Bol. Soc. Mat. Mex.

(2016) 22:213–227
DOI 10.1007/s40590-015-0073-7

ORIGINAL ARTICLE

Toeplitz operators with quasi-separately radial symbols

on the complex projective space

Miguel A. Morales-Ramos1 ·
Armando Sánchez-Nungaray1 ·
Josué Ramírez-Ortega1

Received: 11 January 2015 / Accepted: 2 October 2015 / Published online: 24 October 2015
© Sociedad Matemática Mexicana 2015

Abstract We study the Toeplitz operators with quasi-separately radial symbols over
the complex projective space CPn . We describe such Toeplitz operators and we prove
that each bounded operator is unitarily equivalent to a Toeplitz operator whose symbol
is a finite sum of quasi-separately radial symbols.

Keywords Toeplitz operators · Quasi-separately radial symbols · Complex

projective space

Mathematics Subject Classification 47L80 · 47B35

1 Introduction

The C ∗ -algebras of Toeplitz operators acting on (weighted) Bergman spaces for dif-
ferent domains and properties of such operators have been extensively studied (see, for
example [2,3,5,14]). It turned out that some geometric properties of the symbols play
an essential roll for the Toeplitz operators that generate a commutative C ∗ -algebra.

This work was partially supported by CONACYT Project 236109.

B Armando Sánchez-Nungaray
[email protected]
Miguel A. Morales-Ramos
[email protected]
Josué Ramírez-Ortega
[email protected]

1 Universidad Veracruzana, Xalapa, Ver., Mexico

214 M. A. Morales-Ramos et al.

Under certain conditions on the symbol set, in [4], the authors proved that the C ∗ -
algebra generated by the Toeplitz operators is commutative on each weighted Bergman
space if and only if there is a pencil of hyperbolic geodesics such that the symbols are
constant on the cycles of this pencil. Equivalently, every class of symbols that generates
a commutative C ∗ -algebra of Toeplitz operators arises as the set of functions invariant
under the action of a maximal abelian group of Möbius transformations of the unit disc.
They also proved that there are only three types of such groups: parabolic, hyperbolic
and elliptic. It is important to emphasize that the elliptic case corresponds to radial
functions on the disc.
In [13], it was proved that for the unit ball of dimension n there are n + 2 different
cases of commutative C ∗ -algebras generated by Toeplitz operators, acting on weighted
Bergman spaces. In every case, the bounded measurable symbols for the Toeplitz
operators are invariant under the action of certain commutative abelian groups. The
authors found that there are five different types of such groups: quasi-parabolic, quasi-
hyperbolic, nilpotent, quasi-nilpotent and quasi-elliptic. In particular, the symbols
invariant under the action of the quasi-elliptic group of the unit ball correspond to
separately radial symbols. In the Reinhardt domains on Cn [12] there exist similar
results.
In [10], the authors proved the existence of commutative C ∗ -algebras of Toeplitz
operators on each weighted Bergman space over the complex projective space CPn .
They found that there is only one case, the quasi-elliptic one, for the C ∗ -algebra
of Toeplitz operators to be commutative. The symbols that define such algebra are
separately radial with respect to homogeneous coordinates.
To determine whether the product of two Toeplitz operators is a Toeplitz operator is
a problem that goes back to 1964 [1], where Halmos and Brown solved this question
for Toeplitz operators on the Hardy space. They showed that for f , g in L ∞ (D),
where D is the unit circle in C, T f Tg = T f g if and only if either g is analytic or f is
anti-analytic.
In [9], the authors give necessary and sufficient conditions for the product of Toeplitz
operators on the unit disc to be a Toeplitz operator in special cases. First, they studied
Toeplitz operators with radial symbols. They found an explicit expression for the
symbol of the product of Toeplitz operators with such symbols. Furthermore, they
determined whether the product of two Toeplitz operators with quasi-homogeneous
symbols is equal to a Toeplitz operator.
This paper is structured as follows. In Sect.2, we introduce some geometric pre-
liminaries of CPn . In Sect. 3, following [10], we introduce the weighted Bergman
space for the complex projective space as well as the Bergman projection. Later, we
study the Bergman space using the polar decomposition of complex numbers on local
coordinates. We use Fourier techniques in such a way  that the Bergman space can
be identified with the closed subspace of l2 (Zn ) ⊗ L 2 Rn+ , μm , whose elements are
sequences satisfying integrability and holomorphy conditions. Then, from [10], we
have that each Toeplitz operator with separately radial symbol is unitarily equivalent
to a diagonal operator. In Sect. 4, we use another class of symbols, quasi-separately
radial symbols of order s, and study the Toeplitz operators with such class of symbols.
We prove that every Toeplitz operator with quasi-separately radial symbol of order s
is unitarily equivalent to a s-diagonal matrix (see Definition 12). Finally, in Sect. 5, we
Toeplitz operators with quasi-separately radial symbols… 215

prove a connection between bounded operators and Toeplitz operators. We prove that
every bounded operator equals to a Toeplitz operator whose symbol is a finite sum of
quasi-separately radial symbols.

