Analysis On Fock Spaces, Kehe Zhu (2012)

Graduate Texts in Mathematics 263
Graduate Texts in Mathematics
Series Editors:
Sheldon Axler
San Francisco State University
Kenneth Ribet
University of California, Berkeley
Advisory Board:
Colin Adams, Williams College

Alejandro Adem, University of British Columbia
Ruth Charney, Brandeis University
Irene M. Gamba, The University of Texas at Austin
Roger E. Howe, Yale University
David Jerison, Massachusetts Institute of Technology
Jeffrey C. Lagarias, University of Michigan
Jill Pipher, Brown University
Fadil Santosa, University of Minnesota
Amie Wilkinson, University of Chicago
Graduate Texts in Mathematics bridge the gap between passive study and
creative understanding, offering graduate-level introductions to advanced topics
in mathematics. The volumes are carefully written as teaching aids and highlight
characteristic features of the theory. Although these books are frequently used as
textbooks in graduate courses, they are also suitable for individual study.
For further volumes:

http://www.springer.com/series/136
Kehe Zhu
Analysis on Fock Spaces
123
Kehe Zhu
Department of Mathematics and Statistics
State University of New York
Albany, NY
USA
ISSN 0072-5285
ISBN 978-1-4419-8800-3 ISBN 978-1-4419-8801-0 (eBook)
DOI 10.1007/978-1-4419-8801-0
Springer New York Heidelberg Dordrecht London
Library of Congress Control Number: 2012937293
Mathematics Subject Classification (2010): 30H20
© Springer Science+Business Media New York 2012

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection
with reviews or scholarly analysis or material supplied specifically for the purpose of being entered
and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of
this publication or parts thereof is permitted only under the provisions of the Copyright Law of the
Publisher’s location, in its current version, and permission for use must always be obtained from Springer.
Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations
are liable to prosecution under the respective Copyright Law.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
While the advice and information in this book are believed to be true and accurate at the date of
publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for
any errors or omissions that may be made. The publisher makes no warranty, express or implied, with
respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Preface
Several natural L p spaces of analytic functions have been widely studied in the past
few decades, including Hardy spaces, Bergman spaces, and Fock spaces. The terms
“Hardy spaces” and “Bergman spaces” are by now standard and well established.
But the term “Fock spaces” is a different story. I am aware of at least two other
terms that refer to the same class of spaces: Bargmann spaces and Segal–Bargmann
spaces. There is no particular reason, other than personal tradition, why I use “Fock
spaces” instead of the other variants. I have not done and do not intend to do any
research in order to justify one choice over the others.
Numerous excellent books now exist on the subject of Hardy spaces. Several
books about Bergman spaces, including some of my own, have also appeared in the
past few decades. But there has been no book on the market concerning the Fock
spaces. The purpose of this book is to fill that vacuum. There seems to be an honest
need for such a book, especially when many results are by now complete. It is at
least desirable to have the most important results and techniques summarized in one
book, so that newcomers, especially graduate students, have a convenient reference
to the subject.
There are certainly common themes to the study of the three classes of spaces
mentioned above. For example, the notions of zero sets, interpolating sets, Hankel
operators, and Toeplitz operators all make perfect sense in each of the three cases.
But needless to say, the resulting theories and results as well as the techniques
devised often depend on the underlying spaces. I will not say anything about the
various differences between the Hardy and Bergman theories; experts in these fields
are well aware of them.
What makes Fock spaces a genuinely different subject is mainly the flatness of
the domain on which these spaces are defined: the complex plane with the Euclidean
metric in our setup. Hardy and Bergman spaces are usually defined on curved
spaces, for example, bounded domains or half-spaces with a non-Euclidean metric.
Another major difference between the Fock theory and the Hardy/Bergman theory is
the behavior of the reproducing kernel in the L2 case: the Fock L2 space possesses an
exponential kernel, while the Hardy and Bergman L2 spaces both have a polynomial
kernel.
v
vi Preface
Let me mention a few particular phenomena that are unique to the analysis on
Fock spaces, as opposed to the more well-known Hardy and Bergman space settings.
First, the Fock kernel eα zw is neither bounded above nor bounded below, even
when one of the two variables is fixed. In the Hardy and Bergman theories, the
kernel function (1 − zw)α is both bounded above and bounded below when one of
the two variables is fixed. This makes many estimates in the Fock space setting
much more difficult. On the other hand, the exponential decay of e−α |z| makes it
2
much easier to prove the convergence of certain integrals and infinite series in the
Fock space setting than their Hardy and Bergman space counterparts.
Second, in the Fock space setting, there are no bounded analytic or harmonic
functions other than the trivial ones (constants). Therefore, many techniques in the
Hardy and Bergman space theories that are based on approximation by bounded
functions are no longer valid.
Third, and more technically, in the theory of Hankel and Toeplitz operators on
the Fock space, there is no “cutoff” point when characterizing membership in the
Schatten classes, while “cutoff” exists in both the Hardy and Bergman settings.
Also, for a bounded symbol function ϕ , the Hankel operator Hϕ on the Fock space
is compact if and only if Hϕ is compact. This is something unique for the Fock
spaces.
Fourth, because analysis on Fock spaces takes place on the whole complex
plane, certain techniques and methods from Fourier analysis become available. One
such example is the relationship between Toeplitz operators on the Fock space and
pseudodifferential operators on L2 (R).
And finally, I want to mention the role that Fock spaces play in quantum physics,
harmonic analysis on the Heisenberg group, and partial differential equations. In
particular, the normalized reproducing kernels in the Fock space are exactly the
so-called coherent states in quantum physics, the parametrized Berezin transform
on the Fock space provides a solution to the initial value problem on the complex
plane for the heat equation, and weighted translation operators give rise to a unitary
representation of the Heisenberg group on the Fock space.
I chose to develop the whole theory in the context of one complex variable,
although pretty much everything we do in the book can be generalized to the case
of finitely many complex variables. The case of Fock spaces of infinitely many
variables is a subject of its own and will not be discussed at all in the book.
I have tried to keep the prerequisites to a minimum. A standard graduate course
in each of real analysis, complex analysis, and functional analysis should prepare
the reader for most of the book. There are, however, several exceptions. One is
Lindelöf’s theorem which determines when a certain entire function is of finite type,
and the other is the Calderón–Vaillancourt theorem concerning the boundedness
of certain pseudodifferential operators. These two results are included in Chap. 1
without proof. Used without proof are also a couple of theorems from abstract
algebra when we characterize finite-rank Hankel and Toeplitz operators in Chaps. 6
and 7, and a couple of theorems from the general theory of interpolation when we
describe the complex interpolation spaces for Fock spaces in Chap. 2.
Preface vii
I have included some exercises at the end of each chapter. Some of these are
extensions or supplements to the main text, some are routine estimates omitted in
the main proofs, some are “lemmas” taken out of research papers, while others
are estimates or lemmas that I came up during the writing of the book that were
eventually abandoned because of better approaches found later. I have tried my best
to give a reference whenever a nontrivial result appears in the exercises.
I have tried to include as many relevant references as possible. But I am sure that
the Bibliography is not even nearly complete. I apologize in advance if your favorite
paper or reference is missing here. I did not omit it on purpose. I either overlooked
it or was not aware of it. The same is true with the brief comments I make at the end
of each chapter. I have tried my best to point the reader to sources that I consider
to be original or useful, but these comments are by no means authoritative and are
more likely biased because of my limitations in history and knowledge.
As usual, my family has been very supportive during the writing of this book.
I am very grateful to them—my wife Peijia and our sons Peter and Michael—for
their encouragement, understanding, patience, and tolerance. During the writing of
the book, I also received help from Lewis Coburn, Josh Isralowitz, Haiying Li, Alex
Schuster, Kristian Seip, Dan Stevenson, and Chunjie Wang. Thank you all!
Albany, NY, USA Kehe Zhu

Contents
1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 1
1.1 Entire Functions .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 3
1.2 Lattices in the Complex Plane . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 9
1.3 Weierstrass σ -Functions .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 13
1.4 Pseudodifferential Operators . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 19
1.5 The Heisenberg Group .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 25
1.6 Notes .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 27
1.7 Exercises.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 29
2 Fock Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 31
2.1 Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 33
2.2 Some Integral Operators .. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 43
2.3 Duality of Fock Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 53
2.4 Complex Interpolation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 59
2.5 Atomic Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 63
2.6 Translation Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 75
2.7 A Maximum Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 81
2.8 Notes .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 87
2.9 Exercises.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 89
3 The Berezin Transform and BMO . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 93
3.1 The Berezin Transform of Operators.. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 95
3.2 The Berezin Transform of Functions.. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 101
3.3 Fixed Points of the Berezin Transform.. . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 113
3.4 Fock–Carleson Measures . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 117
3.5 Functions of Bounded Mean Oscillation .. . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 123
3.6 Notes .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 133
3.7 Exercises.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 135
4 Interpolating and Sampling Sequences . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 137
4.1 A Notion of Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 139
4.2 Separated Sequences .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 143
ix
x Contents
4.3 Stability Under Weak Convergence . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 151

4.4 A Modified Weierstrass σ -Function . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 159
4.5 Sampling Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 165
4.6 Interpolating Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 177
4.7 Notes .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 187
4.8 Exercises.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 189
5 Zero Sets for Fock Spaces .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 193
5.1 A Necessary Condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 195
5.2 A Sufficient Condition .. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 197
5.3 Pathological Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 199
5.4 Notes .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 209
5.5 Exercises.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 211
6 Toeplitz Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 213
6.1 Trace Formulas .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 215
6.2 The Bargmann Transform . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 221
6.3 Boundedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 229
6.4 Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 237
6.5 Toeplitz Operators in Schatten Classes . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 245
6.6 Finite Rank Toeplitz Operators .. . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 255
6.7 Notes .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 263
6.8 Exercises.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 265
7 Small Hankel Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 267
7.1 Small Hankel Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 269
7.2 Boundedness and Compactness . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 271
7.3 Membership in Schatten Classes . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 275
7.4 Finite Rank Small Hankel Operators .. . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 281
7.5 Notes .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 283
7.6 Exercises.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 285
8 Hankel Operators .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 287
8.1 Boundedness and Compactness . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 289
8.2 Compact Hankel Operators with Bounded Symbols . . . . . . . . . . . . . . . . . . 293
8.3 Membership in Schatten Classes . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 301
8.4 Notes .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 327
8.5 Exercises.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 329
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 331
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 341
Chapter 1
Preliminaries
In this chapter, we collect several preliminary results about entire functions,

lattices in the complex plane, pseudodifferential operators, and the Heisenberg
group. The purpose is to fix notation and to facilitate references later on. All the
results concerning entire functions, except Lindelöf’s theorem, are well known and
elementary. The section about Weierstrass σ -functions is self-contained, while the
section on pseudodifferential operators is very sketchy.
K. Zhu, Analysis on Fock Spaces, Graduate Texts in Mathematics 263, 1

DOI 10.1007/978-1-4419-8801-0 1,
2 1 Preliminaries
1.1 Entire Functions 3
1.1 Entire Functions
This book is about certain spaces of entire functions and certain operators defined
on these spaces. So we begin by recalling some elementary results about entire
functions. The first few of these results can be found in any graduate-level complex
analysis text, and no proof is included here.
Let C denote the complex plane. If a function f is analytic on the entire complex
plane C, we say that f is an entire function. One of the fundamental results in
complex analysis is the following identity theorem.
Theorem 1.1. If f is entire and the zero set of f ,
Z( f ) = {z ∈ C : f (z) = 0},
has a limit point in C, then f ≡ 0 on C.

Another version of the identity theorem is the following:
Theorem 1.2. Suppose f is an entire function. If there is a point a ∈ C such that
f (n) (a) = 0 for all n ≥ 0, then f ≡ 0 on C.
When we say that {zn } is the zero sequence of an entire function f , we always
assume that any zero of multiplicity k is repeated k times in {zn }. As a consequence
of the identity theorem, we see that the zero set of an entire function that is not
identically zero cannot have any finite limit point and no value occurs infinitely
many times in the sequence. Consequently, the zero sequence {zn } of an entire
function is either finite or satisfies the condition that |zn | → ∞ as n → ∞. In particular,
we can always arrange the zeros so that |z1 | ≤ |z2 | ≤ · · · ≤ |zn | ≤ · · · .
The following result is called the mean value theorem, which follows from the
subharmonicity of the function | f (z)| p in |z − a| < R.
Theorem 1.3. Suppose f is entire and 0 < p < ∞. Then
2π
1
| f (a)| p ≤ | f (a + reiθ )| p dθ (1.1)
2π 0
for all a ∈ C and all r ∈ [0, ∞).

Because r above is arbitrary, we often multiply both sides of (1.1) by some
function of r and then integrate with respect to r. For example, if we multiply both
sides of (1.1) by r and then integrate from 0 to R, the result is

1
| f (a)| p ≤ | f (z)| p dA(z), (1.2)
π R2 |z−a|<R
where z = x + iy and dA(z) = dxdy is the Lebesgue area measure. The inequality in
(1.2) is the area version of the mean value theorem.
The next result is called Liouville’s theorem.
4 1 Preliminaries
Theorem 1.4. A bounded entire function is necessarily constant. More generally,

if a complex-valued harmonic function defined on the entire complex plane is
bounded, then it must be constant.
The lack of bounded entire functions is one of the key differences between the
theory of Fock spaces and the more classical theories of Hardy and Bergman spaces.
A central problem in complex analysis is the study of zeros of analytic functions
in specific function spaces. An important tool in any such study is the classical
Jensen’s formula below:
Theorem 1.5. Suppose that
(a) f is analytic on the closed disk |z| ≤ r,
(b) f does not vanish on |z| = r,
(c) f (0) = 1, and
(d) the zeros of f in |z| < r are {z1 , · · · , zN }, with multiple zeros repeated according
to multiplicity,
Then
N 2π
r 1
∑ log |zk | = 2π 0
log | f (reiθ )| dθ . (1.3)
k=1
If f (0) is nonzero but not necessarily 1, Jensen’s formula takes the form
N 2π
r 1
log | f (0)| = − ∑ log + log | f (reiθ )| dθ , (1.4)
k=1 |zk | 2π 0
where {z1 , · · · , zN } are zeros of f in 0 < |z| < r. More generally, if f has a zero of
order k at the origin, then Jensen’s formula takes the following form:
2π
| f (k) (0)| N
r 1
log + k logr = − ∑ log + log | f (reiθ )| dθ ,
k! k=1 |zk | 2 π 0
where {z1 , · · · , zN } are zeros of f in 0 < |z| < r.

Let f be an entire function. We can factor out the zeros of f in a canonical way,
a process that is usually referred to as Weierstrass factorization. The basis for the
Weierstrass factorization theorem is a collection of simple entire functions called
elementary factors. More specifically, we define
E0 (z) = 1 − z,
and for any positive integer n,

z2 zn
En (z) = (1 − z) exp z + + · · · + .
2 n
If a is any nonzero complex number, it is clear that E(z/a) has a unique, simple zero
at z = a.
Theorem 1.6. Let {zn } be a sequence of nonzero complex numbers such that the
sequence {|zn |} is nondecreasing and tends to ∞. Then it is possible to choose a
sequence {pn } of nonnegative integers such that
∞ pn +1
r
∑ |zn |
<∞ (1.5)
n=1
for all r > 0. Furthermore, the infinite product

∞
z
P(z) = ∏ E pn (1.6)
n=1 zn
converges uniformly on every compact subset of C, the function P is entire, and the
zeros of P are exactly {zn }, counting multiplicity.
Note that the choice pn = n − 1 will always satisfy (1.5). In many cases, however,
there are “better” choices. In particular, if {zn } is the zero sequence of an entire
function f and if there exists an integer p such that
∞
1
∑ |zn | p+1 < ∞, (1.7)
n=1
we say that f is of finite rank. If p is the smallest integer such that (1.7) is satisfied,
then f is said to be of rank p. A function with only a finite number of zeros has
rank 0. A function is of infinite rank if it is not of finite rank.
If f is of finite rank p and {zn } is the zero sequence of f , then (1.7) is satisfied
with pn = p. The product P(z) associated with this canonical choice of {pn } will be
called the standard form.
Theorem 1.7. Let f be an entire function of finite rank p. If P is the standard
product associated with the zeros of f , then there exist a nonzero integer m and
an entire function g such that
f (z) = zm P(z)eg(z) . (1.8)
The integer m is unique, and the entire function g is unique up to an additive constant
of the form 2kπ i.
For an entire function of finite rank, we say that (1.8) is the standard factorization
of f , or the Weierstrass factorization of f .
Let f be an entire function of finite rank p. If the entire function g in the standard
factorization (1.8) of f is a polynomial of degree q, then we say that f has finite
genus. In this case, the number μ = max(p, q) is called the genus of f .
Let f be an entire function. For any r > 0, we write
M(r) = M f (r) = sup{| f (z)| : |z| = r}.

6 1 Preliminaries
We say that f is of order ρ if

log log M(r)
ρ = lim sup .
r→∞ log r
It is clear that 0 ≤ ρ ≤ ∞. When ρ < ∞, f is said to be of finite order; otherwise, f

is of infinite order.
There are two useful characterizations for entire functions to be of finite order,
the first of which is the following:
Theorem 1.8. An entire function f is of finite order if and only if there exist positive
constants a and r such that
| f (z)| < exp(|z|a ), |z| > r.
In this case, the order of f is the infimum of the set of all such numbers a.
The following characterization of entire functions of finite order is traditionally
referred to as the Hadamard factorization theorem.
Theorem 1.9. An entire function f is of finite order ρ if and only if it is of finite
genus μ . Moreover, the order and genus of f satisfy the following relations: μ ≤
ρ ≤ μ + 1.
When 0 < ρ < ∞, we define
log M(r)
σ = lim sup .
r→∞ rρ
If σ < ∞, we say that f is of finite type. More specifically, we say that f is of order
ρ and type σ . If σ = ∞, we say that f is of maximum type or infinite type.
Let {zn } denote the zero sequence, excluding 0, of an entire function f . The
infimum of all positive numbers s such that
∞
1
∑ |zn |s < ∞
n=1
will be denoted by ρ1 = ρ1 ( f ). The smallest positive integer s satisfying the

convergence condition above will be denoted by m + 1.
Theorem 1.10. For any entire function f that is not identically zero, we have the
following relations among the constants defined above:
(a) ρ1 − 1 ≤ m ≤ ρ .
(b) If ρ is not an integer, then ρ = ρ1 .
(c) m = [ρ1 ] if ρ1 is not an integer.
Here, [x] denotes the greatest integer less than or equal to x.
The following result is sometimes called Lindelöf’s theorem. This result is not so
standard in the sense that it does not appear in most elementary complex analysis
texts. See [38] for a proof.
Theorem 1.11. Suppose that ρ is a positive integer, f is an entire function of order
ρ , f (0) = 0, and {zn } is the zero sequence of f . Then f is of finite type if and only
if the following two conditions hold:
(a) n(r) = O(rρ ) as r → ∞, where n(r) is the number (counting multiplicity) of
zeros of f in |z| ≤ r.
(b) The partial sums
1
S(r) = ∑ ρ
|zn |≤r zn
are bounded in r.
Lindelöf’s theorem will be useful for us in Chap. 5 when we study zero sequences
for functions in Fock spaces. The reader should be mindful of the fact that there
are several results in complex analysis that are called Lindelöf’s theorem. In most
cases, these results are certain generalizations of the classical maximum modulus
principle.
8 1 Preliminaries
1.2 Lattices in the Complex Plane 9
1.2 Lattices in the Complex Plane
The complex plane is flat, and lattices in it are easy to describe. We will need to use
rectangular lattices on several occasions later on. In this section, we fix notation and
collect basic facts about lattices in the complex plane.
The simplest lattice in C is the standard integer lattice
Z2 = {m + in : m ∈ Z, n ∈ Z},
where Z = {0, ±1, ±2, · · · , } is the integer group. All lattices we use in the book are
isomorphic to Z2 .
Let ω be any complex number, and let ω1 and ω2 be any two nonzero complex
numbers such that their ratio is not real. For any integers m and n, let ωmn = ω +
mω1 + nω2. The set
Λ = Λ (ω , ω1 , ω2 ) = {ωmn : m ∈ Z, n ∈ Z}
is then called the lattice generated by ω , ω1 , and ω2 .

The initial parallelogram at ω spanned by ω1 and ω2 has vertices
ω, ω + ω1 , ω + ω2 , ω + ω1 + ω2 ,
and is centered at
1
c = ω + (ω1 + ω2).
2
We shift this parallelogram so that the center becomes ω and the vertices become
1 1 1 1
ω − (ω1 + ω2 ), ω + (ω1 − ω2 ), ω + (ω2 − ω1 ), ω + (ω1 + ω2).
2 2 2 2
We denote this new parallelogram by R00 and call it the fundamental region of
Λ (ω , ω1 , ω2 ). For any integers m and n, let Rmn = R00 + ωmn , with ωmn being the
center of Rmn .
Lemma 1.12. Let Λ = Λ (ω , ω1 , ω2 ) be any lattice in C. For any positive number
δ , there exists a positive constant C such that
∑ e−δ |z−w|
2
≤C
z∈Λ
for all w ∈ C.
10 1 Preliminaries
Proof. By translation invariance, it suffices for us to prove the desired inequality for
w in the fundamental region R00 of Λ . If w is in the relatively compact set R00 , then
|w/z| < 1/2 for all but a finite number of points z ∈ Λ . For all such points z, we have
1
|z − w|2 = |z|2 |1 − (w/z)|2 ≥ |z|2 .
4
δ
Since ∑z∈Λ e− 4 |z| is obviously convergent, we obtain the desired result.
2

Lemma 1.13. With notation from above, we have

C= {Rmn : m ∈ Z, n ∈ Z},
and

C
f (z) dA(z) = ∑ f (z) dA(z)
m,n∈Z Rmn
for every f ∈ L1 (C, dA).

Proof. The decomposition of C into the union of congruent parallelograms is
obvious. Since any two different Rmn only overlap on a set of zero area, the desired
integral decomposition follows immediately.

