a

r
X
i
v
:
1
3
1
2
.
6
5
7
1
v
1


[
m
a
t
h
-
p
h
]


2
3

D
e
c

2
0
1
3
FIAN-TD-2012-02
J. Math. Phys. 54, 073517 (2013)
ALGEBRAS WITH CONVERGENT STAR PRODUCTS AND THEIR
REPRESENTATIONS IN HILBERT SPACES
M. A. SOLOVIEV
Abstract. We study star product algebras of analytic functions for which the power
series dening the products converge absolutely. Such algebras arise naturally in
deformation quantization theory and in noncommutative quantum eld theory. We
consider dierent star products in a unifying way and present results on the structure
and basic properties of these algebras, which are useful for applications. Special
attention is given to the Hilbert space representation of the algebras and to the exact
description of their corresponding operator algebras.
1. Introduction
In this paper, we study star product algebras of analytic functions for which the
power series dening the products are absolutely convergent. We discuss the structure
and some basic properties of these algebras, consider their representations in Hilbert
spaces, and describe precisely the corresponding operator algebras. The question of
convergence of star products was previously considered in the context of deformation
quantization and specically for the Weyl-Moyal and Wick products [1][4]. It acquires
a new signicance in noncommutative quantum eld theory (NC QFT), because taking
the nonlocal nature of the star products into account is important for the physical
interpretation of noncommutative models [5][8]. As shown in [6], a causality condition
for NC QFT can be formulated through the use of a star product algebra associated
with the light cone and having the property of absolute convergence, and some simple
models satisfy this condition. Since in the literature there is no consensus on the
equivalence of the Weyl-Moyal and Wick-Voros formulations of noncommutative eld
theory [9][11], it is useful to dene and explore those properties of the convergent star
product algebras that are independent of the specic form of the star product.
We consider dierent quantizations in a unifying way, starting from a general form
of star products dened by a constant matrix whose antisymmetric part is the Poisson
tensor and whose symmetric part species the quantization. In the case of the Weyl-
Moyal product the symmetric part is zero. An important role in our approach is played
by function spaces that are algebras under each of these products and consist of rapidly
decreasing functions. As well known, the usual axiomatic quantum eld theory uses
the Schwartz space S of all rapidly decreasing smooth functions, on which quantum
elds are dened as operator-valued distributions. The space S is an algebra under the
Weyl-Moyal product but not under the Wick product, for which the symmetric part
is pure imaginary diagonal. Moreover, any test function space having the structure
2000 Mathematics Subject Classication. 53D55, 81S30, 46L65, 46H15, 46H30, 46N50.
Key words and phrases. deformation quantization, Moyal star product, twisted convolution, Weyl
correspondence, Wick ordering, Fock-Bargmann representation.
1
2 M. A. SOLOVIEV
of an algebra under the Wick product must consist of entire analytic functions. This
suggests that any attempt to develop a Wick-Voros formulation of noncommutative
QFT should be based on treating it as an essentially nonlocal theory, which can also
be conrmed by the simple but instructive example of the normal ordered star product
square of a free eld, considered for the Weyl-Moyal and Wick products in [8] and [12].
We believe that using appropriate function algebras consisting of rapidly decreasing
functions and universal for dierent star products is very useful in the general theory of
quantization on symplectic linear spaces. In particular, their operator representations
associated with a representation of the translation group are dened in an obvious way
and can be extended by continuity to larger algebras. In [3], a faithful representation
of a Wick star product algebra with absolute convergence is dened by means of the
Gelfand-Naimark-Segal construction. We construct Hilbert space representations of
various star product algebras explicitly and characterize the domain of denition of the
corresponding operator algebras.
The paper is organized as follows. In Sec. 2, we introduce the main denitions and
notation. In Sec. 3, we consider the spaces c

consisting of entire functions of order

2 and of minimal type, and prove that these spaces are topological algebras under
any star product

dened by a constant matrix . In Sec. 4, we show that the algebra

(c

) contains a family of subalgebras associated naturally with cone-shaped regions.

If > 1, then (c

) has a subalgebra W

consisting of functions rapidly decreasing

at real innity. These are just the functions that can serve as test functions for NC
QFT, and we explain the relation of W

with the Gelfand-Shilov spaces S

used in the
previous papers [6], [8], and [12]. In Sec. 5, we describe the Fourier transform of the
algebra (W

), which is used later in Sec. 6 to construct Hilbert space representations

of this algebra. In Sec. 7, we prove that the representations of (W

) can be uniquely
extended by continuity to representations of the algebra (c

). We also characterize
exactly the domain of denition of the corresponding operator algebra in the space
L
2
of square integrable functions. Analogous results for the Fock-Bargmann space
representation are derived in Sec. 8. The nal Sec. 9 contains concluding remarks.
2. Star products on symplectic linear spaces
Let X be be a real 2n-dimensional vector space equipped with a symplectic bilinear
form . Choosing a symplectic basis in X, we identify the form with its matrix
=
_
0 I
n
I
n
0
_
and we denote by the inverse matrix of . Then the Poisson bracket
of two functions on X is written as
f, g =
2n

j,k=1

jk
f
x
j
g
x
k
.
The Weyl-Moyal star product f

g is a noncommutative deformation of the ordinary

(pointwise) product of the functions in the direction of the Poisson bracket and is
dened by
(f

g)(x) = f(x) e
i
2

j

jk

k
g(x), (1)
where the Planck constant plays the role of a noncommutativity parameter, and where
the summation convention for the repeated indices is used. In this paper, we consider
ALGEBRAS WITH CONVERGENT STAR PRODUCTS AND THEIR REPRESENTATIONS 3
star products of the more general form
(f

g)(x) = f(x) e
i

j

jk

k
g(x) =
= f(x)g(x) +

m=1
(i)
m
m!
_

jk

x
j
x
k
_
m
f(x)g(x

x=x

, (2)
where is a constant and, in general, complex matrix. By the classical-quantum corres-
pondence principle, the antisymmetric part of must be equal to
1
2
. Then, and only
then, the product (2) satises the limit relation
lim
0
1
i
(f

g g

f) = f, g.
Therefore, in what follows we assume that
= Q+
1
2
, =
1
, Q
jk
= Q
kj
. (3)
Dierent choices of the symmetric matrix Q correspond to dierent quantizations, i.e.,
to dierent rules of association between functions on X and quantum operators, includ-
ing the standard, the anti-standard, the normal, and the anti-normal ordering rules.
This is demonstrated below by constructing explicitly Hilbert space representations
which convert the star products to the operator product. In Sec. 35 we write

instead of

, assuming that the deformation parameter is included in the matrix .

