Pramana J. Phys.

(2016) 87: 27 
c Indian Academy of Sciences
DOI 10.1007/s12043-016-1230-x

On phase-space representations of quantum mechanics

using Glauber coherent states
DIGENES CAMPOS

Academia Colombiana de Ciencias Exactas, Fsicas y Naturales, ACCEFYN, Bogot, Colombia

E-mail: dcamposr@cable.net.co; dcr257@gmail.com

MS received 21 September 2015; accepted 20 October 2015; published online 16 July 2016

Abstract. A phase-space formulation of quantum mechanics is proposed by constructing two representations

(identified as pq and qp) in terms of the Glauber coherent states, in which phase-space wave functions (probability
amplitudes) play the central role, and position q and momentum p are treated on equal footing. After finding
some basic properties of the pq and qp wave functions, the quantum operators in phase-space are represented
by differential operators, and the Schrdinger equation is formulated in both pictures. Afterwards, the method
is generalized to work with the density operator by converting the quantum Liouville equation into pq and qp
equations of motion for two-point functions in phase-space. A coordinate transformation between those points
allows one to construct a cell in phase-space, whose central point can be treated as a parameter. In this way,
one gets equations of motion describing the evolution of one-point functions in phase-space. Finally, it is shown
that some quantities obtained in this paper are related in a natural way with cross-Wigner functions, which are
constructed with either the position or the momentum wave functions.

Keywords. Phase-space quantum mechanics; coherent states; Husimi function; Wigner function.

PACS Nos 03.65.w; 03.65.Ca; 05.30.d

1. Introduction . . . , qf } and p = {p1 , p2 , . . . , pf }, satisfy the canoni-

Quantum mechanics is a fundamental theory of physics
that supports different formulations [1]: Schrdinger,
Heisenberg, Feynman, . . . . Each formulation has its
own strengths and weaknesses in the sense that it
facilitates the understanding of some aspects of the
theory or the application in specific fields. In this arti-
cle, the focus is on the phase-space formulation of
quantum mechanics, which started with the early con-
tributions of Weyl (1927), Wigner (1932), Groenewold
(1946) and Moyal (1949). As position and momentum
variables, q and p, are placed on equivalent footing,
similarly to what happens in Hamilton mechanics, the
phase-space formulation is a natural bridge for compar-
ing the quantum and classical descriptions, improving
the knowledge about quantum systems and understand-
ing the corresponding classical limit.
In this paper, a quantum system with f degrees of
freedom is considered. The position and momentum
operators in the Hilbert space, denoted by q = {q1 , q2 ,
cal commutation relations [qn , pm ] = i h 1 nm , where
1 is the identity operator, h is the reduced Planck con-
stant, nm is the delta of Kronecker, and n, m=1, 2, . . . ,
f. A multi-index notation will also be used: n := (n1 ,
n2 , . . . , nf ) is an ordered set of f non-negative integers
such that |n| = n1 +n2 + +nf , n! := n1 !n2 ! . . . nf !,
z := (z1 , z2 , . . . , nf ) is a set of f complex numbers
n
and zn := z1n1 z2n2 . . . zff . The dimensionless symbols
a = 0 q +i0 p and a + = 0 q i0 p denote annihila-
tion and creation operators, respectively,
where 0 :=
1/(q0 2) and 0 := 1/(p0 2) are related with the
units of length (q0 ) and momentum (p0 ), which are
restricted by the condition q0 p0 = h. Note that there
is only one independent parameter, either q0 or p0 .
The non-Hermitian operator a satisfies the eigen-
value equation a |z =z |z , for each complex number
z=0 q + i0 p, and the normalized eigenket |z =
|0 q + i0 p  is the so-called Glauber coherent state
[2]. In terms of these states, a phase-space Cf = {z =

1
27 Page 2 of 18 Pramana J. Phys. (2016) 87: 27

0 q +i0 p, i = 1} = R2f = {(q, p)} is associated An alternative to W is the Husimi function [9]
with the quantum system, the or coherent state representation, which is equal to a
 resolution of the identity Gaussian smoothed Wigner function, e.g. [10],
operator takes the form d(z) |z  z|=1, with a
measure d(z)= f d 2f z=(2 h)f dqdp := dqp , H (q, p, t) := |z |(t) |2
where dq=dq1 dq2 . . . dqf and dp= dp1 dp2 . . . dpf ,
and for each one-dimensional variable the integration is = (2 h)f W (q ,p ,t)[M(q  q)M(p  p)]2 dq  dp  .
over the reals, i.e., between and . (3)
Regarding phase-space formulations of quantum
mechanics, the literature may be classified into two For getting (3), one needs the relation exp(zz ) =
general areas, which are briefly described in the fol- ( h)f/2 M(p)M(q), where

lowing subsections. f/2
M(q) := (2 h) w(q, p)M(p)dp
 f/2
= q0 exp( (0 q)2 ) (4)
1.1 Phase-space formulation a la Wigner
and
This research area is guided by the idea of getting a 
f/2
quantum formulation in close analogy with the struc- M(p) := (2 h) w (q, p)M(q)dq
ture of classical statistical mechanics in which (see for  f/2
example [38] and references therein) (i) to the quan- = p0 exp( (0 p)2 ) (5)
tum mechanical state at time t, described either by the are Gaussian functions ([11], eqs (54a) and (54b)), which
ket|(t)  or the density operator (t), corresponds a are related to each other by a Fourier transform pair on
phase-space function of the real-valued position and Rf . In (3), as the Wigner function may display large
momentum variables, q = (q1 , q2 , . . . , qf ) and p = oscillations and adopt negative values in some regions
(p1 , p2 , . . . , pf ), (ii) observables are represented by of phase-space, the Gaussian smoothing function [ ]2
functions in the phase-space, and (iii) equations of the plays the role of a coarse-grained description in which
Liouville-type give account of the quantum dynamics. the finest details of the Wigner function are either
The prototype of this approach is the Wigner function, smoothed over or averaged out,resulting in a non-nega-
W (q, p, t), which can be obtained from the momen- tive probability distribution H q , p , t (see e.g. [10]).
tum and position wave functions, (q, t) and (p, t),
by the transformations
1.2 Phase-space formulation a la Bargmann
    
i  1
W (q  , p  , t) = exp qp  p  p, t A more recent line of development for the formulation
h 2 of quantum mechanics on the phase-space has emerged
 
1 in the last decades, starting with the work of Bargmann
 p  + q, t dp
2 [12] in 1961. This research is guided by the idea of rep-
     resenting the quantum states |(t) by phase-space wave
i   1
= exp qp  q q, t

functions, in analogy with the Schrdingers approach
h 2
  used in constructing the position and momentum rep-
1
 q  + q, t dq, (1) resentations, see e.g. [11,1322]. It is postulated that
2 there are phase-space representations in which: (i)
the state of the system |(t)  is described by a
where means complex conjugate. In general, the complex phase-space amplitude or wave function
Wigner function is always real, but fails to act as a (q, p, t), (ii) the position and momentum operators
probability distribution because it may take negative q and p are described by differential operators Q
values, although it accurately reproduces the marginal and P of the phase-space coordinates, q and p, (iii)
distributions of position and momentum: these operators preserve the Heisenberg commutation
 relations, and (iv) the Schrdinger or the quantum
W (q, p, t)dp = |(q, t)|2 , Liouville equations can be expressed in terms of phase-
 space variables. As a matter of notation, the circumflex

W (q, p, t)dq = |(p, t)|2 . (2) () denotes operators acting on functions defined in the
phase-space.
Pramana J. Phys. (2016) 87: 27
Instead of starting with the formulation based on the
Wigner function W (q, p, t), the approach presented in
this paper constructs phase-space wave functions using
the Glauber coherent states [2,2325] and the Husimi
amplitude, z |(t) . Thus, in order to put into con-
text the present paper, some previous literature in this
area of research (a la Bargmann) will now be briefly
reviewed.
(1) Bargmann [12] introduces the analytical repre-
sentation of quantum states, in which wave func-
tions (q, t) are transformed from the position
 
