Physica A 393 (2014) 655–668

Contents lists available at ScienceDirect

Physica A
journal homepage: www.elsevier.com/locate/physa

Classical limit of quantum mechanics induced by

continuous measurements
Adélcio C. Oliveira ∗
Departamento de Física e Matemática, Universidade Federal de São João Del Rei, C.P. 131, Ouro Branco, MG, 36420 000, Brazil

highlights
• We investigate the quantum–classical transition problem in the newtonian regime.
• We show that the Newtonian regime occurs when the system is strongly monitored.
• We show that the Liouvillian regime is mimicked, for the position observable for a weak monitoring.
• We studied the quartic oscillator and our the numerical simulations confirm the analytical results.

article info abstract

Article history: We investigate the quantum–classical transition problem. The main issue addressed is how
Received 12 November 2012 quantum mechanics can reproduce results provided by Newton’s laws of motion. We show
Received in revised form 7 September 2013 that the measurement process is critical to resolve this issue. In the limit of continuous
Available online 17 September 2013
monitoring with minimal intervention the classical limit is reached. The Classical Limit of
Quantum Mechanic, in Newtonian sense, is determined by two parameters: the semiclassi-
Keywords:
cal time (τsc ) and the time interval between measurements (∆τu ). If is ∆τu small enough,
Ehrenfest time
Classical limit
comparing with the τsc , then the classical regime is achieved. The semiclassical time for
Quasideterminism Gaussian initial states coincides with the Ehrenfest time. We also show that the classical
Overlap limit of an ensemble of Newtonian trajectories, the Liouville regime, is approximately ob-
Quantum breaking time tained for the quartic oscillator model if the number of measurements in the time interval
is large enough to destroy the revival and small enough to not reach the Newtonian regime.
Namely, the Newtonian regime occurs when τsc ≫ ∆τu and the Liouvillian regime is mim-
icked, for the position observable, if ∆τu ∈ [τsc , TR ], where TR is the revival time.
© 2013 Elsevier B.V. All rights reserved.

1. Introduction

Historically, immediately after the birth of the Quantum Mechanics many physicists believed that Quantum Mechan-
ics was a universal theory, i.e. applicable to all Physics problems. In fact this idea was in the core of the Bohr–Sommerfeld
quantization rule. However, later, it became clear that this was not the case. A different approach for high energies was
necessary, as well as in the macroscopic ‘‘World’’. In 1917, Einstein [1,2] presented a reformulation of the Bohr–Sommerfeld
quantization rules of the old quantum theory. The paper also offered an insight on the limitations of the old quantum theory
when applied to a mechanical system that is nonintegrable. However, it had and in fact still it has, a trend to believe that
Newtonian Mechanics is a particular case of the Quantum Mechanics, hypothesis that still opened, although great advances
have occurred, some examples can be found in Refs. [3–29]. In the core of this question, the main aspect is the ‘‘status’’ of

∗ Tel.: +55 3137416999.

E-mail addresses: adelcio@ufsj.edu.br, adelcio.olv@gmail.com.

0378-4371/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.physa.2013.09.025
656 A.C. Oliveira / Physica A 393 (2014) 655–668

the Quantum Mechanics while the fundamental theory, at least for low energies. This is the Classical Limit of Quantum
Mechanics problem (CLQM).
In the earlier times, the CLQM was studied in terms of Ehrenfest theorem [15–17,13,14,9,25–28] which states that, un-
der certain conditions, the centroid of a wave-packet state will follow a classical trajectory. In a collection of papers [5,3,4],
Ballentine and collaborators stated that the Ehrenfest’s theorem is not adequate to characterize the classical regime, in
their own words ‘‘A quantum state may behave essentially classically, even when Ehrenfest’s theorem does not apply, if it
yields agreement with the results calculated from the Liouville equation for a classical ensemble’’. and ‘‘We have shown that
Ehrenfest’s theorem is neither necessary nor sufficient to characterize the classical regime in quantum theory’’. Those results
were later confirmed by others [22,8–10,18,19,24]. Zurek and Paz [7,21] have argued that no quantum system is isolated
from their environment, thus we must consider a bigger system that includes the environment. In opposition, Wiebe and
Ballentine have considered a coarse grained measurement [6] and concluded that ‘‘for all practical purposes, the quantum
theory of the chaotic tumbling motion of Hyperion will agree with the classical theory, even without taking into account the
effect of the environment. Decoherence aids in reducing the quantum–classical differences, but it is not correct to assert that
the environmental decoherence is the root cause of the appearance of the classical world’’. In a recent paper [19], Oliveira
and coworkers have shown that, in fact, there is a combination of factors. For a non-linear oscillator, they observed that ‘‘the
agreement between quantum and classical mechanics is achieved through the convergence of three factors: large actions,
the interaction with the environment and experimental observation limitations. Large classical actions are important for
classicality during the whole dynamics, but it is not a sufficient condition: diffusion plays an essential role due to the pres-
ence of the revival in the quantum case. Deviations between the quantum and classical dynamics are not detectable if we
take the experimental limitations into account’’. The experimental resolution and its relation with CLQM have been studied
by many authors, see Refs. [23,30,11,3,29,31,19,32]. The decision whether a system is classical or quantum depends on the
experimental apparatus which is essentially classical.
Recently, Angelo [11,29] has returned with the Einstein’s question ‘‘inevitable conception that Physics must ferment a
realistic description of only one system.’’ Indeed, ‘‘nature as a whole may be thought of as an individual system existing
only one time, with no need for repetitions and not as an ‘ensemble of systems’’ see Refs. [11,29] and references there in.
He concludes that the classical behavior only exists as an approximated notion derived from low-resolution measurements,
‘‘A scenario of quasideterminism may then be defined, within which the motion is experimentally indistinguishable from
the truly deterministic motion of Newtonian mechanics. Beyond this time scale, predictions for individual systems can be
given only statistically and, in this case, it is shown that diffusive decoherence is indeed a necessary ingredient to establish
the quantum–classical correspondence’’.
There are distinct notions of classicality [33]. In this contribution, our aim is to shed a light under the Newtonian
Classical Limit of Quantum Mechanics issue and contrast it with Liouvillian Classical Limit of Quantum Mechanics [19]. The
fundamental point is the action of the measurement apparatus on the system, in fact this is closely related with decoherence
program [31,34,25–28]. We consider a situation where the environment action is negligible and thus the system is isolated
or the environment acts like a phase reservoir [11,29,18]. The essential idea here is that if one tries to make continuous
simultaneous measurement of position and momentum (CMPM), then the information about the quantum nature of the
particle will be lost.
Our procedure is similar to Refs. [25–28], they consider a quantum system entangled with the environment or the detec-
tor, formally they treat the system as a Wiener process, and in some circumstances, the purity of the state is preserved [27].
They numerically recovered the classical trajectories of regular and chaotic systems, they argue that the measurement pro-
cess can be modeled by a stochastic master equation. According to Bhattacharya and collaborators [27] there is no reason
to believe that the particular measurement model would affect the results and ‘‘any measurement or interaction which
produces a localization in the phase space should lead to the classical behavior’’, and our results corroborate with that sup-
position. In this contribution, we investigate analytically and numerically this procedure. We show that in a CMPM the
Quantum dynamics is indistinguishable of the classical Newtonian dynamics and we also show that the classical limit of
an ensemble of Newtonian trajectories, the Liouville regime, is approximately obtained for the quartic oscillator model if
the number of measurements in the time interval is large enough to destroy the revival and small enough to not reach the
Newtonian regime. The Newtonian regime is related with a higher precision measurement in CMPM, which leads to a strong
localization and in this case, the dynamics is governed by the semiclassical approximation, see next section and Refs. [9,20,
35,36], this result is model independent.
This paper is organized as follows: in Section 2 we present the method, and an alternative view of Ref. [9]. In Section 3 we
define a semiclassical time in terms of scalar products, overlap is considered. Section 4 we analyze the Newtonian Classical
Limit of Quantum Mechanics problem. We show in an analytical way that the Newtonian regime is a natural consequence
of using the CMPM. In Section 5 we use the quartic oscillator as model. Section 6 contains conclusions.