We note that the general form (2) of star product was used in recent papers by
Blaszak and Doma nski [13] and Omori et al. [14], but the attention of these papers was
focused on other aspects. The rst of them presents the formalism of the phase space
quantum mechanics in a very systematic way, but does not consider problems with
convergence of the innite series representing the star products, whereas the second
paper focuses mainly on the important case of the -exponential functions of quadratic
forms and their related Cliord algebra and does not concern the representation of star
product algebras in Hilbert spaces.
3. Convergence of the star products
We will consider spaces of functions for which the star product (2) is absolutely
convergent for any matrix . Such algebras were previously studied [1], [2] mostly for
the case of Weyl-Moyal product. The generalization to arbitrary star products in not
dicult, but we give it for completeness in the form convenient for deriving our main
results in Secs. 48.
Let 0 < < and let c

(R
d
) denote the space of all smooth functions on R
d
satisfying the inequalities
[

f(x)[ C
L
L
||
(!)
11/
e
|x/L|

, (4)
where L > 0 and can be chosen arbitrarily large, C
L
is a constant depending on f and
L, [x[ = max
1jd
[x
j
[, and = (
1
, . . . ,
d
) Z
d
+
. Here and hereafter, we use the standard
multi-index notation

=

||
x

1
1
. . . x

d
d
, [[ =
1
+ +
d
, ! =
1
! . . .
d
!.
The topology of c

(R
d
) is dened by the family of norms
|f|
L
= sup
x,
L
||
[

f(x)[
(!)
11/
e
|x/L|

, L > 0. (5)
4 M. A. SOLOVIEV
We note that ! in (4), (5) can be replaced with

1
1
. . .

d
d
, because by Stirlings
formula we have
e
||

! C

(1 +)
||
e
||

for any > 0. (6)

However the denition in form (4), (5) is more convenient for what follows.
The denition of c

(R
d
) can be reformulated in terms of complex variables. Let
E

(C
d
) be the space of all entire functions of order and minimal type, i.e. satisfying
the condition
[f(z)[ C

e
|z|

, (7)
where > 0 and can be chosen arbitrarily small, z = x + iy, and [z[ = max
1jd
[z
j
[. The
natural topology on E

(C
d
) is dened by the countable system of norms
|f|

= sup
z
[f(z)[e
|z|

, = 1,
1
2
,
1
3
, . . . . (8)
and is hence metrizable. It follows from the elementary properties of analytic functions
of several variables that E

(C
d
) is complete and therefore a Frechet space.
Proposition 1. The space E

(C
d
) is canonically isomorphic to c

(R
d
).
Proof. The Taylor series expansion of f c

(R
d
) about a point x R
d
has an innite
radius of convergence and denes an analytic continuation of f to C
d
which satises (7).
Indeed, using the rst of inequalities (6) and choosing L

< L, we obtain
[f(x +iy)[

[y

[
!
[

f(x)[ |f|
L
e
|x/L|

[y

[
L
||
(!)
1/

|f|
L
e
|x/L|

sup

[y

[e
||/
L

||
(

)
1/

_
L

L
_
||
C
L,L
|f|
L
e
|x/L|

+(d/)|y/L

C
L,L
|f|
L
e
(1+d/)|z/L

. (9)
Therefore, f(x +iy) E

(C
d
) and |f|

C
,L
|f|
L
if (1 +d/)L

< , and thus the

corresponding map from c

(R
d
) to E

(C
d
) is continuous.
Conversely, let f E

(C
d
) and let D
r
= C
d
: [
j
[ r
j
, j = 1, . . . , d, where all
r
j
s are positive. It follows from the Cauchy inequality that
[

f(z)[
!
r

sup
Dr
[f(z )[
!
r

|f|

e
|2z|

+|2r|

, (10)
where [r[

= max
j
r

j
r

j
. A simple computation gives
inf
r
e

j
r

j
r
k
=
(e)
||/
(

)
1/
. (11)
Combining (10) and (11), we obtain
[

f(z)[ |f|

2
||
(e)
||/
!
(

)
1/
e
|2z|

. (12)
Using the second of inequalities (6), this can also be written as
[

f(z)[ C

|f|

2
||
(

)
||/
(k!)
11/
e
|2z|

for any

> . (13)
We infer that the restriction f[
R
d of the function to the real space belongs to c

(R
d
)
and satises
|f[
R
d|
L
C

,L
|f|

for L <
1
2
1/
min
_
1,
1

1/
_
,
which completes the proof.
ALGEBRAS WITH CONVERGENT STAR PRODUCTS AND THEIR REPRESENTATIONS 5
Remark 1. We prefer to say that E

(C
d
) is isomorphic to c

(R
d
) instead of saying
that these spaces coincide, because the elements of c

(R
d
) can also be continued as
anti-analytic functions.
We now characterize the innite order dierential operators that are endomorphisms
of c

(R
d
).
Theorem 1. Let > 1 and

= /( 1). If G(s) =

is an entire function of
order

and nite type, then the dierential operator G() =

maps c

(R
d
)
continuously into itself and the series

f converges absolutely in every norm of

this space for any f c

(R
d
). If 0 < 1, then analogous statements hold for each
operator of the form G(), where G(s) is an entire function.
Proof. In deriving (12), we have already reproduced the well-known estimate of the
Taylor coecients of entire functions with a given order of growth. When applied to a
function G(s) of order

and type < a, a similar estimate shows that

[c

[ =
[

G(0)[
!
C
(a

)
||/

(!)
1/

, (14)
where C is a positive constant. If 1/

= 11/ and

, then it follows from (13)

and (14) that

| c

f|

[c

[ sup
z
[

f(z)[e

|z|

|f|

2
||
(

)
||/
(a

)
||/

,
where

> and can be taken arbitrarily close to . Because can be chosen arbitrarily
small, we conclude that the series

f converges absolutely in E

(C
d
) c

(R
d
)
and the operator G() is continuous in the topology of this space. If 1, then it
suces to take into account that the Taylor coecients of any entire function satisfy
the inequality [c