2 Geometric preliminaries on CPn

As usual, CPn is the complex n-dimensional manifold that consists of all elements
[w] = Cw\{0}, where w ∈ Cn+1 \{0}. For every j = 0, . . . , n we have an open set
 
U j = [w] ∈ CPn : w j = 0

and a holomorphic chart ϕ j : U j → Cn given by

1  
ϕ j ([w]) = w0 , . . . , wˆ j , . . . , wn = (z 1 , . . . , z n ).
wj

The numbers z k are known as the homogeneous coordinates with respect to the
map ϕ j . The collection of all such maps yields a holomorphic atlas of CPn .
For each j = 0, . . . , n, consider the function f j : U j → C given by


n
wk w̄k 
n
f j ([w]) = =1+ z k z̄ k ,
w j w̄ j
k=0 k=1

where the z k ’s are the homogeneous coordinates with respect to ϕ j . Then, it is easily
seen that

∂ ∂¯ log f j = ∂ ∂¯ log f k

on U j ∩ Uk for every j, k = 0, . . . , n. In particular, there is a well-defined closed

(1, 1)-form w on CPn given by

ω = i∂ ∂¯ log f j

on U j . This yields the canonical Kähler structure on CPn , which turns out to be a
Hermitian symmetric space with constant positive holomorphic sectional curvature.
The corresponding Riemannian metric is known as the Fubini–Study metric.
With respect to the chart ϕ0 , we have the following induced Kähler form on Cn
n n
(1 + |z|2 ) k=1 dz k ∧ dz̄ k − k,l=1 z̄ k z l dz k ∧ dz̄l
ω0 = (ϕ0−1 )∗ (ω) =i .
(1 + |z|2 )2

The volume element on CPn with respect to the Fubini–Study metric is defined by

1
= ωn .
(2π )n
216 M. A. Morales-Ramos et al.

A consequence of the definition of the volume element of a Riemannian metric

and its properties for Hermitian symmetric spaces is that the volume element on Cn
induced by the Fubini–Study metric of CPn is given by

1 1 dV (z)
= ωn = n
(2π ) n π (1 + |z 1 | + · · · + |z n |2 )n+1
2

where dV (z) = dx1 ∧ dy1 ∧ · · · ∧ dxn ∧ dyn is the Lebesgue measure on Cn .

3 Bergman spaces on CPn and Toeplitz operators with separately radial

symbols

All the results employed in this work are extracted from [10], although these are
presented here briefly, it is important to familiarize us with the notation and techniques
employed since we will use them later.
The tautological or universal line bundle of CPn is given by

T = ([w], z) ∈ CPn × Cn+1 : z ∈ Cw ,

and assigns to every point in CPn the line in Cn+1 that such point represents. T is a
holomorphic line bundle. Furthermore, T has a natural Hermitian metric h 0 inherited
from the usual Hermitian inner product on Cn+1 . Let H = T ∗ denote the dual line
bundle with the corresponding induced metric h dual to the metric h 0 on T . The line
bundle H is called the hyperplane line bundle.
For every m ∈ Z+ and with respect to the coordinates given by ϕ0 , the m-weighted
measure on CPn is given by

(n + m)! (z)
dνm (z) =
m! (1 + |z 1 |2 + · · · + |z n |2 )m
(n + m)! dV (z)
= ,
π m! (1 + |z 1 | + · · · + |z n |2 )n+m+1
n 2

where dV (z) = dx1 ∧ dy1 ∧ · · · ∧ dxn ∧ dyn is the Lebesgue measure on Cn . To

simplify the notation, we use the same symbol dνm to denote the weighted measures
for both CPn and Cn .
Let H m = H ⊗· · ·⊗ H be the tensorial product of H with itself taken m times. The
line bundle H can be trivialized over each subset U j ⊂ CPn so that the corresponding
transition functions for H m are given as follows

gkmj : U j ∩ Uk → C∗
wkm
[w] → .
w mj
Toeplitz operators with quasi-separately radial symbols… 217

In particular, there is a trivialization of H m over U0 . Then, every section ζ of H m

restricted to U0 can be considered as a map ζ |U0 : U0 → C. Since ϕ0−1 : Cn → U0
defines a biholomorphism, the composition ζ̂ = ζ |U0 ◦ ϕ0−1 maps Cn into C.
Then, we have the following well-known result. The map given by

0 : L 2 (CPn , H m ) → L 2 (Cn , νm )
ζ →

ζ, (3.1)

is an isometry.