2. The semiclassical expansion

ρ . Its time evolution is given by

We consider a general quantum density operator 

d
ih̄ ρ = −[
 ρ, 
H] (1)
dt
 A.C. Oliveira / Physica A 393 (2014) 655–668 657

where  H is the Hamiltonian. We define a semiclassical Hamiltonian,  Hsc (t ), in a such way that for an initial coherent state,
all classical1 expectation values for any observable will be precisely reproduced, then we write the total Hamiltonian  H as

Hsc (t ) + 
H =
 δ(t ), (2)
then we have
d
ih̄ ρ = −[
 ρ,  ρ ,
Hsc ] − [ δ(t )], (3)
dt
where δ(t ) =  Hsc (t ), and we considered it as a perturbation. Using the interaction picture
H −
Ď
ρI = 
 Usc (t )
ρ
Usc (t ) (4)

Usc (t ) is the semiclassical time evolution operator, then we have

where 
d
ih̄ ρI = −[
 ρI , ∆
s ]. (5)
dt
where
 s (t ) = Ď
∆ Usc (t )
δ(t )
Usc (t ), (6)

Usc (t ) is defined as
and 
∂
ih̄ Usc (t ) = 
Hsc (t )
Usc (t ). (7)
∂t
The solution of Eq. (5) can be formally written as
 t
 2  t  t1
i i
ρI (t ) = 
 ρ (0) + ρ (0), ∆
[ s1 ]dt1 + ρ (0), ∆
[[ s2 ], ∆
s1 ]dt2 dt1
h̄ 0 h̄ 0 0
 3  t  t1  t2
i
+ ρ (0), ∆s3 ], ∆
[[[ s1 ]dt3 dt2 dt1 + · · · .
s2 ], ∆
h̄ 0 0 0

Then we finally obtain

  t  2  t  t1
i i
ρ (t ) = Usc (t ) 
  ρ (0) + ρ (0), ∆s1 ]dt1 +
[  ρ (0), ∆
[[ s1 ], ∆
s1 ]dt2 dt1
h̄ 0 h̄ 0 0

 3  t  
t1  t2
i Ď
+ ρ (0), ∆
[[[ s3 ], ∆
s2 ], ∆
s1 ]dt3 dt2 dt1 + · · · 
Usc (t ) (8)
h̄ 0 0 0

where we have used the short notation ∆sk = ∆s (tk ).

Hsc (t ), for initial coherent states, the time evolution of the first order term
Now we will impose another restriction on 
in the expansion (8) is always a Gaussian. Thus if we choose an initial coherent state |α0 ⟩ and considering one degree of
freedom, we have

Usc (t ) |α0 ⟩ = eiφ(t ) |α(t )⟩

 (9)
where α(t ) obeys the classical Newtonian equation of motion, thus
d 1 ∂ Hcls
α= (10)
dt ih̄ ∂α ∗
d 1 ∂ Hcls
α∗ = − . (11)
dt ih̄ ∂α
The phase φ(α(t )) is
 t
1
φ(t ) = − L[α, α ∗ ]dt ′ (12)
h̄ 0

and

H |α0 ⟩ ,
Hcls = ⟨α0 |  (13)

1 Classical in Newtonian sense.

658 A.C. Oliveira / Physica A 393 (2014) 655–668

where L[α, α ∗ ] is the classical Lagrangian of the system. Thus, after some straightforward algebra, see (Appendix B), we
obtain
Ď
Usc (α(t )) = eiφ(t ) e−iωa at 
 D(α(t ))
D(α0 )−1 (14)
where 
D is the well known displacement operator
Ď ∗
D(y) = eya −y a ,

and

ω= k/m,

(15)
∂ 2 V (x) 
k=  ,
∂ x 2 x =0
V is the potential energy. The generalization for the N dimensional case is
N
Ď Ď Ď Ď
−iωj  aj t αj (t )
aj −αj∗ (t ) aj +αj∗
Usc (α(t )) = eiφ(t )
 aj  aj −αj aj
 e e e .
j

N
= eiφ(t )

j
Usc
 (αj (t )) (16)
j

j j
where Usc is the semiclassical evolution operator related to the k-th degree of freedom. Note that  Usc depends on αj (t ) solely,
but in general we have αj (t ) = f (α1 (t ), α2 (t ), . . . , αN (t )). If we write Hcls as a Taylor expansion, we obtain
 m
Hcls = h̄ω α ∗ α + Am,n α ∗ αn ,

m,n

where A1,1 = 0. The semiclassical Hamiltonian, for one degree of freedom [9], is
m−1 m
aĎ aĎ
 
Hsc (α(t )) = h̄ω m Am,m α ∗ α m−1 ( a − α ∗ α) + Am,n α ∗ αn
 
 a+
m̸=0 m,n
m−1 m
aĎ − α ∗ ) +
 
m Am,n α ∗ α n ( n Am,n α ∗ α n−1 (
a − α).
 