[ C

||
with arbitrarily small > 0. Combining this inequality
with (13), we see that in this case, any operator of the form G(), with G(s) an entire
function, maps c

(R
d
) continuously into itself. The theorem is proved.
Proposition 2. The set of polynomials is dense in E

(C
d
).
Proof. We show that the Taylor series expansion of f E

(C
d
) converges absolutely
in E

(C
d
). Let

> 2

d. Using the inequality d[z[

j
[z
j
[

together with (11) and

(12), we get

_
_
_
_
z

f(0)
_
_
_
_

f(0)[
!
sup
z
[z

[e
(

/d)

j
|z
j
|

|f|

2
||
(d/

)
||/
C
,
|f|

.
Because the topology of E

(C
d
) is ner than the topology of simple convergence, we
conclude that the Taylor series converges absolutely in the topology of E

(C
d
) to the
function f. The proposition is proved.
Theorem 2. If 2, then for any matrix , the space c

(R
d
) is an associative
unital topological algebra under the star product f

g. The series dening this product

converges absolutely in c

(R
d
).
Proof. According to the denition (2), the product f

g is obtained by applying the

operator exp(i
jk

j

k
) to the function (f g)(x, x

) c

(R
2d
) and then restricting
to the diagonal x = x

. The entire function exp(i

jk
s
j
s

k
) is of order 2 and nite type.
From the isomorphism c

(R
d
) E

(C
d
), we immediately infer that the restriction map
6 M. A. SOLOVIEV
c

(R
2d
) c

(R
d
) is continuous. Therefore Theorem 1 implies that the series dening
f

g is absolutely convergent. The bilinear map (f, g) f g from c

(R
d
) c

(R
d
)
to c

(R
2d
) is obviously jointly continuous. Therefore the bilinear map (f, g) f

g
is also jointly continuous. To prove the associativity of this product we use the identity

x
f(x, x) = (
x
+
x
)

f(x, x

x=x

, (15)
which holds for any f c

(R
2d
). With the help of (15) we obtain
(f

(g

h))(x) = e
i
jk

x
j
(
x
k
+
x
k
)
f(x)
_
e
i
jk

x
j

x
k
g(x

)h(x

)
_

x=x

=x

(16)
((f

g)

h)(x) = e
i
jk
(
x
j
+
x
j
)
x
k
_
e
i
jk

x
j

x
k
f(x)g(x

)
_
h(x

x=x

=x

(17)
The right-hand sides of (16) and (17) coincide by Theorem 1 applied to the operator
exp(i
jk

x
j (
x
k +
x
k) +i
jk

x
j
x
k) acting to c

(R
3d
). Clearly, the function f(x)
1 is the unit element of the algebra (c

(R
d
),

). The theorem is proved.

4. Subalgebras associated with cone-shaped regions
If > 1, then the space c

(R
d
) has a nontrivial subspace W

(R
d
) consisting of
functions rapidly decreasing at real innity. More precisely, W

(R
d
) consists of smooth
functions on R
d
satisfying the condition
[

f(x)[ C
N,L
L
||
(!)
11/
(1 +[x[)
N
for all L > 0, N = 0, 1, 2, . . . . (18)
Correspondingly, its topology of is dened by the system of norms
|f|
N,L
= sup
x,
L
||
(!)
(11/)
(1 +[x[)
N
[

f(x)[, (19)
and under this topology W

(R
d
) is a Frechet space. By reasoning similar to the proof
of Preposition 1, this space is canonically isomorphic to the space of entire functions
satisfying the inequalities
(1 +[x[)
N
[f(x +iy)[ C
N,
e
|y|

, > 0, N = 0, 1, 2, . . . , (20)
and the system of norms (19) is equivalent to the system
|f|
N,
= sup
x,y
(1 +[x[)
N
[f(x +iy)[e
|y|

. (21)
A similar space can be associated with each open cone V R
d
. We dene W

(V ) to
be the space of all entire functions with the nite norms
|f|
V,N,
= sup
x,y
(1 +[x[)
N
[f(x +iy)[e

V
(x)|y|

, (22)
where
V
(x) = inf
V
[x [ is the distance of x from V . We note that
V
(tx) =
V
(x)
for all t > 0, because cones are invariant under dilations. Clearly, if V
1
V
2
, then
W

(V
2
) is canonically embedded in W

(V
1
). The space W

(R
d
) is embedded in every
W

(V ) and all of these spaces are embedded in c

(R
d
). If K is a closed cone and a
functional v (W

(R
d
))

has a continuous extension to each W

(V ), where V K0,
then the cone K can be thought of as a carrier of v, because this property implies that
v decreases in all directions outside K more rapidly than any test function.
Theorem 3. If 1 2, then for any open cone V R
d
, the space W

(V ) is an
associative topological algebra under the star product f

g and the series dening this

product converges absolutely in W

(V ).
ALGEBRAS WITH CONVERGENT STAR PRODUCTS AND THEIR REPRESENTATIONS 7
Proof. Let f, g W

(V ). It is easy to see that f g W

(V V ) and
|f g|
V V,N,2
|f|
V,N,
|g|
V,N,
,
if [y[ in (22) is dened by [y[ = max
j
[y
j
[. It is clear that if h(z, z

) W

(V V ), then
the function

h(z) = h(z, z) belongs to W

(V ) and |

h|
V,N,
|h|
V V,N,
. Therefore,
it suces to prove an analogue of Theorem 1 for W

(V ). To this end we derive an

estimate similar to (13). Using Cauchys inequality and the elementary inequality
(1 +[x[) (1 +[[)(1 +[x [), we obtain
(1 +[x[)
N
[

f(z)[
!
r

sup
=+iDr
(1 +[x[)
N
[f(z )[

!
r

(1 +[r[)
N
|f|
V,N,
e

V
(2x)+|2y|

sup
=+iDr
e
|2|

+|2|

C
N,

!
r

|f|
V,N,
e

V
(2x)+|2y|

+2

j
(2r
j
)