In several situations later, we will need to decompose a given lattice into several
sparse sublattices. The following lemma tells us how to do it.
Lemma 1.14. Let Λ = Λ (ω , ω1 , ω2 ) be a lattice in C. For any positive number R,
there exists a positive integer N such that we can decompose Λ into the disjoint
union of N sublattices,
Λ = Λ1 ∪ · · · ∪ ΛN ,
such that the distance between any two points in each of the sublattices is at least R.
Proof. Fix a positive integer k such that k|ω1 | > R and k|ω2 | > R. For each j =
( j1 , j2 ) with 0 ≤ j1 ≤ k and 0 ≤ j2 ≤ k, let
Λ j = Λ (ω + j1 ω1 + j2 ω2 , kω1 , kω2 )
= {(ω + j1 ω1 + j2 ω2 ) + (mkω1 + nkω2 ) : m ∈ Z, n ∈ Z}.
Then each Λ j is a sublattice of Λ ; the distance between any two points in Λ j is

at least R, and Λ = ∪Λ j . There are a few duplicates among Λ j caused by points
from the boundary of the parallelogram at ω spanned by kω1 and kω2 . After these
duplicates are deleted, we arrive at the desired decomposition for Λ .

1.2 Lattices in the Complex Plane 11
Most lattices we use in the book are square ones. More specifically, for any given
positive parameter r, we consider the case when ω = 0, ω1 = r, and ω2 = ir. The
resulting lattice is
rZ2 = {rm + irn : m ∈ Z, n ∈ Z}.

We mention two particular cases. First, for r = π /α , where α is a positive
parameter, the resulting lattices are used in the next section when we introduce the
Weierstrass σ -functions. Second, for r = 1/N, where N is a positive integer, the
resulting lattices will be employed in Chaps. 6–8 when we characterize Hankel and
Toeplitz operators in Schatten classes.
For any two points z = x + iy and w = u + iv in rZ2 , we let γ (z, w) denote the
following path in rZ2 : we first move horizontally from z to u + iy and then vertically
from u + iy to u + iv. When z = 0, we write γ (w) in place of γ (0, w). The path γ (z, w)
is of course discrete. We use |γ (z, w)| to denote the number of points in γ (z, w) and
call it the length of γ (z, w).
The following technical lemma will play a critical role in Chap. 8.
Lemma 1.15. For any positive r and σ , there exists a positive constant C = Cr,σ
such that
∑ ∑ e−σ |z−w| χγ (z,w) (u) ≤ C

2
z∈rZ2 w∈rZ2
for all u ∈ rZ2 , where χγ (z,w) is the characteristic function of γ (z, w).
Proof. Without loss of generality, we may assume that r = 1. Adjusting the constant
σ will then produce the general case.
Also, it is obvious that
u + γ (z, w) = γ (u + z, u + w),
which implies that the sum
∑ ∑ e−σ |z−w| χγ (z,w) (u)

2
S=
z∈Z2 w∈Z2
is actually independent of u. For convenience, we will assume that u = 0.

For any z and w, the path γ (z, w) consists of a horizontal segment and a vertical
segment (one or both are allowed to degenerate). From the definition of γ (z, w), we
see that the origin 0 lies on the horizontal segment of γ (z, w) if and only if one of
the following is true:
(1) z is on the negative x-axis and w is in the first or fourth quadrant: z = −n,
w = m + ki, where n and m are nonnegative integers and k is an integer.
(2) z is on the positive x-axis and w is in the second or third quadrant: z = n, w =
−m + ki, where n and m are nonnegative integers and k is an integer.
12 1 Preliminaries
Similarly, 0 lies on the vertical segment of γ (z, w) if and only if one of the following
is true:
(3) w is on the positive y-axis and z is in the third or fourth quadrant: w = ni,
z = k − mi, where n and m are nonnegative integers and k is an integer.
(4) w is on the negative y-axis and z is in the first or second quadrant: w = −ni,
z = k + mi, where n and m are nonnegative integers and k is an integer.
In each of the cases above, we have
|z − w|2 = (n + m)2 + k2 ≥ n2 + m2 + k2 .
Therefore,
∞ ∞ ∞
S≤4∑ ∑ ∑ e−σ (n
2 +m2 +k2 )
n=0 m=0 k=−∞

∞ ∞ ∞
= 4 ∑ e− σ n ∑ e− σ m ∑ e−σ k < ∞.
2 2 2
n=0 m=0 k=−∞
This proves the lemma.

1.3 Weierstrass σ -Functions 13
1.3 Weierstrass σ -Functions
In this section we introduce several Weierstrass functions on the complex plane and
prove their periodicity or quasiperiodicity. In particular, the Weierstrass σ -function
will serve as a prototype for functions in Fock spaces and will play an important role
in our characterization of interpolating and sampling sequences for Fock spaces.
Lattices in this section are all based at the origin:
Λ = Λ (0, ω1 , ω2 ) = {ωmn }, ωmn = mω1 + nω2 .
To every such lattice, we associate a function P(z) = PΛ (z) as follows:

1 1 1
P(z) = 2 + ∑
− 2 , (1.9)
m,n (z − ωmn ) ωmn
z 2
where the summation (with a prime) extends over all integers m and n with (m, n) =
(0, 0).
Proposition 1.16. The function P is an even meromorphic function in the complex
plane whose poles are exactly the points in the lattice Λ . Furthermore, P is doubly
periodic with periods ω1 and ω2 :
P(z + ω1 ) = P(z), P(z + ω2 ) = P(z), (1.10)
for all z ∈ C − Λ .
Proof. For any small δ > 0, let
Uδ = {z ∈ C : d(z, Λ ) > δ , |z| < 1/δ }.
It is clear that for z ∈ Uδ we have

1 1 1
− = O
(z − ωmn )2 ωmn
2 |ωmn |3
when |ωmn | is large. Since

1
∑ |ωmn |3
< ∞,
(m,n)=(0,0)
the series in (1.9) converges uniformly and absolutely to an analytic function in Uδ .

Since δ is arbitrary, the series in (1.9) converges to an analytic function P on C − Λ .
At each point ωmn , it is clear that P has a double pole. So P is meromorphic with
double poles at precisely the points of Λ .
14 1 Preliminaries
To see that P is doubly periodic with periods ω1 and ω2 , we differentiate

the defining equation (1.9) term by term, which is permissible because the series
converges uniformly on compact subsets of C − Λ . Thus,
1
P (z) = −2 ∑ .
m,n (z − ωmn )
3
Since {−ωmn : m ∈ Z, n ∈ Z} represents the same lattice Λ and the series above
converges absolutely (so its terms can be rearranged in any way we like), we see
that P is an odd function, and so the original function P is even.
On the other hand, for each k = 1, 2, we have
1
P (z + ωk ) = −2 ∑ .
m,n (z − ωmn + ωk )3
Since {ωmn − ωk : m ∈ Z, n ∈ Z} represents the same lattice Λ and the above series
converges absolutely for any z ∈ C − Λ , we see that P (z + ωk ) = P (z), so P is
doubly periodic with periods ω1 and ω2 .
If we integrate the equation P (z + ωk ) = P (z) on the connected region C − Λ ,
we will find a constant Ck such that P(z + ωk ) = P(z) + Ck for k = 1, 2 and all
z ∈ C − Λ . Setting z = −ωk /2 and using the fact that P is even, we obtain Ck = 0
for k = 1, 2. This shows that P is doubly periodic with periods ω1 and ω2 .

To every lattice Λ = Λ (0, ω1 , ω2 ) = {ωmn }, we associate another function ζ (z) =
ζΛ (z) as follows:

1 1 1 z
ζ (z) = + ∑ + + 2 . (1.11)
z m,n z − ωmn ωmn ωmn
The following proposition lists some of the basic properties of this function, which
should not be confused with the famous Riemann ζ -function.
Proposition 1.17. Each ζ is an odd meromorphic function with simple poles at
precisely the points of Λ . Furthermore, for k = 1, 2, we have
ζ (z + ωk ) = ζ (z) + ηk , z ∈ C −Λ, (1.12)
where ηk = 2ζ (ωk /2).

Proof. Again we fix any small positive number δ and consider the region Uδ defined
in the proof of the previous proposition. It is clear that

1 1 z 1
+ + 2 =O , z ∈ Uδ ,
z − ωmn ωmn ωmn |ωmn |3
as |ωmn | → ∞. It follows that the series in (1.11) converges to an analytic function

in C − Λ , and the convergence is uniform and absolute on the relatively compact set
Uδ . It is clear that the resulting function ζ has a simple pole at (and only at) each
point of Λ .
A rearrangement of terms in the series (1.11) easily shows that ζ is an odd
function on C − Λ . Differentiating the series (1.11) term by term shows that the two
Weierstrass functions P and ζ are related by the differential equation ζ (z) = −P(z)
coupled with the condition

1
lim ζ (z) − = 0.
z→0 z
If we integrate the equation P(z + ωk ) = P(z) on the connected region C − Λ , we

obtain a constant ηk such that ζ (z + ωk ) = ζ (z) + ηk for k = 1, 2 and all z ∈ C − Λ .
Setting z = −ωk /2 and using the fact that ζ is odd, we obtain ηk = 2ζ (ωk /2). This
completes the proof of the proposition.

Because of the relations in (1.12), we say that the Weierstrass function ζ is
quasiperiodic.
Lemma 1.18. The periods ωk and the constants ηk are related by the following
equation:
η1 ω2 − η2 ω1 = 2π i. (1.13)
Proof. If we pull the center c = (ω1 + ω2 )/2 of the parallelogram spanned by ω1

and ω2 to the origin, the result is another parallelogram R = RΛ with the following
vertices:
1 1 1 1
− (ω1 + ω2 ), (ω2 − ω1 ), (ω1 − ω2 ), (ω1 + ω2 ).
2 2 2 2
Recall that R = RΛ is the fundamental region of the lattice Λ .
It is clear that ζ is analytic on R, up to the boundary, except a simple pole at the
center of R (which is the origin) with residue 1. Therefore,

ζ (z) dz = 2π i.
∂R
Break this into integration over the four sides of R and use the quasiperiodicity of
ζ . We obtain the desired result.

To every lattice Λ = Λ (0, ω1 , ω2 ) = {ωmn }, we associate yet another function
σ (z) = σΛ (z) as follows:

z z z2
σ (z) = z ∏ 1 − exp + . (1.14)
m,n ωmn ωmn 2ωmn 2
The following proposition lists some of the basic properties of the Weierstrass
σ -functions.
16 1 Preliminaries
Proposition 1.19. Each σ is an entire function whose zero set is exactly the lattice
Λ = {ωmn }. Furthermore, σ is odd and quasiperiodic in the following sense:
σ (z + ωk ) = −eηk (z+(ωk /2)) σ (z), (1.15)
where k = 1, 2 and ηk are the constants from the previous proposition.

Proof. It follows from a standard argument involving the Weierstrass product (see
Sect. 1.1) that the infinite product in (1.14) converges to an entire function σ and the
convergence is uniform and absolute on any compact subset of the complex plane.
It is also clear that the zero set of σ is exactly the lattice Λ = {ωmn }.
Replace z by −z in (1.14) and observe that {−ωmn : m ∈ Z, n ∈ Z} is exactly the
same lattice Λ (arranged differently). We see that the function σ is odd.
To prove the quasiperiodicity of σ , we note that the Weierstrass functions σ and
ζ are related by the differential equation
d
log σ (z) = ζ (z),
dz
coupled with the condition
σ (z)
lim = 1.
z→0 z
If we integrate the equation
ζ (z + ωk ) = ζ (z) + ηk
in the connected region C − Λ and then exponentiate the result, we obtain a constant
ck such that
σ (z + ωk ) = ck eηk z σ (z), z ∈ C.
Let z = −ωk /2 and use the fact that σ is odd. We get ck = −eηk ωk /2 .

Finally, in this section, we consider the special case of square lattices.
For any
positive
parameter α , we consider the lattice Λ = Λ α given by ω 1 = π /α and
ω2 = π /α i. Thus,

Λα = { π /α (m + in) : m ∈ Z, n ∈ Z}.
In this particular case, we will compute the constants ηk and relate the quasiperiod-
icity of σ to a certain isometry on Fock spaces.
Proposition 1.20. Suppose σ is the Weierstrass σ -function associated
to the square
lattice
Λ α = { ω mn }, where ω mn = π / α (m + in), so that ω = π /α and ω2 =
√ √ 1
π /α i. Then η1 = πα and η2 = − πα i. Furthermore,
α
eαω mn z− 2 |ωmn | σ (z − ωmn ) = (−1)m+n+mn σ (z)
2
(1.16)
for all z ∈ C and ωmn ∈ Λα .

Proof. In this particular case, we have

ωmn = π /α (m + in) = i π /α (n − im) = iωnm ,
where m = −m. It follows that

1 1 1 iz
ζ (iz) = + ∑ + + 2
iz m,n iz − ωmn ωmn ωmn

1 1 1 iz
= +∑
+ +
iz m,n iz − iωnm iωnm (iωnm )2

1 1 1 1 z
= +∑
+ + 2
i z m,n z − ωnm ωnm ωnm
1
= ζ (z).
i
Therefore,
2 η1
η2 = 2ζ (ω2 /2) = 2ζ (iω1 /2) = ζ (ω1 /2) = .
i i
√ √
This, along with (1.13), gives η1 = πα and η2 = − πα i.
To prove the translation relation in (1.16), observe that
ωmn = mω1 + nω2.
It follows from (1.15) and induction that
σ (z + mω1 ) = (−1)m σ (z)emη1 z+ 2 m

1 2η ω
1 1
for all positive integers m. Since σ is an odd function, it is then easy to see that the
above equation also holds for negative integers m. Similarly,
σ (z + nω2 ) = (−1)n σ (z)enη2 z+ 2 n

1 2η ω
2 2
for all integers n. Therefore,
σ (z + ωmn ) = (−1)n enη2 (z+mω1 )+ 2 n

1 2η ω
2 2 σ (z + mω1 )
nη2 (z+mω1 )+ 12 n2 η2 ω2
emη1 z+ 2 m
1 2η ω
= (−1)m+n e 1 1 σ (z)
= (−1)m+n e(nη2 +mη1 )z+nmη2 ω1 + 2 (n

1 2 η ω +m2 η ω )
2 2 1 1 σ (z).
Plug in
18 1 Preliminaries
√ √
ω1 = π /α , ω2 = π /α i, η1 = πα , η2 = − πα i.
We obtain
α
σ (z + ωmn ) = (−1)m+n+mn eαω mn z+ 2 |ωmn | σ (z)
2
for all z ∈ C and all ωmn ∈ Λα . Replacing ωmn by −ωmn , we obtain

α
eαλ mn z− 2 |ωmn | σ (z − ωmn ) = (−1)m+n+mn σ (z)
2
for all z ∈ C and all ωmn ∈ Λα .

Corollary 1.21. For any α > 0, the Weierstrass function σ associated to Λα has
the following properties:
α 2
(a) The function |σ (z)|e− 2 |z| is doubly periodic with periods π /α and i π /α .
α 2
(b) |σ (z)|e− 2 |z| ∼ d(z, Λα ), where d(z, Λα ) denotes the Euclidean distance from z
to the lattice Λα .
Proof. Property (a) follows from the quasiperiodicity of σ ; see (1.15) and (1.16).
Property (b) then follows from (a) and the fact that each point in Λα is a simple zero
of σ .

As a consequence of condition (b) above, we see that the Weierstrass σ -function
associated to Λα is of order 2 and of type α /2.
1.4 Pseudodifferential Operators 19
1.4 Pseudodifferential Operators
One of the tools we will employ in Chap. 6 when we study Toeplitz operators is
the notion of pseudodifferential operators. More specifically, Toeplitz operators on
the Fock space are unitarily equivalent to a class of pseudodifferential operators on
L2 (R). In this section, we introduce the concept of pseudodifferential operators on
the real line and collect several results in this area that will be needed later. The
references for this section are Folland’s books [92] and [93].
We begin with two well-known operators D and X defined on the space of smooth
functions on R by
1
X f (x) = x f (x), D f (x) = f (x), (1.17)
2α i
where α is any fixed positive constant. The introduction of a parameter α at this
point will facilitate and simplify our computations later in association with the Fock
spaces. The number h = π /α plays the role of Planck’s constant in quantum physics.
It is easy to verify that, as densely defined unbounded operators on L2 (R, dx),
both D and X are self-adjoint. This is an easy consequence of integration by parts.
The operators
Z = X + iD, Z ∗ = X − iD, (1.18)
will also be useful in our discussions.

If f is a sufficiently good function on R, it is clear how to define f (D) and f (X),
respectively. For example, if f (x) = ∑ ak xk is a polynomial, then
f (D) = ∑ ak Dk , f (X) = ∑ ak X k
are perfectly and naturally defined. This easily extends to a large class of symbol
functions f . What results in are symbol calculi for the self-adjoint operators D
and X.
The notion of pseudodifferential operators arises when we try to establish a
symbol calculus for the pair of operators D and X. In other words, if we are given a
good function f (ζ , x) on R × R, we wish to define an operator f (D, X) in a natural
way. If f = aζ + bx is linear, obviously we should just define f (D, X) = aD + bX.
But we already run into problems when f is just a second-degree polynomial, say
f (ζ , x) = ζ x = xζ ,
because now we have two natural choices,
f (D, X) = DX or f (D, X) = XD.

20 1 Preliminaries
The operators D and X do not commute, so the two products above are not equal. In
fact, it is easy to verify the following commutation relation:
1
[D, X] = DX − XD = I, (1.19)
2α i
where I is the identity operator.

If f (ζ , x) is a polynomial in ζ and x, then there are several canonical ways to
define f (D, X). For example, if we want the differentiations to come before any
multiplication, then we write
f (ζ , x) = ∑ amn xm ζ n
and define
f (D, X) = ∑ amn X m Dn .
Similarly, if we want to perform multiplications before differentiations, then we

write
f (ζ , x) = ∑ amn ζ n xm
and define
f (D, X) = ∑ amn Dn X m .
Again, the resulting operators are generally different.

It is also possible to carry out the above constructions using the operators Z and
Z ∗ from (1.18) and think of a function of two real variables as depending on z and
z. More specifically, if
σ (z, z) = ∑ amn zn zm = ∑ amn zm zn
is a polynomial in z and z, then we can define
σw (Z ∗ , Z) = ∑ amn Z ∗m Z n , σaw (Z, Z ∗ ) = ∑ amn Z n Z ∗m . (1.20)
The functional calculi defined this way are called Wick and anti-Wick correspon-
dences. They have been studied extensively in analysis and mathematical physics.
There is another important functional calculus for D and X, the John–Nirenberg
correspondence, which is especially important in partial differential equations.
We will not pursue any of the above correspondences. Instead, we focus on the
so-called Weyl pseudodifferential operators. This approach depends on a particular,
but natural, choice for the definition of σ (D, X) when σ (ζ , x) = e2π i(pζ +qx), where p
and q are real constants. Once this is done, the definition of σ (D, X) for more general
symbol functions σ (ζ , x) can be given with the help of Fourier and inverse Fourier
transforms.
Definition 1.22. For any real coefficients p and q, we define
e2α i(pD+qX) f (x) = e2α iqx+α ipq f (x + p), (1.21)
or, equivalently,
π 2 ipq
π p
e2π i(pD+qX) f (x) = e2π iqx+ α f x+ . (1.22)
α
To see the rationale behind the definition above, let

g(x,t) = e2α it(pD+qX) f (x)
denote the formal solution to the differential equation
∂g
= 2α i(pD + qX)g, (1.23)
∂t
subject to the initial condition g(x, 0) = f (x). Rewrite the equation in (1.23) as
∂g ∂g
−p = 2α iqxg, (1.24)
∂t ∂x
and let G(t) = g(x(t),t) with x(t) = x − pt. Then by the chain rule,
∂g ∂g
G (t) = −p ,
∂t ∂x
so G(t) satisfies the following equations:
G (t) = 2α iq(x − pt)G(t), G(0) = f (x).
It is elementary to solve the above equation and obtain
G(t) = f (x)e2α qixt−α it

2 pq
.
Let t = 1. We have
g(x − p, 1) = f (x)e2α iqx−α ipq .
Replace x by x + p. We arrive at
e2α i(pD+qX) f (x) = g(x, 1) = e2α iqx+α ipq f (x + p).
This gives a justification for the definition in (1.21).

22 1 Preliminaries
More generally, if σ (ζ , x) is regular enough so that we can perform the Fourier

and inverse Fourier transforms on it, then

σ (ζ , x) = σ (p, q)e2π i(pζ +qx) dp dq, (1.25)
R R
and we define

σ (D, X) = σ (p, q)e2π i(pD+qX) dp dq. (1.26)
R R
Here, the integral is an ordinary Bochner integral whenever σ , the Fourier transform
of σ , is in L1 (R × R).
Theorem 1.23. If σ (ζ , x) and f (x) are regular enough, then we have

α x + y 2α i(x−y)ζ
σ (D, X) f (x) = σ ζ, e f (y) dy dζ . (1.27)
π R R 2
Proof. The Fourier inversion formula

e2π i(u−v)ζ f (v) dv dζ = f (u)
R R
can be expressed in the language of distributions as

e2π ixζ dζ = δ (x), (1.28)
R
where δ (x) is classical δ -function. Therefore,
σ (D, X) f (x)

= σ (p, q)e2π i(pD+qX) f (x) dp dq
R R
π p 2π iqx+ π 2 pqi
= σ (p, q) f x + e α dp dq
R R α
π p
π 2 pq
= σ (ζ , w)e−2π i(pζ +qw) e2π iqx+
f x+ dp dq dζ dw
α
R R R R α
π p −2π ipζ π p
= σ (ζ , w)δ x − w + e f x+ dp dζ dw
R R R 2α α
π p −2π ipζ π p
= σ ζ,x + e f x+ dp dζ
R R 2α α

α 1
= σ ζ , (x + y) e−2α i(y−x)ζ f (y) dy dζ ,
π R R 2
which is the desired formula.