+ (17)
m̸=n m,n

The generalization for N degrees of freedom is

−−→ (1) −−→ (2) −−→
Hsc (α(t )) = Hcls + 
 Hsc (α(t )) ⊗ 
12 · · · ⊗ 
1N + 
1⊗
Hsc (α(t )) ⊗  13 + · · · ,
13 · · · ⊗  (18)

(k) −−→
where Hsc is the semiclassical Hamiltonian operator related to the k-th degree of freedom and (α(t )) = (α1 (t ), α2 (t ), . . . ,
αN (t )). In the Schrödinger picture, the time evolution of the state is
  t  2  t  
t1
1 1
|Ψ ⟩ (t ) = Usc (t ) 1 +
 s (t )dt1 +
∆ s (t1 )∆
∆ s (t2 )dt1 dt2 + · · · |Ψ ⟩ (0). (19)
ih̄ 0 ih̄ 0 0

The convergence of the method was demonstrated for the quartic oscillator [9] and for the chaotic Dickie model [35].

3. Scalar product of states

Once we are working in a Hilbert space with a Hermitian Hamiltonian, it is well known that the scalar product of any two
states which evolves with the same Hamiltonian must remain constant. This ‘‘constancy’’ can be a test of our semiclassical
approximation; moreover it teaches us something about when quantum corrections become dominant. The question we ad-
dress is the following: consider two neighboring states in the sense that ⟨Ψ |Φ ⟩ ≈ 1. We know that ⟨Ψ |Φ ⟩ (t ) = ⟨Ψ |Φ ⟩ (0),
so the important issue relies on which of the ingredients of the quantum evolution are the ones that most affect this rela-
tion. We write the states |Ψ ⟩ and |Φ ⟩ in the form (19), thusit is straightforward to obtain the semiclassical expansion for
⟨Ψ |Φ ⟩ (t ), for initial states |Ψ (0)⟩ = |α0 , β0 ⟩ and |Φ (0)⟩ = α0′ , β0′ . For short times we get


 1
 t
Ď ′  ′ Ď ′  ′
⟨Ψ |Φ ⟩ (t ) ≈ ⟨β0 , α0 |  Usc α0 , β0′ − dt1 ⟨β0 , α0 | ∆ Usc α0 , β0′
  
Usc Usc
1 
ih̄ 0

t
 
1 Ď ′   ′
dt1 ⟨β0 , α0 |  Usc ∆1 α0 , β0′
 
+ Usc (20)
ih̄ 0
 A.C. Oliveira / Physica A 393 (2014) 655–668 659

Ď Ď
where 
Usc ′
Usc Usc (t , α, β)
= Usc (t , α ′ , β ′ ), and ∆ s (t1, α, β). The first term reads
1 = ∆
   
αt − α ′ 2

 Ď ′  ′ ′ t
β0 , α0  Usc  α0 , β0 = exp − + iIm αt αt′∗ + i(φ(α ′ (t )) − φ(α(t )))
  
Usc 
2
   
βt − β ′ 2

t
+ iIm βt βt′∗ + i(φ(β ′ (t )) − φ(β(t ))) .
 
× exp − (21)
2

The last term (21) is relevant for the expansion (20) along a time (t ) where the conditions αt ≈ αt′ and βt ≈ βt′ hold. Re-
membering that αt , αt′ , βt and βt′ follow classical trajectories, then this time will be longer (almost always) in the case when
the system is regular.
Consider the time evolution of an operator  a. We can compare it with his classical analog, acl . We consider the case
where acl (0) = ⟨a⟩q (0) ̸= 0. We assume a short time dynamics, which means that we can define a reference frame where2
acl (t ) ̸= 0. Then we can write the quantum expectation value as a function of his classical analog, such as

⟨a⟩q (t ) = acl (t ) 1 + Fq (t )
 
(22)
where, Fq (t ) is the quantum correction. Writing Fq (t ) as
∞
∂ Fq (t ) 
 m 
 1
Fq (t ) = tm (23)
∂ m

m=1
m ! t t =0

then one defines the Ehrenfest’s time as

− 1n
1 ∂ n Fq (t ) 
 
TE = , (24)
∂tn

n! t =0

here n represents the first non zero term in the summation (23). The Ehrenfest’s time as described above, can be interpreted
as the time when the quantum corrections are of the same order of the classical value, i.e. the lifetime of Classical Newtonian
approach. It cannot be associated directly with the classical limit, see Ref. [19] for more details. The Ehrenfest time was
derived in several papers [9,15–17,13,11,14]. For integrable systems the Ehrenfest time has a general form
1
TE ≈ (S /h̄)δ , (25)
Ω
where S is some constant, usually related with the classical action, Ω the typical frequency and δ is a power factor. For
chaotic systems
1
TE ≈ ln(S /h̄), (26)
λ
where λ is the Lyapunov exponent. In analogy with the Ehrenfest time we define a semiclassical time in a similar way. We
write (20) as
Ď ′  ′
2 2
⟨β0 , α0 |  Usc α0 , β0′  =  β0 (t ), α0 (t )|α0′ (t ), β0′ (t ) 
  
Usc
2
=  β0 , α0 |α0′ , β0′  [Λ(t )]

(27)
then the semiclassical time is
Λ(τsc ) ≈ 1/2. (28)
The generalization of τsc to n-degrees of freedom system is straightforward.  semiclassical time (τsc ) can be interpreted
This
Ď ′  ′
as a lifetime of the approximation ⟨Ψ (t ) | Φ (t )⟩ ≈ ⟨β0 , α0 |  Usc α0 , β0′ . Now consider two classical different initial

Usc 
conditions. We define D(t ) as the distance between these trajectories in phase space. In the chaotic regime the distance
D(t ) grows as D(t ) ≈ D0 eλt , where λ corresponds to the largest short time Lyapunov exponent. This exponent is calculated
for short time series [37]. Under this assumption, we have
 
1 DM
τsc ≈ ln , (29)
λ D0
where DM = D(τsc ). See also Appendix A. If we write DM in terms of the canonically conjugate variables (x0 , p0 ), then
 2 
mω2  p(τsc ) − p′ (τsc )

2 1 2
DM = α(τsc ) − α (τsc ) = x(τsc ) − x (τsc ) + .
2
 ′ ′
  (30)
h̄ω 2 2m

2 It is always possible if the system is stable in Lagrange sense, i.e. |a (t )| ≤ C , C ∈ R for  t ∈ Dom(a ).
cl cl
660 A.C. Oliveira / Physica A 393 (2014) 655–668

 2 
Ď ′  ′ 
Fig. 1. O(t ) = ⟨α0 |  Usc α0  , for the driven oscillator, ẍ + x3 = β sin(ωt ), with β = 1. Chaotic initial conditions (dotted line), ω = 1.88 rad/s,

Usc 


(x1 , ẋ1 ) = (0, 0), (x2 , ẋ2 ) = (0.002, 0). Regular, same initial conditions (continuous line) for ω = 3.88 rad/s. X axis corresponds to time in seconds.