Definition 1 The weighted Bergman space on CPn , with weight m ∈ Z+ , is defined

by
 
A2m (CPn ) = ζ ∈ L 2 (CPn , H m ) : ζ is holomorphic
=
hol (CPn , H m ).

It is also known that for every m ∈ Z+ , the Bergman space A2m (CPn ) satisfies the
following properties.
1. With respect to the homogeneous coordinates of CPn , the Bergman space can be
identified with the space P (m) (Cn+1 ) of all homogeneous polynomials of degree
m over Cn+1 .
2. For the isometry 0 given in (3.1), we have

0 (A2m (CPn )) = Pm (Cn ),

where Pm (Cn ) denotes the space of all polynomials on Cn of degree less than or
equal to m.
For more details, we refer the reader to [8]. These two properties allow us to make
computations on L 2 (Cn , νm ) and Pm (Cn ) instead of L 2 (CPn , H m ) and A2m (CPn ),
respectively.
Recall the following notation for multi-indices p, q ∈ Zn+ and z ∈ Cn :

p! = p1 ! . . . pn !,
| p| = p1 + · · · + pn ,
p p
z p = z1 1 . . . zn n ,
δ p,q = δ p1 ,q1 . . . δ pn ,qn .

The Bergman space A2m (Pn (C)) has a basis consisting of the polynomials z p =
p p
z1 1. . . z k k , where p ∈ Zn+ and | p| ≤ m. Hence, we will consider the set

Jn (m) = { p ∈ Zn+ : | p| ≤ m}.

218 M. A. Morales-Ramos et al.

More precisely, an easy computation shows that the set


α
αmp z : α ∈ Jn (m) (3.2)

is an orthonormal basis of A2m (Pn (C)), where

 1/2
m!
αm
p = .
p!(m − | p|)!

For ψ ∈ L 2 (Pn (C), H m ), and considering the identification , we define the

Bergman projection by

(n + m)! ψ(w)K (z, w)
(Bm ψ)(z) = dV (w),
π n m! Cn (1 + w1 w 1 + · · · + wn w n )n+m+1

where

K (z, w) = (1 + z 1 w 1 · · · + z n w n )m .

Note that the above integral is a polynomial of degree at most m in the variable z, and
so it defines an element of A2m (Pn (C)) with respect to the identification . Also, the
operator Bm satisfies the well-known reproducing property.
Proposition 2 If ψ ∈ L 2 (Pn (C), H m ), then Bm (ψ) belongs to the weighted Bergman
space A2m (Pn (C)). Also, Bm (ψ) = ψ for ψ ∈ A2m (Pn (C)).
We define the Toeplitz operator Ta on A2m (Pn (C)) with bounded symbol a ∈
L ∞ (Pn (C)) by

Ta (ϕ) = Bm (aϕ),

for every ϕ ∈ A2m (Pn (C)).

Now, consider the polar coordinates for the complex numbers: z j = t j r j , where
t j ∈ T, T is the unitary circle in C, and r j ∈ R+ for each j ∈ {1, . . . , n}. Thus, for
z ∈ Cn , we have

z = (z 1 , . . . , z n ) = (t1r1 , . . . , tn rn ) = (t, r ),

where t = (t1 , . . . , tn ) ∈ Tn , r = (r1 , . . . , rn ) ∈ Rn+ . In particular, Cn = Tn × Rn+

and the associate volume form can be written as

n
dt j 
n
dV (z) = r j dr j .
it j
j=1 j=1

We have the decomposition

L 2 (Cn , νm ) = L 2 (Tn ) ⊗ L 2 (Rn+ , μm ),

Toeplitz operators with quasi-separately radial symbols… 219

where


n  
dt j
L 2 (Tn ) = L 2 T,
2πit j
j=1

and the measure dμm in Rn+ is given by

2n (n + m)!  n
dμm = (1 + r12 + · · · + rn2 )−n−m−1 r j dr j .
m!
j=1

With respect to the local coordinates given by ϕ0 , the Bergman space A2m (CPn )
consists of all functions in L 2 (Cn , νm ) satisfying the equations
 