It is thus also natural to simply take (1.27) as the definition of the Weyl
pseudodifferential operator σ (D, X). We remind the reader that there is a positive
parameter α built into our definition of pseudodifferential operators. To see the pre-
cise relationship between our rescaled σ (D, X) and the classical pseudodifferential
operators (as defined in Folland’s book [92], for example), we change variables and
rewrite (1.27) as follows:

x + y 2π i(x−y)ζ
σ (D, X) f1/r (rx) = σr ζ , e f (y) dy dζ , (1.29)
R R 2

where r = π /α . Here, fr (x) = f (rx) denotes the dilation of f by a positive
number r. The integral on the right-hand side of (1.29) is the classical definition
of the Weyl pseudodifferential operator with symbol σr .
The results in the three theorems below are all invariant under dilation. Therefore,
our rescaling does not alter the validity of these classical results.
The pseudodifferential operator σ (D, X) is so far only loosely defined. If σ
is sufficiently regular and f is compactly supported on R, then the integral in
(1.27) converges. For general σ , the integral in (1.27) may or may not converge,
and the definition of σ (D, X) f may only be defined for f in a certain class.
Our main concern here is the following problem: for which functions σ can the
pseudodifferential operator σ (D, X) be extended to a bounded or compact operator
on L2 (R, dx)?
Theorem 1.24. Suppose σ (ζ , x) is a function on R × R of class C3 and there exists
a positive constant C such that
n+m
∂ σ
∑ ∂ ζ n ∂ xm
(ζ , x) ≤C

n+m≤3
for all ζ and x in R. Then the pseudodifferential operator σ (D, X) is bounded on

L2 (R, dx).
The above result is usually referred to as the Calderón–Vaillancourt theorem.
Let C0 (C) = C0 (R × R) be the space continuous functions f on C = R × R such
that f (z) → 0 as z → ∞. The following is the compactness version of the Calderón–
Vaillancourt theorem.
Theorem 1.25. Suppose σ (ζ , x) is a function on R × R of class C3 and
∂ n+m σ
∈ C0 (R × R)
∂ ζ n ∂ xm
for every pair of nonnegative integers m and n with n + m ≤ 3. Then the pseudodif-
ferential operator σ (D, X) is compact on L2 (R, dx).
24 1 Preliminaries
There is also a result concerning membership of the pseudodifferential operators

σ (D, X) in Schatten classes. We refer the reader to [250] for a brief discussion of
Schatten class operators on a Hilbert space.
Theorem 1.26. Suppose 1 ≤ p < ∞ and there exists a positive constant k = k(p)
such that
∂ n+m σ
∈ L p (R × R, dxdζ )
∂ ζ n ∂ xm
for all nonnegative integers m and n with n + m ≤ k. Then the pseudodifferential

operator σ (D, X) belongs to the Schatten class S p .
1.5 The Heisenberg Group 25
1.5 The Heisenberg Group
Although we will not use the Heisenberg group in a critical way anywhere in the
book, it is interesting to show how it fits nicely in the theory of Fock spaces. In this
brief section, we give its definition and produce a unitary representation based on
pseudodifferential operators.
The Heisenberg group H is the set C × R (or R2 × R) with the following group
operation:
(z, s) ⊕ (w,t) = (z + w, s + t − Im (zw)),
where z and w are complex and s and t are real.

More generally, if n is any positive integer, the Heisenberg group Hn is the set
Cn × R with the group operation
(z, s) ⊕ (w,t) = (z + w, s + t − Im (z, w)),
where z = (z1 , · · · , zn ), w = (w1 , · · · , wn ), and
z, w = z1 w1 + · · · + zn wn .
There is a natural representation of the Heisenberg group as unitary operators on

the Hilbert space L2 (R, dx). To simplify notation, let us write
ρ (p, q) = e2α i(pD+qX)
for real p and q.

Lemma 1.27. We have
ρ (p1 , q1 )ρ (p2 , q2 ) = eα i(p1 q2 −p2 q1 ) ρ (p1 + p2 , q1 + q2)
for all real numbers p1 , q1 , p2 , and q2 .

Proof. This follows directly from the definition of ρ (p, q) in (1.21). Details are left
to the reader.

Lemma 1.28. We have
ρ (p1 , q1 )ρ (p2 , q2 ) = e2α i(p1 q2 −p2 q1 ) ρ (p2 , q2 )ρ (p1 , q1 )
for all real numbers p1 , q1 , p2 , and q2 .

Proof. This is a direct consequence of Lemma 1.27.

Theorem 1.29. Suppose α is any positive parameter and pseudodifferential oper-
ators are defined as in the previous section. For any real p and q, the pseudodif-
26 1 Preliminaries
ferential operator e2α i(pD+qX) is a unitary operator on L2 (R, dx). Furthermore, the
mapping
(p + iq,t) → u(p + iq,t) =: eα it e2α i(pD+qX)
is a unitary representation of the Heisenberg group H on L2 (R, dx).

Proof. By (1.21), the action of each u(z,t) on L2 (R, dx), where z ∈ C and t ∈ R, is
a unimodular constant times a certain translation of R. Since any translation of R is
a unitary operator on L2 (R, dx), we see that each u(p + iq,t) is a unitary operator on
L2 (R, dx).
Let z1 = p1 + iq1 and z2 = p2 + iq2 . It follows from Lemma 1.27 that
u(z1 ,t1 )u(z2 ,t2 ) = eα i(t1 +t2 ) ρ (p1 , q1 )ρ (p2 , q2 )

= eα i(t1 +t2 +p1 q2 −p2 q1 ) ρ (p1 + p2 , q1 + q2)
= u(z1 + z2 ,t1 + t2 − Im (z1 z2 )).
This shows that u(z,t) preserves the group operation in the Heisenberg group H.

The mapping u(z,t) is called the Schrödinger representation of the Heisenberg
group H on L2 (R). In the next chapter, we will obtain another representation of H,
a unitary representation on the Fock space based on weighted translations.
1.6 Notes 27
1.6 Notes
The results in the first section, except Lindelöf’s theorem, are all well known and
can be found in any elementary complex analysis book. In particular, these results
can all be found in Conway’s book [67].
Lindelöf’s theorem will be needed in Chap. 3 when we study zero sequences for
Fock spaces. This is probably not a result that can be found in elementary texts. See
2.10.1 of Boas’ book [38] for a detailed proof of this result.
The section about lattices in the complex plane is completely elementary.
Whenever we really use lattices later on, we restrict our attention to square lattices,
although many arguments can easily be adapted to arbitrary lattices, even to
sequences that behave like lattices. Perhaps Lemma 1.15 looks peculiar to the reader,
but it is critical for the study of Hankel operators in Chap. 8.
Pseudodifferential operators constitute an important subject by itself, and there is
extensive literature about them. Of course, we have only touched the surface of this
vast area of modern analysis. The connection between pseudodifferential operators
and Toeplitz operators on the Fock space is both fascinating and useful. Because
of this connection, the study of Toeplitz operators on the Fock space becomes
especially interesting and fruitful. In particular, this provides us with extra and
unique tools to study Toeplitz operators on the Fock space as opposed to Toeplitz
operators on the Hardy and Bergman spaces.
Our presentation in Sect. 1.4 follows Folland’s books [92, 93] very closely. A
slight modification is made in the definition of pseudodifferential operators here in
order to incorporate the weight parameter α into everything. Note that the proof of
Theorem 1.23 depends on certain elementary facts from Fourier analysis that we are
taking for granted. It should be easy for the interested reader to make the arguments
completely rigorous.
The Heisenberg group appears very naturally in many different areas, including
Fourier analysis, harmonic analysis, and mathematical physics. The Heisenberg
group shows up in this book when we study the action of translations on Fock
spaces. Although it is possible for us to avoid the Heisenberg group, we thought
it is nice to put things in the right context.
The Weierstrass σ -functions provide a family of examples that will be very useful
to us later on when we study zero sets, interpolating sets, and sampling sets. The
book [241] contains much more information about the Weierstrass σ -functions as
well as several other important classes of entire and meromorphic functions.
28 1 Preliminaries
1.7 Exercises 29
1.7 Exercises
1. Suppose f is entire, f (z) = 0 for some z ∈ C, and |z| < r. Then log | f (z)| is
equal to
2 iθ
N r − zk z 1 2π re + z
− ∑ log +
2π 0 Re iθ − z
log | f (reiθ )| dθ ,
k=1 r(z − zk ) re
where {z1 , · · · , zN } are the zeros of f in 0 < |w| < r.

2. If u is a bounded (complex-valued) harmonic function on the entire complex
plane, then u must be constant.
3. Show that

1
∑ (n2 + m2) p : n ∈ Z, m ∈ Z < ∞
if and only if p > 1.
4. Suppose rZ2 = {ωmn } is any square lattice and R is any other positive radius.
Show that there exists a positive constant C = C(r, R) such that

∑ f (z) dA(z) ≤ C f (z) dA(z)
m,n |z−ωmn |<R C
for all nonnegative functions f on C. Here dA is area measure.

5. Verify that H with the operation defined in Sect. 1.5 is indeed a group.
6. Show that the Heisenberg group is nonabelian.
7. Show that ρ1 ≤ ρ . See Sect. 1.1 for definitions of these numbers.
8. Suppose f is entire, 0 R
if and only if f is identically zero.

9. Show that both X and D are self-adjoint operators on L2 (R, dx).
10. Justify every interchange of the order of integration in the proof of
Theorem 1.23.
11. Discuss the continuity of the Schrödinger representation, namely, the unitary
representation of the Heisenberg group given in Theorem 1.29.
12. Show that for any lattice Λ = {ωmn }, we have
1
∑ |ωmn | p < ∞
m,n
if and only if p > 2, where the summation is to exclude the possible occurrence
of 0 in the denominator.
13. Prove the commutation relation (1.19).
14. Convince yourself that the formal identity (1.28) is equivalent to the Fourier
inversion formula.
Chapter 2
Fock Spaces
In this chapter, we define Fock spaces and prove basic properties about them.
The following topics are covered in this chapter: reproducing kernel, integral
representation, duality, complex interpolation, atomic decomposition, translation
invariance, and a version of the maximum modulus principle.

DOI 10.1007/978-1-4419-8801-0 2,
32 2 Fock Spaces
2.1 Basic Properties 33
2.1 Basic Properties
For any positive parameter α , we consider the Gaussian measure
α −α |z|2
dλα (z) = e dA(z),
π
where dA is the Euclidean area measure on the complex plane. A calculation with
polar coordinates shows that dλα is a probability measure.
The Fock space Fα2 consists of all entire functions f in L2 (C, dλα ). It is easy
to show that Fα2 is a closed subspace of L2 (C, dλα ). Consequently, Fα2 is a Hilbert
space with the following inner product inherited from L2 (C, dλα ):

f , gα = f (z) g(z) dλα (z).
C
Proposition 2.1. For any nonnegative integer n, let

αn n
en (z) = z .
n!
Then the set {en } is an orthonormal basis for Fα2 .

Proof. A calculation with polar coordinates shows that {en } is an orthonormal set.
Given f ∈ Fα2 and n ≥ 0, we have

f , en α = lim f (z) en (z) dλα (z).
R→∞ |z|<R
Since the Taylor series

∞
f (z) = ∑ ak zk
k=0
converges uniformly on |z| < R, we have

∞
|z|<R
f (z) en (z) dλα (z) = ∑ ak |z|<R
zk en (z) dλα (z).
k=0
Using polar coordinates again, we obtain

f , en α = lim an zn en (z) dλα (z) = an zn en (z) dλα (z).
R→∞ |z|<R C
Therefore, the condition that f , en = 0 for all n ≥ 0 implies that an = 0 for all
n ≥ 0 which in turn implies that f = 0. This shows that the system {en } is complete
in Fα2 .

34 2 Fock Spaces
As a consequence of the above proposition, the Taylor series of every function f

in Fα2 converges to f in the norm topology of Fα2 .
For any fixed w ∈ C, the mapping f → f (w) is a bounded linear functional on
Fα2 . This follows easily from the mean value theorem. By the Riesz representation
theorem in functional analysis, there exists a unique function Kw in Fα2 such that
f (w) = f , Kw α for all f ∈ Fα2 . The function Kα (z, w) = Kw (z) is called the
reproducing kernel of Fα2 .
Proposition 2.2. The reproducing kernel of Fα2 is given by
Kα (z, w) = eα z w , z, w ∈ C.
Proof. For any f ∈ Fα2 , we have

f (0) = f , e0 α = f (z) dλα (z).
C
Fix any w ∈ C and replace f (z) by f (w − z). We obtain

α
f (w − z)e−α |z| dA(z)
2
f (w) =
π C

α
f (z)e−α |z−w| dA(z)
2
=
π C

= e−α |w| f (z)eα zw+α zw dλα (z).
2
Replace f (z) by f (z)e−α zw . The result is

f (w) = f (z)eα zw dλα (z).
C
The desired result then follows from the uniqueness in Riesz representation.

Recall that every closed subspace X of a Hilbert space H uniquely determines an
orthogonal projection P : H → X.
Corollary 2.3. The orthogonal projection
Pα : L2 (C, dλα ) → Fα2
is an integral operator. More specifically,

Pα f (z) = Kα (z, w) f (w) dλα (w)
C
for all f ∈ L2 (C, dλα ) and all z ∈ C.

Proof. Fix f ∈ L2 (C, dλα ) and z ∈ C. We have
Pα f (z) = Pα f , Kz α = f , Pα Kz α = f , Kz α

= f (w)Kα (z, w) dλα (z).
C
This proves the integral representation for Pα .

For any z ∈ C, we let
Kα (w, z) α 2
kz (w) = = eα zw− 2 |z|
Kα (z, z)
denote the normalized reproducing kernel at z. Each kz is a unit vector in Fα2 . The
following change of variables formula will be used many times later in the book.
Corollary 2.4. Suppose f ≥ 0 or f ∈ L1 (C, dλα ). Then for any z ∈ C, we have

f (z ± w) dλα (w) = f (w)|kz (w)|2 dλα (w),
C C
and

f [±(z − w)]|kz (w)|2 dλα (w) = f (w) dλα (w).
C C
Proof. It is clear that

α
f (z ± w)e−α |w| dA(w)
2
f (z ± w) dλα (w) =
C π C

α
f (w)e−α |z−w| dA(w)
2
=
π C

f (w)e−α |z|
2 +α zw+α zw
= dλα (w)
C

= f (w)|kz (w)|2 dλα (w).
C
The assumption that f ≥ 0 or f ∈ L1 (C, dλα ) ensures that all integrals above make
sense. The proof of the other identity is similar.

Corollary 2.5. Suppose α > 0 and β is real. Then

β zā
e dλα (z) = eβ |a| /4α
2 2
for all a ∈ C.
36 2 Fock Spaces
Proof. It follows from the definition of the reproducing kernel that

Kα (a, a) = |Kα (a, z)|2 dλα (z), a ∈ C.
C
Replacing a by β a/(2α ), we obtain the desired result.

For α > 0 and p > 0, we use the notation Lαp
to denote the space of Lebesgue
measurable functions f on C such that the function f (z)e−α |z| /2 is in L p (C, dA).
2
p
For f ∈ Lα , we write

pα α 2 p
f pp,α = f (z)e− 2 |z| dA(z). (2.1)
2π C
Similarly, for α > 0 and p = ∞, we use the notation L∞

α to denote the space of
Lebesgue measurable functions f on C such that

f ∞,α = esssup | f (z)|e−α |z| /2 : z ∈ C < ∞.
2
(2.2)
Obviously, we have Lαp = L p (C, dλ pα /2) for 0 0 and 0 < p ≤ ∞ we let Fαp denote the space of entire functions in
Lα . We will call Fαp Fock spaces. It is elementary to show that Fαp is closed in Lαp .
p
Therefore, Fαp is a Banach space when 1 ≤ p ≤ ∞, and it is a complete metric space

when 0 < p < 1.
Note that the measure associated with the Fock space Fαp , dλ pα /2, depends on
both α and p. This is a bit unusual and unnatural at first glance, but there are
underlying reasons why Fock spaces should be defined this way, and plenty of
past experience suggests that this way of defining the Fock spaces will make the
statement of many results a lot easier and a lot more natural.
Lemma 2.6. Suppose α > 0, ζ ∈ C − {0}, and 0 < p ≤ ∞. Then the dilation
operator f (z) → f (ζ z) is an isometry from Lαp onto L|pζ |2 α , and it is an isometry
from Fαp onto F|ζp |2 α .
Proof. This follows from a simple change of variables.

The following result gives the optimal rate of growth for functions in Fock
spaces.
Theorem 2.7. For any 0 < p ≤ ∞ and z ∈ C, we have
sup{| f (z)| : f p,α ≤ 1} = eα |z|

2 /2
.
Furthermore, for 0 < p < ∞, any extremal function is of the form:

α
f (w) = eα zw− 2 |z|
2 +iθ
,
where θ is a real number.

Proof. We first assume that 0 < p < ∞.
The case z = 0 follows from the subharmonicity of the function | f | p and
integration in polar coordinates:
p

pα f (w)e−
α |w|2
dA(w) = f pp,α .
| f (0)| p ≤
2

2π C
Equality occurs if and only if f is constant.

More generally, for any z ∈ C and f ∈ Fαp , we consider the function
F(w) = f (z − w)eα wz−(α |z|

2 /2)
.
From the inequality |F(0)| p ≤ F pp,α we deduce that

pα 2 p
| f (z)| p e−α p|z| f (z − w)eα zw e−α |z| /2 e−α |w| /2 dA(w)
2 /2 2
≤
2π C

pα 2 p
= f (z − w)e−α |z−w| /2 dA(w)
2π C

pα 2 p
= f (w)e−α |w| /2 dA(w)
2π C
p
= f p,α .
This shows that

| f (z)| ≤ f p,α eα |z|
2 /2
.
Furthermore, equality is attained if and only if F is constant. This shows that the
extremal functions are of the form
f (w) = eα zw−(α |z|

2 /2)+iθ
.
This proves the desired results for 0 < p < ∞.

If p = ∞, it follows from the definition of f ∞,α that | f (z)| ≤ eα |z| /2 for all f
2
with f ∞,α ≤ 1. Therefore,
sup{| f (z)| : f ∞,α ≤ 1} ≤ eα |z|

2 /2
.
38 2 Fock Spaces
On the other hand, the function f (w) = kz (w) is a unit vector in Fα∞ and kz (z) =
eα |z| /2 . Thus, we actually have
2
sup{| f (z)| : f ∞,α ≤ 1} = eα |z|

2 /2
.
This proves the case for p = ∞.

When p = ∞, the extremal functions in Theorem 2.7 consist of more than
constant multiples of reproducing kernels. For example, if f is any polynomial
normalized so that f ∞,α = 1, then
1 = sup | f (z)|e−α |z| = | f (z0 )|e−α |z0 |

2 /2 2 /2
z∈C
for some z0 ∈ C because in this case we have
lim f (z)e−α |z|

2 /2
= 0.
z→∞
Therefore, this polynomial f is an extremal function for the extremal problem in

Theorem 2.7 when p = ∞ and z = z0 .
Corollary 2.8. Let f ∈ Fαp and 0 < p ≤ ∞. Then
| f (z)| ≤ f p,α eα |z|

2 /2
for all z ∈ C and the estimate is sharp.

When 0 < p < ∞, the estimate above can be somewhat improved. More
specifically, we can actually show that
lim f (z)e−α |z|

2 /2
=0
z→∞
for every function f ∈ Fαp . This will follow from the next proposition.
Proposition 2.9. Suppose 0 < p < ∞, f ∈ Fαp , and fr (z) = f (rz). Then:
(a) fr − f p,α → 0 as r → 1− .
(b) There is a sequence {pn } of polynomials such that pn − f p,α → 0 as n → ∞.
Proof. Suppose {gn } and g are functions in L p (X, d μ ) such that
gn (x) → g(x), n → ∞,
almost everywhere. Then it is well known that

lim |gn − g| p dμ = 0
n→∞ X
if and only if

lim |gn | p dμ = |g| p dμ .
n→∞ X X
This is a simple consequence of Fatou’s lemma; see Lemma 3.17 of [119] for
example. Given f ∈ Fαp , we have

pα 2 p
fr pp,α = f (rz)e−α |z| /2 dA(z)
2π C
p
pα −α |z|2 /2 −pα |z|2 (r−2 −1)/2
= f (z)e e dA(z).
2π r 2 C
Since
2 (r−2 −1)/2
e−pα |z| ≤1
for all z ∈ C and 0 < r < 1, an application of the dominated convergence theorem
shows that fr p,α → f p,α , and hence fr − f p,α → 0 as r → 1− . This proves
part (a).
Part (b) follows from part (a) if we can show that for every r ∈ (0, 1), the function
fr can be approximated by its Taylor polynomials in the norm topology of Fαp .
To this end, we fix some r ∈ (0, 1) and fix some β ∈ (r2 α , α ). It follows from
Corollary 2.8 that fr ∈ Fβ2 . Similarly, it follows from Corollary 2.8 that Fβ2 ⊂ Fαp
and there exists a positive constant C such that g p,α ≤ Cg2,β for all g ∈ Fβ2 .
Now, if pn is the nth Taylor polynomial of fr , then by Proposition 2.1,
fr − pn p,α ≤ C fr − pn 2,β → 0
as n → ∞. This proves part (b).

Let fα∞ denote the space of entire functions f (z) such that
lim f (z)e−α |z|

2 /2
= 0.
z→∞
Obviously, fα∞ is a closed subspace of Fα∞ . In fact, fα∞ is the closure in Fα∞ of the set
of all polynomials. Thus, the space fα∞ is separable while the space Fα∞ is not.
p q
Theorem 2.10. If 0 < p < q < ∞, then Fα ⊂ Fα , and the inclusion is proper and
continuous. Moreover, Fα ⊂ fα∞ , and the inclusion is proper and continuous.
p
Proof. For any entire function f , we consider the integral

qα
| f (z)e−α |z|
2 /2
f qq,α = |q dA(z).
2π C
40 2 Fock Spaces
It follows from the pointwise estimate in Corollary 2.8 that

qα
| f (z)| p | f (z)|q−p e−qα |z| /2 dA(z)
q 2
f q,α =
2π C

qα
| f (z)| p e−pα |z| /2 dA(z)
q−p 2
≤ f p,α
2π C
q
= f qp,α .
p
This shows that Fαp ⊂ Fαq with f q,α ≤ (q/p)1/q f p,α for all f ∈ Fαp .
To see that the inclusion Fαp ⊂ Fαq is proper, let us assume that Fαp = Fαq . Then the
identity map I : Fαp → Fαq is bounded, one-to-one, and onto. By the open mapping
theorem, there must exist a constant C > 0 such that
C−1 f p,α ≤ f q,α ≤ C f p,α
for all f ∈ Fαp . On the other hand, a computation with Stirling’s formula shows that
∞
rnp e−pα r
2 /2
zn pp,α = αp r dr
0

1 np/2 np
= Γ +1
αp 2
n np/2 √
∼ n.
αe
Thus,
n n/2 1
zn p,α ∼ n 2p ,
αe
and similarly,
n n/2 1
zn q,α ∼ n 2q .
αe
It is then obvious that there is no positive constant C with the property that
C−1 zn p,α ≤ zn q,α ≤ Czn p,α

q
for all n. This contradiction shows that the inclusion Fαp ⊂ Fα must be proper.
To show that Fαp ⊂ fα∞ , observe that for every polynomial f , we have f ∈ fα∞ , and
it follows from Corollary 2.8 that f ∞,α ≤ f p,α . The desired result then follows
from the density of polynomials in Fαp , the boundedness of the inclusion Fαp ⊂ Fα∞ ,
and the fact that fα∞ is closed in Fα∞ .
Finally, by elementary calculations,

n n/2
zn ∞,α = .
αe
Another appeal to the open mapping theorem then shows that the inclusion Fαp ⊂ fα∞
is proper.