Here ω is defined by (20) and m is the particle mass. Choosing a reference system that x′ (t ) = 0 and p′ (t ) = 0, we can say
that h̄D2M ≈ Scl , where Scl is the classical action of the system, putting this into the Eq. (29) we have
 
1 Scl
τsc ≈ ln . (31)
λ D0 h̄
For regular regions, where we should have instead a power law separation of neighboring trajectories, D(t ) ≈ D0 + Ft γ ,
then we obtain
 1/γ
Scl
τsc ≈ . (32)
2h̄Re(D0 F )
We call the reader’s attention on the similarities between Eqs. (26) and (31), also (25) and (32). Also, from the above expres-
sion, we can conclude that the semiclassical time is obtained as a function of classical dynamics solely. We know that, in
general, classical trajectories remain closer for longer times in the case of integrable systems, then we can say that integrable
systems are more ‘‘robust’’ regarding quantum corrections, i.e., the next terms in the expansion. On the other hand, chaotic
systems will very soon need quantum corrections for an adequate description. Suppose that we have a mixed dynamics
system, e.g. Ref. [36]. In this case, we know that for the regular regions, neighboring  trajectories keep close for a longer time
Ď ′  ′
than trajectories located in the ergodic region. The product ⟨β0 , α0 |  Usc α0 , β0′ will tend more rapidly to zero when it is in

Usc 
a region where there is chaos[36,38]. Fig. 1 shows the semiclassical square modulus of the overlap between two neighboring
Ď ′  ′ 2
states, Osc (t ) = | ⟨α0 |  Usc α0 | , for the driven conservative oscillator [39]. As we can see in this figure, the behavior of

Usc 
the semiclassical overlap is strongly dependent of the classical regime, as expected. Thus we can say that the validity of the
semiclassical approximation is longer in the classical regular initial conditions. We consider a Newtonian classical limit, in
this situation, a distance between two coherent states |x⟩ and |y⟩ is D = |x − y|, which corresponds to the classical coun-
 relation D = −ln(Osc (t )),
2
terpart [40]. In the semiclassical regime it is related with the semiclassical overlap Osc  ′by the
Ď
thus the classical limit for this observable is reached when ⟨Ψ |Φ ⟩ (t ) ≈ ⟨β0 , α0 |  Usc α0 , β0 , it is known that the classical
Usc ′ ′

regime depends on the observable [10].

4. The high accurate position and moment measurement and classical limit of quantum mechanics

We are considering a quantum state, initially in a coherent state, that evolves by a time ∆t (not necessarily small). During
the time evolution ∆t we make simultaneous approximated position and momentum measurements with independent
intervention. This procedure is carried out n times, in this situation the best approach for the state after measurement is
a Gaussian state [31]. A similar procedure can be found in Refs. [25–28]. The measurement has an accuracy δ x for position
and δ p for momentum. The interaction with the environment is weak enough such that can be ignored, then we can assume
that during the time interval ∆nt the quantum state system undergoes unitary evolution between measurements and suffers
a nonunitary change at the instant of measurement. From now on, we label the unitary evolution time interval as ∆τu ,
where ∆τu = ∆nt . In the standard formalism of Quantum Mechanics the change in the state of a system produced by an
instantaneous measurement may be calculated using projection operators. In the case of position, its precise measurement
requires a strong coupling and a large amount of energy [31]. Since we are measuring the position and momentum
 A.C. Oliveira / Physica A 393 (2014) 655–668 661

simultaneously, but with independent intervention [41–44], each measurement to precision δ x produces a variance of
momentum δ p which then feeds back into the uncertainty of the position, at the time of the next measurement. A detailed
description of the measures with independent intervention can be found in Ref. [41]. This result was experimentally
confirmed by Rozema and co-workers [43], and by Erhart and co-workers [44].
Quantum Mechanics imposes a restriction on the acceptable states, they have to be normalizable and the probability
density must be positive, the position eigenstate |x⟩ and momentum eigenstate |p⟩ are not normalized and the associated
probability density are Dirac delta functions. Dirac delta function is not a function, it has to be understood in terms of
distributions, in a similar way we must deal with |x⟩ and |p⟩.
If we assume that just after measurement the state is |x⟩, and we consider an ensemble of N identical particles prepared in
the same initial state, then the result of the measurement of position for i-th particle is xi we write it in terms the ensemble
xi = r, thus xi = r + ξi , we assume that ξi is N (0, s2 ), i.e. normal random variable with variance s2 , thus the state after the
measurement is

|ϕ⟩ = |x⟩, (33)

 
S (λ) = exp − h̄i 
the overbar represents the ensemble mean. Using the position displacement operator,  pλ , and the
resolution of unity in |p⟩ bases we have
  
i
|ϕ⟩ = dpexp − pξi |p⟩ ⟨p|r ⟩ . (34)
h̄
   2 2
This ensemble mean can be easily computed [45], it is exp − h̄i pξi = exp − p2 h̄s2 , also we have ⟨p|r ⟩ = (2π h̄)−1/2
 
exp −
ipr
h̄
, them

p 2 s2
    
−1/2 ipr
|ϕ⟩ = (2π h̄) dp exp − exp − |p⟩ (35)
2 h̄2 h̄

and

p 2 s2 ip(r − x)
    
ϕ(x) = ⟨x|ϕ⟩ = (2π h̄)−1 dp exp − exp − (36)
2 h̄2 h̄

integrating (36) we obtain

(x − r )2
 
1
ϕ(x) = √ exp − . (37)
2π s2 2s2
 
T (λ) = exp
The momentum measurement can be studied in analogous way, just observing that now  λ is the momen- i
h̄
q
tum displacement operator, again we use a normal random variable ζi , and write the momentum pi as pi = d + ζi . Since ζi
is N (0, a2 ) it is straightforward to show that

(p − d)2
 
1
ϕ(p) = √ exp − , (38)
2π a2 2a2

where d is the ensemble mean of pi . The state at the time ti , just after the measurement is a Gaussian state centered in
(r , d), i.e., then the state is a Gaussian state in Wigner representation centered at (r , d). It is straightforward to show that
r = Tr [ ρ (ti−1 , ti )x] and d = Tr [
ρ (ti−1 , ti )x], where 
ρ (ti−1 , ti ) is given by

ρ (ti−1 , ti ) = 
 U (ti−1 , ti ) U Ď (ti−1 , ti )
ρ (ti−2 , ti−1 ) (39)