Hilbert space L2 Rf onto a Hilbert subspace
F (the so-called Bargmann or FockBargmann
space) of the entire holomorphic functions on
Cf that are square integrable with respect to the
Gaussian measure dg (z) = f exp(zz )d2f z.
A quantum state (z ) = z |  = exp( 12 zz )
g(z ) is represented in F by an analytical
entire function g(z ) of the complex variable
z = 0 q i0 p, and creation and annihilation
operators play the role of multiplication and
differentiation with respect to z , respectively
(see e.g. [26]): z|a + | = z (z , t) and
z|a| = (/z ) (z , t).
(2) Torres-Vega and Frederick [13,14] postulate the
existence of a Hermitian phase-space operator
 of unknown identity such that  |  =  | ,
and choose the orthonormal eigenvectors |  =
|p, q  as the basis of the representation. This
approach is based on a complex-valued wave
function () := |(t) , and the operators
of position and momentum [14] are chosen in
the form Q(q, p) = (q + i h/p)(q, p)
and P (q, p) = (p + i h/q)(q, p), where
, , and are some real parameters such that
= 1. The application of a certain crite-
rion leads to the canonical values = = 1/2
and = = 1.
(3) Harriman [15] connects with the works of Torres-
Vega and Frederick [13,14] and studies a linear
transformation between the phase-space and the
position (momentum) space that is implemented
via integral transforms, in which the scalar prod-
uct  |(t)  between the state |(t)  and a
Gaussian wave packet (or coherent state) |(t) 
is used. He finds that the integral kernels for
the maps from the position and momentum rep-
resentations to the phase-space representation
must be a constant C = 1/( 2) times a nor-
malized Gaussian wave packet. The parameter
defines an overall scaling of the phase-space
Page 3 of 18 27
variables q and p, and for = 1/ 2, a phase-
space representation equivalent to Bargmanns
formulation is obtained. In 1997, Mller et al
[16] argued that the state-vector representation
formulated in [14] and analysed in [15] com-
prises the class of all coherent state representa-
tions for the HeisenbergWeyl group.
(4) Wlodarz [17] introduces a wave function (p,
q, t) by decomposing the Wigner distribution
function in the form W (p, q, t) = ( )(p,
q, t), where (p, q, t) belongs to the space
L2 R2 of square-integrable phase-space func-
tions, and (a b)(p, q) is the star product of
phase-space functions. He finds that the phase-
space Schrdinger equation is equivalent to the
Liouville equation for the Wigner distribution,
and that (p, q, t) describes the state of a quan-
tum system, while at the same time |(p, q, t)|2
seems to describe its classical limit (h 0).
Wlodarzs approach coincides with [14], when
neglecting operator ordering problems.
(5) The relative-state formulation proposed by Ban
[19] is obtained by enlarging the Hilbert space
(H) of a quantum system with the Hilbert space
(Hr ) of an auxiliary reference quantum system,
i.e., H = HHr . He follows [14] and concludes
that under certain conditions, the relative-state
formulation is equivalent to those obtained by
Torres-Vega and Frederick.
(6) Because the ordinary coherent states in Hilbert
space cannot satisfy the two requirements of the
phase-space representations proposed in [14]
and [15], Smith [21] follows Bans formalism
and in the augmented space H builds a com-
plete (but not overcomplete) orthonormal set
of state vectors |(r, k; s) , which is overcom-
plete in H; here (r, k) are points in phase-space
and s is a real parameter. In this formalism,
 (p, q; t) is the Weyl symbol of the projector
|  |, where | is any fixed state and Q =
q i(h/2)/p and P = pi(h/2)/q rep-
resent the phase-space position and momentum
operators ([21], see eqs (41) and (42)). These
pairs of operators (Q+ , P+ ) and (Q , P ) act
on functions in the phase-spaces associated with
H and Hr , respectively.
(7) Bracken and Watson [22] address several ques-
tions concerning the implications of the projec-
tor |  0 | introduced by Smith [21] to extend
the phase-space formalism. They use the Weyl
Wigner transform W ( ) to define phase-space
27 Page 4 of 18
wave functions (q, p, t) := (2 )1/2 W (| 
 0 |)(q, p, t), where the arbitrary normalizable
state  0 | plays the role of a window vector.
In the coordinate representation, these functions
can be written as
   pq
1 1 
(q, p, t) = 0 q + q
2 2
 
1
 q q  , t dq  , (6)
2
a result that is closely related to Gabors win-
dowed Fourier transform of the function (q, t)
with respect to the window 0 (q). When a
 1/4  2
Gaussian function 0 (x) = 2 / exp
  2 
( )/4 exp (x ) /2 is used as a
2 definitions of coherent states pq and qp, and the impo-
window, a generalized Bargmann representa-
is obtained in which, after defining z =
tion
2 (q ip/), the position and momentum
operators
by Q = (z + /z)/
are described
( 2) and P = i(/ 2)(z /z) i 2 . particular, because they allow the calculation of posi-
1.3 Purpose of this work
As noted previously, there are important contributions
related to the formulation of quantum mechanics using
phase-space wave functions. In this work, as the vector
|z  = |0 q + i0 p  is a continuous function of the
label z, or equivalently of q and p, one can associate
to the state |(t) , the joint positionmomentum wave
functions
 
1
( |q,p,t) := w (2 1) q,p 0 q + i0 p|(t)
 2
 (q, p, t) if = 0
= (7)
+ (q, p, t) if = 1,
defined on the phase-space  = {(q, p)} = R2f , and
using the auxiliary quantity
 
i
w(q, p) := exp qp . (8)
h contradict the belief that the overall phase of the
With this choice: (i) Given a coherent state |z  = |0 q
+i0 p, the introduction of a phase factor w(q/2, p)
provides a means for handling the position and momen-
tum operators q and p on equal footing. (ii) Two phase-
space wave functions can be associated with the state
|(t) , which are linked to each other by the relation
+ (q, p, t) = w(q, p) (q, p, t), and when the squared
moduli of the functions  (q, p, t) are calculated,
one gets the probability of finding the system in
a particular Glauber coherent state |z , i.e., the
Husimi function [9]. (iii) As summarized in table 1,
Pramana J. Phys. (2016) 87: 27
Table 1. Representatives of the state |(t) , and the posi-
tion and momentum operators (q and p) in the pq- and qp-
representations.
Representation Operator q Operator p State |(t) 
i h /p p i h /q  (q, p, t) 0
qp q + i h /p i h /q + (q, p, t) 1
the phase-space representation of the position and
momentum operators, q and p, does not involve the
use of parameters, unlike the approaches given in
[14,15,22].
This work is arranged as follows: In 2, the basis
of the proposed method is introduced, which includes
sition of a symmetry condition between the variables q
and p. In 3, it is shown that phase-space wave func-
tions  (q, p, t), defined in (7), offer a suitable frame-
work for the formulation of quantum mechanics, in
tion and momentum wave functions along with their
() (p, t). In 4, functions
derivatives,  () (q, t) and 
of the quantum position and momentum operators are
converted to the language of the phase-space, and in
5, the Schrdinger and Liouville equations are refor-
mulated in that picture. In 6, the relation between
the treatment used in this paper and the method of
Wigner for the formulation of quantum mechanics in
phase-space is considered. Finally, in 7, summary and
conclusions are given.
2. Basis of the proposed method
This section deals with the phase-space wave func-
tions  (q, p, t) defined by (7), whose properties are
given in 2 and 3. The wave functions  (q, p, t)
and + (q, p, t) = w(q, p) (q, p, t) only differ
in the phase factor w(q, p), a fact that seems to
wave function plays no role in the probabilistic pre-
dictions of quantum mechanics. Naturally, the phase
factors w( 12 q, p) do not change the expectation val-
ues of Hermitian operators or the Husimi function
(3), but introducing them is an original approach to
ensure an overall consistency between the phase-space,
position and momentum representations of quantum
mechanics.
With regard to relations (7), the motivations to incor-
porate the phase factors w( 12 q, p) are explained in
the following two subsections.
Pramana J. Phys. (2016) 87: 27 Page 5 of 18 27

2.1 Definitions of pq and qp coherent states the Glauber coherent state |z = D(q, p) |0 then, as
According to Glauber [2,23], each complex number the q and p operators do not commute, moving the
z = 0 q + i0 p corresponds a normalized coherent operator exp iq p/
 h to the
 right (+) or to the left ()
state |z = |0 q + i0 p given by of the operator exp ipq/h requires a compensation by
means of the phase factor w( 12 q, p).
|z = D(q, p) |0