∂ϕ 1 ∂ ∂
= +i ϕ = 0, j = 1, . . . , n,
∂ z̄ j 2 ∂x j ∂yj

or, in terms of polar coordinates

 
∂ϕ tj ∂ tj ∂
= − ϕ = 0, j = 1, . . . , n.
∂ z̄ j 2 ∂r j r j ∂t j

Let F denote the discrete Fourier transform F : L 2 (T) → l2 = l2 (Z) defined by

 
dt
F : f → c j = f (t)t − j .
S1 2πit j∈Z

Of course F is an unitary operator with inverse given by


F−1 = F∗ : {c j } j∈Z → f = cjt j.
j∈Z

Now, consider the unitary operator

U = F(n) ⊗ I : L 2 (Tn ) ⊗ L 2 (Rn+ , μm ) → l2 (Zn ) ⊗ L 2 (Rn+ , μm )

where F(n) = F ⊗ · · · ⊗ F (n times). Thus, A2m = U (A2m (CPn )) is the closed subspace
of l2 (Zn )⊗ L 2 (Rn+ , μm ) whose elements are sequences {c p (r )} p∈Zn , where each c p (r )
satisfies the equations
 
1 ∂ pj
− c p (r ) = 0, j ∈ {1, . . . , n}.
2 ∂r j rj

The general solution of this system of equations has the form

c p (r ) = c p α m
pr ,
p
220 M. A. Morales-Ramos et al.

where c p ∈ C and α m
p is given by

 p1 pn
− 1
(n + m)! t . . . t n t 1 dt 1 . . . t n dt n
2
αm
p =
1
Rn+ (1 + t1 + · · · + tn )
m! n+m+1

 1
m! 2
= .
p!(m − | p|)!

Since we must have c p (r ) ∈ L 2 (Rn+ , μm ), it follows that c p = 0 for p ∈ Zn \Jn (m).

Consider the isometric embedding

R0 : l2 (Jn (m)) → l2 (Zn ) ⊗ L 2 (Rn+ , μm )

defined by

c p αm
pr ,
p p ∈ Jn (m)
R0 : {c p } p∈Jn (m) → c p (r ) = .
0 p ∈ Zn \Jn (m)

Its adjoint operator R0∗ : l2 (Zn ) ⊗ L 2 (Rn+ , μm ) → l2 (Jn (m)) is given by

 
R0∗ : { f p (r )} p∈Zn → αm
p r f p (r )dμm (r )
p
.
Rn+ p∈Jn (m)

It is known that ([10])

R0∗ R0 = I : l2 (Jn (m)) → l2 (Jn (m))

R0 R0∗ = P1 : l2 (Zn ) ⊗ L 2 (Rn+ , μm ) → l2 (Jn (m)),

where P1 is the orthogonal projection from l2 (Zn ) ⊗ L 2 (Rn+ , μm ) onto the space
l2 (Jn (m)) = A2m . This yields the following theorem.

Theorem 3 The operator R = R0∗ U maps L 2 (CPn , νm ) onto A2m , and

R|A2m (CPn ) : A2m (CPn ) → l2 (Jn (m))

is an isometric isomorphism. The adjoint operator

R ∗ = U ∗ R0 : l2 (Jn (m)) → A2m (CPn ) ⊂ L 2 (CPn , νm )

is an isomorphism of l2 (Jn (m)) = A2m onto the subspace A2m (CPn ). Furthermore,

R R ∗ = I : l2 (Jn (m)) → l2 (Jn (m))

R ∗ R = Bm : L 2 (CPn , νm ) → A2m (CPn )
Toeplitz operators with quasi-separately radial symbols… 221

where Bm is the Bergman projection of L 2 (CPn , νm ) onto A2m (CPn ).

An explicit calculation yields the following results.

Corollary 4 The isometric isomorphism R ∗ : l2 (Jn (m)) → A2m (CPn ) ⊂ L 2 (CPn , νm )

is given by 
R ∗ : {c p } p∈Jn (m) → c p αm
pz .
p
(3.3)
p∈Jn (m)

Corollary 5 The isometric isomorphism R|A2m (CPn ) : A2m (CPn ) → l2 (Jn (m)) is
given by  
R : ψ → α pm
ψ(z)z̄ dνm (z)
p
. (3.4)
Cn p∈Jn (m)

Next, we will consider the following family of symbols for the study of Toeplitz
operators.

Definition 6 We call a function a([w]) : CPn → C separately radial if it can be

written in the form ã(|w0 |, . . . , |wn |) for some function ã : [0, ∞)n+1 → C which is
homogeneous of degree 0.