U (ti−1 , ti ) = exp −i H (ti − ti−1 )/h̄ . The maximum precision is achieved when δ xδ p = h̄/2, this is the standard
 
where 
quantum limit (SQL) [42,46], then after measurement the system is a bosonic coherent state. Even if one measures the state
with greater precision than SQL [43], one cannot use it as an initial state, as pointed by Ballentine [47] ‘‘the uncertain princi-
ple restricts the degree of statistical homogeneity which is possible to achieve in an ensemble of similarly prepared systems
and thus it limits the precision for any system can made’’, then we assume that the state cannot be narrower than a coherent
state. A similar approach can be found in Refs. [25–28], the main difference is that we can switch from Liouville regime to
Newtonian regime as we show in the following sections. As we show in the following sections, the Liouville regime is mim-
icked for the position observable, the semiclassical state is always a gaussian, while the classical distribution is not [10,24].
The final Gaussian state can be a consequence of dispersive interaction [48] in some specific situations.
662 A.C. Oliveira / Physica A 393 (2014) 655–668

4.1. The classical limit: Newtonian regime

We consider the above procedure with a maximum accuracy achieved, then the system state is a coherent state when
t = tk . Now we use the semiclassical expansion (8) for the time interval (tk , tk+1 ), them the quantum state is given by
  t  2  t  t1
i i
ρ (tk , t ) = Usc (α(tk ), t ) 
  ρ (tk ) + ρ (tk ), ∆
[ s1 ]dt1 + ρ (tk ), ∆
[[ s1 ], ∆
s1 ]dt2 dt1
h̄ tk h̄ tk tk

 3  t  
t1  t2
i Ď
+ ρ (tk ), ∆s3 ], ∆s2 ], ∆s1 ]dt3 dt2 dt1 + · · · 
[[[    Usc (α(tk ), t ). (40)
h̄ tk tk tk

Using the fact that tk+1 − tk = ∆t /n, where n is the number of measurements carried out during the time interval ∆t, then we
write t (τ ) = tk +τ ∆t /n, τ ∈ (0, 1) then all integration variable can be defined in a similar way, thus ti (τi ) = tk +τi ∆t /n , and
τi ∈ (0, 1) is the new integration variable, then we obtain
τ τ τ1
   2  
i i
ρ (tk , t ) = 
 Usc (α(tk ), t ) 
ρ (0) + ∆t /n ρ (tk ), ∆
[( s1 )]dτ1 + ∆t /n ρ (tk ), ∆
[[ s1 ]dτ2 dτ1
s1 ], ∆
h̄ 0 h̄ 0 0

τ τ1 τ2
 3    
i Ď
+ ∆t /n ρ (tk ), ∆s3 ], ∆s2 ], ∆s1 ]dτ3 dτ2 dτ1 + · · · 
[[[    Usc (α(tk ), t ). (41)
h̄ 0 0 0

If n ≫ 1 then
  τ 
i Ď
ρ (tk , t ) ≈ Usc (α(tk ), t ) 
  ρ (0) + ∆t /n ρ (tk ), ∆s1 ]dτ1 
[  Usc (α(tk ), t ). (42)
h̄ 0

Using the mean value theorem for integrals we obtain

 
i Ď
  ρ (0) + τ (∆t /n) [
ρ (tk , t ) ≈ Usc (α(tk ), t )  ρ (tk ), ∆s1 ] 
 Usc (α(tk ), t ), (43)
h̄
the overbar represents time mean, now if we take the limit n −→ ∞ of (41) we have
Ď
ρ (tk , t ) = 
lim  Usc (α(tk ), t )
ρ (0)
Usc (α(tk ), t )
n−→∞

ρsc (α(tk ), t )
= (44)
where ρsc (α(tk ), t ) is the semiclassical density function of a system in time t that was in a coherent state in time tk . The time
ρsc is such that all classical observables are recovered by it [9] then the classical limit is achieved for any system,
evolution of 
the unique restriction is that we should represent the potential in Laurent series. As we will see in the following, we do not
need to take limit n −→ ∞ to achieve the classical regime, in fact it is sufficient to have τsc ≫ ∆τu . We call the reader’s
attention that the limit n −→ ∞ is fundamentally different of taking h̄ −→ 0 or Sh̄ −→ 0 where Scl is the classical action,
cl
n −→ ∞ corresponds to continuous monitoring the system [34], thus the Classical Newtonian regime is a special kind of
Dynamic Quantum Zeno Effect [49]. We know that classical Newtonian systems are not continuously monitored, however,
our result means that the information about its trajectory is someway available.

5. The quartic oscillator revisited

Now we numerically investigate the results of the previous section. How large needs ‘‘n’’ to be in order to achieve the
classical regime? As a model, we consider the following classical Hamiltonian
2
p2 kx2 p2 kx2

m
Hcls = + +g + . (45)
2m 2 k 2m 2
Here p is the particle momentum, x is the position, m is the mass, k is the string constant and g is the nonlinearity constant.
This model has been considered by several authors in the context of classical limit of observables [15,16,10,9,19,11,18,24,12].
It is non-linear thus it has strong quantum features such as revival. We make a change of variables
α − α∗ α + α∗
p=  ; x=  , (46)
2mω
i m2ωh̄ h̄
√
where ω = k/m, them we write the Hcls as

Hcls = h̄ω α ∗ α + g h̄2 (α ∗ )2 α 2 . (47)

 A.C. Oliveira / Physica A 393 (2014) 655–668 663

Fig. 2. Time evolution of classical (dotted line) position and quantum (full line) position expectation values for ω = 100 g and four values of n: (A) n = 1,
(B) n = 10, (C) n = 25 and (D) n = 125. The classical initial state is q(t0 ) = 3, p(t0 ) = 0 and the quantum initial state is a coherent state with amplitude
α(t0 ) = q(t0 ) + ip(t0 ). We set with p(t0 ) = 0 and q(t0 ) = 3. X axis corresponds to gt and t0 = 0.

The classical Newtonian solution for equations of motion of the model is

α(t ) = α(t0 ) exp{−i(ω + 2g h̄ |α(t0 )|2 )(t − t0 )}, (48)

and the exact quantum solution for an initial coherent state |α(t0 )⟩ is

a⟩ (t ) = α(t0 )e−iω(t −t0 ) e−|α(t0 )|

2 [1−exp(−2i(t −t
0 )g h̄)]
⟨ , (49)
where t0 is the initial time. The Ehrenfest’s time for this model is
√
tE ∼ (h̄g |α(t0 )| 2)−1 , (50)
for more details see Refs. [9,10], note that the classical action Scl ≈ h̄ |α(t0 )| , thus tE ∝ 2 √1
Scl
. If the quantum system is
monitored as we state above, for t ∈ (tk , tk+1 ), them the quantum solution is

a⟩ (tk )e−iω(t −tk ) e−|(⟨a⟩(tk ))|

2 [1−exp(−2i(t −t
k )g h̄)]
a⟩ (t ) = ⟨
⟨ (51)
again we use t = tk + τ ∆t /n, then

a⟩ (tk )e−iωτ ∆t /n e−|(⟨a⟩(tk ))|

2 [1−exp(−2iτ g h̄∆t /n)]
a⟩ (τ ) = ⟨
⟨ . (52)
√
We know that τsc ≈ tE , then for τsc ≫ ∆τu =⇒ 2g h̄∆t /n ≪ |α(t2)| . Thus, for sufficient large |α(t0 )|, we can say that in the
0
semiclassical regime (t ≪ τsc ) we have 2g h̄∆t /n ≪ 1 then

a⟩ (τ ) ≈ ⟨
a⟩ (tk )exp −i ω + 2g h̄ |⟨
a⟩ (tk )|2 (t − tk ) .
   