1/2 
zn 2.2 Implementation of the q p symmetry condition
= ( h) M(p)M(q)
f/2
|n , (9)
n=0
n! From a purely conceptual point of view, one should
treat the phase-space position and momentum q and p
where |n are the eigenkets of the number operator on equal footing (briefly, q p symmetry condition),
N = a + a, and except for the implications due to the commutation
 
i  relations [qn , pm ] = i h 1 nm , i.e., i i, because
D(q, p) = D(z) := exp p q q p [pm , qn ] = i h 1 nm . To account for this fact, the
  h
1 coordinate (q  ) and momentum (p  ) representations of
= w q, p D(0, p)D(q, 0) the Glauber coherent state |z = |0 q + i0 p are taken
 2  as
1
= w q, p D(q, 0)D(0, p) (10) z (q  ) := q  | z = M(q  q)w(q  , p) exp(i/h),
2
z (p  ) := p  | z = M(p  p)w (q, p  p)
is the Weyl operator [27]. The notation emphasizes that
exp(i/h), (13)
D(q, p) depends parametrically on the points (q, p) in
a 2f -dimensional real phase-space ( = R2f ) whose where the arbitrary phase factor exp(i/h) is chosen
elements are given by all the possible position and so that exp(i/h) = w (q, p/2), i.e.,
 
momentum expectation values: w(q  , p) exp(i/h) = w q  q/2, p
q0   and
q := z|q|z = z + z ,  
2 w (q, p  p) exp(i/h) = w q, p  p/2 .
p0  
p := z|p|z = i z z . (11) Thus, after exchanging the roles of (q, p) (q  , p  )
2
(dummy variables), the scalar products linking coordi-
Now, in quantum mechanics, a well-behaved func- nate, momentum and coherent-state representations are
tion F (a, a + ) of the operators a and a + can be written given by
in many equivalent ways, for example, if F (a, a + ) q|p = p| q = (2 h)f/2 w(q, p),
admits a convergent power-series expansion, the func-
q|p , q   = w(q, p  )M(q q  ),
tion is normal (antinormal) ordered if all the operators
a + stand always at the left (right) of the operators a. q|q  , p   = w (q  , p  ) q|p , q  
However, due to the properties of the Weyl operator, = w(q q  , p  )M(q q  ),
one may also use the standard (antistandard) ordering p|p , q  = w (q  , p p  )M(p p  ),
 

for functions of the q and p operators [2830]: all p|q  , p   = w (q  , p  ) p|p  , q  

powers of the operator q (p) precede all powers of = w (q  , p)M(p p  ). (14)
the operator p (q). Thus, one can define pq and qp
coherent states by the relations
2.3 Relation with other phase-space representations

|p, q := D(0, p)D(q, 0) |0 ,

2.3.1 Bargmanns approach. Based on the expansion
if = 0
|, q, p = (12) (9) for the Glauber coherent state |z , any quantum

|q, p = D(q, 0) D(0, p) |0 , state |(t)  can be represented by the wave function


if = 1.  
1
(z , t) = z |(t)  = exp zz g(z , t),

These states differ from the Glauber coherent state |z, 2
and each other, by a phase factor, in particular |p, q = where the function
w (q, p) |q, p. From (12), one also concludes that 
(z )n
the phase factors w( 12 q, p) are a consequence of the g(z , t) =

n|(t) , (15)
properties of the Weyl operator. In fact, starting with n=0
n!
27 Page 6 of 18
is an element of the Bargmann space of the entire
functions [12]. By using the transformation (q, p)
(z, z ) given after (10), the functions  (q, p, t) of the
real variables q and p used in this work can be written
as
 
1 2
 (q, p, t)|(q,p)(z,z ) = exp [z (z ) ] 2
4
(z , t).
(16)
Then, one concludes that  (q, p, t) are not com-
plex analytic (or holomorphic) functions in z or z function of the complementary variable v (either p or q)
[31], because the normalization factor exp( 12 zz ) and
the phase factor exp( 14 [z2 (z )2 ]) depend on both
complex variables, z and z .
2.3.2 Husimi function. The expectation value of the
projector (t) = |(t)  (t)| with respect to the
Glauber coherent state |z  = |0 q + i0 p  is given by
H (q, p, t) = z|(t)|z = | (q, p, t)|2 . (17)
As commented in (3), the quantity z|(t)|z is known
as Husimi function, a name given in recognition
of the work of Husimi published in 1940 [9], two
decades before the work of Glauber [2]. The non-
holomorphic phase factors w( 12 q, p) included in
 (q, p, t) do not alter the Husimi function but
instead of the marginal property of the Wigner function
given by (2), they will provide relationships between
the phase-space wave functions  (q, p, t) and the
position and momentum Schrdinger wave functions.
2.3.3 -basis representation. In this work, instead of
considering the -basis generated by the eigenvalue
equation  |p, q  =  |p, q  of an unknown Hermitian
operator  proposed in [14], the wave functions
 (q, p, t) and + (q, p, t) given in (7) are constructed
using the eigenkets |z  of the non-Hermitian annihila-
tion operator, which are duly modified by phase factors
arising out of the physical and mathematical consid-
erations. Roughly, the present treatment is linked to
the work of Mller et al [16] in which the Weyl opera-
tor D(q, p) is applied to a normalized fiducial vector
|  and the resulting set of coherent states is used
as a basis in a state-vector representation of quan-
tum mechanics. They use the operator mapping Q =
2 q + i h/p and P = 2 p i h/q, discuss its origin
1 1
and the implications on the possibilities of doing wave
mechanics in phase-space, whereas in the present work
the resulting mappings (see table 1) are a consequence
of introducing the wave functions  (q, p, t) defined
by (7).
Pramana J. Phys. (2016) 87: 27
3. Properties of the phase-space wave functions,
 (q, p, t)
In this section, some properties of the phase-space
wave functions  (q, p, t) are deduced: (i) As shown
in eqs (20) and (23) below, the functions  (q, p, t)
are separable in the sense that they can be expressed
as the sum of an infinite number of terms, where each
one is the product of a time-dependent function of a
variable u (either q or p) only and a time-independent
only. (ii) If H (0 v ) is the Hermite polynomial of
order , and 0 v is a dimensionless variable, where
0 is either 0 or 0 , integration of  (q, p, t)
H (0 v ) over v results in the th derivative of the
position or momentum wave function at the point u .
(iii) In particular, when = 0, the position and
momentum probability amplitudes can be obtained
from the phase-space wave functions  (q, p, t). In
3.1 and 3.2, results of the preliminary investigations
of the author a few years ago have been discussed [32].
3.1 Wave function + (q, p, t), = 1
The completeness relation for the position operator q,
the expression q  |q, p given by (14) and the posi-
tionspace wave function (q, t) := q| (t) lead to
+ (q, p, t) := q, p|(t)

   
= w q  q, p M q  q (q  , t)dq  . (18)
Comparison of this equation with the relation + (q,
p, t) = w(q, p) (q, p, t) gives
 (q, p, t) := p, q|(t)

   
= w q  , p M q  q (q  , t)dq  . (19)
The transformations (18) and (19) pass from the posi-
tion wave function (q, t) to the phase-space wave
functions + (q, p, t) = w(q, p) (q, p, t) and  (q,
p, t). All these functions
  constitute a subspace of the
Hilbert space L2 Rf formed by pairs of functions,
which differ from each other by a phase factor w(q,
p) and they represent the state |(t)  in the basis of the
rephased coherent states defined by (12).
In (18), changing the integration variable to q  :=
q  q, and expanding
   the position wave function (q 

+ q, t) = exp q /q (q, t) in powers of q , one finds
 1 (m)
+ (q, p, t) = (2 h) f/2
 (q, t)Jm (p). (20)
m!
m=0
Pramana J. Phys. (2016) 87: 27 Page 7 of 18 27

Here m = (m1 , m2 , . . ., mf ) is a multi-index,  (m) gives the momentumspace wave function (p, t) :=
(q, t) := m (q, t)/q m is the mth derivative of the (0) (p, t).
space wave function at point q, In general, from (22) and (25), integration of + (q,
 p, t) and  (q, p, t) over the variables p and q,

Jm (p) = (2 h) f/2
w (q  , p)M(q  )(q  )m dq  respectively, leads to the position and momentum wave
functions, (q, t) and (q, t). Therefore, when the
 m
q0 squared moduli of these functions are calculated, one
= i Hm (0 p)M(p) (21)
2 gets the proper marginal density in the position and
momentum representations, |(q, t)|2 and |(p, t)|2 ,
are time-independent coefficients and Hm (0 p) = which is equivalent to the property of the Wigner func-
Hm1 (0 p1 )Hm2 (0 p2 ) . . . Hmf (0 pf ) is a product of tion given by (2). The formulations [14,17] do not lead
Hermite polynomials. Now, as the one-dimensional to proper marginal density in the position and momen-
Hermite polynomials are orthogonal in the range (, tum representations, except if an indirect link with the
), with respect to the weighting function exp(y 2 ), Wigner function is used (comment annotated in [17]).
(20) can be inverted, and the th derivative of the An additional feature of  (q, p, t) is their ability to
position wave function ( = 0, 1, 2, . . .) is given by generate derivatives of the wave functions (q, t) and
 f/2
(p, t) to any order .
 () (q, t) = 2p0 (iq0 2) (2 h)f/2
 After noting the similarity between (22) and (25),
H (0 p)+ (q, p, t)dp. (22) one concludes that the q p symmetry condition
introduced in the paragraph before eq. (13) is equiv-
alent to a transformation between symbols: in fact,
In particular, for = 0, H (0 p) = 1, thus when
one integrates + (q, p, t) over the momentum vari- the equations only differ from each other in the sign
able p, the positionspace wave function (q, t) := of i and in renaming the variables. Based on the
 (0) (q, t) is obtained. foregoing text, one concludes that  (q, p, t) and
+ (q, p, t) exhibit greater affinity or closeness with
either the momentum or the position wave function,
respectively.
3.2 Wave function  (q, p, t), = 0
A procedure similar to that of 3.1, using now the com- 3.3 Relation with the eigenfunctions of the harmonic
pleteness relation for the momentum operator p and the oscillator
wave function  (q, p, t) := p, q| (t), gives
Using the multi-index notation n = (n1 , n2 , . . . , nf ),