The next result gives another useful dense subset of Fαp .
Lemma 2.11. For any positive parameters α and γ , the set of functions of the form
n n
f (z) = ∑ ck Kγ (z, wk ) = ∑ ck eγ zwk ,
k=1 k=1
is dense in Fαp and fα∞ , where 0 < p < ∞.

Proof. Since the points wk are arbitrary, we may assume that γ = α .
The result is obvious when p = 2. In fact, if a function h in Fα2 is orthogonal to
each function f (z) = Kα (z, w), then h(w) = 0 for every w.
In general, with the help of Corollary 2.8, we can find a positive parameter β
such that Fβ2 ⊂ Fαp continuously, say f p,α ≤ C f 2,β for all f ∈ Fβ2 . In fact, any
β ∈ (0, α ) works. Now, if f is a polynomial and {w1 , · · · , wn } are points in the
complex plane, then
n n
f − ∑ ck Kα (z, wk ) p,α ≤ C f − ∑ ck Kα (z, wk )2,β
k=1 k=1
n
= C f − ∑ ck Kβ (z, α wk /β )2,β .
k=1
Combining this with the density of the functions ∑nk=1 ck Kβ (z, uk ) in Fβ2 , we
conclude that every polynomial can be approximated in the norm topology of Fαp
by functions of the form ∑nk=1 ck Kα (z, wk ). Since the polynomials are dense in Fαp ,
we have proved the result for Fαp , 0 < p < ∞.
The proof for fα∞ is similar.

Finally, in this section, as a consequence of the pointwise estimates, we establish
the maximum order and type for functions in the Fock spaces.
Theorem 2.12. Let f ∈ Fαp with 0 < p ≤ ∞. Then f is of order less than or equal
to 2. When f is of order 2, it must be of type less than or equal to α /2.
Proof. By Corollary 2.8, there exists a positive constant C such that
| f (z)| ≤ Ceα |z|

2 /2
42 2 Fock Spaces
for all z ∈ C. In particular, M(r) ≤ Ceα r /2 for all r > 0. It follows that the order ρ
2
of f satisfies
log log M(r)
ρ = lim sup ≤ 2.
r→∞ log r
Also, if the order of f is actually 2, then its type σ satisfies
log M(r) α
σ = lim sup ≤ ,
r→∞ r2 2
completing the proof of the theorem.

2.2 Some Integral Operators 43
2.2 Some Integral Operators
In this section, we consider the boundedness of certain integral operators on L p

spaces associated with Gaussian measures. More specifically, for any α > 0, we
consider the integral operators Pα and Qα defined by

Pα f (z) = Kα (z, w) f (w) dλα (w), (2.3)
C
and

Qα f (z) = |Kα (z, w)| f (w) dλα (w), (2.4)
C
respectively.
We need two well-known results from the theory of integral operators. The first
one concerns the adjoint of a bounded integral operator.
Lemma 2.13. Suppose 1 ≤ p < ∞ and 1/p + 1/q = 1. If an integral operator

T f (x) = H(x, y) f (y) dμ (y)
X
is bounded on L p (X, d μ ), then its adjoint
T ∗ : Lq (X, d μ ) → Lq (X, d μ )
is the integral operator given by

T ∗ f (x) = H(y, x) f (y) d μ (y).
X
Proof. This is a standard result in real analysis. See [113] for example.

The second result is a useful criterion for the boundedness of integral operators
on L p spaces, which is usually referred to as Schur’s test.
Lemma 2.14. Suppose H(x, y) is a positive kernel and

T f (x) = H(x, y) f (y) dμ (y)
X
is the associated integral operator. Let 1 < p < ∞ with 1/p + 1/q = 1. If there exist
a positive function h(x) and positive constants C1 and C2 such that

H(x, y)h(y)q dμ (y) ≤ C1 h(x)q , x ∈ X,
X
44 2 Fock Spaces
and

H(x, y)h(x) p dμ (x) ≤ C2 h(y) p , y ∈ X,
X
then the operator T is bounded on L p (X, d μ ). Moreover, the norm of T on L p (X, d μ )
1/q 1/p
does not exceed C1 C2 .
Proof. See [250] for example.

We now consider the action of the operators Pα and Qα on the space L p (C, dλβ ).
Thus, we fix two positive parameters α and β for the rest of this section and rewrite
the integral operators Pα and Qα as follows:

α
eα zw̄+β |w|
2 −α |w|2
Pα f (z) = f (w) dλβ (w),
β C
and

α
|eα zw̄+β |w|
2 −α |w|2
Qα f (z) = | f (w) dλβ (w).
β C
It follows from Lemma 2.13 that the adjoint of Pα and Qα with respect to the integral
pairing

f , gβ = f (z)g(z) dλβ (z)
C
is given, respectively, by

α (β −α )|z|2
Pα∗ f (z) = e eα zw̄ f (w) dλβ (w), (2.5)
β C
and

α (β −α )|z|2
Q∗α f (z) = e |eα zw̄ | f (w) dλβ (w). (2.6)
β C
We first prove several necessary conditions for the operator Pα to be bounded on

L p (C, dλβ ).
Lemma 2.15. Suppose 0 0, and β > 0. If Pα is bounded on
L p (C, dλβ ), then pα ≤ 2β and p ≥ 1.
Proof. Consider functions of the following form:
fx,k (z) = e−x|z| zk ,

2
z ∈ C,
where x > 0 and k is a positive integer. We have

β β Γ ((pk/2) + 1)
|z| pk e−(px+β )|z| dA(z) =
2
| fx,k | p dλβ = .
C π C px + β (px + β ) pk/2
On the other hand, it follows from the reproducing formula in Fα2+x that

α
eα zw̄ wk e−(α +x)|w| dA(w)
2
Pα ( fx,k )(z) =
π C

α
= e(α +x)[α z/(α +x)w] wk dλα +x (w)
α +x C
k 1+k
α αz α
= = zk .
α +x α +x α +x
Therefore,
p(1+k)
α
|Pα ( fx,k )| p dλβ = |z| pk dλβ (z)
C α +x C
p(1+k)
α Γ ((pk/2) + 1)
= .
α +x β pk/2
Now, if the integral operator Pα is bounded on L p (C, dλβ ), then there exists a
positive constant C (independent of x and k) such that
p(1+k)
α Γ ((pk/2) + 1) β Γ ((pk/2) + 1)
≤C ,
α +x β pk/2 px + β (px + β ) pk/2
or
p(1+k) 1+(pk/2)
α β
≤C .
α +x β + px
Fix any x > 0 and look at what happens in the above inequality when k → ∞. We
deduce that 2
α β
≤ .
α +x β + px
Cross multiply and simplify. The result is
pα 2 ≤ 2αβ + β x.
Let x → 0. Then pα 2 ≤ 2αβ , or pα ≤ 2β .

Similarly, if we let k = 0 and let x → ∞ in the previous paragraph, the result is
p ≥ 1. This completes the proof of the lemma.

Since the operator Pα (and hence Qα ) is never bounded on L p (C, dλβ ) when
0 β .
46 2 Fock Spaces
Proof. If p > 1 and Pα is bounded on L p (C, dλβ ), then Pα∗ is bounded on Lq (C, dλβ ),
where 1/p + 1/q = 1. Applying the formula for Pα∗ from (2.5) to the constant
function f = 1 shows that the function e(β −α )|z| is in Lq (C, dλβ ). From this, we
2
deduce that
q(β − α ) < β ,
which is easily seen to be equivalent to β < pα .

Lemma 2.17. If Pα is bounded on L1 (C, dλ β ), then α = 2β .
Proof. Fix any a ∈ C and consider the function
eα zā
fa (z) = , z ∈ C.
|eα zā |
Obviously, fa ∞ = 1 for every a ∈ C. On the other hand, it follows from (2.5) and
Corollary 2.5 that

α (β −α )|a|2
Pα∗ ( fa )(a) = e |eα wā | dλβ (w)
β C
α
= e(β −α )|a| eα |a| /(4β ) .
2 2 2
Since Pα∗ is bounded on L∞ (C), there exists a positive constant C such that
α (β −α )|a|2 α 2 |a|2 /(4β )

e e ≤ Pα∗ ( fa )∞ ≤ C fa ∞ = C
β
for all a ∈ C. This clearly implies that
α2
β −α + ≤ 0,
4β
which is equivalent to (2β − α )2 ≤ 0. Therefore, we have α = 2β .

Lemma 2.18. Suppose 1 0 and k is a positive integer. It follows from (2.5) and the reproducing
property in Fα2+x that

α (β −α )|z|2
Pα∗ ( fx,k )(z) = eα zw̄ wk e−(β +x)|w| dA(w)
2
e
π C

α (β −α )|z|2
= e e(β +x)[α z/(β +x)]w̄wk dλβ +x(w)
β +x C

α (β −α )|z|2 αz k
= e
β +x β +x
1+k
α
e(β −α )|z| zk .
2
=
β +x
Suppose 1 < p ≤ 2 and 1/p + 1/q = 1. If the operator Pα is bounded on

L p (C, dλβ ), then the operator Pα∗ is bounded on Lq (C, dλβ ). So there exists a positive
constant C, independent of x and k, such that

|Pα∗ ( fx,k )|q dλβ ≤ C | fx,k |q dλβ .
C C
We have

β Γ ((qk/2) + 1)
| fx,k |q dλβ = .
C qx + β (qx + β )qk/2
On the other hand, it follows from Lemma 2.16 and its proof that
β − q(β − α ) > 0,
so the integral

I= |Pα∗ ( fx,k )|q dλβ
C
can be evaluated as follows:
q(1+k)
α β
|z|qk e−(β −q(β −α ))|z| dA(z)
2
I=
β +x π C
q(1+k)
α β
= |z|qk dλβ −q(β −α )(z)
β +x β − q(β − α ) C
q(1+k)
α β Γ ((qk/2) + 1)
= .
β +x β − q(β − α ) (β − q(β − α ))qk/2
Therefore,
q(1+k)
α β Γ ((qk/2) + 1)
β +x β − q(β − α ) (β − q(β − α ))qk/2
is less than or equal to
Cβ Γ ((qk/2) + 1)
,
qx + β (qx + β )qk/2
48 2 Fock Spaces
which easily reduces to

q(1+k) 1+(qk/2)
α β − q(β − α )
≤C .
β +x β + qx
Once again, fix x > 0 and let k → ∞. We find out that

2
α β − q(β − α )
≤ .
β +x β + qx
Using the relation 1/p + 1/q = 1, we can change the right-hand side above to
pα − β
.
(p − 1)β + px
It follows that
α 2 (p − 1)β + α 2 px ≤ (pα − β )(β 2 + 2β x + x2),
which can be written as
(pα − β )x2 + [2β (pα − β ) − α 2 p]x + β 2(pα − β ) − α 2(p − 1)β ≥ 0.
Let q(x) denote the quadratic function on the left-hand side of the above inequality.
Since pα − β > 0 by Lemma 2.16, the function q(x) attains its minimum value at
pα 2 − 2β (pα − β )
x0 = .
2(pα − β )
Since 2 ≥ p, the numerator above is greater than or equal to
pα 2 − 2pαβ + pβ 2 = p(α − β )2 .
It follows that x0 ≥ 0 and so h(x) ≥ h(x0 ) ≥ 0 for all real x (not just nonnegative x).
From this, we deduce that the discriminant of h(x) cannot be positive. Therefore,
[2β (pα − β ) − pα 2]2 − 4(pα − β )[β 2(pα − β ) − α 2(p − 1)β ] ≤ 0.
Elementary calculations reveal that the above inequality is equivalent to
(pα − 2β )2 ≤ 0.
Therefore, pα = 2β .

Lemma 2.19. Suppose 2 < p < ∞ and Pα is bounded on L p (C, dλβ ). Then
pα = 2β .
Proof. If Pα is a bounded operator on L p (C, dλβ ), then Pα∗ is also bounded on
Lq (C, dλβ ), where 1 < q < 2 and 1/p + 1/q = 1. It follows from (2.5) that there
exists a positive constant C, independent of f , such that
q

e(β −α )|z|2 eα zw̄ f (w)e(α −β )|w| dλα (w) dλβ (z)
2
C
C

C | f (w)|q dλβ (w),
C
where f is any function in Lq (C, dλβ ). Let
f (z) = g(z)e(β −α )|z| ,

2
where g ∈ Lq (C, dλβ −q(β −α )). Recall from Lemma 2.16 that
β − q(β − α ) > 0.
We obtain another positive constant C (independent of g) such that

|Pα g|q dλβ −q(β −α ) ≤ C |g|q dλβ −q(β −α ),
C C
for all g ∈ Lq (C, dλβ −q(β −α )). So the operator Pα is bounded on Lq (C, dλβ −q(β −α )).
Since 1 < q < 2, it follows from Lemma 2.18 that
qα = 2[β − q(β − α )].
It is easy to check that this is equivalent to pα = 2β .

We now prove the main result of this section. Recall that Pα and Qα are never
bounded on L p (C, dλβ ) when 0 0, β > 0, and 1 ≤ p < ∞. Then the following
conditions are equivalent:
(a) The operator Qα is bounded on L p (C, dλβ ).
(b) The operator Pα is bounded on L p (C, dλβ ).
(c) The weight parameters satisfy pα = 2β .
Proof. When p = 1, that (b) implies (c) follows from Lemma 2.17, that (c) implies
(a) follows from Fubini’s theorem and Corollary 2.5, and that (a) implies (b) is
obvious.
50 2 Fock Spaces
When 1 < p < ∞, that (b) implies (c) follows from Lemmas 2.18 and 2.19, and
that (a) implies (b) is still obvious.
So we assume 1 < p < ∞ and proceed to show that condition (c) implies (a). We
do this with the help of Schur’s test (Lemma 2.14).
Let 1/p + 1/q = 1 and consider the positive function
h(z) = eδ |z| ,
2
z ∈ C,
where δ is a constant to be specified later.

Recall that

Qα f (z) = H(z, w) f (w) dλβ (w),
C
where
α α zw̄ (β −α )|w|2
H(z, w) = |e e |
β
is a positive kernel. We first consider the integrals

I(z) = H(z, w)h(w)q dλβ (w), z ∈ C.
C
If δ satisfies
α > qδ , (2.7)
then it follows from Corollary 2.5 that

α
|eα zw̄ | e−(α −qδ )|w| dA(w)
2
I(z) =
π C

α
= |eα zw̄ | dλα −qδ (w)
α − qδ C
α
eα |z| /4(α −qδ ) .
2 2
=
α − qδ
If we choose δ so that
α2
= qδ , (2.8)
4(α − qδ )
then we obtain

α
H(z, w)h(w)q dλβ (w) ≤ h(z)q (2.9)
C α − qδ
for all z ∈ C.
We now consider the integrals

J(w) = H(z, w)h(z) p dλβ (z), w ∈ C.
C
If δ satisfies
β − pδ > 0, (2.10)
then it follows from Corollary 2.5 that

α
|eα zw̄ e(β −α )|w| | h(z) p dλβ (z)
2
J(w) =
β C

α
= e(β −α )|w| |eα zw̄ | e−(β −pδ )|z| dA(z)
2 2
π C
α (β −α )|w|2 α 2 |w|2 /4(β −pδ )
= e e
β − pδ
α
e[(β −α )+α /4(β −pδ )]|w| .
2 2
=
β − pδ
If we choose δ so that
α2
β −α + = pδ , (2.11)
4(β − pδ )
then we obtain

α
H(z, w)h(z) p dλβ (z) ≤ h(w) p (2.12)
C β − pδ
for all w ∈ C. In view of Schur’s test and the estimates in (2.9) and (2.12), we
conclude that the operator Qα would be bounded on L p (C, dλβ ) provided that
we could choose a real δ to satisfy conditions (2.7), (2.8), (2.10), and (2.11)
simultaneously.
Under our assumption that pα = 2β , it is easy to verify that condition (2.8) is
the same as condition (2.11). In fact, we can explicitly solve for qδ and pδ in (2.8)
and (2.11), respectively, to obtain
α 2β − α
qδ = , pδ = .
2 2
The relations pα = 2β and 1/p + 1/q = 1 clearly imply that the two resulting δ ’s
above are consistent, namely,
α 2β − α
δ= = . (2.13)
2q 2p
Also, it is easy to see that the above choice of δ satisfies both (2.7) and (2.10). This
completes the proof of the theorem.

Theorem 2.21. If 1 ≤ p < ∞ and pα = 2β , then

|Pα f | p dλβ ≤ |Qα f | p dλβ ≤ 2 p | f | p dλβ
C C C
for all f ∈ L p (C, dλβ ).

52 2 Fock Spaces
Proof. With the choice of δ in (2.13), the constants in (2.9) and (2.12) both reduce
to 2. Therefore, Schur’s test tells us that, in the case when 1 0 and 1 ≤ p ≤ ∞, the operator Pα is a bounded
p p p
projection from Lα onto Fα . Furthermore, Pα f p,α ≤ 2 f p,α for all f ∈ Lα .
Proof. The case 1 ≤ p < ∞ follows from Theorem 2.21. The case p = ∞ follows
from Corollary 2.5.

2.3 Duality of Fock Spaces 53
2.3 Duality of Fock Spaces
It follows easily from the usual duality of L p spaces that for any 1 ≤ p < ∞, we have
(Lαp )∗ = Lqβ , where 1/p + 1/q = 1, α and β are any positive parameters, and the
duality pairing is given by

γ
f (z)g(z)e−γ |z| dA(z).
2
f , gγ =
π C
Here, γ = (α + β )/2 is the arithmetic mean of α and β .

In this section, we are going to identify all bounded linear functionals on the Fock
space Fαp , where 0 < p < ∞. We will also do the same for the space fα∞ . Somewhat
surprisingly, the duality of Fock spaces depends on the geometric mean of α and β
instead of their arithmetic mean. Let us begin with the case p > 1.
Theorem 2.23. Suppose β > 0, 1 < p < ∞, and 1/p + 1/q = 1. Then the dual space
q
of Fαp can be identified with Fβ under the integral pairing

γ
f (z)g(z) e−γ |z| dA(z),
2
f , gγ = lim
R→∞ π |z|<R

where γ = αβ is the geometric mean of α and β .
Proof. First, assume that g ∈ Fβq and F is defined by

γ
f (z)g(z)e−γ |z| dA(z).
2
F( f ) = lim
R→∞ π |z|<R
We proceed to show that F gives rise to a bounded linear functional on Fαp . To avoid
the use of limits all over the place, we appeal to Lemma 2.11 and further assume
that g is a finite linear combination of kernel functions.
If f (z) = eγ za for some a ∈ C, then by the reproducing property of the kernel
functions Kγ (z, w) and Kα (z, w), we have

γ
f (z)g(z)e−γ |z| dA(z),
2
g(a) =
π C
and

α γ
g(a) = g a
β α

α α
eα (γ a/α )z g
z e−α |z| dA(z)
2
=
π
C β

α α
z e−α |z| dA(z).
2
= f (z) g
π C β
54 2 Fock Spaces
Therefore,
α α
z e−α |z| dA(z).
2
f g dλγ = f (z)g (2.14)
C π C β
This shows that

α α
z e−α |z| dA(z)
2
F( f ) = f (z)g
π C β
α
α α 2
− α2 |z|2
= f (z)e g z e− 2 |z| dA(z)
π C β
for all functions f of the form
N
f (z) = ∑ ck eγ zak ,
k=1
which are dense in Fαp by Lemma 2.11.

It is clear that g ∈ Lqβ is equivalent to the condition that

ϕ (z) = g( α /β z) ∈ Lqα .
An application of Hölder’s inequality then gives
|F( f )| ≤ C f p,α ϕ q,α = C f p,α gq,β , (2.15)
where f is any finite linear combination of kernel functions, and C and C are
positive constants. This shows that F defines a bounded linear functional on Fαp .
Next, assume that F : Fαp → C is a bounded linear functional. Define a function
g on the complex plane by
g(w) = Fz eγ zw .
It is easy to show that g is entire. We are going to show that g ∈ Fβq and F( f ) =
p
f , gγ for all f in a dense subset of Fα .
To show that g ∈ Fβ , we need to show that the function g(w)e−β |w| /2 is in
q 2
Lq (C, dA). To this end, we consider the integrals

h(w)g(w)e−β |w|
2 /2
Φ (h) = dA(w), h ∈ L p (C, dA).
C
It suffices for us to show that Φ defines a bounded linear functional on the space
L p (C, dA). Without loss of generality, we may assume that h has compact support
in C. In this case, the integral

h(w)eγ zw e−β |w|
2 /2
dA(w)
C
converges in the norm topology of Fαp , and we have

h(w)Fz eγ zw e−β |w| /2 dA(w)
2
Φ (h) =
C

h(w)eγ zw e−β |w|
2 /2
=F dA(w)
C

α α
w eα zw e−α |w| /2 dA(w)
2
= F h
β C β
π
= F (Pα (ϕ )) ,
β
where
α α 2
ϕ (z) = h z e 2 |z| .
β
Since h ∈ L p (C, dA) is equivalent to ϕ ∈ Lαp and since the projection Pα maps Lαp
boundedly into Fαp , we conclude that
π
|Φ (h)| ≤ FPα (ϕ ) p,α ≤ Ch,
β
where h denotes the usual norm in L p (C, dA). This shows that the function g is
in Fβq .
Finally, if f (z) = eγ za for some a ∈ C, then by the remarks immediately following
this proof and the reproducing property in Fγ2 ,

γ
eγ za g(z)e−γ |z| dA(z) = g(a) = F( f ).
2
f , gγ = lim
R→∞ π |z|<R
It follows that F( f ) = f , gγ whenever f is a finite linear combination of kernel

functions. This, along with Lemma 2.11, finishes the proof of the theorem.