⟨ (53)
As we can easily verify, (53) is the classical solution for α(t0 ) = ⟨
a⟩ (tk ) and it is worth for any k, thus we are in the classical
regime. In Fig. 2 we show the time evolution of classical position and quantum position expectation values for the quartic os-
cillator for α(0) = 3. In Fig. 2(A) the is the unitary quantum solution (n = 1, τsc /∆τu ≈ 0.5), in Fig. 2(B) n = 10, τsc /∆τu ≈
5, in Fig. 2(C) n = 50, τsc /∆τu ≈ 12 and in Fig. 2(D) n = 125, τsc /∆τu ≈ 60. As we can observe, the quantum solution
664 A.C. Oliveira / Physica A 393 (2014) 655–668

Fig. 3. Time evolution of classical (dotted line) position and quantum (full line) position expectation values for ω = 100 g and four values of n: (A) n = 1,
(B) n = 10, (C) n = 50 and (D) n = 1000. The classical initial state is q(t0 ) = 10, p(t0 ) = 0 and the quantum initial state is a coherent state with amplitude
α(t0 ) = q(t0 ) + ip(t0 ). We set with p(t0 ) = 0 and q(t0 ) = 10. X axis corresponds to gt and t0 = 0.

approaches the classical one as we increase n. In Fig. 3 we have the same as Fig. 2 for α(t0 ) = 10 for n = 1, 10, 50, 1000.
Again, the quantum solution approaches the classical one as we increase n. As observed by others [9,10,29,12,18], this model
has become ‘‘more classical’’ in Newtonian sense, for small α , and it is more classical in Liouvillian sense for large α .
This model has a collapse, it occurs at time TC = π /2g h̄, but it is not a quantum signature as pointed out by Ballentine
and co-workers [5,3,4], an ensemble of classical orbits also collapses at this time [10], the collapse is actually√effect the
enlargement of the package. The classical solution for an initial Gaussian state centered in α0 = (x0 + ip0 ) / 2, can be
easily computed [10], then the expectation value of position x is given by

x20 + p20
    
1 −iωt x0 + ip0
⟨xcle ⟩ = Re e × exp (1 − A(t )) , (54)
2 A (t )2 A(t )

where A (t ) = igt.
In Fig. 4(A) we show the time evolution of classical position expectation values. In Fig. 4(B) we have the time evolution
of quantum position expectation values for n = 1 (then we have τsc /∆τu ≈ 0.014 and TC /∆τu ≈ 0.3). In Fig. 4(C) we use
n = 5 (then we have τsc /∆τu ≈ 0.71, TC /∆τu ≈ 1.6), and in Fig. 4(D) n = 80 (then τsc /∆τu ≈ 1 and TC /∆τu ≈ 25). The
classical model is an ensemble of classical orbits and the initial state is a Gaussian centered at (10, 0) and the quantum initial
state is a coherent state centered at (10, 0). As we can see from Fig. 4, the quantum expectation value follows the classical
expectation value if τsc / ∆τu / TC . The contra-intuitive behavior observed in Fig. 4(D) is due to the fact that when we
increase n, we approach the Newtonian classical limit, in this case, there is no ‘‘damping’’ and we recover the ‘‘Newtonian
regime’’ as observed in Fig. 3.
The quartic oscillator has another important quantum time scale, the revival time, for this model it is given by TR = sπ /g h̄,
s ∈ N. In Fig. 5 we show the time evolution of classical position expectation values and the time evolution of quantum
position expectation values for α(t0 ) = 10. In Fig. 5(A) we use n = 1, then τsc /∆τu ≈ 0.014 and TC /∆τu ≈ 0.3. In Fig. 5(B)
we set n = 10 then τsc /∆τu ≈ 0.14 and TC /∆τu ≈ 3. Fig. 5(C) is the same as 5(A) and (D) is the same as 5(B), both for t in
the neighborhood of TR . As we can see in Fig. 5(D), the revival disappears when we consider CMPM, it does not mean that
CLQM, in the Liouville sense, has been achieved, in fact the quantum states is a Gaussian at time t = tk , but the classical
state is not, as we can see in Refs. [10,24].
In order to estimate the difference between the semiclassical state given by (44) and the quantum state (39) in a CMPM
we compute the fidelity. The fidelity (Fd) between two states can be defined as

Fd(ρ1 (t ), ρ2 (t )) = Tr [ρ1 (t )ρ2 (t )] . (55)

If ρ1 (t ) and ρ2 (t ) are pure states, then Fd = 1 when ρ1 (t ) = ρ2 (t ). In Fig. 6 we show the fidelity Fd between ρsc (t ) and
ρn (t ), for α(t0 ) = 3, n = 1, 50, 100, 500, as we can observe in this figure, the fidelity becomes close to unity as we increase
n, this means that the quantum state ρn (t ) ≈ ρsc (t ) for n large enough.
 A.C. Oliveira / Physica A 393 (2014) 655–668 665

Fig. 4. (A) Time evolution of classical position expectation values. (B) Time evolution of quantum position expectation values for n = 1. (C) Same as (B) for
n = 5. (D) Same as (B) for n = 80. We use, for all figures, a gaussian as initial state centered in q(t0 ) = 10, p(t0 ) = 0. X axis corresponds to gt, ω = 100 g
and t0 = 0.

6. Conclusions

We define a semiclassical time τsc , which determines the ‘‘classical’’ Newtonian regime. If the system is under contin-
uous monitoring then it is driven to classicality, the monitoring is a simultaneous approximate position and momentum
measurements with independent intervention. As a general remark, we can say that the Newtonian classical limit of Quan-
tum Mechanics is reached when τsc ≫ ∆τu , where ∆τu is the time interval between the two consecutive measurements.
We observe that, in this regime, the continuous measurement of position and momentum acts like a dynamic Zeno effect
forcing the system to behave as a classical Newtonian system, this is the quasideterminism scenario. We studied the quartic
oscillator and the numerical simulations confirm the analytical results, the Newtonian regime occurs when τsc ≫ ∆τu and
the Liouvillian regime is mimicked, for the position observable, if ∆τu ∈ [τsc , TR ]. The classical Liouvillian regime is only
reached when we include the action of the environment, then we can take into account other observables than the position,
see Refs. [12,10,24,18,19]. This result suggests that there is an intermediate regime that may occur if the system is strongly
coupled with the environment and is subjected to CMPM.