 1 (m) the position and momentum eigenfunctions of an
 (q, p, t) = (2 h)f/2 (p, t)Jm (q),(23)

m! f -dimensional harmonic oscillator can be written as
m=0
1
with coefficients n (q) = Hn (q/q0 )M(q),
 2n n!
f/2
Jm (q) = (2 h) dp  w(q, p  )M(p  )(p  )m (i)n
n (p) = Hn (p/p0 )M(p). (26)
  2n n!
p0 m
= +i Hm (0 q)M(q). (24)
2 Therefore,
recalling that 0 q = q/(q0 2) and 0 p =
p/(p0 2), and using the identity
Similarly, the th derivative of the momentum wave
function is given by ( = 0, 1, 2, . . .)   [m/2]
x 1  m!
  Hm = m/2 (1)k Hm2k (x), (27)
() (p, t) = 2q0 f/2 (+ip0 2)
 2 2 k!(m2k)!
 k=0
f/2
(2 h) H (0 q) (q, p, t)dq. (25)
the functions Jm (p) and Jm (q), given by (21) and
Again, as a particular case, one obtains that the (24), can be expressed as the following linear combina-
phase-space function  (q, p, t) integrated over q tions of the momentum and position eigenfunctions of
27 Page 8 of 18 Pramana J. Phys. (2016) 87: 27

the f -dimensional harmonic oscillators, n (p, t) and where the last equality is based on the completeness
n (q, t), respectively: relations for the eigenkets |p , and (p, t) is the
  [m/2] momentum wave function.
q0 m  m! By following a procedure similar to the foregoing
Jm (p)= + k
m2k (p) (28) text, one also finds that
2 k=0 2 k! (m2k)!

and (t)| q |(t) = (i h)

 (p, t)
() (p, t)dp
  [m/2]
p0 m  m! 
Jm (q) = + i k
= q |(q, t)|2 dq. (35)
2 k=0 2 k! (m2k)!
(1)k m2k (q). (29) for  = 0, if (t)| (t) = 1, then
In particular
Now, because the wave functions n (p, t) and |(p, t)|2 dp = |(q, t)|2 dq = 1.
n (q, t) are connected by a Fourier transforms pair, At this point, it is pertinent to note that, according to
the Fourier transforms of Jm (p) and Jm (q) are (2), the expectation values given by (34) and (35) can
given by also be written in terms of the Wigner function:


Jm (q) := (2 h)f/2
w(q, p)Jm (p)dp (t)|p |(t) = p W (q, p, t)dqdp,

(36)
  [m/2] 
q0 m  m!
= + (t)|q |(t) = q W (q, p, t)dpdq. (37)
2 k=0
2k k! (m 2k)!
m2k (q) (30)
4. Phase-space representation of quantum
and
 operators
Jm (p) := (2 h)f/2 w (q, p)Jm (q)dq
As mentioned in 1.2, quantum mechanics on phase-
  [m/2]
p0 m  m! space requires the description of position and momen-
= +i tum operators q and p by means of differential
2 2 k! (m 2k)!
k
k=0 operators Q and P that act over phase-space func-
(1)k m2k (p). (31) tions,  (q, p, t). In general, consider an operator
B(t) := B(q, p, t) that admits a power-series expan-
Note that the sums over k on the right side of (28)
sion in q and p, and ask for the representations
and (31) differ only by the sign factor (1)k , which
in terms of the qp- and pq-coherent states. Unlike
is also the case for eqs (29) and (30). Then, the Fourier
other proposal in which Q and P are parametrized
transforms of Jm (p) and Jm (q) imply that
 (e.g. [1315,1719,21,22]) in this work, no particular
f/2 assumption about the form of these operators is
(2 h) w(q, p)Jm (p)dp = Jm (q) (32)
required.
and

(2 h)f/2
w (q, p)Jm (q)dq = Jm (p). (33) 4.1 Permutation rules

Consider F(x, t) = n=0 an (t)(x x0 ) , centred at the
n

point x0 , as the convergent power series defining a

3.4 Position and momentum expectation values given function F (x). Then, given the relation
By using the completeness relation for the eigenkets  (q,p,t) = w (q,p)+ (q,p,t),
|q  and recalling that i h/q is the momentum oper-
ator p in position representation, one can see that (for one gets the identity
= 0, 1, 2, . . .)  n

(t)|p |(t) = (i h)  (q, t) () (q, t)dq i h x0 1  (q, p, t)
p
  n

= p |(p, t)|2 dp, (34) = w (q, p) q + i h

x0 1 + (q, p, t), (38)
p
Pramana J. Phys. (2016) 87: 27 Page 9 of 18 27

where n is a multi-index and 1 is the unit operator in (41a)(41c) on the left by the factor w (q, p) gives (i.e.,
the phase-space. One thus obtains for = 1)
   

F i h , t  (q, p, t) q, p|F (q, t)| = w(q, p)F i h , t  (q, p, t)
p p
   

= w (q, p)F q + i h , t + (q, p, t)
= F q +i h , t + (q, p, t). (42)
p p
  
Similarly, using the eigenkets p   of the momentum
= w (q, p)F q +i h , t w(q, p) (q, p, t). (39)

p operator p, one gets ( = 1)
  
Similarly, for a given function G(y, t) = m=0 bm (t)

q, p|G(p, t)|(t) = G i h , t + (q, p, t)
(y y0 )m , one gets q
 
 
= w(q, p)G p i h , t
G i h , t + (q, p, t) q
q
   (q, p, t)
 
= w(q, p)G p i h , t  (q, p, t)
q = w(q, p)G p i h , t
  q

= w(q, p)G p i h , t w (q, p)  (q, p, t). (43)
q
+ (q, p, t). (40) This expression can be rewritten as ( = 0)
 

In (39) and (40), the first equalities can be inter- p, q|G(p, t)|(t) = w (q, p)G i h , t
q
preted as intertwining relations between the functions
+ (q, p, t)
 (q, p, t) and + (q, p, t), whereas the second  
ones establish an equivalence between phase-space
= G p i h , t
operators. q
 (q, p, t). (44)
4.2 Functions of the operators q and p
4.2.1 Product of operators. The preceding equations can
From the power series expansion of F (x, t), one
be generalized for dealing with products of operators.
defines the Hilbert-space operator F (q, t). In this
For example, consider p, q|F (q, t)G(p, t)|(t).
way, using the eigenvalue equation q|q   = q  |q  ,
Defining |(t) := G(p, t) |(t), one writes
the completeness relation of the eigenkets |q  , the
scalar product q  |p, q = w(q  , p)M(q  q), the p, q|F (q, t)G(p, t)|(t) = p, q|F (q, t)|(t)
identity F (q  , t)w (q  , p) = F (i h/p, t)w (q  , p), = F(i h/p, t)(q, p, t),
and the relations (19) and (39), one obtains
where the relation (41c) was used. Afterwards, as

p, q|F (q, t)|(t)  (q, p, t) = p, q|G(p, t)|(t),


eq. (44) implies ( = 1)
= F(q , t)w (q ,p)M(q  q)(q  , t)dq  (41a)
  p, q|F (q, t)G(p, t)|(t)
   
=F i h , t w (q , p)M(q  q)(q , t)dq  (41b)
= F i h , t G p i h , t  (q, p, t). (45)
p p q
 

= F i h , t  (q, p, t) Performing a similar procedure, one also finds ( = 0)
p
  q, p|G(
 p, t)F (q,t)|(t)
 
= w (q, p)F q + i h , t + (q, p, t).