Note that (2.14) was proved under the assumption that both f and g are finite
linear combinations of kernel functions. By (2.15), the right-hand side of (2.14)
converges for all f ∈ Fαp and g ∈ Fβq , and the integral is dominated by f p,α gq,β .
An approximation argument with the help of Lemma 2.11 then shows that

α
lim f (z)g(z) dλγ (z) = f (z)g z dλα (z) (2.16)
R→∞ |z|<R C β
for all f ∈ Fαp and g ∈ Fβq . In particular, the limit on the left-hand side of (2.16)
exists for all f ∈ Fαp and g ∈ Fβq .
Alternatively, the identity in (2.16) can be proved with the help of Taylor
expansions. Details are left to the interested reader. We now consider the case of
small exponents.
56 2 Fock Spaces
Theorem 2.24. Suppose 0 0. Then the dual space of Fαp can be
identified with Fβ∞ under the integral pairing

γ
2
f , gγ = lim
R→∞ π |z|<R

where γ = αβ and the limit above always exists.
Proof. First, assume that g ∈ Fβ∞ and F is defined by F( f ) = f , gγ . To show that
p
F extends to a bounded linear functional on Fα , we use (2.16) to rewrite

α
f (z)ϕ (z)e−α |z| dA(z)
2
F( f ) =
π C

α α 2
α 2

= f (z)e− 2 |z| ϕ (z)e− 2 |z| dA(z),
π C
where

α
ϕ (z) = g z
β
is in Fα∞ . It follows from this and the embedding in Theorem 2.10 (and its proof)
that
2
|F( f )| ≤ 2ϕ ∞,α f 1,α ≤ ϕ ∞,α f p,α .
p
So F extends to a bounded linear functional on Fαp , and an approximation argument
shows that the limit in the statement of the theorem always exists.
Next, suppose that F is a bounded linear functional on Fαp . As in the proof of
Theorem 2.23, we consider the function g defined on C by

g(w) = Fz eγ zw .
It follows from the boundedness of F on Fαp and the integral formula in Corollary 2.5
that

pα F p
|eγ zw e−α |z|
2 /2
|g(w)| p ≤ | p dA(z)
2π C

pα F p
|e pγ zw |e−pα |z|
2 /2
= dA(z)
2π C
= F p e pβ |w|
2 /2
.
This shows that g ∈ Fβ∞ with g∞,β ≤ F.

Finally, as in the proof of Theorem 2.23, we have F( f ) = f , gγ for all functions
f of the form
N
f (z) = ∑ ck eγ zuk .
k=1
Since the set of functions of the above form is dense in Fαp , we have completed the
proof of the theorem.

Setting β = α in Theorems 2.23 and 2.24, we obtain the following special case.
Corollary 2.25. If 1 ≤ p < ∞, then the dual space of Fαp can be identified with Fαq
under the integral pairing f , gα , where 1/p + 1/q = 1. If 0 < p < 1, then the dual
space of Fαp can be identified with Fα∞ under the integral pairing f , gα .
It is interesting to observe that under the same integral pairing f , gα , the dual
space of each Fαp , 0 0 and γ = αβ . Then the dual space of fα∞ can be
identified with Fβ1 under the integral pairing f , gγ .
Proof. If g ∈ Fβ1 , then by Theorem 2.24, F( f ) = f , gγ defines a bounded linear

functional on fα∞ .
Now, suppose F is any bounded linear functional on fα∞ . Since the set of finite
linear combinations of kernel functions is dense in fα∞ (but not in Fα∞ ), we can
proceed as in the proof of Theorem 2.23 to obtain

γ
f (w)g(w)e−γ |w| dA(w)
2
F( f ) = lim
R→∞ π |w|<R
for f in a dense subset of fα∞ , where

g(w) = Fz eγ zw .
It remains for us to show that g ∈ Fβ1 .

Since the dual space of Fβ1 is identified with Fα∞ under the integral pairing f , gγ ,
it suffices to show that there exists a constant C > 0 such that
| f , gγ | ≤ C f ∞,α
for all f ∈ Fα∞ . For any positive integer n, consider the function:

n
fn (z) = f z , z ∈ C.
n+1
58 2 Fock Spaces
It is clear that f ∈ Fα∞ implies that each fn ∈ fα∞ with fn ∞,α ≤ f ∞,α for all n.
Now,

γ
f (w)Fz eγ zw e−γ |w| dA(w)
2
f , gγ = lim
R→∞ π |w|<R

γ
fn (w)Fz eγ zw e−γ |w| dA(w)
2
= lim lim
n→∞ R→∞ π |w|<R

γ γ zw −γ |w|2
= lim F fn (w)e e dA(w)
n→∞ π C
= lim F( fn ).
n→∞
Since |F( fn )| ≤ F fn ∞,α ≤ F f ∞,α for all n, we conclude that | f , gγ | ≤
F f ∞,α for all f ∈ Fα∞ . This shows that g ∈ Fβ1 and completes the proof of the
theorem.

2.4 Complex Interpolation 59
2.4 Complex Interpolation
We assume that the reader is familiar with the basic theory of complex interpolation,
including the complex interpolation of L p spaces. The book [250] provides an
elementary introduction to the subject. We will begin with the following well-known
interpolation theorem of Stein and Weiss.
Theorem 2.27. Suppose w, w0 , and w1 are positive weight functions on the complex
plane. If 1 ≤ p0 ≤ p1 ≤ ∞ and 0 ≤ θ ≤ 1, then
[L p0 (C, w0 dA), L p1 (C, w1 dA)]θ = L p (C, wdA)
with equal norms, where

1−θ
1 1−θ θ 1 p0
θ
p
= + , w p = w0 w1 1 .
p p0 p1
This result is very useful and widely known. See [216] for a proof.
Recall that Lαp is the space of Lebesgue measurable functions f on the complex
plane such that the function f (z)e−α |z| /2 is in L p (C, dA). The norm of f in Lα
2 p
p p
was defined in Sect. 2.1. With the inherited norm, Fα is the closed subspace of Lα
consisting of entire functions.
Specializing to exponential weights, we obtain the following special case of the
Stein–Weiss interpolation theorem.
Corollary 2.28. Suppose 1 ≤ p0 ≤ p1 ≤ ∞ and 0 ≤ θ ≤ 1. Then for any positive
weight parameters α0 and α1 , we have
p
Lα00 , Lαp11 θ
= Lαp ,
where
1 1−θ θ
= + , α = α0 (1 − θ ) + α1θ .
p p0 p1
Proof. Since Lαp = L p (C, dλ pα /2), it follows from the Stein–Weiss interpolation
theorem that
p0 p1
Lα1 , Lα2 θ = L p0 (C, dλ p0 α1 /2 ), L p1 (C, dλ p1 α2 /2 ) θ
= L p (C, dλ pα /2) = Lαp ,
where
1 1−θ θ
= + , α = α0 (1 − θ ) + α1θ .
p p0 p1
This proves the desired result.

60 2 Fock Spaces
Although Fαp is a closed subspace of Lαp , the Fock spaces interpolate in a way
that is much different from the containing spaces Lαp . In some sense, the Lebesgue
spaces Lαp interpolate “arithmetically,” while the Fock spaces Fαp interpolate “geo-
metrically.” We begin with the case when the weight parameter α is fixed.
Theorem 2.29. Suppose 1 ≤ p0 ≤ p1 ≤ ∞ and 0 ≤ θ ≤ 1. Then
p0 p1 p
Fα , Fα θ = Fα ,
where
1 1−θ θ
= + .
p p0 p1
Proof. The inclusion
p0 p1 p
Fα , Fα θ ⊂ Fα
p
follows from the definition of complex interpolation, the fact that each Fα k is a
pk p0 p1
closed subspace of Lα , and the fact that Lα , Lα θ = Lαp .
pOn the
other hand, if f ∈ Fαp ⊂ Lαp , then f is entire, and it follows from
Lα0 , Lα1 θ = Lαp that there exist a function F(z, ζ ) (z ∈ C and 0 ≤ Re ζ ≤ 1) and a
p
positive constant C such that:

(a) F(z, θ ) = f (z) for all z ∈ C.
(b) F( · , ζ ) p0 ,α ≤ C for all Re ζ = 0.
(c) F( · , ζ ) p1 ,α ≤ C for all Re ζ = 1.
Define a function G(z, ζ ) by

α
F(w, ζ )eα zw̄ e−α |w| dA(w).
2
G(z, ζ ) =
π C
Then it follows from Corollary 2.22 that:

(a) G(z, θ ) = f (z).
(b) G( · , ζ ) p0 ,α ≤ 2C for all Re ζ = 0.
(c) G( · , ζ ) p1 ,α ≤ 2C for all Re ζ = 1.
Since each function z → G(z, ζ ) is entire, we conclude that f ∈ [Fα 0 , Fαp1 ]θ . This
p

We now consider the case when there are different weight parameters present.
Note that α is an arithmetic mean of α0 and α1 in Corollary 2.28, but α is a
geometric mean of α0 and α1 in the following theorem.
Theorem 2.30. Suppose 1 ≤ p0 ≤ p1 ≤ ∞ and 0 ≤ θ ≤ 1. Then for any positive
weight parameters α0 and α1 , we have

Fαp00 , Fαp11 θ
= Fαp ,
2.4 Complex Interpolation 61
where
1 1−θ θ
= + , α = α01−θ α1θ .
p p0 p1
Proof. For any ζ ∈ C, consider the dilation operator Sζ defined by

(ζ −θ )/2
α0
Sζ f (z) = f z .
α1
According to Lemma 2.6, Sζ is an isometry from Fα00 onto Fα 0 whenever Re ζ = 0,

p p
and Sζ is an isometry from Fαp11 onto Fαp1 whenever Re ζ = 1. Furthermore, both

Sζ f and Sζ−1 f are analytic in ζ when f is analytic. Therefore, by the abstract
Stein interpolation theorem (see [215]), the operator Sθ must be an isometry from
p p p p
[Fα00 , Fα11 ]θ onto [Fα 0 , Fα 1 ]θ . Since Sθ = I is the identity operator, we must have

Fαp00 , Fαp11 θ
= Fαp0 , Fαp1 θ = Fαp ,
where the last step follows from Theorem 2.29.

As a consequence of the above interpolation theorem, we obtain the following
sharp result concerning the action of the Fock projection on L p spaces.
Theorem 2.31. Suppose 1 ≤ p ≤ ∞. Then for any positive weight parameters α , β ,
and γ , we have:
(a) Pα Lβp ⊂ Fγp if and only if α 2 /γ ≤ 2α − β .
(b) Pα Lβp = Fγp if and only if α 2 /γ = 2α − β .
Proof. It is easy to see that a necessary condition for Pα Lβp ⊂ Fγp , 1 ≤ p ≤ ∞, is that
2α > β . So for the rest of the proof, we always assume that 2α > β .
If α 2 /γ ≤ 2α − β , it follows from Corollary 2.5 that Pα maps L∞ ∞
β into Fγ .
Similarly, it follows from Fubini’s theorem and Corollary 2.5 that Pα maps L1β into
Fγ1 . By complex interpolation, Pα maps Lβp into Fγp for all 1 ≤ p ≤ ∞.
If α 2 /γ = 2α − β and f ∈ Fγp , then the function

α α
z e(β −α )|z|
2
g(z) = f
γ γ
belongs to Lβp and Pα g = f . Therefore, Pα Lβp = Fγp for 1 ≤ p ≤ ∞.

If α 2 /γ > 2α − β , then there exists some γ > γ such that α 2 /γ = 2α − β
(here, we used the assumption that 2α > β ). By what was proved in the previous
paragraph, Pα Lβp = Fγp . Since Fγp is strictly contained in Fγp , we see that Pα cannot
possibly map Lβp into Fγp . A similar argument shows that if α 2 /γ < 2α − β , then
Pα L∞ ∞
β = Fγ . This completes the proof of the theorem.

62 2 Fock Spaces
2.5 Atomic Decomposition 63
2.5 Atomic Decomposition
Recall from Lemma 2.11 that the set of finite linear combinations of kernel functions
is dense in Fαp , 0 < p < ∞. In this section, we improve upon this result. We show
that every function in Fαp can actually be decomposed into an infinite series of kernel
functions.
We begin with a basic estimate for integral averages of functions in Fock spaces.
Lemma 2.32. For any positive parameters α , p, and R, there exists a positive
constant C = C(p, α , R) such that

2 p C 2 p
f (a)e−α |a| /2 ≤ 2 f (z)e−α |z| /2 dA(z)
r B(a,r)
for all entire functions f , all complex numbers a, and all r ∈ (0, R]. Here, B(a, r) is
the Euclidean disk centered at a with radius r.
Proof. Let I denote the integral above. Then

| f (z)| p e−pα |z|
2 /2
I= dA(z)
B(a,r)

| f (w + a)| p e−pα |w+a|
2 /2
= dA(w)
|w|<r

| f (w + a)e−α wa| p e−pα (|w|
2 +|a|2 )/2
= dA(w).
|w|<r
Writing the integral in polar coordinates and using the subharmonicity of the
function | f (w + a)e−α wa | p , we obtain

e−pα (|w|
2 +|a|2 )/2
I ≥ | f (a)| p dA(w)
|w|<r
r
te−pα (t
2 +|a|2 )/2
= 2π | f (a)| p dt
0
r2
= π | f (a)e−α |a| e−pα s/2 ds
2 /2
|p
0
2π
(1 − e−pα r /2 )| f (a)e−α |a| /2 | p .
2 2
=
pα
This proves the desired estimate.

Recall that for any positive number r,
rZ2 = {nr + imr : n ∈ Z, m ∈ Z}

64 2 Fock Spaces
is a square lattice in the complex plane. The fundamental region of rZ2 , if we ignore
the boundary points, is the square
Sr = {z = x + iy : −r/2 ≤ x < r/2, −r/2 ≤ y < r/2}.
We also consider the square
Qr = {z = x + iy : −r ≤ x < r, −r ≤ y < r}.
It is clear that the complex plane admits the following decomposition:

C= {Sr + z : z ∈ rZ2 }.
Moreover, the use of half-open and half-closed squares makes the decomposition
above a disjoint union. Thus,

C
f (z) d μ (z) = ∑ f (z) d μ (z),
w∈rZ2 Sr +w
whenever f ∈ L1 (C, d μ ). Furthermore, there exists a positive integer N such that

every point in the complex plane belongs to at most N of the squares Qr + w.
Therefore,

C
f (z) d μ (z) ≤ ∑ f (z) d μ (z) ≤ N f (z) d μ (z)
w∈rZ2 Qr +w C
whenever f is a nonnegative measurable function.

Also, recall that for each a ∈ C, the normalized reproducing kernel of Fα2 at the
point a is given by

K(a, a) = eα za− 2 α |a| .
1 2
ka (z) = K(z, a)/
This is of course a unit vector in Fα2 . The following result is a pleasant surprise.
p
Lemma 2.33. Each ka is also a unit vector in Fα , where 0 < p ≤ ∞.
Proof. It follows from the definition of the norm in Fαp and the reproducing formula
in Fp2α /2 that

pα 2 p
ka (z)e− 2 α |z| dA(z)
1
ka pp,α =
2π C

pα − 1 pα |a|2 pα za 2 − pα |z|2
= e 2 e 2 e 2 dA(z)
2π C
pα pα
= e− 2 |a| 2 |a|
2 2
e = 1,
which proves the desired result for 0 < p < ∞. For p = ∞, observe that
α α
|ka (z)|e− 2 |z| = e− 2 |z−a| .
2 2
It follows that
α
sup |ka (z)|e− 2 |z| = 1,
2
z∈C
and the proof of the lemma is complete.

The main result of this section is the following:
Theorem 2.34. Let 0 < p ≤ ∞. There exists a positive constant r0 such that for any
p
0 < r < r0 , the space Fα consists exactly of the following functions:
f (z) = ∑ cw kw (z), (2.17)

w∈rZ2
where {cw : w ∈ rZ2 } ∈ l p . Moreover, there exists a positive constant C (independent

of f ) such that
C−1 f p,α ≤ inf {cw }l p ≤ C f p,α
for all f ∈ Fαp , where the infimum is taken over all sequences {cw } that give rise to
the decomposition (not unique) in (2.17).
Proof. If 0 < p ≤ 1 and f is given by (2.17) with {cw } ∈ l p , then by Hölder’s
inequality,
| f (z)e−α |z| ∑ |cw | p |kw (z)e−α |z|
2 /2 2 /2
|p ≤ |p.
w∈rZ2
It follows from this and Lemma 2.33 that
f pp,α ≤ ∑ |cw | p .
w∈rZ2
Thus, f ∈ Fαp and

f pp,α ≤ inf ∑ |cw | p .
w∈rZ2
If {cw } ∈ l ∞ and f is given by (2.17), then

α α
| f (z)|e− 2 |z| ≤ {cw }∞ ∑ e− 2 |z−w| .
2 2
w∈rZ2
By Lemma 1.12, there exists a positive constant C such that
f ∞,α ≤ C inf {cw }∞ ,
where the infimum is taken over all sequences {cw } in (2.17).

66 2 Fock Spaces
After interpolating between p = 1 and p = ∞, we have now shown that, for all
p ∈ (0, ∞] and {cw } ∈ l p , the function f given by (2.17) is in Fαp . Furthermore,
f p,α ≤ C inf {cw }l p ,
where C = C(p, α , r) is a positive constant and the infimum is taken over all
sequences {cw } that give rise to the representation of f in (2.17). It is interesting
to note that this part of the proof works for any positive r.
To prove the other part of the theorem, we assume that 0 < r < 1 and consider
the linear operator Tr defined on the space of entire functions as follows:

α α α
∑ eα zw− 2 |w|
2 +α iIm (wu)
f (u)e− 2 |u|
2
Tr f (z) = dA(u).
π w∈rZ2 Sr +w
We proceed to show that Tr is a bounded linear operator on Fαp and to estimate

I − Tr , the norm of I − Tr on Fαp , in terms of r, where I is the identity operator.
Let Dr = I − Tr . If f is in Fαp , then

f (z) = f (u)eα zu dλα (u)
C

α α α
∑ f (u)eα zu− 2 |u|
2 −α iIm (wu) 2 +α iIm (wu)
= e− 2 |u| dA(u).
π w∈rZ2
Sr +w
It follows that

α
Dr f (z) =
π ∑ f (u)H(z, w, u) dA(u), (2.18)
w∈rZ2 Sr +w
where
α α
α 2
H(z, w, u) = eα zw− 2 |w| − eα zu− 2 |u| −α iIm (wu) e− 2 |u| +α iIm (wu) .
2 2
We now estimate the norm of the operator Dr on Fα∞ and on Fα1 .

By (2.18),
α 2 α
|Dr (z)|e− 2 |z| ≤ f ∞,α Jr (z),
π
where
α 2
α zu− α2 |u|2 −α iIm (wu) α 2
Jr (z) = ∑ e − eα zw− 2 |w| e− 2 |z| dA(u).
w∈rZ2 Sr +w
Elementary calculations show that

α
− 2 |z−w|2 α
∑ − e− 2 |z−u| +α iIm (z−w)(u−w) dA(u)
2
Jr (z) = e
w∈rZ2 Sr +w

α α
∑ e− 2 |z−w| 1 − e− 2 |u−w| +α (z−w)(u−w) dA(u)
2 2
=
w∈rZ2 Sr +w

α α
∑ e− 2 |z−w| 1 − e− 2 |u| +α (z−w)u dA(u).
2 2
=
w∈rZ2 Sr
Since |u| < r for all u ∈ Sr and

∞ ζk ∞
| ζ |k

|1 − e | = ∑ ≤ ∑
ζ
= e|ζ | − 1
k=1 k! k=1 k!
for all complex numbers ζ , we have

α α 2 α
1 − e− 2 |u| +α (z−w)u ≤ eα |z−w|r+ 2 r − 1 ≤ r(eα |z−w|+ 2 − 1)
2
α
≤ Cre 4 |z−w|
2
for all u ∈ Sr , where C is a positive constant that only depends on α . Here, we used
the additional assumption that 0 < r < 1. It follows that there exists another positive
constant C, independent of r and z, such that
α
∑ e− 4 |z−w|
2
Jr (z) ≤ Cr3
w∈rZ2
for all z ∈ C and 0 < r < 1. Since
α α 2 α
α

2 2
e− 4 |z−w| = e− 4 |z| e 4 wz e− 8 |w| ,
2
an application of Lemma 2.32 shows that there is yet another positive constant C,
independent of z and r, such that

α
∑ e− 4 |z−u| dA(u)
2
Jr (z) ≤ Cr
w∈rZ2 Sr +w

α
e− 4 |z−u| dA(u)
2
= Cr
C

α 4π Cr
e− 4 |u| dA(u) =
2
= Cr .
C α
68 2 Fock Spaces
This shows that there exists another positive constant C, independent of r, such that
Dr f ∞,α ≤ Cr f ∞,α .
Consequently, the norm of Dr on Fα∞ satisfies
Dr ∞,α ≤ Cr, 0 < r < 1. (2.19)
To estimate the norm of Dr on Fα1 , first note that |Dr f (z)| is less than or equal to

α α zw− α2 |w|2 α α
∑ − eα zu− 2 |u| −α iIm (wu) | f (u)|e− 2 |u| dA(u).
2 2
e
π w∈rZ2 Sr +w
By Fubini’s theorem, the integral

α
|Dr f (z)|e− 2 |z| dA(z)
2

α α
∑ | f (u)|e− 2 |u| H(w, u) dA(u),
2
π w∈rZ2 Sr +w
where

α 2 α α
e− 2 |z| eα zw− 2 |w| − eα zu− 2 |u| −α iIm (wu) dA(z)
2 2
H(w, u) =
C

− α2 |z−w|2 α
− e− 2 |z−u| +α iIm (z−w)(u−w) dA(z)
2
= e
C

− α2 |z|2 α
− e− 2 |z−(u−w)| +α iIm z(u−w) dA(z)
2
= e
C

α 2 α 2
= e− 2 |z| 1 − eα z(u−w)− 2 |u−w| dA(z).
C
Since |u − w| < r for u ∈ Sr + w and |1 − eζ | ≤ e|ζ | − 1 for all complex numbers ζ ,

we have

α 2 α 2 α
1 − eα z(u−w)− 2 |u−w| ≤ eα |z|r+ 2 r − 1 ≤ r eα |z|+ 2 − 1 .
It is now clear that we can find a positive constant C = C(α ) such that H(w, u) ≤ Cr
for all w and u. It follows that

α α
|Dr f (z)|e− 2 |z| dA(z) ≤ Cr | f (u)|e− 2 |u| dA(u)
2 2
C C
for all f ∈ Fα1 . Thus, the norm of Dr on Fα1 satisfies
Dr 1,α ≤ Cr, 0 < r < 1. (2.20)
By (2.19) and (2.20), if r is sufficiently small, then Dr ∞,α < 1 and Dr 1,α < 1.
By complex interpolation, we also have Dr p,α < 1 for all 1 ≤ p ≤ ∞. This shows
p
that if r is small enough, the operator Tr is invertible on Fα for all 1 ≤ p ≤ ∞. When
Tr is invertible and f ∈ Fα , we can write f = Tr g with g = Tr−1 f and obtain the
p
atomic decomposition (2.17) with

α α 2 +α iIm (wu)
cw = g(u)e− 2 |u| dA(u).
π Sr +w
A simple argument with the help of Lemma 2.32 shows that the above sequence
{cw } is in l p whenever g ∈ Fαp . This completes the proof of the theorem in the case
1 ≤ p ≤ ∞.
We will complete the proof of the case 0 < p < 1 after we have proved the
following three lemmas.