Acknowledgments

The author acknowledges, E.W. Dias, A.R. Bosco de Magalhães, and the referee, for helpful comments, suggestions and
criticism, and also acknowledges FAPEMIG for the partial financial support.

Appendix A. Semiclassical expansion and the Lyapunov exponent

The zeroth order term of the semiclassical expansion is always a coherent state or a tensor product of coherent states,
with labels described by classical dynamics. For a general state we have

⟨ψ|φ⟩ = ⟨ψ|φ⟩sc + corrections.

666 A.C. Oliveira / Physica A 393 (2014) 655–668

Fig. 5. Time evolution of classical (dotted line) position and quantum (full line) position expectation values for ω = 100 g and two values of n: (A) n = 1,
(B) n = 10, (C) Same as (A) for a different time interval. (D) Same as (B) for a different time interval. The classical and quantum initial state is gaussian
centered at point q(t0 ) = 10, p(t0 ) = 0. X axis corresponds to gt and t0 = 0.

Fig. 6. Time evolution of fidelity between the quantum state (39) from bottom to top n = 1, 50, 100, 200, 500 and the semiclassical state (44). The initial
state is a coherent state with α(t0 ) = 3. We use for ω = 100 g and X axis corresponds to gt and t0 = 0.

Consider initially that the states are |ψ(0)⟩ = i |αi (0)⟩ and |φ(
 0)⟩ = i |βi (0)⟩, where |αi (0)⟩ and |βi (0)⟩ are coherent
 
states. Then the zeroth semiclassical term is ⟨ψ(t )|φ(t )⟩sc = i ⟨αi (t )|βi (t )⟩. We know [50] that the overlap of coherent
states is

|⟨αi |βi ⟩|2 = exp(− |αi − βi |2 ) for h(3); (56)

2J
|αi − βi |2

= 1− for su(2). (57)
(1 + |βi |2 )(1 + |αi |2 )
 A.C. Oliveira / Physica A 393 (2014) 655–668 667

√
The Eq. (57) coincides with (56) if we make αi = αi / 2J and taking the limit J → ∞. We can also say that |αi − βi |2 =
(xi − yi )2 + (pxi − pyi )2 ≡ ∆q2 + ∆p2 where
αk = xk + ipxk (58)
βk = yk + ipyk . (59)
The Lyapunov exponent is defined as

1 ∆x(t )
λ = lim lim ln (60)
t →∞ ∆x(0)→0 t ∆x(0)
where ∆x(t ) = |x1 (t ) − x2 (t )|, where xi (t ) is the classical evolution for xi (0) as the initial condition. In the above limit, we get
|⟨αi (t )|βi (t )⟩|2 = exp − ∆q(0)2 e2λq t + ∆p(0)2 e2λp t .
  

As we have i λi = 0, the biggest Lyapunov exponent (λmax ) is approximately



   −1 
 ⟨αi (t )|βi (t )⟩
 
ln
1
   
i
λmax = lim lim ln   −1  . (61)
 
t →∞ βi (0)→αi (0) 2t
ln  ⟨αi (0)|βi (0)⟩
   
i

From Eq. (61) we can conclude that the Lyapunov exponent is related with the quantum nature of the system. The faster the
quantum corrections are needed, i.e. how faster the product |⟨αi (t )|βi (t )⟩| → 0, the bigger is the Lyapunov exponent. This
behavior has already been pointed by many others [16,15] using different methods.

Appendix B. Semiclassical time evolution operator

First we observe that the quantum dynamics is given by

∂
ih̄ H |Ψ ⟩ .
|Ψ ⟩ = 
∂t
U (t ) is defined by
The time evolution operator 

∂
U (t ) |α0 ⟩ =  Hsc (t , α0 ) + 
δ(t , α0 ) U (t ) |α0 ⟩
 
ih̄ (62)
∂t
H =
since  Hsc (t , α0 ) + 
δ(t , α0 ) where we choose |Ψ (0)⟩ = |α0 ⟩, thus we have
∂
U (t ) = 
Hsc (t , α0 ) + 
δ(t , α0 ) U (t ).
 
ih̄ (63)
∂t
U (t ) = 
Usc (t , α0 ) 1 + 
G(t , α0 ) then
 
We can use the Baker–Campbell–Hausdorff formula and write 

∂
U (t ) = 
Hsc (t , α0 )
Usc (t , α0 ) + 
δ(t , α0 )
Usc (t , α0 ) 1 + 
G(t , α0 )
 
ih̄ (64)
∂t
since Usc (t , α0 ) is (by definition) the time evolution operator generated by 
Hsc (t , α0 ), it cannot depend on 
G(t , α0 ) or
δ(t , α0 ), then we have

∂
ih̄ Usc (t , α0 ) = 
Hsc (t , α0 )
Usc (t , α0 ) (65)
∂t
 t  2  t  t1
1 1
Usc (t , α0 ) = 1 +
 Hsc (t1 , α0 )dt1 + +
 Hsc (t1 , α0 )
 Hsc (t2 , α0 )dt2 dt1 + · · ·
 (66)
ih̄ 0 ih̄ 0 0

an alternative deduction of (65) and (66) can be found in Ref. [9].

Whereas the coherent states are the only pure quantum states which behave classically [51], so that we have imposed
that an initial coherent state must remain a coherent state, under the semiclassical dynamics, then we can conclude that
Ď
Usc (α(t )) |α⟩ = eiφ(t ) eiΩ (t )a a |α(t )⟩

where Ω ∈ R, and |α⟩ is a coherent state, thus we have
Ď
 D(α) = eiφ(t ) eiΩ (t )a a
Usc (α(t )) D(α(t )) (67)
668 A.C. Oliveira / Physica A 393 (2014) 655–668

Ď
using that D(α) = eα a −α a into (67), we obtain
∗

Ď Ď
D(α) = eiφ(t ) eiΩ (t )a a eα(t )a −α (t )a
∗
Usc (α(t ))
 (68)
and finally
Ď Ď Ď
(α(t )) = e−iφ(t ) e+αa −α a eα (t )a−α(t )a e−iΩ (t )a a .
Ď ∗ ∗
Usc (69)

Usc (α(t )) as
Finally, we can write 
Ď
Usc (α(t )) = eiφ(t ) eiΩ (t )a a
 D(α(t ))
D−1 (α(0)) (70)
where Ω (t ) and φ(t ) are chosen such that ih̄ ∂∂t 
Usc (t ) Hsc (t , α)
=  Usc (t , α). We also have that Hcls = ⟨α0 | 
H |α0 ⟩ =
Hsc (t , α0 ) |α0 ⟩, thus
⟨α0 | 
⟨α0 |  δ(t , α0 ) |α0 ⟩ = 0.