Note that a separately radial function can be expressed in the local coordinates of
ϕ0 , by the expression a([1, z]) = a(z) = a(r ), i.e., a(z) depends only on the radial
components of z = (t, r ) ∈ Cn . When we speak about a separately radial symbol, we
will refer mostly to the expression of the symbol in local coordinates.

Theorem 7 Let a be a bounded measurable separately radial function. Then, the

Toeplitz operator Ta acting on A2m (CPn ) is unitarily equivalent to the multiplication
operator γa,m I = RTa R ∗ acting on l2 (Jn (m)), where R y R ∗ are given by (3.3) and
(3.4), respectively. The sequence {γa,m ( p)} p∈Jn (m) is given by

2 p +1 2 p +1
2n m! a(r1 , r2 , . . . , rn )r1 1 . . . rn n dr1 . . . drn
γa,m ( p) = . (3.5)
p!(m − | p|)! Rn+ (1 + r12 + r22 + · · · rn2 )n+m+1

Hence, we can diagonalize the corresponding Toeplitz operators.

Corollary 8 The Toeplitz operator Ta , with bounded measurable separately radial

symbol a(r ), is diagonal with respect to the orthonormal basis given in (3.2). More
precisely, we have

Ta (ϕ p ) = γa,m ( p)ϕ p ,

for all p ∈ Jn (m).

222 M. A. Morales-Ramos et al.

4 Toeplitz operators with quasi-separately radial symbols

Let k = (k0 , . . . , kl ) ∈ Zl+1

+ be a multi-index so that |k| = n + 1. We will call such
multi-index k a partition of n + 1. For the sake of definiteness, we will always assume
that k0 ≤ · · · ≤ kl . This partition provides a decomposition of the coordinates of
w ∈ Cn+1 as w = (w(0) , . . . , w(l) ), where

w( j) = (wk0 +···+k j−1 +1 , . . . , wk0 +···+k j ) = (w j,1 , . . . , w j,k j ),

for every j = 0, . . . , l, and the empty sum is 0 by convention. For w ∈ Cn+1 , we

define r j = |w( j) | and
w( j)
ξ( j) =
rj

whenever w( j) = 0. Besides the quasi-radii (r0 , . . . , rl ), this provides a set of coordi-

nates (ξ(0) , . . . , ξ(l) ) ∈ Sk0 × · · · × Skl .
Recall that if p, q ∈ Zn+1 + satisfy p · q = 0 and | p| = |q|, then with respect to the
canonical embedding Cn → CPn the k-quasi-homogeneous symbol ϕ associated to
p, q restricted to Cn satisfies

l 
   
z ( j) p( j) z ( j) q( j)
ϕ([1, z]) = ,
|z ( j) | |z ( j) |
j=1

where k ∈ Zl+1 is a partition of n + 1 with k0 = 1.

On the other hand, we know that any bounded measurable function a on CPn is
called k-quasi-radial if it depends only on r0 , . . . , rl . Thus, we know that a k-quasi-
homogeneous quasi-radial symbol is a function CPn → C of the form aϕ, where a is
k-quasi-radial and ϕ is k-quasi-homogeneous.
Now, we introduce the family of quasi-separately radial symbols. These kind of
symbols are an extension of both the separately radial functions and the k-quasi-
homogeneous quasi-radial functions which were used in [11,15].
Definition 9 Let a : CPn → C a separately radial and homogeneous function of
degree 0. Let s ∈ Zn+1 such that |s| = 0. A function as is called quasi-separately
radial of order s if it has the form
n 
 
wj sj
as ([w]) = a([w]) .
|w j |
j=0

Since a is separately radial homogeneous of degree 0 and |s| = 0, the symbol as is

independent of the choice of homogeneous coordinates.
With respect to the local coordinates from ϕ0 , we have that the quasi-separately
radial function of order s as can be rewritten by

as ([1, z]) = as (z) = a(r )t s ,

Toeplitz operators with quasi-separately radial symbols… 223

where z = (t, r ) ∈ Cn , a(r ) is a separately radial function and s ∈ Zn .

For s = (s1 . . . , sn ) ∈ Zn , there exist s+ , s− ∈ Zn+ , such that s = s+ − s− . Thus,
we define
ŝ = max{|s+ |, |s− |}. (4.1)
Remark 10 In particular, for the case k = (1, . . . , 1), the family of k-quasi-
homogeneous quasi-radial symbols associated to p, q coincides with the family of
quasi-separately radial symbols of order p − q.
Next, we introduce a partial order in Zn in the following way: for p, q ∈ Zn , p ≤ q
if q − p ∈ Zn+ .
Let Ss be the shifting operator acting on l2 (Jn (m)) as follows. For c =
{c( p)} p∈Jn (m) ,

