   [38] E.H. Kerner, Canadian Journal of Physics 36 (1958) 371.
= 2i dq q(q)(q) [39] F. Bopp, Zeitschrift fur Angewandte Physik 14 (1962) 699.
dq [40] D.M. Gitman, V.G. Kupriyanov, Journal of Physics A: Mathematical and Theoret-

  ical 40 (33) (2007) 10071.
= 2i lim q (q)(q) + (q)(q) . [41] D.M. Gitman, V.G. Kupriyanov, Action principle for so-called non-Lagrangian
q systems, in: 5th International Conference on Mathematical Methods in Physics
(IC 2006), Rio de Janeiro, 2006, p. 16.
Therefore, functions such that limx q(q) = 0 are not in the [42] D.M. Gitman, I.V. Tyutin, Quantization of Fields with Constraints, Springer-
domains of A. On the other hand, we know that for large values Verlag, 1991.
of n the pseudostationary functions have the asymptotic form (31). [43] K.W.H. Stevens, Proceedings of the Physical Society 72 (6) (1958) 1027.
asym asym [44] H. Dekker, Physics Reports 80 (1) (1981) 1.
Thus, clearly limq qn (q) = 0 and n
/ D ( A ). Note that [45] A. Isar, Fortschritte der Physik 47 (78) (1999) 855.
the closure of A is not affected by the exclusion of the asymptotic [46] V.V. Dodonov, Parametric Excitation and Generation of Nonclassical States in
functions, since they cant be the limit of any sequence. Linear Media, 1st edition, CRC Press, 2002, Chapter 3, 153 pp., Number 2.
[47] M.V. Berry, J.P. Keating, SIAM Review 41 (2) (1999) 236.
[48] G. Sierra, New Journal of Physics 10 (3) (2008) 033016.
Proposition. A is self-adjoint. [49] V. Alfaro, S. Fubini, G. Furlan, Nuovo Cimento A 34 (4) (1976) 569.
[50] C. van Winter, Journal of Mathematical Physics 39 (7) (1998) 3600.
It suces to show that the equation A = i does not have [51] E. Onofri, Lettere Al Nuovo Cimento Series 2 21 (14) (1978) 489.
[52] H. Dekker, Physica A: Statistical and Theoretical Physics 95 (2) (1979) 311.
solutions in D ( A ) [54]. The solutions are of the form [53] V.I. Smirnov, A Course of Higher Mathematics, vol. 3, part 2, Pergamon Press,
1964.
1
= q , = , [54] M. Reed, B. Simon, Methods of Modern Mathematical Physics, I: Functional
2 Analysis, vol. 1, Academic Press Inc., 1981.

You might also like