.
Therefore, (13) is replaced by
(1 +[x[)
N
[

f(z)[ C
N,
|f|
V,N,
2
||
(2

)
||/
(k!)
11/
e

V
(2x)+|2y|

, (23)
which holds for any

> . We now can state that if G(s) =

is an entire function
of order

= /(1) and nite type, then the dierential operator G() =

maps W

(V ) continuously into itself and the series

f converges absolutely in
this space. Indeed, if

> 2

, then it follows from (14) and (23) that

| c

f|
V,N,

[c

[ sup
z=x+iy
(1 +[x[)
N
[

f(z)[e

V
(x)

|y|

|f|
V,N,

2
||
(2

)
||/
(a

)
||/

,
where

> and can be taken arbitrarily close to . In particular, the operator

exp(i
jk

j

k
) is a continuous automorphism of W

(V V ), and so the theorem

is proved.
Remark 2. The family of subalgebras W
2
(V ) of the maximal algebra c
2
(R
d
) with the
absolute convergence property plays a special role. In [6] and [12], we used the notation
S
1/2
instead of W
2
, because this space can be regarded as the projective limit of the
Gelfand-Shilov spaces S
1/2,b
as b 0. The notation W
2
is here more convenient and
agrees with the notation in [15], where the isomorphism between S
,b
and W
1/(1),b
is
established for < 1. The spaces W

b0
W
,b
should be distinguished from the
spaces W

b
W
,b
considered by Gelfand and Shilov [15]. The space S
1/2
= W
2
was proposed in [6] as a universal test function space for noncommutative quantum
eld theory. It was argued there that the spaces S
1/2
(V ) associated with cones can
be used to formulate a generalized causality condition and shown that some simple
noncommutative models satisfy this condition.
5. Deformed convolution product
We now turn to describing the Fourier transform of the algebra (W

). As in the
case [15] of the spaces W

, the behavior of the Fourier transforms of the elements of

W

is characterized by an indicator function which is the Young dual of y

.
8 M. A. SOLOVIEV
Let, as before, > 1 and let

be the dual exponent of , i.e., 1/

+ 1/ = 1. We
dene W

(R
d
) to be the space of all smooth functions on R
d
with the nite norms
|g|
N,a
= max
||N
sup

g())[ e
a||

, a > 0, N = 0, 1, 2, . . . . (24)
In what follows we use the notation x =

j

j
x
j
, t =

j

j

jk
t
j
. The Fourier
transform is dened by

f() = (2)
d/2
_
e
ix
f(x)dx.
Theorem 4. If

2, then for any complex matrix , the space W

(R
d
) is an asso-
ciative topological algebra under the operation g
1
, g
2
g
1

g
2
, where
(g
1

g
2
)() =
1
(2)
d/2
_
g
1
( )g
2
()e
i()
d =
1
(2)
d/2
_
g
1
()g
2
( )e
i()
d. (25)
The Fourier transformation is an isomorphism of (W

(R
d
),

) onto (W

(R
d
),

),
where

= ( 1).
Proof. We note that the bilinear operation dened by (25) is a noncommutative defor-
mation of the ordinary convolution (up to a factor (2)
d/2
). In the case when is a real
skewsymmetric matrix, it is called the twisted convolution, and this case corresponds
to the Weyl-Moyal star product. For the properties of the twisted convolution product
and its relation to the Weyl-Heisenberg group, we refer the reader to [16] and [17]. Let
us now show that the Laplace transformation
g f(x +iy) = (2)
d/2
_
e
i(x+iy)
g()d (26)
maps W

(R
d
) onto W

(R
d
). Under the condition

2, the integral in (26) is well

dened and analytic in C
d
, and we have the chain of inequalities
(1 +[x[)
N
[f(x +iy)[ C
N
max
|N
[x

f(x +iy)[ C

N
max
|N
_
[e
i(x+iy)

g()[d
C

N
|g|
N,(d+1)a

_
e
y(d+1)a

||

d C
N,a
|g|
N,(d+1)a
sup

e
ya

j
|
j
|

.
A standard calculation gives
sup

_
_
y a

j
[
j
[

_
_
= a

j
[y
j
[

ad[y[

,
where
(a

(a)

= 1. (27)
We infer that f W

(R
d
) and |f|
N,
C
N,a
|g|
N,(d+1)a
for ad. Conversely, let f
belong to W

(R
d
). Then the integral (2)
d/2
_
e
i(x+iy)
f(x + iy) dx is independent
of y. Denoting it by g() and assuming that [[ N, we have
[

g()[ (2)
d/2
inf
y
_
e
y
[(x +iy)

f(x +iy)[dx
(2)
d/2
inf
y
_
e
y
(1 +[y[)
N
(1 +[x[)
N
[f(x +iy)[dx
C
N,a
|f|
N+d+1,
inf
y
e
y+a

j
|y
j
|

C
N,a
|f|
N+d+1,
e
a

||

,
ALGEBRAS WITH CONVERGENT STAR PRODUCTS AND THEIR REPRESENTATIONS 9
where a > can be taken arbitrarily close to , and a

and

are dened by (27). We

see that g W

(R
d
) and the map f g =

f from W

(R
d
) to W

(R
d
) is continuous.
The Fourier transformation converts the operator exp(i
jk

j

k
) into the multi-
plication by the function expi
j

jk

k
which is obviously a pointwise multiplier of
W

(R
2d
) for

2. Let h(, ) be an arbitrary function in W

(R
2d
). The formula
1
(2)
d/2
_ _
1
(2)
d/2
_
h( , )d
_
e
ix
d =
1
(2)
d
_
h(, )e
ix+ix

dd

x=x

shows that the operation of restriction to the diagonal is converted by the Fourier
transformation into the operation
W

(R
2d
) W

(R
d
): h(, )
1
(2)
d/2
_
h( , )d.
Consequently,

f
1

f
2
=

f
1

f
2
, (28)
which completes the proof.
6. Operator representations of star products
Let T be a unitary projective representation of the translation group of a symplectic
space (X, ) and let exp
i
2
be its multiplier, so that we have the relation
T
x
T
x
= e
i
2
(x,x