(Ss (c))( p) = c( p − s) if p, p − s ∈ Jn (m),

and (Ss (c))( p) = 0 if p ∈ Jn (m) but p − s ∈ / Jn (m). If cq = {δ pq } p∈Jn (m) , with δ

the Kronecker delta function, then Ss (cq ) = cq+s whenever q, q + s ∈ Jn (m).
Theorem 11 Let as a bounded measurable quasi-separately radial function of order
s. If s is such that ŝ ≤ m, then the Toeplitz operator Tas acting on A2m (CPn ) is
unitarily equivalent to the operator das ,m Ss = RTas R ∗ acting on l2 (Jn (m)). The
sequence das ,m = {das ,m ( p)} p∈Jn (m) is given by

das ,m ( p) = αq+s
m
αm
+ q+s−
a(r )r 2q+s+ +s− dμm (r ), q = p − s+ (4.2)
Rn+

for p ∈ (Jn (m) + s) ∩ Jn (m), while das ,m ( p) = 0 for p ∈ Jn (m)\(Jn (m) + s).
Proof Under the hypotheses of the Theorem, the Toeplitz operator Tas is unitarily
equivalent to the operator RTas R ∗ , then
   
RTas R ∗ {c( p)} p∈Jn (m) = R0∗ U Ma(r )t s U ∗ R0 {c( p)} p∈Jn (m)
⎛ ⎞

= R0∗ U Ma(r )t s ⎝ c( p)α m
pz
p⎠

p∈Jn (m)
⎛ ⎞

= R0∗ U ⎝ p p+s ⎠
p a(r )r t
c( p)α m
p∈Jn (m)
 
= R0∗ {c( p − s)α m
p−s a(r )r
p−s
} p∈Jn (m)+s
= {c( p − s)das ,m ( p)} p∈Jn (m) ,

where das ,m ( p) = 0 for p ∈ Jn (m)\(Jn (m) + s), and

das ,m ( p) = αm
p α p−s
m
r p a(r )r p−s dμm (r )
Rn+
224 M. A. Morales-Ramos et al.

for p ∈ (Jn (m) + s) ∩ Jn (m). Of course p ∈ (Jn (m) + s) ∩ Jn (m) if and only if
p, p − s ∈ Jn (m). Since s = s+ − s− and s j · s̃ j = 0, we have that p, p − s+ + s− ∈
Jn (m) if and only if p ≥ s+ , | p| ≤ m and | p| − |s+ | + |s− | ≤ m. This inequality is
equivalent to | p|−|s+ | ≤ m −|s− |. Thus, (Jn (m)+s)∩ Jn (m) is not empty if and only
s+ , s− belong to Jn (m). On the other hand, if p = q +s+ , then p ∈ (Jn (m)+s)∩ Jn (m)
if and only if q ≥ 0, |q| ≤ m − |s+ | and |q| ≤ m − |s− |, i.e., q ∈ Jn (m − ŝ).
So, for p − s = q + s− with q ∈ Jn (m − ŝ),

das ,m ( p) = αq+s
m
αm
+ q+s−
a(r )r 2q+s+ +s− dμm (r ).
Rn+

Thus,
 
RTas R ∗ {c( p)} p∈Jn (m) = {das ,m ( p)c( p − s)} p∈Jn (m)
= {das ,m ( p)(Ss (c))( p)} p∈Jn (m) .

Definition 12 For s ∈ Zn , a matrix A = (ai j )i, j∈Jn (m) is called s-diagonal if ai j = 0

for i − j = s.

The matrix A is a linear operator on l2 (Jn (m)): for c = {c( p)} p∈Jn (m) , the sequence
b = {b( p)} p∈Jn (m) = A (c) is given by

b( p) = a pj c( j).
j∈Jn (m)

If A is s-diagonal, then b( p) = a p, p−s c( p − s) for p − s ∈ Jn (m); otherwise,

b( p) = 0. In particular, for cq = {δ p,q } p∈Jn (m) we have aq+s,q cq+s = A(cq ), with
q, q + s ∈ Jn (m).

Corollary 13 Let as a bounded measurable quasi-separately radial function of order

s. The Toeplitz operator Tas is s-diagonal with respect to the orthonormal basis given
in (3.2). Actually,
 
Ta(r )t s α m
pz
p
= das ,m ( p + s)α m
p+s z
p+s
, p + s ∈ Jn (m),

 
but Ta(r )t s α m
pz
p = 0 for p + s ∈
/ Jn (m).