Lemma 2.35. Suppose 0 < r < 1, 0 < p ≤ 1, and m is a nonnegative integer. For
any entire function f , we define a sequence
{(S f )w,k : w ∈ rZ2 , 0 ≤ k ≤ m}
by

α α (z − w)k α 2
eα iIm z(z−w)− 2 |z−w| f (z)e− 2 |z| dA(z).
2
(S f )w,k =
π Sr +w k!
Then S maps Fαp boundedly into l p .
Proof. For any w ∈ rZ2 , z ∈ Sr + w, and 1 ≤ k ≤ m, we have

αp α 2 (z − w)
k
|(S f )w,k | p = p eα iIm z(z−w)− 2 |z−w| f (z)e−α (z−w)w
π Sr +w k!
p
α 2 α
e− 2 |z−w| − 2 |w| +α iIm (z−w)w dA(z)
2
p
− α2 |w|2 −α (z−w)w
≤ C1 r e
pk
f (z)e dA(z)
Sr +w
pα
C1 r p(2+k) e− 2 |w| sup{| f (z)e−α (z−w)w | p
2
≤ : z ∈ Sr + w}
p
pα
≤ C2 r p(2+k)−2 e− 2 |w| f (z)e−α (z−w)w dA(z)
2
Qr +w

α 2 α 2 p
= C2 r p(2+k)−2 f (z)e 2 |z−w| − 2 |z| dA(z)
Qr +w

α 2 p
≤ C3 r p(2+k)−2 f (z)e− 2 |z| dA(z).
Qr +w
70 2 Fock Spaces
Let C4 = C3 (m + 1). Then

m
α 2 p
∑ ∑ |(S f )w,k | p
≤ C4 r 2(p−1)
∑ f (z)e− 2 |z| dA(z)
w∈rZ2 k=0 w∈rZ2 Qr +w

α 2 p
≤ C5 r 2(p−1)
f (z)e− 2 |z| dA(z).
C

Lemma 2.36. Suppose 0 < r < 1, 0 < p ≤ 1, and m is a nonnegative integer. For
every sequence
c = {cw,k : w ∈ rZ2 , 0 ≤ k ≤ m},
define a function T c by
m
α
∑ ∑ cw,k [α (z − w)]k eα zw− 2 |w| .
2
T c(z) =
w∈rZ2 k=0
Then T is a bounded linear operator from l p into Fαp .

Proof. It is obvious that the series converges to an entire function f (z) uniformly
on compact subsets of C. Since 0 < p < 1, it follows from Hölder’s inequality that
m p
pk α zw− α2 |w|2
| f (z)| p ≤ ∑ ∑ w,k
|c | p
[ α |z − w|] e .
w∈rZ2 k=0
Thus p
m
α 2 p α
f (z)e− 2 |z| ≤ ∑ ∑ |cw,k |k [α |z − w|] pk e− 2 |z−w|
2
,
w∈rZ2 k=0
and hence
m p
α 2 p α
f (z)e− 2 |z| dA(z) ≤ ∑ ∑ |cw,k | p |α z|k e− 2 |z|
2
dA(z)
C w∈rZ2 k=0 C
m
≤C ∑ ∑ |cw,k | p .
w∈rZ2 k=0

Lemma 2.37. Let r0 be the number from Theorem 2.34 in the case p = ∞. Suppose
0 < r < r0 and 0 < p ≤ 1. Then every monomial zk can be represented as
zk = ∑ cw kw (z),
w∈rZ2
where {cw } ∈ l p .
Proof. Fix ρ ∈ (r, r0 ). By the already-proved case p = ∞ of Theorem 2.34, every

monomial zk can be represented as
zk = ∑ cw kw (z),
w∈ρ Z2
where {cw } ∈ l ∞ . For w ∈ ρ Z2 , we can write w = ρ (m + in) for some integers m and
n. Since
α
kw (z) = eα zw− 2 |w| ,
2
we have for w = r(m + in) that

α 2 α 2 −|w|2 )
kw ((r/ρ )z) = eα zw − 2 |w | e 2 (|w | .
It follows that

r k
z = ∑ cw kw (z),
ρ w ∈rZ2
where
cw = cw e−(ρ
2 −r2 )(n2 +m2 )
is clearly a sequence in l p . This proves the desired decomposition for monomials.

We can now finish the proof of Theorem 2.34 in the case 0 < p < 1.
Fix a sufficiently small r ∈ (0, 1), let m be the integer part of 2(1 − p)/p, and let
S and T be the operators defined in the previous two lemmas. We have

α α
∑ G(z, w, u) f (u)e− 2 |u| dA(u),
2
(I − T S) f (z) =
π w∈rZ2 Sr +w
where

α iIm u(u−w)− α2 |u−w|2
m
[α (z − w)(u − w)]k
G = ku (z) − e ∑ k!
kw (z).
k=0
It is elementary to check that

∞
α iIm u(u−w)− α2 |u−w|2 [α (z − w)(u − w)]k
G=e ∑ k!
kw (z).
k=m+1
For u ∈ Sr + w, we have |u − w| < r. Therefore,
∞
(α r|z − w|)k
|G| ≤ |kw (z)| ∑ k!
,
k=m+1
72 2 Fock Spaces
and so by Hölder’s inequality, |(I − T S) f (z)| p is less than or equal to

p p
∞
αp (α r|z − w|)k α
∑ |kw (z)| p ∑ | f (u)|e− 2 |u| dA(u) .
2
π p w∈rZ 2 k=m+1 k! Sr +w
It follows from this and Fubini’s theorem that

α 2 p
(I − TS) f (z)e− 2 |z| dA(z)
C

p
αp α
∑ C(w) | f (u)|e− 2 |u| dA(u) ,
2
π p w∈rZ 2 Sr +w
where

p
∞
− p2α |z−w|2 (α r|z − w|)k
C(w) = e
C
∑ k!
dA(z)
k=m+1

p
∞
− p2α |z|2 (α r|z|)k
= e
C
∑ k! dA(z)
k=m+1

p
∞
pα 2 (α |z|)k
≤ r(m+1)p e− 2 |z|
C
∑ k! dA(z)
k=m+1

pα 2
≤r (m+1)p
e− 2 |z| +pα |z| dA(z).
C
So there is a constant C > 0 such that

α 2 p
(I − TS) f (z)e− 2 |z| dA(z)
C

p

− α2 |u|2
Cr (m+1)p
∑ Sr +w
| f (u)|e .
w∈rZ2
On the other hand,

p
α 2 α 2 p
| f (u)|e− 2 |u| ≤ r2p sup f (u)e− 2 |u| ,
S +w
r u∈Sr +w
and an application of Lemma 2.32 produces another constant C > 0 (independent of

r ∈ (0, 1)) such that
p

− α2 |u|2 α 2 p
| f (u)|e ≤ Cr
2p−2
f (u)e− 2 |u| dA(z).
Sr +w Qr +w
Thus,

α 2 p
(I − TS) f (z)e− 2 |z| dA(z)
C

α 2 p
Cr(m+1)p+2p−2 ∑ f (u)e− 2 |u| dA(u).
w∈rZ2 Qr +w
So we can find another constant C > 0, independent of r ∈ (0, 1), such that
I − TS p,α ≤ Cr(m+3)−(2/p), 0 < r < 1.
Since
2 2(1 − p) 2
m+3− > − 1 + 3 − = 0,
p p p
we see that there exists some r0 ∈ (0, 1) such that I − TS p,α < 1 whenever r ∈
(0, r0 ). This shows that the operator T S is invertible on Fαp whenever r ∈ (0, r0 ).
Consequently, for any r ∈ (0, r0 ), the operator T is onto, and so every function
f ∈ Fαp can be written as
m
α
∑ ∑ cw,k (z − w)k e−α zw− 2 |w| .
2
f (z) = (2.21)
w∈rZ2 k=0
Furthermore, the coefficients cw,k in (2.21) all depend on f linearly, and

m
∑ ∑ |cw,k | p ≤ C f pp,α ,
w∈rZ2 k=0
where C is a positive constant independent of f .

Given any δ > 0 and any r ∈ (0, r0 ), it follows from Lemma 2.37 that there exist
coefficients cw,k , 0 ≤ k ≤ m, w ∈ rZ2 , |w| ≤ N, such that

k α
z − ∑ cu,k e α zu− 2 |u|
2
<δ
u∈rZ2 ,|u|≤N
p,α
for all 0 ≤ k ≤ m. By a change of variables, the norm of

α α 2 +α zw− α |w|2
(z − w)k eα zw− 2 |w| − ∑ cu,k eα (z−w)u− 2 |u|
2
2
u∈rZ2 ,|u|≤N
74 2 Fock Spaces
in Fαp is less than δ for all 0 ≤ k ≤ m and w ∈ rZ2 . Define an operator Ar on Fαp by
α
∑ ∑ cu,k eα iIm (wu) eα z(w+u)− 2 |w+u|
2
Ar f (z) = cw,k
w∈rZ2 ,0≤k≤m u∈rZ2 ,|u|≤N
and observe that

α 2 α α
α (z − w)u − |u| + α zw − |w|2 = α iIm (wu) + α z(w + u) − |w + u|2.
2 2 2
It then follows from Hölder’s inequality that
f − Ar f pp,α ≤ ∑ |cw,k | p δ p ≤ Cδ p f pp,α

w∈rZ2 ,0≤k≤m
for all f ∈ Fαp . If we choose δ such that Cδ p < 1, then I − Ar p,α < 1, and so the
operator Ar is surjective on Fαp . Since w + u ∈ rZ2 whenever w ∈ rZ2 and u ∈ rZ2 ,
the proof of Theorem 2.34 is now complete.
2.6 Translation Invariance 75
2.6 Translation Invariance
In this section, we consider the action of translations on Fock spaces and determine
three spaces that are unique under such actions: the space Fα∞ is maximal among
translation invariant Banach spaces of entire functions, the space Fα1 is minimal
among translation invariant Banach spaces of entire functions, and the space Fα2 is
the only Hilbert space of entire functions invariant under translations.
For any point a ∈ C, we define three analytic self-maps of the complex plane as
follows:
ta (z) = z + a, τa (z) = z − a, ϕa (z) = a − z.
The map ta is naturally called the translation by a, and it is clear that τa = t−a = ta−1 .
The map ϕa is the composition of the translation ta with the reflection z → −z. Note
that ϕa is its own inverse.
When making a change of variables, observe that

f ◦ ta (z) dλα (z) = f ◦ ϕa (z) dλα (z)
C C

α
f (w)e−α |a−w| dA(w)
2
=
π C

= f (w)|ka (w)|2 dλα (w).
C
On the other hand,

α
f (w)e−α |w+a| dA(w)
2
f ◦ τa (z) dλα (z) =
C π C

= f (w)|k−a (w)|2 dλα (w).
C
Similarly,

f ◦ τa (z)|ka (z)|2 dλα (z) = f ◦ ϕa (z)|ka (z)|2 dλα (z)
C C

= f (z) dλα (z),
C
while

f ◦ ta (z)|ka (z)|2 dλα (z) = f (z + 2a) dλα (z).
C C
See Corollary 2.4. These are some of the subtle differences that can easily be
overlooked.
We can use τa and ϕa to define certain unitary operators on Fα2 . Although there is
an obvious temptation to use only one of these maps in the book, we have found that
there are situations in which one choice is more convenient than the other. Therefore,
we are going to use both in the book.
76 2 Fock Spaces
For a fixed weight parameter α and a ∈ C, we define two operators Wa and Ua as

follows:
Wa f = f ◦ τa ka , Ua f = f ◦ ϕa ka ,
where ka is the normalized reproducing kernel of Fα2 at a. We will consider the
action of these operators on both Lαp and Fαp . The focus in this section is their action
on Fock spaces.
These are weighted translation operators. In some of the literature, the operators
Wa are called Weyl (unitary) operators. We first show that both Wa and Ua are
isometries on the Fock spaces Fαp .
Proposition 2.38. Let 0 < p ≤ ∞. We have
Wa f p,α = Ua f p,α = f p,α
for all a ∈ C and f ∈ Fαp . Furthermore, both Wa and Ua are invertible on Fαp with
Wa−1 = W−a and Ua−1 = Ua . Consequently, Wa and Ua are both unitary operators on
Fα2 with Wa∗ = W−a and Ua∗ = Ua .
Proof. It is easy to check that
α α
e− 2 |z| |Wa f (z)| = e− 2 |z−a| | f (z − a)|,
2 2
and
α α
e− 2 |z| |Ua f (z)| = e− 2 |a−z| | f (a − z)|.
2 2
The identities
Wa f p,α = Ua f p,α = f p,α
then follow from a change of variables. See Corollary 2.4.
To see that Wa is invertible with Wa−1 = W−a , take any f ∈ Fαp and note that
α
W−aWa f (z) = e−α az− 2 |a| (Wa f )(z + a)
2
α α
= e−α az− 2 |a| eα a(z+a)− 2 |a| f (z + a − a)
2 2
= f (z).
A similar argument shows that Ua is invertible with Ua−1 = Ua . This completes the
proof of the proposition.

Although the operators Wa and Ua behave similarly in many situations, there are
sometimes reasons to pick one over the other. For example, the operators Wa almost
have a semigroup property with respect to a, while the operators Ua are all self-
adjoint. In particular, we can use the Weyl operators to obtain the following unitary
representation of the Heisenberg group. Recall that another unitary representation
was given in Chap. 1 based on Weyl pseudodifferential operators.
Theorem 2.39. The mapping (a, θ ) → eiθ Wa is a unitary representation of the

Heisenberg group H on the Fock space Fα2 .
Proof. For any two points a and b in C, we easily check that
WaWb = e−α iIm (ab)Wa+b = eα iIm (ab)Wa+b . (2.22)
This shows that (a, θ ) → eiθ Wa is a group embedding of H into the group of unitary
operators on Fα2 .

In the rest of this section, we work with the Weyl unitary operators Wa . A similar
theory can be developed with the unitary operators Ua , which is left to the reader as
an exercise.
Proposition 2.40. The Fock space Fα∞ is maximal in the sense that if X is any
Banach space of entire functions with the following properties:
(a) Wa f X = f X for all a ∈ C and f ∈ X,
(b) the point evaluation f → f (0) is a bounded linear functional on X,
then X ⊂ Fα∞ and the inclusion is continuous.
Proof. Condition (a) implies that Wa f ∈ X for every f ∈ X and every a ∈ C.
Combining this with condition (b), we see that for every a ∈ C, the point evaluation
f → f (a) is also a bounded linear functional on X, and
α
e− 2 |a| | f (a)| = |W−a f (0)| ≤ CW−a f X = C f X ,
2
where C is a positive constant that is independent of a ∈ C and f ∈ X. Since a is

arbitrary, we conclude that f ∈ Fα∞ with f ∞,α ≤ C f X for all f ∈ X.

Proposition 2.41. The Fock space Fα1 is minimal in the sense that if X is a Banach
space of entire functions with the following properties:
(a) Wa f X = f X for all a ∈ C and f ∈ X,
(b) X contains all constant functions,
then Fα1 ⊂ X and the inclusion is continuous.
Proof. Since X contains all constant functions, applying Wa to the constant function
1 shows that for each a ∈ C, the function
α
ka (z) = eα az− 2 |a|
2
belongs to X. Furthermore, ka X = Wa 1X = 1X for all a ∈ C.

Let {zn } denote a sequence in C on which we have atomic decomposition for Fα1 .
If f ∈ Fα1 , there exists a sequence {cn } ∈ l 1 such that
∞
f= ∑ c n k zn . (2.23)
n=1
78 2 Fock Spaces
Since each kzn belongs to X and ∑ |cn | < ∞, we conclude that f ∈ X with
∞ ∞
f X ≤ ∑ |cn |kzn X = C ∑ |cn |,
n=1 n=1
where C = 1X > 0. Taking the infimum over all sequences {cn } satisfying (2.23),
we obtain another constant C > 0 such that
f X ≤ C f Fα1 , f ∈ Fα1 .

Proposition 2.42. Suppose H is a nontrivial separable Hilbert space of entire
functions with the following properties:
(a) Wa f H = f H for all a ∈ C and f ∈ H.
(b) f → f (0) is a bounded linear functional on H.
Then H = Fα2 and there exists a positive constant c such that f , gH = c f , gα for
all f and g in H.
Proof. Since H contains at least one function that is not identically zero, it follows
from conditions (a) and (b) that for any z ∈ C, the mapping f → f (z) is a nonzero
bounded linear functional on H. Furthermore, for any compact subset S of C, there
exists a positive constant C such that | f (z)| ≤ C f H for all f ∈ H and all z ∈ S.
Consequently, the space H possesses a reproducing kernel KH (z, w). Moreover,
if {en } is an orthonormal basis of H, then
∞
KH (z, w) = ∑ en(z) en (w), (2.24)
n=0
and the convergence is uniform when z and w are restricted to compact subsets of C.
In particular, the series representation for KH (z, w) in (2.24) is independent of the
choice of the orthonormal basis {en }.
It is easy to see from condition (a) and the proof of Proposition 2.38 that each Wa
is a unitary operator on H. Fix any a ∈ C and let σn = Wa en , n ≥ 1. Then {σn } is
also an orthonormal basis of H. Therefore, by (2.24), we have
∞
KH (z, w) = ∑ σn (z) σn (w)
n=1
∞
= ka (z) ka (w) ∑ en (z − a) en(w − a)
n=1
= ka (z) ka (w)KH (z − a, w − a),

where ka is the normalized reproducing kernel of Fα2 at a. Let z = w = a. We obtain
KH (z, z) = eα |z| KH (0, 0) = Kα (z, z)KH (0, 0),

2
z ∈ Cn ,
where Kα (z, w) is the reproducing kernel of Fα2 .

By a well-known result in the function theory of several complex variables, any
reproducing kernel is uniquely determined by its values on the diagonal. See [142].
Therefore, we must have KH (z, w) = cK(z, w) for all z and w, where c = KH (0, 0) > 0
as H contains functions that do not vanish at the origin. This shows that, after an
adjustment of the inner product by a positive scalar, the two spaces H and Fα2 have
the same reproducing kernel, from which it follows that H = Fα2 . This completes the
proof of the proposition.

80 2 Fock Spaces
2.7 A Maximum Principle 81
2.7 A Maximum Principle
The classical maximum principle asserts that if f is an entire function and | f (z)| ≤
M for all |z| = R, then | f (z)| ≤ M for all |z| ≤ R. The purpose of this section is to
prove the following version of the maximum principle for Fock spaces.
Theorem 2.43. For any α > 0 and p ≥ 1, there exists a positive radius R = R(α , p)
such that f p,α ≤ g p,α for all entire functions f and g satisfying
| f (z)| ≤ |g(z)|, |z| ≥ R.
Proof. Without loss of generality, we may assume that g ∈ Fαp . Otherwise, the
desired result is obvious. Under this assumption, we also have

| f | p dλα ≤ |g| p dλα < ∞,
|z|≥R |z|≥R
which easily implies that f ∈ Fαp as well.

For any positive radius r and any function F in the complex plane, we write
2π
I(r, F) = F(reiθ ) dθ .
0
We fix some R > 0 and assume that | f (z)| ≤ |g(z)| for all R ≤ |z| < ∞.
We will try to compare I(r, | f | p − |g| p ) for 0 < r < R to I(ρ , |g| p − | f | p ) for
R < ρ < ∞. To this end, we let ω (z) = f (z)/g(z), which is analytic and has modulus
less than or equal to 1 in the region R < |z| < ∞. We may assume that |ω (z)| <
1 for all R < |z| < ∞. In fact, if |ω (z0 )| = 1 for some R < |z0 | < ∞, then by the
classical maximum modulus principle, the analytic function ω on R < |z| < ∞ must
be constant, which would then imply that f and g differ by a constant multiple in
the whole complex plane, from which the desired result clearly follows.
For any ρ ∈ (R, ∞), pick a point ζ (ρ ) such that |ζ (ρ )| = ρ and
|ω (ζ (ρ ))| = max{|ω (z)| : |z| = ρ }.
We may assume that f is not identically 0, for otherwise the desired result is trivial.
Thus, 0 < |ω (ζ (ρ ))| < 1 for all ρ ∈ (R, ∞). To simplify notation, let us write ωρ =
ω (ζ (ρ )).
Since p ≥ 1, it follows from elementary calculus that
py p−1(x − y) ≤ x p − y p ≤ px p−1 (x − y), (2.25)

82 2 Fock Spaces
for all x ≥ 0 and y ≥ 0. We deduce from the second inequality in (2.25) that for any
0 ≤ r ≤ R < ρ < ∞, we have
I(r, | f | p − |g| p) ≤ I(r, | f | p − |ωρ g| p )

≤ I(r, p| f | p−1 (| f | − |ωρ g|))
≤ I(r, p| f | p−1 | f − ωρ g|).
The function p| f | p−1 | f − ωρ g| is subharmonic on the complex plane, so its integral

mean on |z| = r is an increasing function of r (see [76] for example). Thus,
I(r, | f | p − |g| p) ≤ I(ρ , p| f | p−1 | f − ωρ g|)

= I(ρ , p|ω | p−1|ω − ωρ |(|g| p − | f | p )/(1 − |ω | p)).
Taking x = 1 and y = |ω | in the first inequality of (2.25), we get
p|ω | p−1 1 1 + |ω | 2
≤ = < .
1 − |ω | p 1 − |ω | 1 − |ω | 2 1 − | ω |2
Therefore,
I(r, | f | p − |g| p) ≤ 2I(ρ , |ω − ωρ |(|g| p − | f | p )/(1 − |ω |2)) (2.26)
for all 0 ≤ r ≤ R < ρ < ∞.