(41c)
p = G i h , t F q + i h , t + (q, p, t). (46)
q p
From (12), because |p, q = w (q, p)|q, p Table 1 summarizes the results obtained. The qp-
and  q, p| = w(q, p) p, q|, then multiplication of representation of the position and momentum operators
27 Page 10 of 18 Pramana J. Phys. (2016) 87: 27

corresponds to the quantization mapping introduced in p  , t), and compare the results. Then one arrives
1951 by L Van Hove in the research on the problem at the following rule linking the pq and qp
of getting a quantum formulation of a system starting representations of the operator B(q, p, t):
from its classical description. These operators were
 
also obtained in [33] via considerations based on the
unitary implementation of the Galilei group. B i h , p i h , t
p  q 
In general, for any operator
= w (q,p)B q +i h , i h , t w(q, p) .

 p q
B(q, p, t) = an (t)An (t), (51)
n=0
with real or complex valued coefficients an (t), and (b) In terms of the shorthands Q ( ) = Q ( |q, p)
operators An (t) of the form and P ( ) = P ( |q, p), one writes the commu-
tation relations [Q(0), P (0)] = i h1, [Q(1),
An (t) = F0 (q, t)G0 (p, t) F1 (q, t)G1 (p, t) Q(1)] = i h1, [P (0), P (1)] = 0 and [Q(0),
Fn (q, t)Gn (p, t), (47) Q(1)] = 0, where 1 is the unit operator in the
phase-space. Therefore, these results are consis-
and for any coherent state |, q  , p  , one finds tent with the Heisenberg uncertainty principle.
, q  , p  |B(q, p, t)|(t) (c) Comparison of table 1 with the expressions
    Q(q, p) = (q + ih/p)(q, p) and P (q,
= B (Q |q  , p  , P |q  , p  , t)
p) = (p + i h/q)(q, p) given in [14], eqs
 
 |q  , p , t (48) (2.4) and (2.5) leads to the following results: (i)
in the pq-representation, one has = 0, = 1,
with phase-space operators (compare with table 1) = 1 and = 1, and (ii) in the pq-representation,
  it follows that = 1, = 1, = 0 and = 1.
Q | q  , p  := q  + i h  , In both cases = 1, but the values of
p
  the parameters are different from the canonical
P |q  , p  := (1 )p i h  , (49) values given in [14], eqs (4.4) and (4.5).
q
where the order of the factors must be strictly preserved
as given in (47). Naturally, before the application of 5. Quantum dynamics
rule (48), the operator An (t) may be rearranged in
a particular ordering, either the antistandard or the In this section, the Schrdinger and the quantum
standard ordering. Liouville equations are written using the pq and qp
For future use, it is convenient to consider the adjoint phase-space representations given by eqs (12).
of (48), namely,
(t)|B + (q, p, t)|, q  , p  
    5.1 Schrdinger equation
=  |q  , p  , t B + (Q |q  , p  ,

  
 Consider a quantum system with Hamiltonian H (t) =
P |q , p , t). (50) H (q, p, t), represented at time t0 by the state |(t0 ) ,
In (48) and (50), the notation B ( ) and B ( ) denote with the time evolution of the state |(t)  governed by

that the phase-space operators act to the right and left, the Schrdinger equation,
respectively. For a Hermitian operator,
 B +(q, p,
 t) = d
+
B(q, p, t), one has B (Q |q , p  , P |q ,
 i h |(t)  = H (q, p, t) |(t) ,
dt
p  ), t) = B (Q ( | q  , p  ), P ( |q  , p  ), t). lim |(t)  = |(t0 ) . (52)
tt0
Some additional comments:
The solution of this equation can be written in
(a) In (48), consider in explicit form the cases = 0 terms of the time evolution operator U (t, t0 ), which
   = 1, use
and the relations |p  , q   = w transforms the initial state |(t0 )  to the state
q , p |q , p , q  , p  | = w(q  , p  ) p  ,
 |(t)  = U (t, t0 )|(t0 ) . Thus, using the notation
q  | and  (q  , p  , t) = w (q  , p  )+ (q  , (7), one has ( |q, p, t) = , q, p||U (t, t0 )|(t0 )
Pramana J. Phys. (2016) 87: 27 Page 11 of 18 27

and, for time-independent (q, p), the Schrdinger using (12). Afterwards, define the matrix elements of
equation (52) takes the form (t) in the pq and qp representations,
( |qa ,pa ,qb ,pb ,t):= |qb ,pb|(t)| | qa , pa . (57)
i h  ( | q, p, t)
t
= H (Q ( |q, p) , P ( |q, p) , t)( |q, p, t) , (53) In particular, for a pure state (t) = |(t)  (t)|,
one has

with the initial condition limtt0 ( |q, p, t) = (|qa , pa , qb , pb , t) = ( |qa , pa , t) ( |qb ,pb , t)
( |q, p, t0 ). Assuming a Hermitian Hamiltonian,
and using (50), the complex conjugate of (53) can be and, by convention, [( )] is positioned at the left
written as of ( ). This convention explains the order of the
points (qa , pa ) and (qb , pb ) adopted in the left and

i h  (| q, p, t) = H(Q ( | q, p) , right sides of (57).
t
Inserting the expression (55) for the density operator
P (| q, p) , t) ( |q, p, t) . into (57), one gets
(54)
N
 
Hereafter, in (53) and (54) the arrow below H ( ) will ( | qa , pa , qb , pb , t) = W[k] [k] ( | qa , pa , t)

be suppressed. k=1

5.2 Quantum Liouville equation
Consider a mixture of N independent states |[k] (t)
listed by the index k = 1, 2, . . . , N , where each state
[k] ( | qb , pb , t) , (58)
where ( | qa , pa , qb , pb , t) has 4f independent vari-
ables besides t, namely, qa , pa , qb and pb . Then,
one may derive the equation of motion for the den-
sity
the fact that the wave functions
is normalized to unity [k] (t)|[k] (t) = 1 and has
associated a statistical weight 0 < W[k] 1 and
N
k=1 W[k] = 1. In the Hilbert space, the density
operator describing the mixture is then defined as
N

(t) := W[k] |[k] (t)[k] (t)|,
k=1
where the square bracket [k] is used as a mnemonic
that states |[k] (t) is an element of a statistical mix-
ture. Given the initial mixture at time t0 , described by
(t0 ), the time development of (t) is ruled by the the
quantum Liouville equation,
 operator using
[k] ( |qa , pa , t) and [k] ( |qb , pb , t) satisfy the
Schrdinger equations (54) and (53), with (q, p)
(qa , pa ) and (q, p) (qb , pb ), respectively.
In accordance with (49), one associates with
the independent variables (qb , pb ) and (qa , pa ), the
(55) phase-space operators Q ( | qb , pb ), P ( | qb , pb ),
Q( | qa , pa ) and P ( | qa , pa ). Consequently, the
phase-space representations of the Liouville equation
can be written as
i h ( | qa , pa , qb , pb , t)
t  
= H Q ( |qb , pb ) , P ( |qb , pb ) , t
 
d(t) 
H Q ( |qa , pa ) , P ( |qa , pa ) , t
i h = H (q, p, t), (t) , lim (t) = (t0 ), (56)
dt tt0 ( |qa , pa , qb , pb , t) , (59)
where [H , ] is the commutator  between  the Hermi- with initial condition limtt0 ( |qb , pb , qa , pa , t) =
tian Hamiltonian H (t) = H q, p, t and the density ( |qb , pb , qa , pa , t0 ).
operator (t) = U (t, t0 )(t0 )U 1 (t, t0 ). At this point, using the expression for  ( | q, p, t)
given in (7), one may write


5.2.1 A generalized (qa , pa , qb , pb ) representation. [k] ( |qa , pa , t) [k] ( |qb , pb , t)
In order to get the phase-space representations of the
Liouville equation (56), consider any pair of phase-
space points, (qa , pa ) and (qb , pb ), the complex vari-
ables za = 0 qa + i0 pa and zb = 0 qb + i0 pb ,
the Glauber coherent states |za  and |zb  and the cor-
responding pq and qp coherent states constructed by

= F za |[k] (t) zb |[k] (t).
Except by the phase factor F = exp (i(2 1)(qb pb
qa pa )/(2h)), the kth summand in (58) is the product
of the probability amplitudes of finding simul-
taneously the system in the Glauber coherent states
27 Page 12 of 18 Pramana J. Phys. (2016) 87: 27

|za  = |0 qa + i0 pa  and |zb  = |0 qb + i0 pb , when the points on the border induce an uncertainty equiva-
the system is in the pure state |[k] (t). Thus, ( | qa , lent to qp = ( 12 h)f in the volume of the cell.
pa , qb , pb , t) measures the weighted contributions of Now, consider the set {(qa , pa ), (qb , pb )} of all
the pure states taking part in the mixture (58). points (qa , pa ) and (qb , pb ) having the same middle
point (q  , p  ), so that this point can be treated as a
parameter. The transformations (60) imply that
5.2.2 Parametric representation of Liouville equation.
1 1
Let me start this subsection with figure 1. Given any /qb = /q, /pb = /p,
two points (qa , pa ) and (qb , pb ) in the phase-space, 2 2
one can introduce the transformation 1 1
/qa = /q, /pa = /p,
2 2
1 1
qb = q  + q, pb = p  + p, and operators involved in the Liouville equation (59)
2 2 can be rewritten as
1 1
qa = q  q, pa = p  p. (60) Qb (, q  |q, p) := Q( |qb , pb ),
2 2
Pb (, p  |q, p) := P ( | qb , pb ),
From (60), one finds new phase-space points (q  , p  )
Qa (, q  |q, p) := Q( |qa , pa )
and (q, p) given by
1 1 and
q = (qb + qa ) , p = (pb + pa ) ,
2 2 Pa (, p  |q, p):=P (|qa , pa ).
q = qb qa , p = pb pa , (61)
Therefore, with the help of (60) and (49), it follows
which can be reinterpreted geometrically: the com- that
ponents of (q  , p  ) are formed by the mean position 1
Qb (, q  |q, p) = q  + Q( |q, p),
and momentum of the points (qb , pb ) and (qa , pa ), 2
whereas (q, p) measures the separation between those 1
Pb (, p  |q, p) = (1 )p  + P ( |q, p),
points along the position and momentum axes, respec- 2
tively. Thus, one may contemplate (61) as defining   1
Qa (, q |q, p) = q Q( |q, p),

a 2f -dimensional cell in phase-space, which is cen- 2
tred at the point (q  , p ), and has edge lengths q and 1
Pa (, p |q, p) = (1 )p P ( |q, p).