)
T
x+x
. (29)
We let H denote the Hilbert space on which this representation acts. The Weyl quan-
tization associates Hilbert space operators with functions on X R
2n
in the following
way
f f =
1
(2)
n
_

f(s) T
s
ds, (30)
where

f is the symplectic Fourier transform of f, dened by

f(s) =
1
(2)
n
_
f(x)e
i(x,s)
dx. (31)
Letting, as before, denote the matrix of the symplectic form , we have (x, s) =
x s = x
j

jk
s
k
. Then

f(s) =

f(s), and after the change of variables = s,
the formula (28) can be written as

f
1

f
2
=

f
1

f
2
, where

=
t
= . (32)
The integral in (30) is well dened if

f is integrable, and the Weyl transform takes the
star product (1) into the operator product.
Let now the matrix be of the form given by (3), and consequently,

=

Q
1
2
, where

Q = Q. (33)
Then the Weyl correspondence (30) should be replaced by
f f =
1
(2)
n
_

f(s) e
i
2
s

Qs
T
s
ds. (34)
Theorem 5. If 1 2, then for any of the form (3), the map (34) is a contin-
uous homomorphism of the algebra (W

(R
2n
),

) into the algebra B(H) of bounded

operators on the Hilbert space H.
10 M. A. SOLOVIEV
Proof. It follows from Theorem 4 and from the invariance of W

(R
2n
) under linear
change of variables that the symplectic Fourier transformation is a topological and
algebraic isomorphism of (W

(R
2n
),

) onto (W

(R
2n
),

). Because

2, we
have
|f|
1
(2)
n
_
[

f(s)[e

2
|

Q||s|
2
ds C
a
|

f|
0,a
,
where a > (/2)[

Q[. It remains to show that the map

W

(R
2n
) B(H):

f f =
1
(2)
n
_

f(s) e
i
2
s

Qs
T
s
ds
is an algebra homomorphism. Using (29), (33), the symmetry of the matrix

Q, and the
Fubini theorem, we obtain
f =
1
(2)
n
_
(

f
1

f
2
)(s) e
i
2
s

Qs
T
s
ds =
=
1
(2)
2n
__

f
1
(t)

f
2
(s t)e
i t

(st)+
i
2
s

Qs
T
s
dtds =
=
1
(2)
2n
__

f
1
(t)

f
2
(t

)e
i t

+
i
2
(t+t

Q(t+t

)
T
(t+t

)
dtdt

=
=
1
(2)
2n
__

f
1
(t)

f
2
(t

)e
i
2
t

Qt+
i
2
t

Qt

T
t
T
t
dtdt

= f
1
f
2
.
The theorem is proved.
In quantum mechanics on phase space, the following notation is usually used: x =
(p
1
, . . . , p
n
; q
1
, . . . , q
n
), where q
j
are the coordinate variables and p
j
are their conjugate
momentums. The standard projective representation of the phase space translations in
the Hilbert space L
2
(R
n
) is written as
T
p,q
= e
i

(pqqp)
, (35)
where q
j
is the operator of multiplication by the jth coordinate function and
i

p
j
is
the operator of dierentiation with respect to the same coordinate. Let s = (u, v) and

Q =
1
2
_
0 I
n
I
n
0
_
. Then s

Qs = u v and the correspondence (34) takes the form
f f =
1
(2)
n
__

f(u, v)e
i

2
uv+i(uqvp)
dudv =
1
(2)
n
__

f(u, v)e
iuq
e
ivp
dudv, (36)
which denes the so-called qp-quantization, i.e., the standard ordering of the operators.
We let j
st
denote the map dened by (36). In the theory of pseudodierential operators,
it is usually called the Kohn-Nirenberg correspondence. Under this correspondence the
monomial p

transforms into the operator q

. Formally, this follows from the

relations

= (2)
n
(i)
||
i
|

(v)

(u),

u
e
iuq
e
ivp
= (i)
||
i
|
q

e
iuq
p

e
ivp
,
but a rigorous proof is based on the extension of map (36) to tempered distributions [18]
An alternative proof, based on extending it to the space c
2
, is given in Sec. 7. The star
product corresponding to the standard ordering is written as
(f
st
g)(p, q) = f(p, q) e
i

p
j

q
j
g(p, q). (37)
ALGEBRAS WITH CONVERGENT STAR PRODUCTS AND THEIR REPRESENTATIONS 11
Let us point out some other important cases. The matrix

Q =
1
2
_
0 I
n
I
n
0
_
denes
the correspondence
f f =
1
(2)
n
__

f(u, v)e
ivp
e
iuq
dudv,
under which the monomial p

transforms to p

, and so we have the anti-standard

ordering. Now, let

Q =
i
2
_

2
I
n
0
0
2
I
n
_
, where the dimensional parameter is such
that q and
1
p are of the same dimension. In this case, it is convenient to use the
holomorphic variables
z
j
=
1

2
(q
j
+i
1
p
j
), z
j
=
1

2
(q
j
i
1
p
j
). (38)
By 1, the elements of c

(R
2n
) and W

(R
2n
) can be regarded as entire functions of these
variables, considered as independent variables. The restriction to the real space R
2n
amounts then to identifying z with the complex conjugate of z. With these variables,
the formula (31) becomes

f(w, w) =
1
(2i)
n
__
f(z, z)e
z w zw
dz d z,
where w = (u i
1
v)/

2. As usual, we introduce the annihilation and creation

operators
a
j
=
1

2
(q
j
+i
1
p
j
), a

j
=
1

2
(q
j
i
1
p
j
).
Then the representation operator (35) can be written as
T
z
= e
1

(za

za)
= e

1
2
z z
e
1

za

za
.
In terms of the new variables, the quadratic form s

Qs becomes iw w, and the
correspondence (34) takes the form
f f =
1
(2i)
n
__

f(w, w) e
wa

e
wa
dw d w. (39)
Therefore, in this case we have the Wick quantization, or in other words, the normal
ordering, under which the monomial z

corresponds to the operator (a

. The
corresponding star product is
(f
Wick
g)(z, z) = f(z, z)e

z
g(z, z).
By changing the sign of

Q, we obtain the quadratic form iw w, which denes the
anti-Wick quantization and the anti-normal ordering a

(a

.
7. Extensions of representations
We now show that the map (34) can be naturally extended to functions in c