References

[1] A. Einstein, Dtsch. Phys. Ges. Verh. 19 (1917) 82.

[2] A. Engel, A Translation of the Paper Appears, the Collected Papers of Albert Einstein, vol. 6, Trans., Princeton U. Press, Princeton, NJ, 1997, p. 434.
[3] L.E. Ballentine, S.M. McRac, Phys. Rev. A 58 (1998) 1799.
[4] L.E. Ballentine, Phys. Rev. A 63 (2001) 31.
[5] L.E. Ballentine, Y. Yang, J.P. Zibin, Phys. Rev. A 50 (1994) 2854.
[6] N. Wiebe, L.E. Ballentine, Phys. Rev. A 72 (2005) 022109.
[7] W.H. Zurek, J.P. Paz, 1996. arXiv:quant-ph/9612037v1.
[8] W.H. Zurek, Rev. Mod. Phys 75 (2003) 715.
[9] A.C. Oliveira, M.C. Nemes, K.M.Fonseca Romero, Phys. Rev. E 68 (2003) 036214.
[10] A.C. Oliveira, J.G. Peixoto de Faria, M.C. Nemes, Phys. Rev. E 73 (2006) 046207.
[11] R.M. Angelo, Phys. Rev. A. 76 (2007) 052111.
[12] R.M. Angelo, E.S. Cardoso, K. Furuya, Phys. Rev. A. 73 (2006) 062107.
[13] A. Iomim, G.M. Zaslavsky, Phys. Rev. E 63 (2001) 047203.
[14] A. Iomim, G.M. Zaslavsky, Phys. Rev. E 67 (2003) 027203.
[15] G.P. Berman, G.M. Zaslavsky, Physica A (Amsterdam) 91 (1977) 450.
[16] G.P. Berman, A.M. Iomin, G.M. Zaslavsky, Physica D 4 (1981) 113.
[17] G.P. Berman, V. Yu Rubaev, G.M. Zaslavsky, Nonlinearity 4 (1991) 543.
[18] Adélcio C. Oliveira, A.R. Bosco de Magalhães, Phys. Rev. E 80 (2009) 026204.
[19] Adélcio C. Oliveira, A.R. Bosco de Magalhães, J.G. Peixoto Faria, Physica A 391 (2012) 5082.
[20] A.C. Oliveira, J. Modern Phys. 3 (2012) 694.
[21] W.H. Zurek, Phys. Scr. T 76 (1998) 186.
[22] W.H. Zurek, Los Alamos Science 27 (2002) 86.
[23] T. Bhattacharya, S. Habib, K. Jacobs, The emergence of classical dynamics in a quantum world, Los Alamos Science 27 (2002) 110.
[24] J.G. Peixoto de Faria, Eur. Phys. J. D 42 (2007) 153.
[25] S. Ghose, B.C. Sanders, Phys. Can. 63 (2007) 173.
[26] Kurt Jacobs, Daniel A. Steck, Contemp. Phys. 47 (2006) 279.
[27] T. Bhattacharya, S. Habib, K. Jacobs, Phys. Rev. Lett. 85 (2000) 4852.
[28] T. Bhattacharya, S. Habib, K. Jacobs, Phys. Rev. A 67 (2003) 042103.
[29] R.M. Angelo, Low-resolution measurements induced classicality, 2008. arXiv:0809.4616.
[30] Kofler, J. Brukner, Phys. Rev. Lett. 99 (2007) 180403.
[31] A.M. Caves, G.J. Milburn, Phys. Rev. A 36 (1987) 5543.
[32] A.C. Oliveira, Z.T. Oliveira Jr., N.S. Correia, Complementarity and Classical Limit of Quantum Mechanics: Energy Measurement Aspects, 2013,
arXiv:1307.0528.
[33] M. Schlosshauer, Found Phys. 38 (2008) 796.
[34] L.F. Lopes Oliveira, R.B. Rossi Jr, A.R. Bosco de Magalhães, J.G. Peixoto de Faria, M.C. Nemes, Phys. Lett. A 376 (2012) 1786.
[35] M. Reis, M.C. Nemes, J.G. Peixoto de Faria, Phys. Rev. E 78 (2008) 036220.
[36] K.M. Fonseca Romero, M.C. Nemes, J.G. Peixoto de Faria, A.F.R. de Toledo Piza, Phys. Lett. A. 327 (2004) 129.
[37] X. Zeng, R. Eykholt, R.A. Pielke, Phys. Rev. Lett. 66 (1991) 3229.
[38] K.M. Fonseca Romero, Júlia E. Parreira, L.A.M. Souza, M.C. Nemes, W. Wreszinski, J. Phys. A: Math. Theor. 41 (2008) 115303.
[39] H.P.W. Gottlieb, J.C. Sprott, Phys. Lett. A. 291 (2001) 385.
[40] Zyczkowski, Slomczynski, J. Phys. A 31 (1998) 9095.
[41] M. Ozawa, Phys. Rev. A 67 (2003) 042105.
[42] R. Lynch, Phys. Rev. Lett. 54 (1985) 1599.
[43] Lee A. Rozema, Ardavan Darabi, Dylan H. Mahler, Alex Hayat, Yasaman Soudagar, Aephraim M. Steinberg, Phys. Rev. Lett. 109 (2012) 100404.
[44] J. Erhart, S. Sponar, G. Sulyok, G. Badurek, M. Ozawa, Y. Hasegawa, Nat. Phys. 8 (2012) 185–189.
[45] B. Oksendal, Stochastic Differential Equations: An Introduction with Applications, fifth ed., Springer-Verlag, Heidelberg, New York, 2000.
[46] D.M. Appleby, Internat. J. Theoret. Phys. 37 (1998) 1491.
[47] L.E. Ballentine, Rev. Modern Phys. 42 (1970) 358.
[48] J.G. Peixoto de Faria, M.C. Nemes, Phys. Rev. A 59 (1999) 3918.
[49] B. Misra, C.G. Sudarshan, J. Math. Phys. 18 (1977) 756.
[50] W. Zhang, H. Feng, R. Gilmore, Rev. Modern Phys. 62 (1990) 867.
[51] D.M. Davidovic, D. lalovic, J. Phys. A: Math. Gen. 31 (1998) 2281.