(62)
p. However, when the system is in a coherent state, 2
the position and momentum uncertainties are given Based on the above, one can define a function
by q = (q0 / 2)f and p = (p0 / 2)f , respec-
(, q  , p  |q, p, t)
tively. Therefore, the coherent states associated with  
 1  1  1  1
:= |q q, p p, q + q, p + p, t
p
2 2 2 2
 1   

p + 2 p, q + 2 q|(t)|p 2 p, q 12 q
1 1
p
if = 0
(qb , pb)
= (63)
q zb
q  + 12 q, p  + 12 p|(t)|q  12 q, p  12 p


if = 1,
p p
which exhibits a parametric dependence on and (q  ,
p  ) and, apart from t, has 2f independent variables:
(qa , pa)
q and p. Thus, from the point of view of an observer
za q located at the point (q  , p  ), the Liouville equation (59)
q q can be rewritten as

Figure 1. The phase-space points (qb , pb ) and (qa , pa )
i h (, q  , p  |q, p, t)
defined by (60) are located on a 2f -dimensional hypercell t
represented in the graphics by the box with the central point
= [Hb (, q  , p  |q, p, t) Ha (, q  , p  |q, p, t)]
(q  , p ) and the edges q and p. The role of an arbitrary
phase-space point (q  , p ) will be explicit in (71) and (72). (, q  , p  |q, p, t), (64)
Pramana J. Phys. (2016) 87: 27 Page 13 of 18 27

with initial condition limtt0 (, q  , p  |q, p, t) = parametric role of (q  , p  ), and the fact that the pq
(, q  , p  |q, p, t0 ), and operators and qp representations are selected by choosing = 0
  and = 1, respectively. (ii) Equation (63) provides
Hb (, q  , p  |q, p, t) 1
:= H q  Q( |q, p), information on the allowed functions for solving the
Ha (, q  , p  |q, p, t) 2 
quantum Liouville equation (64), which has no ana-
1
(1 )p  P ( |q, p), t . (65) logues in the Bopps and Kubos formulations. Thus,
2 the present treatment provides a novel method for the
These operators are associated with the Hamiltonian description of quantum dynamics in the phase-space.
of the system, and the phase-space points (qb , pb )
and (qa , pa ), respectively. Both points contribute to 5.3 On the matrix elements ( |qa , pa , qb , pb , t)
the dependence of (, q  , p  |q, p, t) on the vari- and the Husimi function
ables q and p, as seen by writting (63) in explicit
form: In order to gain insight about the functions ( |qa ,
pa , qb , pb , t) and (, q  , p  |q, p, t), defined by (57)
 
1  1  1  1 and (63), consider a pure state (t) = |(t)  (t)|
| q q, p p, q + q, p + p, t and note that (57) reduces to ( |qa , pa , qb , pb , t) =
2 2 2 2
N    [( |qa , pa , t)] ( |qb , pb , t). Then, in the general
 1  1
= W[k] [k,a] | q q, p p, t case of a mixed state, using (23) and (20) one arrives at
2 2 the expansions
k=1
 
1 1  
[k,b] | q  + q, p  + p, t . (66) 1
2 2 (0|qa , pa , qb , pb , t) = (2 h)f
! !
=0 =0
The subscripts have been complemented, [k] [k, a] N 
and [k] [k, b], in order to trace the genealogy of   
W[k]  () (pa , t)  () (pb , t) J (qa )J (qb )
the functionss arguments due to the transforms (60). [k] [k]
Thus, similar to what was noted with respect to (59), k=1
the action of Hb (, q  , p  |q, p, t) and Ha (, q  , (67)
p  |q, p, t) on ( |q  12 q, p  12 p, q  + 12 q, p  +
1 and
2 p, t), in (66), is restricted to the functions denoted
by the subindices [k, b] and [k, a], respectively.  
1
(1|qa , pa , qb , pb , t) = (2 h) f
! !
=0 =0
5.2.3 Some comments. Bopp in 1961 [34] interpreted N 
the quantum formalism as an extension of classical   
W[k] [k] (qa , t) [k] (qb , t) J (pa ) J (pb) ,
() ()
statistical mechanics by describing the disturbance
the observer causes on a system as an ensemble in k=1
phase-space, i.e., as a cloud of points in phase-space (68)
with a density represented by a distribution func-
tion (q, p, t). Similarly, Kubo in 1964 [35] used where J ( ) and J ( ) are reference functions
Wigner representation to study the motion of elec- associated with the f -dimensional harmonic oscillator
trons in a magnetic field. In these works, opera- (3.3), whereas the contributions of the specific sys-
tors denoted by Bopp [34] as Q = q + 12 i h/p, tem under study are the ones enclosed within the square
brackets.
P = p 12 i h/q, Q = q 12 i h/p and P =
The Husimi function at the point (q  , p  ) can be
p + 12 i h/q allowed to write the equation of obtained as a particular case from (67) and (68), when
motion for (q, p, t) in the form i h d(q, p, t)/dt = the points (qa , pa ) and (qb , pb ) are arbitrarily close. In
[H (Q, P ) H (Q , P )](q, p, t). fact, from (60), (57) and (63), it follows that
Despite the apparent similarity between the results
of Bopp and Kubo and the ones obtained in this paper,  
lim ( |qa,pa,qb,pb,t) = lim ,q ,p  |q,p,t
one notes that: (i) Qb (, q  |q, p), Pb (, p  |q, p), q 0 q 0
Q a (, q  |q, p) and Pa (, p  |q, p) given in (62) dif- p0 p0

fer from Bopps operators. This can be seen by the = H (q  , p  , t). (69)
27 Page 14 of 18 Pramana J. Phys. (2016) 87: 27

Finally, because the expectation value of an observable 6.1 Structure of W ( |q  , p  , t), when = 0
B(t) is given by B(t) = Tr(B(t)(t)), and the trace
of any operator may be expressed as an integral over From (67), (63) and (60), one arrives at
 
the pq or the qp coherent states, then 1  
 W (0|q , p , t) = q , p , t
1   ! !
B(Q( ), P( ), t)H q  , p  , t
=0 =0
B(t) =   
(2 h)f
J q , p (74)
dq  dp  , (70) with (see Appendix A)
where Q( ) and P ( ) are given by (49), with the N
 
substitution (q, p) (q  , p  ).  
(q , p , t) := W[k] w(q  , p)
k=1
    
6. Relationship between pq and qp 1 1
 () p  p, t () p  + p, t dp

representations, and Wigner function [k] [k]
2 2
      
The purpose of this section is to connect the treat- i i   1
= + w (q, p ) q q

ment in this paper with the Wigner function. After h h 2
recalling the transformations (60) and (63), one may    
 1  1  1
multiply both sides of (67) and (68) by w (q, p  )w(q  , q + q q + q | (t) |q q dq (75)
p) and integrate the resulting expressions over the f - 2 2 2
and
dimensional variables p and q. Because the point (q  ,    
p  ) is a parameter, it is convenient to define a dou-   1
J (q , p ) :=
  
w q, p J q q

ble Fourier transform (p  , q  ) (q, p) through the 2
relation (figure 1)  
 1
1 J q  + q dq
   
W(, q , p |q , p , t) := w(q  , p)w (q, p  ) 2
(2 h)f    
  1
(, q  , p  |q, p, t)dqdp, = w(q , p) J p p
2
(71)  
1
whose inverse (p  , q  ) (q, p) is given by J p  + p dp. (76)
 2
  1
(, q , p |q, p, t) = w (q  , p)w(q, p  ) For getting the last equality, one uses (32) to express
(2 h)f J (q  12 q) and J (q  + 12 q) as Fourier transforms of
W (, q , p |q , p , t)dq dp  . J (y) and J (Y ), performs the integration over q and
(72) reorganizes the resulting integral over y, introducing a
As a particular situation, in (71) and (72), a coa- change of variable of the form 12 p = p y.
lescence of (q  , p  ) with (q  , p  ) can happen, so that Using the expression