Set

|ω (z) − ωρ |
γ (ρ ) = max : |z| = ρ
1 − |ω (z)|2
for ρ ∈ (R, ∞). By (2.26),
I(r, | f | p − |g| p) ≤ 2γ (ρ )I(ρ , |g| p − | f | p ) (2.27)
for all 0 ≤ r ≤ R < ρ < ∞. Fix ρ and integrate both sides of (2.27) over [0, R] against
the measure re−α r dr. The result is
2

1 − e− α R
2
(| f | p − |g| p) dλα ≤ γ (ρ )I(ρ , |g| p − | f | p ) (2.28)

|z|≤R π
for all R < ρ < ∞. Divide both sides of (2.28) by γ (ρ ) and integrate both sides over
(R, ∞) against the measure ρ e−αρ dρ . The result is
2

(| f | p − |g| p) dλα ≤ CR (|g| p − | f | p ) dλα ,
|z|≤R |z|≥R
where
1 − e− α R
2
CR = ∞ .
ρ e−αρ
2
α dρ
R γ (ρ )
If R is a positive radius such that CR < 1, then the integral

J= (| f | p − |g| p) dλα
C
satisfies the following estimates:

J= + (| f | p − |g| p) dλα
|z|≤R |z|≥R

≤ CR (|g| p − | f | p ) dλα + (| f | p − |g| p) dλα
|z|≥R |z|≥R

≤ (|g| − | f | ) dλα +
p p
(| f | p − |g| p) dλα
|z|≥R |z|≥R
= 0,
which proves the desired result.

We will actually show that CR < 1 for all sufficiently small positive radius R. To
this end, let d denote the pseudohyperbolic metric in the unit disk D, namely,

z−w

d(z, w) = .
1 − zw
Since

|a − b| d(a, b) 1 − |b|2
=
1 − |a|2 1 − d 2(a, b) 1 − |a|2
for all a and b in the unit disk, we see that for all z with |z| = ρ ,

|ω (z) − ωρ | d(ω (z), ωρ ) 1 − | ω ρ |2
=
1 − |ω (z)| 2
1 − d (ω (z), ωρ ) 1 − |ω (z)|2
2
d(ω (z), ω (ζ (ρ )))

≤ .
1 − d 2(ω (z), ω (ζ (ρ )))
It follows that
d(ω (z), ω (ζ (ρ )))

γ (ρ ) ≤ sup , R < ρ < ∞. (2.29)
|z|=ρ 1 − d 2(ω (z), ω (ζ (ρ )))
84 2 Fock Spaces
The function H(z) = ω (R/z) is analytic from the punctured disk 0 < |z| < 1 into
the unit disk. Since H is bounded near z = 0, it has a removable singularity at z = 0.
Thus, we can think of H as analytic self-maps of the unit disk. By the classical
Schwarz lemma, we have
d(H(z), H(w)) ≤ d(z, w), z, w ∈ D.
It follows that
d(ω (z), ω (ζ (ρ ))) = d (H(R/z), H(R/ζ (ρ ))) ≤ d(R/z, R/ζ (ρ ))
for all |z| = ρ . Combining this with (2.29), we obtain
d(R/z, R/ζ (ρ ))
γ (ρ ) ≤ sup .
|z|=ρ 1 − d 2(R/z, R/ζ (ρ ))
By symmetry of the unit disk,
2Rρ
sup d(R/z, R/ζ (ρ )) = d(−R/ζ (ρ ), R/ζ (ρ )) = .
|z|=ρ ρ 2 + R2
From this, we deduce that
2Rρ
γ (ρ ) ≤ , R < ρ < ∞.
ρ 2 − R2
Plugging this into the formula for CR , we obtain the estimate
2R(1 − e−α R )
2
CR ≤ ∞ .
(ρ 2 − R2)e−αρ dρ
2
α
R
The quotient above tends to 0 as R → 0+ . Therefore, CR < 1 for all sufficiently small
positive radius R. This completes the proof of the theorem.

If 0 < p < 1, the inequalities in (2.25) are replaced by
px p−1 (x − y) ≤ x p − y p ≤ py p−1 (x − y), (2.30)
and a similar sequence of estimates leads to
I(r, | f | p − |g| p) ≤ I(ρ , p|ωρ | p−1 |ω − ωρ |(|g| p − | f | p )/(1 − |ω | p))

for all 0 ≤ r ≤ R < ρ < ∞. So in this case, we need to consider the function

p|ωρ | p−1 |ω (z) − ωρ |
γ (ρ ) = max : |z| = ρ , R < ρ < ∞.
1 − |ω (z)| p
Note that the function γ depends on f and g. We just need to bound γ from above by
a function that is independent of f and g. By the left inequality in (2.30), we have
|ωρ | p−1 |ω (z) − ωρ | |ω (z) − ωρ |

γ (ρ ) ≤ sup ≤ 2|ωρ | p−1 sup .
|z|=ρ 1 − | ω (z)| |z|=ρ 1 − |ω (z)|
2
Therefore, we just need to bound |ωρ | from below by a positive function that is
independent of f and g. But this is impossible, for we may have a situation like
f (z) = g(z)/N, where N is large; in this case, we have H = 1/N, and we can choose
N to be arbitrarily large.
86 2 Fock Spaces
2.8 Notes 87
2.8 Notes
There are two ways to define the Fock spaces. One way is to consider subspaces
L p (C, dλα ) consisting of entire functions. This would be similar to the definitions of
the more classical Hardy and Bergman spaces. It turns out that this is not a good way
to define the Fock spaces. The seemingly cumbersome definition of Fαp as the space
of entire functions f such that f (z)e−α |z| /2 belongs to L p (C, dA) will make the
2
statements and proofs of many results much easier and more convenient later on.
The constant α in Fαp is not essential in our theory. No generality is lost if
we choose to develop the theory with a particular choice of α , say α = 1. This
weight parameter plays the role of Planck’s constant in mathematical physics, and
it provides us with an extra level of freedom that is useful in several situations.
Although the Fock space Fα2 is a central subject in quantum physics, this book is
focused on purely mathematical analysis on Fock spaces. No serious effort is made
to show any connections or applications to physics. We refer the interested reader to
books such as [177] for applications of the Fock space in physics.
The characterization of the boundedness of Pα and Qα on L p spaces was obtained
in [74], where more precise norm estimates can also be found. See [96] for an
even more elaborate study of similar integral operators. The boundedness of the
projection Pα on Lαp for 1 ≤ p ≤ ∞ can be found in [138]. The papers [214] and
[217] also study the boundedness of Pα and Qα on L p spaces.
The study of the Heisenberg group is a small industry by itself. This is especially
so in quantum physics and harmonic analysis, where the connection of Fock spaces
to the Heisenberg group is evident. But we will not use the Heisenberg group in any
way other than the special elements Wa in it.
The paper [138] by Janson, Peetre, and Rochberg is a key reference throughout
this book. In particular, the duality, atomic decomposition, and complex inter-
polation for the Fock spaces Fαp , where 1 ≤ p ≤ ∞, were proved in [138]. Our
presentation of the case 0 < p < 1 follows [231] very closely.
The translation invariance of the Fock spaces was first considered in [138], where
it was shown that Fα1 is minimal and Fα∞ is maximal among Banach spaces of entire
functions whose norm is invariant under the action of the Heisenberg group. The
uniqueness of Fα2 among Hilbert spaces of entire functions whose norm is invariant
under the action of Wa was proved in [255].
The version of the maximum modulus principle in Sect. 2.7 was first proved in
[194], based on a technique introduced in [122] to tackle the corresponding problem
for Bergman spaces on the unit disk. That such a maximum modulus principle might
be true for the Bergman space was first conjectured by Korenblum in [141] and was
proved in [117] in the case p = 2 and in [122] when 1 ≤ p < ∞. See [232–239] for
other work concerning Korenblum’s maximum principle.
88 2 Fock Spaces
2.9 Exercises 89
2.9 Exercises
1. Show that the Fock space Fαp is a closed subspace of L p (C, dλ pα /2).
2. Show that Mz , the operator of multiplication by the coordinate function z, is a
densely defined unbounded linear operator on Fα2 . Show that the adjoint of Mz
on Fα2 is essentially the operator of differentiation. More specifically, Mz∗ f (z) =
(1/α ) f (z) for all f ∈ Fα2 .
3. Let 0 < p ≤ ∞, S be a compact subset of C, and k be a positive integer. Show
that there exists a positive constant C such that
| f (k) (z)| ≤ C f p,α
for all z ∈ S and f ∈ Fαp .

4. Let ϕ be an entire function. Show that the composition operator Cϕ defined by
Cϕ f = f ◦ ϕ is bounded on Fαp if and only if ϕ (z) = az + b, where |a| < 1 or
|a| = 1 and b = 0. Characterize compact composition operators on Fαp . See [46]
and [110].
5. Suppose 1 < p < ∞ and f ∈ Fαp . Show that the Taylor polynomials of f converge
to f in the norm topology of Fαp .
6. Suppose 0 < p ≤ 1. Are there functions f ∈ Fαp such that the Taylor polynomials
of f do not converge to f in the norm topology of Fαp ? See [256] for the
corresponding problem in the context of Hardy and Bergman spaces.
7. Show that fα∞ is a closed subspace of Fα∞ .
8. Show that the set of polynomials is dense in fα∞ .
9. Characterize the space Pα C0 (C), where C0 (C) is the space of continuous
functions on C that vanish at ∞.
10. If 1 < p < ∞ and f (z) = ∑ an zn is a function in Fαp , then

α n 41 − 2p1
an = o n , n → ∞.
n!
See [224] for this and the next few problems.

11. If f (z) = ∑ an zn is a function in Fα1 , then

αn − 1
an = O n 4 , n → ∞.
n!
12. Let 1 ≤ p < ∞ and let {δn } be any sequence of positive numbers decreasing
to 0. Then there exists a function f (z) = ∑ an zn in Fαp such that

α n 14 − 2p1
an = O n δn , n → ∞.
n!
90 2 Fock Spaces
13. If 0 < p ≤ 2 and

∞ p
n! 2
∑ |an |
p 1
p
n− 4 + 2 < ∞,
n=0 αn
then the function f (z) = ∑ an zn belongs to Fαp .
14. If 0 < p ≤ 2 and the function f (z) = ∑ an zn belongs to Fαp , then
∞ p
n! 2
∑ |an| p
3p 3
n 4 − 2 < ∞.
n=0 αn
15. If 2 ≤ p < ∞ and

∞ p
n! 2
∑ |an|
3p 3
p
n 4 − 2 < ∞,
n=0 αn
then the function f (z) = ∑ an zn belongs to Fαp .
16. If 2 ≤ p < ∞ and the function f (z) = ∑ an zn is in Fαp , then
∞ p
n! 2
∑ |an |
p 1
p
n− 4 + 2 < ∞.
n=0 αn
17. Let 1 < p ≤ 2 with 1/p + 1/q = 1 and f (z) = ∑ an zn is in Fαp . Then
∞ q
n! 2
∑ |an |
q 1
q
n 4 − 2 < ∞.
n=0 αn
18. If f (z) = ∑ an zn is in Fαp , where 0 < p < ∞, then

α e n2
|an | ≤ f p,α ,
n
for all n ≥ 1.
19. Suppose 2 ≤ p ≤ ∞, 1/p + 1/q = 1, and
∞ q
n! 2
∑ |an |q
q 1
n 4 − 2 < ∞,
n=0 αn
then the function f (z) = ∑ an zn is in Fαp .

20. Suppose (X, μ ) is a measure space and fn ∈ L p (X, d μ ) for n ≥ 0, where 0 <
p < ∞. Show that

lim | fn − f0 | p dμ = 0
n→∞ X
2.9 Exercises 91
if and only if fn → f0 pointwise and

lim | fn | p dμ = | f0 | p dμ .
n→∞ X X
21. Let R be a positive radius and let
α R2
(R2 − |z|2 )α R −1 dA(z)
2
dAR (z) =
πR 2 α R 2
denote the normalized weighted area measure on the disk B(0, R), where dA is
area measure. For any entire function f , show that

lim | f (z)| p dAR (z) = | f (z)| p dλα (z).
R→∞ B(0,R) C
Therefore, we can think of the Fock space as a certain limit of weighted

Bergman spaces.
22. Show that the norm of the operator Qα on Lαp is exactly 2, where 1 ≤ p ≤ ∞.
See [74].
23. If we define

α α α
∑ eα zw− 2 |w| f (u)e− 2 |u| dA(u),
2 2
Tr f (z) =
π w∈rZ2
Sr +w
show that Tr − I∞,α → 0 as r → 0 and (Tr − I) f 1,α → 0 as r → 0 for any

f ∈ Fα1 . Do we have Tr − I1,α → 0 as r → 0?
24. Determine the interpolation space [ fα∞ , Fβp ]θ , where 1 ≤ p ≤ ∞.
25. Use the mean value theorem and Hölder’s inequality to show that there exists a
positive constant C = C(α , p) such that Dr p,α ≤ Cr3−(2/p) for all 0 < p ≤ 1.
This shows that the method employed to prove atomic decomposition for Fα1
can be extended to the range 2/3 < p < 1.
p
26. Let f be an entire function and 0 < p ≤ ∞. Show that f ∈ Fα if and only if there
exists a complex Borel measure μ such that

α
eα az− 2 |a| dμ (a)
2
f (z) =
C
and {|μ |(Sr + w) : w ∈ rZ2 } ∈ l p .

27. If μ is a positive Borel measure on C and 0 < p ≤ ∞, show that the condition
{ μ (Sr + w) : w ∈ rZ2 } ∈ l p is equivalent to the condition that the function z →
μ (B(z, r)) is in L p (C, dA).
p
28. Suppose f ∈ Fα . Then there are constants a, b, and c such that f (z) =
k az 2 +bz+c
z P(z)e , where k is the order of zero of f at the origin and P(z) is the
Weierstrass product associated with the zeros (excluding the origin) of f .
29. Suppose T is a bounded linear operator on Fα2 and it commutes with every
operator Wa . Show that T is a constant multiple of the identity operator. This
result is called Schur’s lemma in mathematical physics. See [177] for example.
92 2 Fock Spaces
30. Show that the main atomic decomposition theorem remains valid if we replace
the square lattice rZ2 by any sequence {wk } in the complex plane with the
following properties: C = ∪k B(wk , r), |wk − w j | ≤ r, and |wk − w j | ≥ r/4
whenever k = j. Here, r is any sufficiently small positive radius.
31. Show that harmonic conjugation is a bounded linear operator on Lαp for
1 ≤ p ≤ ∞.
32. Characterize lacunary series in Fαp . See [226].
33. Prove an atomic decomposition for the space fα∞ .
34. Suppose f ∈ Fαp and f (a) = 0. Show that there exists a positive integer N and
at most one more point b such that

α 2 p α 2 p
f pp,α = N f (a)e− 2 |a| + f (b)e− 2 |b| .
35. Prove the analogs of Propositions 2.40, 2.41, and 2.42 when the operators Wa
are replaced by the operators Ua .
36. Suppose ω1 and ω2 are strictly positive and Lebesgue measurable weight
functions on the complex plane. If 1 ≤ p < ∞ and 1/p + 1/q = 1, then
[L p (C, ω1 dA)]∗ = Lq (C, ω2 dA),
with equal norms, where the duality pairing is given by the integral

f , gω = f (z)g(z) ω (z) dA(z),
C
and 1 1
ω (z) = ω1 (z) p ω2 (z) q
is a geometric mean of ω1 (z) and ω2 (z).
37. Suppose 1 ≤ p < ∞ and 1/p + 1/q = 1. For any positive parameters α and β ,
show that (Lαp )∗ = Lqβ under the integral pairing

γ
2
f , gγ =
π C
where γ = (α + β )/2 is the arithmetic mean of α and β .

38. If 0 < β < α , show that Fβp ⊂ Fαq for all 0 < p ≤ ∞ and 0 < q ≤ ∞.
39. If F is a bounded linear functional on Fαp or fα∞ , show that the function
g(w) = Fz (eγ zw )
is entire.
p p
40. Suppose 1 ≤ p ≤ ∞. Show that Fα is a complemented subspace of Lα , that is,
p p p p p
there exists a closed subspace Xα of Lα such that Lα = Fα ⊕ Xα . Study the case
when 0 < p < 1.
Chapter 3
The Berezin Transform and BMO
In this chapter, we study the Berezin transform on Fα2 and certain spaces of functions
of bounded mean oscillation (BMO) on the complex plane. We first consider the
Berezin symbol of a bounded linear operator on Fα2 and show that this is a Lipschitz
function in the Euclidean metric. We then consider the Berezin transform of a
function and show that there is a semigroup property with respect to the parameter
α . We also consider the action of the Berezin transform on L p spaces and the
behavior of the Berezin transform when it is iterated.
For every exponent p ∈ [1, ∞), we define a space BMO p of functions of bounded
mean oscillation, based on Euclidean disks of a fixed radius, and study the structure
of these spaces. When 1 < p < ∞, we will show that the Berezin transform of every
function in BMO p is Lipschitz in the Euclidean metric.
As is well known, the Berezin transform is closely related to the notion of
Carleson measures. So we include the discussion of Fock–Carleson measures in
this chapter as well.

DOI 10.1007/978-1-4419-8801-0 3,
94 3 The Berezin Transform and BMO
3.1 The Berezin Transform of Operators 95
3.1 The Berezin Transform of Operators
Recall that for each z ∈ C, we use kz to denote the normalized reproducing kernel
at z, namely,
α 2
kz (w) = K(w, z)/ K(z, z) = eα wz− 2 |z| .
These are unit vectors in Fα2 .

If T is any linear operator on Fα2 whose domain contains all the normalized
reproducing kernels, then we can define a function T on C as follows:
T(z) = T kz , kz , z ∈ C, (3.1)
where , is the inner product in Fα2 . We are going to call T the Berezin transform
(or sometimes the Berezin symbol) of T . In particular, if T is a bounded linear
operator on Fα2 , then the Berezin transform T is well defined and is actually real
analytic in C.
Proposition 3.1. Let L(Fα2 ) be the Banach space of all bounded linear operators on
Fα2 . Then T → T is a bounded linear mapping from L(Fα2 ) into L∞ (C). Furthermore,
the mapping is one-to-one and order preserving.
Proof. Everything is obvious except the one-to-one part. To see this, assume that
T is a bounded linear operator on Fα2 and that T kz , kz = 0 for all z ∈ C. Then
T Kz , Kz = 0 for all z ∈ C, where Kz (w) = K(w, z). The function F(z, w) =
T Kz , Kw is real analytic on C × C, holomorphic in w, and conjugate holomorphic
in z. Also, F vanishes on the diagonal of C × C. It follows from a well-known
theorem in several complex variables (see [142] for example) that F is identically
zero on C × C. Consequently, T Kz (w) = 0 for all z and w, or T Kz = 0 for all z ∈ C.
Since the set of finite linear combinations of kernel functions is dense in Fα2 , we
conclude that T = 0.

Note that the proof above concerning the one-to-one property of the Berezin
transform works for certain unbounded operators as well. More specifically, if T is
an unbounded linear operator on Fα2 such that its domain contains all finite linear
combinations of kernel functions and T Kz , Kw is real analytic, then T = 0 implies
that T = 0.
Proposition 3.2. If T is compact on Fα2 , then T(z) → 0 as z → ∞.
Proof. It is easy to see that kz → 0 weakly in Fα2 as z → ∞. This gives the desired
result.

It is a classical result in functional analysis that if T is positive and compact on a
Hilbert space H, then there exists an orthonormal set {en } in H and a nonincreasing
sequence {sn } of positive numbers such that
T (x) = ∑ sn x, en en , x ∈ H.

n
The numbers sn are uniquely determined by T and are called the singular values
of T .
Let T be a positive and compact operator with singular values {sn }, and let
0 < p < ∞. We say that the operator T belongs to the Schatten class S p if the
sequence {sn } belongs to l p . For a more general operator T , we say that it belongs
to the Schatten class S p if |T | = (T ∗ T )1/2 belongs to S p . If {sn } is the sequence of
singular values for |T |, we write
1/p
T S p = ∑ snp .
n
Two special cases are worth mentioning: S1 is called the trace class, and S2 is called
the Hilbert–Schmidt class. We refer the reader to [250] for more information about
the Schatten classes.
Proposition 3.3. If S is a trace-class operator or a positive operator, then

α dA(z).
tr (S) = S(z) (3.2)
π C
Furthermore, a positive operator S belongs to the trace class if and only if the
integral in (3.2) converges.
Proof. First, assume that S is positive, say S = T 2 for some T ≥ 0. Then for any
orthonormal basis {en }, it follows from Fubini’s theorem that
∞ ∞ ∞
tr (S) = ∑ Sen, en α = ∑ Ten22,α = ∑ |Ten (z)|2 dλα (z)
n=1 n=1 n=1 C

∞ ∞
=
C
∑ |Ten (z)| 2
dλα (z) =
C
∑ Ten , Kz 2α dλα (z)
n=1 n=1

∞
=
C
∑ en , T Kz 2α dλα (z) =
C
T Kz 22,α dλα (z)
n=1

= SKz , Kz α dλα (z) =
S(z)K(z, z) dλα (z)
C C

α dA(z).
= S(z)
π C
Next, assume that S is self-adjoint and belongs to the trace class. Then we can
write
|S| + S |S| − S
S= − ,
2 2
where each of the two quotients above is a positive operator in the trace class. The
desired trace formula then follows from the corresponding ones for positive trace-
class operators.
Finally, an arbitrary trace-class operator S can be written as
S + S∗ S − S∗
S= +i ,
2 2i
where each of the two quotients above is a self-adjoint operator in the trace class.
The desired trace formula for S follows from the corresponding ones for self-adjoint
trace-class operators.