Proof Let s = s+ − s− , with s+ , s− ∈ Jn (m). We define the matrix Aa(r )t s =

(ai j )i, j∈Jn (m) as follows. Let i, j ∈ Jn (m). We take ai, j = 0 if i − j = s. Suppose
that i − j = s. Then, i = q + s+ and j = q + s− , with q ∈ Jn (m − ŝ). We define ai j
as

ai j = aq+s+ ,q+s− = das ,m ( p), p = i = q + s+ .

Of course, Aa(r )t s is s-diagonal.

Toeplitz operators with quasi-separately radial symbols… 225

In A2m (CPn ), we have Tas = R ∗ RTas R ∗ R. Then, by Theorem 11,

  
R ∗ RTas R ∗ R(α m
p̃ z p̃
) = R ∗
RT a s R ∗
δ p̃, p p∈Jn (m)
  
∗
=R das ,m ( p)Ss (δ p̃, p ) p∈J (m)
n
  
∗
=R das ,m ( p)δ p̃, p−s p∈J (m)
n
  
∗
=R das ,m ( p)δ p̃+s, p p∈J (m) .
n

  
If p̃ + s ∈ Jn (m), then R ∗ das ,m ( p)δ p̃+s, p p∈J (m) = das ,m ( p̃ + s)α m
p̃+s z
p̃+s . If
 
n

/ Jn (m), then R ∗ das ,m ( p)δ p̃+s, p p∈J (m) = R ∗ (0) = 0.

p̃ + s ∈ 

n

5 Connection between operators in A2m (CPn ) and Toeplitz operators

We know that Am 2 (CP ) is a finite-dimensional space, thus every operator in this space
n

can be identified with a square matrix. Actually, we will prove that every operator in
this space is a Toeplitz operator.

Definition 14 For a matrix A = (ai j )i, j∈Jn (m) , we define its s-projection by

0 if i − j = s
As = (5.1)
ai j if i − j = s.

We also define the set E ⊂ Zn by

E = {s = s+ − s− : s+ , s− ∈ Jn (m), s+ ⊥ s− }. (5.2)

Let i 0 , j0 ∈ Jn (m). Note that

|(i 0 − j0 )+ | ≤ |i 0 | ≤ m,

then (i 0 − j0 )+ ∈ Jn (m). A similar argument shows that (i 0 − j0 )− ∈ Jn (m), therefore,


i 0 − j0 ∈ E. It is clear from Definition 14 that As = As if and only if s = s  thus, for
a matrix A = (ai j )i, j∈Jn (m) each ai j belongs to one and only one As , with s ∈ E. It
follows that A = s∈E As .

Lemma 15 Let s = s+ − s− , with s+ , s− ∈ Jn (m). Consider {bq }q∈Jn (m−ŝ) , where

bq ∈ C. Then, there exists a bounded measurable function a(r ) such that

a(r )r 2q+s+ +s− dμm (r ) = bq , ∀q ∈ Jn (m − ŝ). (5.3)
Rn+
226 M. A. Morales-Ramos et al.

Proof Consider a(r ) having the following form

 r 2u+s+ +s−
a(r ) = cu .
(1 + r 2 )2(n+m+1)
u∈Jn (m−ŝ)

Let v = u + s+ and p = q + s+ . The condition q, u ∈ Jn (m − ŝ) means that

p, p − s, v, v − s ∈ Jn (m). Furthermore, 2q + s+ + s− = 2 p − s and 2u + s+ + s− =
2v − s. Thus, the function r v /(1 + r 2 )n+m+1 is bounded, and r 2v−s /(1 + r 2 )n+m+1
belongs to L 2 (Rn+ , μm ). Introduce the system of linear equations with respect to the
scalars cu :

bq = a(r )r 2q+s+ +s− dμm (r )
Rn+

r 2u+2q+2s+ +2s−
= cu dμm (r ),
Rn+ (1 + r )
2 2(n+m+1)
u∈Jn (m−ŝ)
  
r 2u+s+ +s− r 2q+s+ +s−
= cu , ,
(1 + r 2 )n+m+1 (1 + r 2 )n+m+1
u∈Jn (m−ŝ)

where we consider the inner product in the space L 2 (Rn+ , μm ). The set of functions
f u (r ) = r 2u+s+ +s− /(1+r 2 )n+m+1 , with u ∈ Jn (m − ŝ), is linearly independent, hence
the corresponding system has the associated matrix ( f u , f q )u,q∈Jn (m−ŝ) , whose
determinant is different from zero (it is a Gram determinant). This means that the
system has a unique solution. 


Theorem 16 Let A be s-diagonal. Then, there is a quasi-separately radial function

a(r )t s such that RTa(r )t s R ∗ = A.