W (, q  , p  |q  , p  , t) reduces to
(x x  ) = (2 h)1 w(x x  , y)dy
W ( |q  , p  , t) := W (, q  , p  |q  , p  , t)
 for the integral representation of the Dirac delta, it
1
= w(q  , p)w (q, p  ) follows that
(2 h)f 
1       N  

(, q  , p  |q, p, t)dqdp. (73) q , p , t dq = W [k] () p  , t

[k]
(2 h)f
k=1
In this case, W ( | q  , p  , t) depends on (q  , p  ) as a (p  , t)
()
[k] (77)
blend of two contributions, the first one is associated
with the role of (q  , p  ) as parameter, whereas the sec- and
ond one arises from the Fourier transform (p  , q  )   
1   
  q  +
(q, p). In what follows, it will be noted that the func- q , p , t dp = (+i)

tion W ( | q  , p  , t) is directly related to the Wigner (2 h)f h

function. q | (t)|q  . (78)

Pramana J. Phys. (2016) 87: 27 Page 15 of 18 27

Similarly, one finds that Similarly, from (83), the following properties of
 J (q  , p  ) can be derived:
1  
J q  , p  dp  = [J (q  )] J (q  ) (79)
(2 h)f 
1
and J (q  , p  )dq  = [J (p  )] J (p  ) (86)
 (2 h)f
1
J (q  , p  )dq  = [J (p  )] J (p  ). (80)
(2 h)f and

6.2 Structure of W ( | q  , p  , t), when = 1 1
J (q  , p  )dp  = [J (q  )] J (q  ). (87)
(2 h)f
Similarly, from (68), (63) and (60), the function
W ( 1| q  , p  , t) can be written as


1     6.3 Reorganization of W ( |q  , p  , t)
 
W(1|q ,p ,t) = q ,p ,t J q ,p  ,
! !
=0 =0 From (12), one notes that |p, q = w (q, p) |q, p
(81) and, therefore, the matrix elements involved in (57) are
with (see Appendix A) related as
N
    
  ()  1
 pb , qb |(t)|pa , qa  = w (qb , pb )w(qa , pa )
(q ,p ,t) := W[k] w (q,p ) [k] q q,t

2 qb , pb |(t)|qa , pa . (88)
k=1  
1
[k] q  + q, t dq
()
Consequently, using (60) and (63) one may write
2
      
i i 1
= + w(q , p) p  p (0, q  , p  |q, p, t) = w (q, p  )w (q  , p)
h h 2 (1, q  , p  |q, p, t).
   (89)
1 1
p + p p  + p| (t) |p
 2 2 Therefore, the integrands of W ( | q  , p  , t) in the
1 left hand sides of (74) and (81) can be writ-
p dp (82)
2 ten as w (q, 2p  )(1, q  , p  |q, p, t) and w(2q  , p)
and (0, q  , p  |q, p, t), respectively. Thus, inserting (68)
    and (67) into the integrals W ( |q  , p  , t), one finds
 1 

J (q , p ) := w q , p J p p
  
2 

  1
1 W(0|q ,p ,t) = (q ,2p ,t)J (0, p  )
J p  + p dp ! !
2 =0 =0
    (90)
1
= w (q, p  ) J q  q
  2
1 and
J q  + q dq. (83)
2 

1
W(1|q ,p ,t) = (2q , p , t)J (q ,0).
Equation (82) implies that ! !
=0 =0
 N

(91)
1    () 
(q , p , t)dp = W[k] [k] (q , t)
(2 h)f
k=1 A comparison between (74) and (90), and between
[k] (q  , t)
()
(84) (81) and (91), shows a change in the value of the
and functions arguments together with an interchange in
  the role of the functions, namely: (i) for = 0,
1    p  + (q  , p  , t) (q  , 2p  , t) and J (q  , p  )
(q , p , t)dq = (i)
(2 h)f h J (0, p  ) and (ii) for = 1, (q  , p  , t)
p  | (t)|p . (85) (2q  , p  , t) and J (q  , p  ) J (q  , 0).
27 Page 16 of 18 Pramana J. Phys. (2016) 87: 27

6.4 Some comments on the above defined functions pure state has an associated weight W[k] , the function
(q  , p  , t) given in (75) may be interpreted as a
At this point one may conclude weighted measure of the interference pattern gener-
ated at the point (q  , p ) and time t by the cumula-
(1) In (75) and (82), when = = 0, the functions
tive contributions of all N pure states in the mixture
00 (q  , p  , t) and 00 (q  , p  , t) coincide with
(t). Using (92), one can calculate the average on the
the standard definition of the Wigner function
statistical mixture of states, namely:
W (q  , p  , t), given by (1) in the case of a pure
state, |(t) . N
  
(2) Comparing (76) and (83), it follows that the [k] p  + p  , t |2
W[k] |
change of the pq- to the qp-representation (i.e., k=1
N
from = 0 to = 1) is associated with a 

(p  )+   
change of the roles of the functions Jm (q) and = () p  , t ]
W[k] [ [k]
! !
Jm (q), and Jm (p) and Jm (p), and vice versa. =0 =0 k=1
(3) The functions Jm (q), Jm (q), Jm (p) and Jm (p)
have a universal nature, in the sense that (p  , t) .

()
(93)
[k]
they are independent of the particular quan-
tum system under study, except by the number
The sum inside the parenthesis is nothing but the right
of degrees of freedom, f. These functions allow
side of expression (77), i.e., the product of the probabil-
the calculation of J (q  , p  ) and J (q  , p  ),
ity amplitudes for observing the system at time t with
which escort the cross-Wigner transforms,
momentum p , due to the components  () (p  , t) and
(q  , p  , t) and (q  , p  , t), in the expan- [k]
sions (74), (81), (90) and (91).  () (p  , t) of the momentum wave function  [k] (p  +
[k]
(4) The general transforms (75) and (82) belong to p , t), averaged over the statistical mixture of states.
the class of the so-called cross-Wigner trans- In complete analogy to the above process with the
forms [36], in this case for all the pairs momentum wave function one deals with the posi-
( () , 
[k]
() ) and ( () ,  () ) induced by the
[k] [k] [k]
tion wave function [k] (q  +q  12 q, t), the cross-
pure states involved in the density operator (t). Wigner function q  , p  , t given by (82) and the
expression (84).
According to [36], the physical interpretation of the
cross-Wigner transform of a pair (, ) is that of an
interference term in the Wigner distribution of the sum 7. Summary and conclusion
+. From this point of view, in order to get a glimpse
of the meaning of the cross-Wigner transforms (75) As a general conclusion, the phase-space formulation
and (82), let us consider the kth pure state in (t), the of quantum mechanics based on the Glauber coherent
Taylor series expansion states constitutes an appropriate and consistent method
     of quantum theory, and it is able to bridge the formu-

1 (p  ) ()  1 lations a la Bargmann and a la Wigner summarized in
 [k] p +p p, t =
 
[k] p p, t ,
2 ! 2 1. When the system is in a pure state then, in simi-
=0
(92) larity to the position and momentum wave functions of
the Schrdinger picture, the proposed method uses the
and refer to  (p  1 p, t) as the th component
()
probability amplitudes (phase-space wave functions)
[k] 2
of  [k] (p  + p  1 p, t) around the point p  . These  (q, p, t) defined by (7):
2
components remain unmodified when the phase-space
(1) The functions  (q, p, t) are closely related to
point (q  , p  ) moves up or down along the momen-
those of the Bargmann formulation of quantum
tum axis p (see figure 1). A similar feature exhibits the
mechanics, but they are not elements of the
components of the position wave function [k] (q  +
Bargmann space of the entire analytic functions.
q  12 q, t), when (q  , p  ) is moved left or right along
(2) When a suitable integration is carried out,
the position axis.
the functions  (q, p, t) give the position
When the components  () (p  1 p, t) and
[k] 2 and momentum probability amplitudes together
() 
[k] (p + 2 p, t) overlap, at a given time t, they give
1
with the derivatives of the wave functions,
rise to an interference phenomenon and, because each () (p, t).
 () (q, t) and 
Pramana J. Phys. (2016) 87: 27 Page 17 of 18 27

(3) As a consequence of working with the wave Appendix A. Proof of (75) and (82)
functions  (q, p, t), it is possible to get the
mapping of quantum operators in the pq and qp To demonstrate the second equality in (75), consider
representations (table 1). a pure state described by the momentum wave func-

tion (p, t). The Fourier transform of the th partial
(4) The functions ( | q, p, t) allow the formula-
derivative of (p, t) with respect to p is denoted as
tion of the Schrdinger equation in the phase-
 (q, t) and given by 
space in terms of the pq( = 0) and qp( = 1)
 (q, t) := (2 h)f/2 w(q, p) () (p, t) dp
representations. The resulting equations describe
the evolution of probability amplitudes in a form  
having high similarity with the usual coordi- i
= q (q, t), (A.1)
nate and momentum representations in quantum h
mechanics. which is simply related to the position wave function
(q, t). Using the inverse Fourier transform of (A.1),
When the system is in a mixed state described by the the functions  () (p  1 p, t) and  () (p  + 1 p, t)
2 2
density operator (t), the method is easily generalized are written as Fourier transforms of  (x, t) and
in such a way that the state of the system is represented  (X, t), and inserted in (75), in the definition of
by the function ( | qa , pa , qb , pb , t) defined in (58), (q  , p  , t). One performs the integration over p
which depends on two sets of variables, (qa , pa ) and by invoking the integral representation of the Dirac
(qb , pb ), unlike the standard position representation in delta function, and reorganizes the resulting integral
which the matrix element q  |(t)|q   relies on two over x by introducing a change of variable of the form