(R
2n
),
where 2. For clarity we rst consider the representation in the space of square
integrable functions L
2
(R
n
). In doing so we use another function space whose elements
decrease at innity faster than the elements of c

increase.
Let > 1 and > 1. The space W

(R
n
) consists of all entire analytic functions
satisfying the condition
[(q +iy)[ Ce
a|q|

+b|y|

(40)
where a, b, and C are positive constants depending on . As before, we really deal with
the restrictions of these functions to R
n
, using the analytic continuation as an auxiliary
12 M. A. SOLOVIEV
tool. The space W

coincides with the Gelfand-Shilov space S

11/
1/
and is nontrivial
if and only if . Its natural topology is the inductive limit topology with respect
to the family of Banach spaces W
,b
,a
equipped with the norms
||
a,b
= sup
q,y
[(q +iy)[e
a|q|

b|y|

. (41)
Using reasoning similar to that in the proof of Theorem 4, it is easy to show that W

is
invariant under the Fourier transformation if and only if 1/ +1/ = 1. It is important
that W

is dense in L
2
. Indeed, if f L
2
, W

and
_
(q)dq = 1, then the sequence
f

(q) =
n
_
f(q )()d belongs to W

and converges to f in the norm of L

2
as
.
We let /
W

(L
2
) denote the algebra of operators which are dened on the subspace
W

of the Hilbert space L

2
and map W

continuously into itself, and we endow this

algebra with the topology of uniform convergence on bounded subsets of W

.
Theorem 6. If 1 < 2, then the map (34) dened in Theorem 5 as an algebra ho-
momorphism from W

(R
2n
) to B(L
2
) extends uniquely to a continuous homomorphism
of the algebra (c

(R
2n
),

) to the algebra /
W

(L
2
), where

= /( 1).
Proof. Let f(p, q) c

(R
2n
). This function can be written as
f(p, q) =

,
c
,
p

, (42)
where by the estimate (12) the coecients c
,
=
1
!!

q
f(0, 0) satisfy
[c
,
[ |f|

(2

e)
(||+||)/
(

)
1/
. (43)
We rst extend the transformation (36) that denes the standard ordering. If such a
continuous extension to c

exists, then it is unique because W

is dense in c

by 2.
Indeed, every polynomial P(z) can be approximated in the topology of E

by a sequence
of functions of the form f

(z) = P(z)f
0
(z/), where f
0
W

is such that f
0
(0) = 1.
Let W

(R
n
). We dene an operator f by
(f)(q)
def
=

,
c
,
(i)
||
q

(q) (44)
and show that the series on the right-hand side of (44) converges to an element of W

,
and f is hence well dened as an operator from W

ito itself. To this end, we estimate

the derivatives of in the complex domain, using the Cauchy inequality again. Let
z = q +iy. Taking into account that

, we get
[

(z)[
!
r

sup
=+iDr
[(z )[
!
r

||
a,b
e
a|q/2|

+b|2y|

sup
+iDr
e
a||

+b|2|

!
r

||
a,b
e
a|q/2|

+b|2y|

e
a+(a+2

b)

j
r

j
.
Using (11) with

instead of , we obtain
[

(z)[ C||
a,b
b
||/

1
!
(

)
1/

e
a|q/2|

+b|2y|

, where b
1
= (a + 2

b)

e. (45)
ALGEBRAS WITH CONVERGENT STAR PRODUCTS AND THEIR REPRESENTATIONS 13
It follows from (43) and (45) that

,
[c
,

||
q

(z)[
C|f|

||
a,b
e
a|q/2|

+b|2y|

(2

)
||/
b
||/

1
!
(

)
1/
(

)
1/

[q

[
(2

e)
||/
(

)
1/
,
where (

)
1/
(

)
1/

. Using (6), we see that the sum over does not exceed

(eb
1

)
||/
and is dominated by a constant if < 1/(eb
1

). The sum over is

estimated similar to the sum in (9) and does not exceed C

n|2q|

, where

> and
can be taken arbitrarily close to . We infer that

,
[c
,

||
q

(z)[ C

|f|

||
a,b
e
(a/2)|q/2|

+b|2y|

(46)
if is suciently small compared to a and b. Therefore, f W

(R
n
). The operator
f maps W

into itself continuously, because the right-hand side of (46) contains the
factor ||
a,b
. Moreover, the Kohn-Nirenberg map j
st
: f f is continuous from c

to
/
W

(L
2
), because the right-hand side of (46) contains |f|

.
We now show that if f W

(R
2n
), then the denition of f by (44) is equivalent to
the above denition (36). To accomplish this, we consider the matrix element , f ,
where , W

(R
n
) and the angle brackets denote the inner product of L
2
(R
n
). The
function
(u, v) = , e
iuq
e
ivp
=
_
(q)e
iuq
(q v)dq
belongs to W

(R
2n
), because W

(R
2n
) and this space is invariant under linear
changes of variables and under the partial Fourier transform. From (36), we have
, f =
1
(2)
n
__

f(u, v)(u, v)dudv =
1
(2)
n
__
f(p, q) (p, q)dpdq =
=
1
(2)
n

,
c

__
p

(p, q)dpdq,
where the order of summation and integration can be interchanged because of absolute
convergence. Taking the (inverse) symplectic Fourier transform, we obtain
, f =

,
c

(i)
||
i
|
_

u
,
_
=

,
c

(i)
||
_
(q)q

(q)dq.
Thus, the denitions (36) and (44) are indeed consistent. The star product (37) is
continuous in c

by Theorem 2, and the operator product is separately continuous in

/
W

(L
2
) by the denition of the topology of bounded convergence. Because W

is
dense in c

, we conclude that the constructed extension of correspondence (36) is an

algebra homomorphism from (c

) to /
W

(L
2
).
We let Q
st
denote the matrix
1
2
_
0 I
n
I
n
0
_
dening the map j
st
and let j
Q
denote
the symbol-operator correspondence of general form (34). If f W

, then for any Q,

f(s)e
s

Qs
is the symplectic Fourier transform of e
Q
jk

k
f. Hence we have the relation
j
Q
(f) = j
st
_
e
i
2
(Q
jk
st
Q
jk
)
j

k
f
_
, (47)
14 M. A. SOLOVIEV
which holds for all f W

. We can now extend j

Q
to a map from c

into /
W

(L
2
)
by replacing j
st
in (47) with its extension constructed above. The map dened in this
way is continuous, because the dierential operator on the right-hand side of (47) is a
continuous automorphism of c

by Theorem 1. This completes the proof.