Proof Let A = (m i j )i, j∈Jn (m) be s-diagonal, i.e., m i j = 0 if i − j = s. Let

 −1
bq = αq+s m
+
α m
q+s−
m q+s+ ,q+s− for q ∈ Jn (m − ŝ). Now, by Lemma 15, there
is a bounded measurable function a(r ) satisfying (5.3). Thus, the Toeplitz opera-
tor with quasi-separately radial symbol a(r )t s is represented by a s-diagonal matrix
RTa(r )t s R ∗ A = (ai j )i, j∈Jn (m) , where

aq+s+ ,q+s− = αq+s
m
αm
+ q+s−
a(r )r 2q+s+ +s− dμm (r ).
Rn+

Therefor, aq+s+ ,q+s− = m q+s+ ,q+s− . 



Combining these results, we have the following corollary.

Corollary 17 For each matrix A = (ai j )i, j∈Jn (m) , there is a bounded function a(r, t)
such that RTa(r,t) R ∗ = A.
Toeplitz operators with quasi-separately radial symbols… 227


Proof Let E given by (5.2). We write A = s∈E As . Hence, by Theorem 16,
As = RTas (r )t s R ∗ . Then,
 
A= As = RTas R ∗
s∈E s∈E

=R Tas (r )t s R ∗
s∈E
= RTs∈E as (r )t s R ∗ .

References
1. Brown, A., Halmos, P.R.: Algebraic properties of Toeplitz operators. J. Reine Angew. Math. 213,
89–102 (1963). (MR0160136)
2. Grudsky, S., Karapetyants, A., Vasilevski, N.: Toeplitz operators on the unit ball in Cn with radial
symbols. J. Oper. Theory 49, 325–346 (2003). (MR1991742)
3. Grudsky, S., Karapetyants, A., Vasilevski, N.: Dynamics of properties of Toeplitz operators with radial
symbols. Integral Equ. Oper. Theory 20(2), 217–253 (2004). (Zbl 1120.47305, MR2099791)
4. Grudsky, S., Quiroga-Barranco, R., Vasilevski, N.: Commutative C-algebras of Toeplitz operators and
quantization on the unit disk. J. Funct. Anal. 234(1), 1–144 (2006). (Zbl 1100.47023, MR2214138)
5. Grudsky, S., Vasilevski, N.: Bergman-Toeplitz operators: Radial component influence. Integral Equ.
Oper. Theory 40(1), 16–33 (2001). (Zbl 0993.47023, MR1829512)
6. Helgason, S.: Differential geometry, Lie groups, and symmetric spaces. Pure and Applied Mathematics,
vol. 80. Academic Press, Inc., New York (1978)
7. Kobayashi, S., Nomizu, K.: Foundations of differential geometry, vol. II. Reprint of the 1969 original.
Wiley Classics Library. A Wiley-Interscience Publication, Wiley, New York (1996)
8. Griffiths, P., Harris, J.: Principles of algebraic geometry. Wiley, New York (1978)
9. Louchi, I., Strouse, E., Zakariasy, L.: Products of Toeplitz operators on the Bergman space. Integral
Equ. Oper. Theory 54, 525–539 (2006)
10. Quiroga-Barranco, R., Sánchez-Nungaray, A.: Commutative C ∗ algebras of Toeplitz operators on
complex projective spaces. Integral Equ. Oper. Theory 71, 225–243 (2011). (MR2838143)
11. Quiroga-Barranco, R., Sánchez-Nungaray, A.: Toeplitz operators with quasi-radial quasi-homogeneous
symbols and bundles of Lagrangian frames. J. Oper. Theory 71(1), 199–222 (2014)
12. Quiroga-Barranco, R., Vasilevski, N.: Commutative algebras of Toeplitz operators on the Reinhardt
domains. Integral Equ. Oper. Theory 59, 67–98 (2007). (Zbl 1135.47057, MR2351275)
13. Quiroga-Barranco, R., Vasilevski, N.: Commutative C ∗ -algebras of Toeplitz operators on the unit ball,
I. Bargmann type transforms and spectral representations of Toeplitz operators. Integral Equ. Oper.
Theory 59(3), 379–419 (2007). (Zbl 1144.47024, MR2363015)
14. Vasilevski, N.: Toeplitz operators on the bergman spaces: inside-the-domain effects. Contemp. Math.
289, 79–146 (2001). (Zbl 1054.47512, MR1864540)
15. Vasilevski, N.: Quasi-radial quasi-homogeneous symbols and commutative Banach algebras of Toeplitz
operators. Integral Equ. Oper. Theory 66(1), 141–152 (2010)