2 q = q x (see figure 1). Afterwards, the second
1
position variables. In this case,
equality in (A.1) is used, and the resulting expression
is reorganized by means of the identity 
(1) The quantum Liouville equation is formulated      
 1  1  1 
in terms of the pq and qp representations,  q q, t  q + q, t = q + q  (t)
eq. (59), in terms of a two-point function in 2 2 2
  
phase-space, namely: (qa , pa ) and (qb , pb ).   1

(t) q q .
(2) By using the transforms (60), the state of the 2
system can be represented by a reduced func- (A.2)
tion (, q  , p  |q, p, t), defined in (63), which Finally, by considering all the pure states involved in
is a one-point function in phase-space: (q, p) the density operator (t), given by (55), one obtains
are independent variables and (q  , p  ) is a para- the second equality in (75).
metric point. A similar procedure allows to demonstrate the sec-
(3) The quantum Liouville equation is reformulated ond equality in (82). Instead of (A.1), one now uses
in terms of the reduced function (, q  , p  | the expression 
q, p, t), eq. (64).
 (p, t) := (2 h) f/2
w (q, p) () (q, t) dq
(4) The relationship between the method in this  
i
work and the Wigner function is shown by using = + p (p, t) (A.3)
the fact that the pq and qp phase-space represen- h
and the identity
tations of the density operator are given by novel       
 1  1  1 

expansions involving the position and momen-
 p p, t  p + p, t = p + p  (t)

tum wave functions and their derivatives, as seen 2 2 2
  
in expansions (67)(68) and in 6. From these   1
expansions, we can see that the Wigner func- (t) p p .
2
tion arises in a very natural way by means of the (A.4)
double Fourier transformation described in 6.

References
Acknowledgements
[1] D F Styer, M S Balkin, K M Becker, M R Burns, C E Dudley,
The author is very grateful to the researchers whose S T Forth, J S Gaumer, M A Kramer, D C Oertel, L H Park,
constructive comments have helped in improving the M T Rinkoski, C T Smith and T D Wotherspoon, Am. J. Phys.
theoretical formulation presented in this article. 70(3), 288 (2002), http://dx.doi.org/10.1119/1.1445404
27 Page 18 of 18
[2] R J Glauber, Phys. Rev. 131(6), 2766 (1963), http://dx.doi.
org/10.1103/PhysRev.131.2766
[3] M Hillery, R F OConnell, M O Scully and E P Wigner,
Phys. Rep. 106(3), 121 (1984), http://dx.doi.org/10.1016/
0370-1573(84)90160-1
[4] Y S Kim and M E Noz, Phase space picture of quan-
tum mechanics. Group theoretical approach, Lecture notes in
physics (World Scientific, Singapore, 1991) Vol. 40
[5] D Dragoman, Phys. Scr. 72(4), 290 (2005), http://dx.doi.
org/10.1238/Physica.Regular.072a00290
[6] C K Zachos, D B Fairlie and T L Curtright (Eds), Quantum
mechanics in phase space. An overview with selected papers,
World scientific series in 20th century physics (World
Scientific Publishing Co., 2005) Vol. 34, http://www.world
scientific.com/worldscibooks/10.1142/5287
[7] T L Curtright, D B Fairlie and C K Zachos, A concise trea-
tise on quantum mechanics in phase space (World Scientific
Publishing Co. and Imperial College Press, 2014), ISBN 978-
98114520430
[8] W P Schleich, Quantum optics in phase space (Wiley-VCH
Verlag Berlin, 2001), ISBN: 978-3-527-63500-9
[9] K Husimi, Proceedings of the Physico-Mathematical Society of
Japan, 3rd Series 22(4), 264 (1940), http://goo.gl/yDbV3c
[10] K Takahashi, Prog. Theor. Phys. Suppl. 98, 109 (1989),
http://dx.doi.org/10.1143/PTPS.98.109
[11] D Campos, J D Urbina and C Viviescas, J. Phys. A: Math.
Gen. 33, 6129 (2000), http://iopscience.iop.org/0305-4470/
33/35/304
[12] V Bargmann, Commun. Pure Appl. Math. 14(3), 187 (1961),
http://dx.doi.org/10.1002/cpa.3160140303
[13] G Torres-Vega and J H Frederick, J. Chem. Phys. 93(12), 8862
(1990)
[14] G Torres-Vega and J H Frederick, J. Chem. Phys. 93(4), 3103
(1993)
[15] J E Harriman, J. Chem. Phys. 100(5), 3651 (1994),
http://dx.doi.org/10.1063/1.466353
[16] K B Mller, T G Jrgensen and G Torres-Vega, J. Chem. Phys.
106(17), 7228 (1997), http://dx.doi.org/10.1063/1.473684
[17] J J Wlodarz, J. Chem. Phys. 100(10), 7476 (1994),
http://dx.doi.org/10.1063/1.466891
[18] Q-S Li and X-G Hu, Phys. Scr. 51(4), 417 (1995),
http://dx.doi.org/10.1088/0031-8949/51/4/001
[19] M Ban, J. Math. Phys. 39(4), 1744 (1998), http://dx.doi.org/
10.1063/1.532262
Pramana J. Phys. (2016) 87: 27
[20] M A Gosson, J. Phys. A 38, 9263 (2005), http://stacks.iop.
org/0305-4470/38/i=42/a=007
[21] T B Smith, J. Phys. A 39(6), 1469 (2006), http://dx.doi.org/
10.1088/0305-4470/39/6/019
[22] A J Bracken and P Watson, J. Phys. A 43(39), 395304 (2010),
http://dx.doi.org/10.1088/1751-8113/43/39/395304
[23] W H Louisell, Quantum statistical properties of radia-
tion (John Wiley & Sons, New York, 1973), ISBN-10:
0471547859
[24] W-M Zhang, D H Fen and R Gilmore, Rev. Mod. Phys.
62(4), 867 (1990), http://dx.doi.org/10.1103/RevModPhys.
62.867
[25] N Mukunda, Pramana J. Phys. 56(23), 245 (2001),
http://dx.doi.org/10.1007/s12043-001-0122-9
[26] A D Ribeiro, F Parisio and M A M de Aguiar, J. Phys.
A 42(10), 105301 (2009), http://dx.doi.org/10.1088/1751-
8113/42/10/105301
[27] K E Cahill and R J Glauber, Phys. Rev. 177(5), 1857 (1969),
http://dx.doi.org/10.1103/PhysRev.177.1857
[28] J G Kirkwood, Phys. Rev. 44(1), 31 (1933), http://dx.doi.
org/10.1103/PhysRev.44.31
[29] N L Balazs and B K Jennings, Phys. Rep. 104(6), 347 (1984),
http://dx.doi.org/10.1016/0370-1573(84)90151-0
[30] H-W Lee, Phys. Rep. 259(3), 147 (1995), http://dx.doi.org/
10.1016/0370-1573(95)00007-4
[31] B Tsirelson, Analytic function, Encyclopedia of Mathematics,
http://goo.gl/Uc6iOZ
[32] D Campos, Interconexon entre las funciones de onda en rep-
resentacon de coordenadas, impulsos y estados coherentes
(1996), http://www.bdigital.unal.edu.co/35028/1/35260-
138008-1-PB.pdf
[33] L C Papaloucas, Prog. Theor. Phys. 95(1), 37 (1995),
Ptp.oxfordjournals.org/content/95/1/37.full.pdf
[34] F Bopp, Werner Heisenberg und die Physik unserer
Zeit (Vieweg+Teubner Verlag, 1961), chap.: Statistische
Mechanik bei Strung des Zustands eines phyikalis-
chen System durch die Beobachtung, pp. 128149,
http://dx.doi.org/10.1007/978-3-663-05439-9
[35] R Kubo, J. Phys. Soc. Jpn 19(11), 2127 (1964),
http://doi.org/10.1143/kJPSJ.19.2127
[36] M A de Gosson and S M Gosson, Phys. Lett. A 376(4), 293
(2012), http://dx.doi.org/10.1016/j.physleta.2011.11.007
Copyright of Pramana: Journal of Physics is the property of Springer Science & Business
Media B.V. and its content may not be copied or emailed to multiple sites or posted to a
listserv without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.