8. The Fock-Bargmann space representation
The Wick correspondence is usually realized using a representation in the Fock-
Bargmann space F
2
, which consists of antiholomorphic functions on C
n
satisfying the
condition
_
[( z)[
2
e

z
2
d(z) < ,
where d(z) = (2i)
n
dz z =
n
d(Re z)d(Imz) and |z|
2
= z z =

j
[z
j
[
2
. It is a
Hilbert space with inner product given by
, =
1

n
_
( z)( z)e

z
2
d(z).
The representation (35) on the space L
2
can be transferred to a representation on F
2
by means of the Bargmann transformation
(q) (B)( z)
def
= ()
n/4
e
1
2
z
2
_
e

1
2
(

2 zq)
2
(q)dq, (48)
which is a unitary isomorphism from L
2
onto F
2
(see [18], [19]). When realized on
F
2
, the creation operator is the multiplication by z and the annihilation operator is

z
. Accordingly, the representation of the phase-space translations is realized by the
operators
BT
w
B
1
= T
w
= e
1

w z w z
= e

1
2
w w
e
1

w z
e
w z
,
and the Wick correspondence takes the form
f f =
_

f(w, w) e
w z
e
w z
d(w), (49)
Let 1 < 2 and A > 0. We dene

E
2,A

to be the space of all entire antiholomorphic

functions satisfying the condition
[( z)[ Ce
Az
2
rz

, (50)
where C and r are constants depending on , and we give

E
2,A

the topology of the

inductive limit of Banach spaces with the norms
||
r
= sup
z
[( z)[e
Az
2
+rz

. (51)
Theorem7. Suppose 1 < 2 and

= /(1). Then the Bargmann transformation

dened by (48) maps the space W

isomorphically onto

E
2,A

, where A = 1/(2).
Proof. The transformation (48) is the composition of the following three operations: (1)
convolution by the function e

1
2
z
2
, (2) dilation by a factor of

2, and (3) multiplication

by ()
n/4
e
1
2
z
2
. The rst operation is equivalent to the multiplication of

() by
ce

2
, where
2
=

j

2
j
and the precise value of the constant c is of no importance
for the proof. Let

W

,b
,a
(R
n
) and

1
() =

()e

2
. Then we have
[

1
( +i)[ |

|
a,b
e

2
(
2

2
)a||

+b||

.
ALGEBRAS WITH CONVERGENT STAR PRODUCTS AND THEIR REPRESENTATIONS 15
Taking the inverse Laplace transform, setting z = x+iy, and using the Cauchy-Poincare
theorem, we get
[
1
( z)[ (2)
n/2
|

|
a,b
inf

e
i(xiy)(+i)

1
( +i)

d
C
a
|

|
a,b
inf

e
x+

2
+b||

sup

e
y

2
a

,
where a

< a/n
/2
. Because

2, we have
2
n+

j
[
j
[

, and the inmum over

can hence be estimated in a manner similar to that used in proving Theorem 4. Setting
B

= b +/2, we obtain
inf

_
x +

2

2
+b[[

n
2
+ inf

_
_
x +B

j
[
j
[

_
_

n
2
B[x[

,
where (B

(B)

= 1. We now show that there exists a positive constant such

that
y

2

2
a

||

1
2
y
2
|y|

for ||
_
/2. (52)
After a suitable rescaling of the variables, (52) takes the form
(y )
2

|y|

||

, (53)
where

= (2)
/2
, a

= (2/)
/2
a

, and || 1. Let us consider the ray y = t,

t 0 in R
2n
. Because 2 and || 1, the inequality (53) holds on this ray if
(t 1)
2

.
This condition in turn is satised for all t if (t 1)
2

t
2
a

, or equivalently,

/(1 +a

). We note that for all in the ball || 1, we have (y )

2

|y|

with some constant C

> 0. Thus, there exists a suciently small positive number

r(a, b, ) such that
[
1
(x iy)[ C

a
|

|
a,b
e
1
2
y
2
rx

ry

. (54)
Therefore, the function ( z) = e
1
2
z
2

1
(

2 z) satises (50) with A = 1/(2) and with a

preexponential factor proportional to |

|
a,b
. We conclude that the Bargmann operator
B maps W

continuously to

E
2,1/2

.
Conversely, let

E
2,1/2

and ||
r
< . Then the function
1
( z) = e

1
4
z
2
( z/

2)
satises
[
1
( z)[ ||
r
e
1
2
y
2
r

,
where r

= r/2
1+/2
. Hence its Fourier-Laplace transform satises the estimate
[
1
( +i)[ (2)
n/2
inf
y
_

e
i(xiy)(+i)

1
(x iy)

dx
C
b
||
r
inf
y
e
y+
1
2
y
2
r

sup
x
e
xbx

, (55)
with b < r

. Substituting y = , we obtain
inf
y
_
y +
1
2
y
2
r

|y|

2
r

||

.
16 M. A. SOLOVIEV
Taking the supremum over x gives the conjugate convex function b

||

. Therefore, it
follows from (55) that

2
(+i)
2

1
( +i)

C||
r
e
a||

+b

||

,
where a = r

and b

= b

/2
. We conclude that B
1
W

and the inverse

Bargmann transformation is continuous from

E
2,1/2

to W

. The theorem is proved.

Corollary 1. For any satisfying 1 < 2, the Wick correspondence (49) has
a unique extension to a continuous algebra homomorphism from (c

(R
n
),
Wick
) to
/

E
2,A

(F
2
), where A = 1/(2).
9. Concluding remarks
In this paper, we consider only analytic symbols of operators, but the above method
of extending the star products and their representations by continuity, starting from a
suitable auxiliary space like W

or W

, has a wider eld of application. In particular,

this approach allows us to extend the Weyl symbol calculus beyond the traditional
framework of tempered distributions, see [20], [21]. Such a generalization is also desir-
able for the anti-Wick correspondence because not all bounded operators have anti-Wick
symbols among the usual functions [18].
If > 2, then the space W

is not an algebra under the Wick star product. However

it is a topological algebra with respect to the Weyl-Moyal product

. It follows from a
theorem of [17] that the elements of c

are two-sided multipliers of the algebra (S

1/
1/
,

)
which coincides with the algebra (W

) for > 1. This implies that for all f c

and W

, the products f

and

f belong to W

and are continuous in f

and . Furthermore, the following associativity relations hold:
(f

= f

), (

f)

(f

), (

f =

f).
As shown in [21], the Weyl correspondence denes an isomorphism between the left
multiplier algebra of (W

) and the operator algebra /

W

(L
2
). Therefore, for the
case of Weyl-Moyal product, Theorem 6 can be derived as a consequence of these results.
The direct proof given here is more illuminating and applies to any star product of the
form (2).
Finally, we note that along with the spaces c

, it is natural to consider the spaces of

entire functions of the same order and nite type, i.e., satisfying the condition
[f(z)[ Ce
a|z|

,
where a and C are constants depending on f. We let E

denote this space and endow

it with the inductive limit topology dened by the family of Banach spaces E
,a
with
the norms |f|
a
= sup
z
[f(z)[e
a|z|

. It is easy to see that an analogue of Theorem 2

holds for E

, but only if the strong inequality < 2 holds. It is essential that if

1/

+ 1/ = 1 and 2, then the series dening the star product f

g converges
absolutely in E

for all f c

and g E

. Moreover, it can be proved that the

algebra (c

) acts continuously on the space E

and on its subspace W

by the left
and right

-multiplication. In other words, the function spaces E

and W

(as well
as the analogous spaces associated with cone-shaped regions) are topological bimodules
over the algebra (c

) for any .
ALGEBRAS WITH CONVERGENT STAR PRODUCTS AND THEIR REPRESENTATIONS 17
Acknowledgements. This paper was supported in part by the the Russian Foundation
for Basic Research (Grant No. 12-01-00865)
References
[1] H. Omori, Y. Maeda, N. Miyazaki, and A. Yoshioka, Deformation quantization of Frechet-Poisson
algebras Convergence of the Moyal product, in: Conference Moshe Flato 1999. Quantization,
Deformation, and Symmetries, edited by G. Dito, D. Sternheimer, Math. Physics Studies, no. 22,
(Dordrecht, Kluwer, 2000), Vol. 2, pp. 233-245.
[2] H. Omori, Y. Maeda, N. Miyazaki, and A. Yoshioka, Non-formal deformation quantization of
Frechet-Poisson algebras: The Heisenberg Lie algebra case, Contemp. Math. 434 (2007) 99-124.
[3] S. Beiser, H. R omer, and S. Waldmann, Convergence of the Wick product, Commun. Math.
Phys. 272 (2007) 25-52 [arXiv:math.QA/0506605].
[4] S. Beiser and S. Waldmann, Frechet algebraic deformation quantization of the Poincare disk,
e-print arXiv:1108.2004.
[5] O. W. Greenberg, Failure of microcausality in quantum eld theory on noncommutative space-
time, Phys.Rev. D 73 (2006) 045014 [arXiv:hep-th/0508057].
[6] M. A. Soloviev, Noncommutativity and -locality, J. Phys A: Math. Theor. 40 (2007) 14593-
14604 [arXiv:0708.1151].
[7] M. Chaichian, M. N. Mnatsakanova, A. Tureanu, and Yu. A. Vernov, Test function space in
noncommutative quantum eld theory, JHEP 0809 (2008) 125 [arXiv:0706.1712].
[8] M. A. Soloviev, Failure of microcausality in noncommutative eld theories, Phys. Rev. D 77
(2008) 125013 [arXiv:0802.0997]
[9] S. Galluccio, F. Lizzi, and P. Vitale, Twisted noncommutative eld theory with the Wick-Voros
and Moyal products, Phys. Rev. D 78 (2008) 085007 [arXiv:0810.2095].
[10] A. P. Balachandran and M. Martone, Twisted quantum elds on Moyal and Wick-Voros planes
are inequivalent, Mod. Phys. Lett. A 24 (2009) 1721-1730 [arXiv:0902.1247].
[11] A. P. Balachandran, A. Ibort, G. Marmo, and M. Martone, Inequivalence of QFTs on noncommu-
tative spacetimes: Moyal versus Wick-Voros, Phys. Rev. D 81 (2010) 085017 [arXiv:0910.4779].
[12] M. A. Soloviev, Noncommutative deformations of quantum eld theories, locality, and causality,
Theor. Math. Phys. 163 (2010) 741-752 [arXiv:1012.3536].
[13] M. Blaszak and Z. Doma nski, Phase space quantum mechanics, Ann. of Phys. 327 (2012) 167-
211 [arXiv:1009.0150].
[14] H. Omori, Y. Maeda, N. Miyazaki, and A. Yoshioka, Deformation expression for elements of
algebras (II) (Weyl algebra of 2m-generators), e-print arXiv:1105.1218.
[15] I. M. Gelfand and G. E. Shilov, Generalized Functions, Vol. 3 (Academic, New York 1968).
[16] J. M. Gracia-Bondia and J. C. Varilly, Algebras of distributions suitable for phase-space quantum
mechanics. I. J. Math. Phys. 29 (1988) 869-879.
[17] M. A. Soloviev, Twisted convolution and Moyal star product of generalized functions, Theor.
Math. Phys. 172 (2012) 885-900 [arXiv:1208.1838].
[18] G. B. Folland, Harmonic Analysis in Phase Space, Annals of Math. Studies, Vol. 122 (Princeton
Univ. Press, Princeton, 1989).
[19] F. A. Berezin and M. A. Shubin, Schrodinger Equation (Dordrecht, Kluwer, 1991).
[20] M. A. Soloviev, Moyal multiplier algebras of the test function spaces of type S, J. Math. Phys.
52 (2011) 063502 [arXiv:1012.0669].
[21] M. A. Soloviev, Generalized Weyl correspondence and Moyal multiplier algebras, Theor. Math.
Phys. 173 (2012) 1359-1376.
I. E. Tamm Department of Theoretical Physics, Lebedev Institute of Physics, Leninsky
Prospect 53, 119991 Moscow, Russia
E-mail address: soloviev@lpi.ru