Lemma 3.4. Suppose T is a positive operator on a Hilbert space H and x is a
unit vector in H. Then T p x, x ≥ T x, x p for p ≥ 1 and T p x, x ≤ T x, x p for all
0 < p ≤ 1.
Proof. See Proposition 1.31 of [250].

Proposition 3.5. If p ≥ 1 and T is in the Schatten class S p , then T belongs to
L p (C, dA).
Proof. If T is in the trace class, then we can write
T = T1 − T2 + i(T3 − T4 ),
where each Tk is a positive trace-class operator. By Proposition 3.3 above, the

function
T = T1 − T2 + iT3 − iT4
is in L1 (C, dA).
If T is a bounded linear operator on Fα2 , the function T is in L∞ (C, dA). It follows
from complex interpolation that if T is any operator in the Schatten class S p , 1 <
p < ∞, then the function T is in L p (C, dA).
Alternatively, if 1 ≤ p < ∞ and T is in the Schatten class S p , then by the
decomposition T = T1 − T2 + i(T3 − T4 ), we may assume that T is positive. But
when T is positive, it is in the Schatten class S p if and only if T p is in the trace class,
so the desired result follows from Proposition 3.3 and Lemma 3.4.

Note that we did not need the positivity of T above, while this is necessary in the
next proposition.
Proposition 3.6. Suppose 0 < p ≤ 1 and T is a positive operator on Fα2 . If T ∈
L p (C, dA), then T belongs to the Schatten class S p .
Proof. Since T is positive, it belongs to the Schatten class S p if and only if S p is in
the trace class. The desired result then follows from Proposition 3.3 and Lemma 3.4.

Theorem 3.7. Let T be any bounded linear operator on Fα2 . We have

1/2
|T(z) − T(w)| ≤ 2T 1 − |kz, kw |2
for all z and w in C.

Proof. For any z ∈ C, let Pz denote the rank-one projection from Fα2 onto the one-
dimensional subspace spanned by kz . More specifically,
Pz ( f ) = f , kz kz , f ∈ Fα2 .
It is clear that Pz is a positive operator with tr (Pz ) = 1.

Let {ek } be an orthonormal basis of Fα2 with e1 = kz . Then
∞
tr (T Pz ) = ∑ T Pz en, en = T Pzkz , kz = T kz , kz = T(z).
n=1
It follows that
|T(z) − T(w)| = |tr (T (Pz − Pw ))| ≤ T Pz − PwS1 ,
where S1 denotes the trace class as a Banach space. Note that we have just used the
well-known inequality
|tr (T S)| ≤ T SS1
from operator theory.

For any two different complex numbers z and w, the operator Pz −Pw is a rank-two
self-adjoint operator with trace 0. So there is an orthonormal basis in which Pz − Pw
is diagonal with two nonzero eigenvalues λ and −λ , where λ = Pz − Pw > 0.
Consequently, the positive rank-two operator (Pz − Pw )2 has a single nonzero
eigenvalue λ 2 of multiplicity 2, and its trace equals 2λ 2 . It follows that the positive
operator |Pz − Pw| has a single positive eigenvalue λ with multiplicity 2, and its trace
is 2λ , which is also the value of Pz − Pw S1 .
Since
tr (Pz − Pw)2 = tr (Pz − PzPw − Pw Pz + Pw) = 2 − 2tr(Pz Pw ),
we can expand the unit vector kw to an orthonormal basis of Fα2 and calculate the
trace of Pz Pw with respect to this basis to obtain
tr (Pz Pw ) = Pz Pw kw , kw = Pz kw , kw .
But Pz kw = kw , kz kz , we have

tr (Pz − Pw )2 = 2 1 − |kz, kw |2 .
It follows that λ 2 = 1 − |kz , kw |2 , which gives the desired result.

Corollary 3.8 Let T be any bounded linear operator on Fα2 . Then

√
|T(z) − T(w)| ≤ 2 α T |z − w|
Proof. It is easy to see that
1 − |kz, kw |2 = 1 − e−α |z−w| ≤ α |z − w|2

2
for all z and w. The desired Lispchitz estimate is then obvious.

Every bounded linear operator on Fα2 also induces a function on C × C. More
specifically, if S is a bounded linear operator on Fα2 and z ∈ C, then
S f (z) = S f , Kz α = f , S∗ Kz α
for all f ∈ Fα2 . We then define
KS (w, z) = S∗ Kz (w) = S∗ Kz , Kw α = Kz , SKw α (3.3)
for all z and w in C. It is easy to see that the function KS (w, z) is uniquely determined
by the following two properties:

(a) S f (z) = f (w)KS (w, z) dλα (w) for all f ∈ Fα2 and z ∈ C.
C
(b) KS ( · , z) ∈ Fα2 for all z ∈ C.
We collect in the following proposition some of the elementary properties of the
kernel function KS (w, z) induced by S.
Proposition 3.9. The mapping S → KS has the following properties:
(1) KS+T = KS + KT , KcS = cKS .
(2) KS ( · , z) ∈ Fα2 .
(3) KS∗ (w, z) = KS (z, w).
(4) KI (w, z) = K(w, z).
(5) KSn → KS pointwise whenever Sn → S weakly.
(6) |KS (w, z)| ≤ S K(w, w)K(z, z).
(7) KSn → KS uniformly on compacta whenever Sn → S in norm.
(8) KS (z, z) = K(z, z)S∗ (z).
(9) KS (w, w) ≡ 0 if and only if S = 0.
Proof. Properties (1)–(5) and (8) are direct consequences of the definition of KS
in (3.3) and the definition of the Berezin transform. Property (6) follows from (3.3)
and the Cauchy–Schwarz inequality, and it implies property (7). Since the Berezin
transform S → S is one-to-one, we see that (9) follows from (8).

Proposition 3.10. Let S and T be bounded operators on Fα2 . Then

KST (w, z) = KS (u, z)KT (w, u) dλα (u)
C
for all w and z in C.

Proof. It follows from (3.3) that
KST (w, z) = T ∗ S∗ Kz , Kw α = S∗ Kz , T Kw α

= S∗ Kz (u)T Kw (u) dλα (u)
C

= S∗ Kz , Ku α T ∗ Ku , Kw α dλα (u)
C

= KS (u, z)KT (w, u) dλα (u)
C

Proposition 3.11. If S is a positive or trace-class operator, then

tr (S) = KS (z, z) dλα (z).
C
Proof. This follows from Proposition 3.3 and property (8) in Proposition 3.9.

Corollary 3.12. Let S and T be bounded linear operators on Fα2 such that ST is
trace class. Then

tr (ST ) = dλα (w) KS (z, w) KT (w, z) dλα (z).
C C
Proof. This is a direct consequence of Propositions 3.10 and 3.11.

3.2 The Berezin Transform of Functions 101
3.2 The Berezin Transform of Functions
We say that a Lebesgue measurable function ϕ satisfies condition (I p), where 0 <
p < ∞, if ϕ ◦ ta ∈ L p (C, dλα ) for every a ∈ C. In particular, any function satisfying
condition (I p ) must be in L p (C, dλα ).
By a change of variables, we see that a Lebesgue measurable function ϕ on C
satisfies condition (I p ) if and only if

|K(z, a)|2 |ϕ (z)| p dλα (z) < ∞ (3.4)
C
for all a ∈ C. By the exponential form of the kernel function K(w, z), the above
condition is equivalent to

|K(z, a)||ϕ (z)| p dλα (z) < ∞, a ∈ C. (3.5)
C
We are mostly interested in two particular cases: p = 1 and p = 2. The case p = 1 is

needed in this section, while the case p = 2 will be used in Chap. 6 when we study
Toeplitz operators with unbounded symbols. It is clear that every function in L∞ (C)
satisfies condition (I p ).
Suppose f satisfies condition (I1 ). We can then define a function f on C as
follows:

f(z) = f kz , kz = |kz (w)|2 f (w) dλα (w). (3.6)
C
We will also call f the Berezin transform of f . It is clear that we can write

α
f(z) = f (w)e−α |z−w| dA(w) =
2
f (z ± w) dλα (w). (3.7)
π C C
Sometimes, we will need to emphasize the dependence on α . In such situations,

we will use the notation

α
f (w)e−α |z−w| dA(w),
2
Bα f (z) = z ∈ C. (3.8)
π C
Thus, f = Bα f if no parameter is specified.

Theorem 3.13. Let Ht = B1/t for any positive parameter t. Then we have the
following semigroup property: Hs Ht = Hs+t for all positive parameters s and t.
Proof. We check the semigroup property on L∞ (C). For f ∈ L∞ (C), we have

1 1
f (w)e− t |z−w| dA(w)
2
Ht f (z) = (3.9)
πt C
for z ∈ C and

1 1 1
e− s |z−w| dA(w) f (u)e− t |w−u| dA(u)
2 2
Hs Ht f (z) =
π 2 st C C
for z ∈ C. By Fubini’s theorem,

Hs Ht f (z) = f (u)I(z, u) dA(u), z ∈ C,
C
where

1 1 2 − 1 |w−u|2
I(z, u) = e− s |z−w| t dA(w).
π 2 st C
Since

1 1 1 1 1 1
− |z − w|2 − |w − u|2 = − + |w|2 − |z|2 − |u|2
s t s t s t
z u
z u
+ + w+ + w,
s t s t
we have

1 − 1 |z|2 − 1 |u|2 ( zs + ut )w 2 −( 1s + 1t )|w|2
I(z, u) = e s t e e dA(w)
π 2 st C
− 1s |z|2 − 1t |u|2 2
e 1
+ 1t ( 1s + 1t ) tz+su
s+t w e−( s + t )|w| dA(w).
1 1 2
= · s
e
π (s + t) π C
Applying the reproducing formula in F12+ 1 , we obtain

s t
1
e− s |z| − t |u| +( s + t )| s+t | .
1 2 1 2 1 1 tz+su 2
I(z, u) =
π (s + t)
Elementary calculations then show that
1 1
e− s+t |z−u| .
2
I(z, u) =
π (s + t)
Therefore,

1 1
f (u)e− s+t |z−u| dA(u) = Hs+t f (z).
2
Hs Ht f (z) =
π (s + t) C

Because of the following result, the operator Ht is sometimes called the heat
transform.
Theorem 3.14. The function u(x, y,t) = Ht f (z), where z = x + iy, satisfies the heat
equation
∂ 2u ∂ 2u ∂u
+ =4 . (3.10)
∂ x2 ∂ y2 ∂t
Moreover, if f is bounded and continuous on C, then u also satisfies the initial

condition
lim Ht f (z) = f (z), z ∈ C. (3.11)
t→0+
Proof. With z = x + iy and w = u + iv, we have

1 1
f (u, v)e− t [(x−u)
2 +(y−v)2 ]
u(x, y,t) = du dv.
πt R2
Differentiating under the integral sign, we obtain

∂u 1 1
f (u, v)e− t [(x−u) +(y−v) ] du dv
2 2
=− 2
∂t π t R2

1 1
[(x − u)2 + (y − v)2] f (u, v)e− t [(x−u) +(y−v) ] du dv.
2 2
+ 3
π t R2
Similarly,

∂u 2 1
(x − u) f (u, v)e− t [(x−u)
2 +(y−v)2 ]
=− 2 du dv,
∂x πt R2
and

∂ 2u 2 1
f (u, v)e− t [(x−u) +(y−v) ] du dv
2 2
=− 2
∂ x2 π t R2

4 1
(x − u)2 f (u, v)e− t [(x−u) +(y−v) ] du dv.
2 2
+ 3
π t R2
Combining this with a similar calculation for ∂ 2 u/∂ y2 gives

4 1
f (u, v)e− t [(x−u) +(y−v) ] du dv
2 2
Δu = −
π t R2
2

4 1
[(x − u)2 + (y − v)2] f (u, v)e− t [(x−u) +(y−v) ] du dv
2 2
+ 3
π t R2
∂u
=4 ,
∂t
where
∂ 2u ∂ 2u
Δu = +
∂ x2 ∂ y2
is the Laplacian of u. Thus, u satisfies the heat equation (3.10).

To show that u also satisfies the initial condition (3.11), assume that f is bounded
and continuous on C. Fix a point z ∈ C and write

1 1
( f (w) − f (z))e− t |z−w| dA(w)
2
Ht f (z) − f (z) =
πt C

1 1
= +
π t |w−z|<δ π t |w−z|>δ
=: I1 + I2.
Given any positive ε , we can choose a positive δ such that
| f (w) − f (z)| < ε , w ∈ B(z, δ ).
It follows that

ε 1 ε 1
e− t |z−w| dA(z) < e− t |z−w| dA(w) = ε .
2 2
|I1 | ≤
πt |w−z|<δ πt C
On the other hand,

1 1
e− t |z−w| dA(w)
2
|I2 | ≤ 2 f ∞
πt |z−w|>δ

1 1
e− t |w| dA(w)
2
= 2 f ∞
πt |w|>δ
−δ 2 /t
= 2 f ∞ e →0
as t → 0+ . It follows that
lim sup |Ht f (z) − f (z)| ≤ ε .

t→0+
Since ε is arbitrary, we must have
lim Ht f (z) = f (z),

t→0+
which completes the proof of the theorem.

Note that in the heat equation (3.10), the value u(x, y,t) represents the
temperature at the point (x, y) ∈ C at time t. Thus, the function f (z) represents
the initial temperature distribution in the complex plane at time t = 0. With this
interpretation, the assumption that f be bounded and continuous is reasonable.
However, the initial condition in (3.11) can be shown to hold for certain functions
that are more general than bounded and continuous ones.
The following result is a direct consequence of Theorems 3.13 and 3.14.
Corollary 3.15. For any positive α and β , we have the identities
Bα Bβ = B αβ = Bβ Bα .
α +β
If f is bounded and continuous, then
lim Bα f (z) = f (z)

α →+∞
for every z ∈ C.
We need the following result from Fourier analysis to generalize Proposition 3.1
to the Berezin transform of functions.
Lemma 3.16 Suppose that n is a positive integer and f is a function on Rn such
that the function
x → f (x)e|tx| e−x
2
is integrable on Rn with respect to Lebesgue measure dx for any t ∈ Rn . Here,
x = (x1 , . . . , xn ), t = (t1 , . . . ,tn ), tx = t1 x1 + · · · + tn xn ,
and
x2 = x21 + · · · + x2n , dx = dx1 · · · dxn .
If

f (x)P(x)e−x dx = 0
2
Rn
for every polynomial P, then f = 0 almost everywhere on Rn .

Proof. Since
∞
(itx)k
eitx = ∑
k=0 k!
and

N (itx)k ∞
|tx|k

∑ ≤ ∑ = e|tx|
k=0 k! k=0 k!
for all N ≥ 0, we apply the dominated convergence theorem to partial sums to obtain
∞
ik
eitx f (x)e−x dx = ∑ k! (tx)k f (x)e−x dx = 0
2 2
Rn k=0 Rn
for all t ∈ Rn . By the Fourier inversion theorem, we have f (x)e−x = 0, and hence
2
f (x) = 0 for almost every x ∈ Rn .

Note that the integral condition (3.12) in the next proposition is slightly stronger
than condition (I1 ) which was necessary for the definition of Bα f .
Proposition 3.17. The Berezin transform Bα is linear and order preserving. Fur-
thermore, if Bα f = 0 and f satisfies the condition that

| f (z)|e|tz| e−α |z| dA(z) < ∞
2
(3.12)
C
for all real t, then f (z) = 0 for almost every z ∈ C.

Proof. It is clear that each Bα is linear and order preserving.
If Bα f = 0 and f satisfies the integral condition (3.12), then differentiating under
the integral sign gives

∂ n+m
f (w)wm wn e−α |w| dA(w),
2
Bα f (0) = cm,n
∂ zn ∂ zm C
where cm,n is a nonzero constant. It follows that

f (w)wm wn e−α |w| dA(w) = 0
2
for all nonnegative integers m and n. The result then follows from Lemma 3.16.

In the next few results, we describe some of the mapping properties of the Berezin
transform. In particular, we will compare Bα f and Bβ f in various situations.
Theorem 3.18. Let 1 ≤ p ≤ ∞. Suppose α , β , and γ are positive weight parameters.
Then Bα Lβp ⊂ Lγp if and only if γ (2α − β ) ≥ 2αβ .
Proof. First, assume that γ (2α − β ) ≥ 2αβ . Then, in particular, α > β2 . If f ∈ L∞

β,
we write

α −α |z|2 β − α − β2 |w|2
f (w)e− 2 |w| |eα zw |2 e
2
Bα f (z) = e dA(w).
π C
It follows that
2
α f ∞,β −α |z|2 (α − 2 ) α − β w − α − β2 |w|2
β αz
|Bα f (z)| ≤ e e 2 e dA(w)

π C
2

α (α − β2 ) α zβ
−α |z|2 α−
= f ∞,β e e 2 .
α − β2
Therefore,
γ 2α f ∞,β −(α + 2γ − 2α 2 )|z|2
|Bα f (z)|e− 2 |z| ≤
2
e 2α − β .
2α − β
It is elementary to check that the condition γ (2α − β ) ≥ 2αβ is equivalent to
γ 2α 2
α+ − ≥ 0.
2 2α − β
Thus, Bα maps L∞ ∞
β into Lγ .
If f ∈ L1β , the integral

γ 2
I= Bα f (z)e− 2 |z| dA(z)
C
equals

α γ
e−(α + 2 )|z|2
f (w)|e α zw 2 −α |w|2
| e dA(w) dA(z),
π C C
which by Fubini’s theorem is less than or equal to

α γ
| f (w)|e−α |w| dA(w) |eα zw |2 e−(α + 2 )|z| dA(z).
2 2
π C C
With the help of Corollary 2.5, we obtain

2α 2
− α − 22αα+γ |w|2
I≤ | f (w)|e dA(w).
2α + γ C
Again, it is elementary to check that the condition γ (2α − β ) ≥ 2αβ is equivalent to
2α 2 β
α− ≥ .
2α + γ 2
Thus, Bα maps L1β into L1γ .

By complex interpolation, the Berezin transform Bα maps Lβp into Lγp for all
1 ≤ p ≤ ∞ whenever γ (2α − β ) ≥ 2αβ .
To prove the other direction, observe that
Bα f (z) = e−α |z| Qα f (2z).

2
It follows from this and a change of variables that Bα f ∈ Lγp if and only if Qα f ∈
L pγ + α . Therefore, Bα Lβp ⊂ Lγp is equivalent to Qα Lβp ⊂ L pγ + α , which implies that
4 2 4 2
Pα Lβp ⊂ L pγ + α . Combining this with Theorem 2.31, we conclude that Bα Lβp ⊂ Lγp
4 2
implies that
γ α
α 2 ≤ (2α − β ) + ,
4 2
which is equivalent to γ (2α − β ) ≥ 2αβ . This completes the proof of the theorem.

Corollary 3.19. Let α > 0 and β > 0. For 1 ≤ p ≤ ∞, we have
(a) Bα : Lαp → Lβp if and only if β ≥ 2α .
(b) Bα : Lβp → Lαp if and only if 2α ≥ 3β .
Proposition 3.20. Let α > 0 and 1 ≤ p < ∞. Then

(a) Bα : L∞ (C) → L∞ (C) is a contraction.
(b) Bα : C0 (C) → C0 (C) is a contraction.
(c) Bα : L p (C, dA) → L p (C, dA) is a contraction.
Proof. Part (a) is obvious. If f ∈ Cc (C), namely, if f is a continuous function on C

with compact support, then it is easy to see that Bα f ∈ C0 (C). Thus, part (b) follows
from (a) and the fact that Cc (C) is dense in C0 (C) in the supremum norm.
To prove (c), we first consider the case p = 1. In this case, it follows from Fubini’s
theorem that

α
e−α |z−w| dA(z)
2
|Bα f (z)| dA(z) ≤ | f (w)| dA(w)
C π C C

= | f (w)| dA(w).
C
The case 1 < p < ∞ then follows from complex interpolation.

Proposition 3.21. Let 0 < β < α and 1 ≤ p < ∞. Then
(a) Bα f ∈ L∞ (C) implies Bβ f ∈ L∞ (C) with
Bβ f ∞ ≤ Bα f ∞
for all f .
(b) Bα f ∈ C0 (C) implies that Bβ f ∈ C0 (C).

(c) Bα f ∈ L p (C, dA) implies that Bβ f ∈ L p (C, dA) with

|Bβ f (z)| p dA(z) ≤ |Bα f (z)| p dA(z)
C C
for all f .
Proof. Choose a positive γ such that 1/γ + 1/α = 1/β . By Corollary 3.15, we have
Bβ = Bγ Bα . The desired result then follows from Proposition 3.20.

Proposition 3.22. If 0 < β < α , 0 < p < ∞, and f ≥ 0. Then
α
Bα f (z) ≤ B f (z), z ∈ C.
β β
Consequently:
(a) Bβ f ∈ L∞ (C) implies that Bα f ∈ L∞ (C).
(b) Bβ f ∈ C0 (C) implies that Bα f ∈ C0 (C).
(c) Bβ f ∈ L p (C, dA) implies that Bα f ∈ L p (C, dA).
Proof. Since f ≥ 0 and 0 < β < α , we have

α
f (w)e−α |z−w| dA(w)
2
Bα f (z) =
π C

α
f (w)e−β |z−w| dA(w)
2
≤
π C

α β
f (w)e−β |z−w| dA(w)
2
= ·
β π C
α
= Bβ f (z).
β
This proves the desired results.

Theorem 3.23. Suppose α and β are positive weight parameters and f ≥ 0 on C.
For 0 < p ≤ ∞, we have
(a) Bα f ∈ L p (C, dA) if and only if Bβ f ∈ L p (C, dA).
(b) Bα f ∈ C0 (C) if and only if Bβ f ∈ C0 (C).
Proof. Part (a) in the case 1 ≤ p ≤ ∞ and part (b) follow from Propositions 3.21
and 3.22. Part (a) in the case 0 < p < 1 will be proved in Chap. 6.

Recall that for any a ∈ C, we have
ta (z) = z + a, τa (z) = z − a, ϕa (z) = a − z.
The following result shows that the Berezin transform commutes with each of these
maps.
Proposition 3.24. If f is a function such that the Berezin transform Bα f is well
defined, then for any a ∈ C, we have
(i) Bα ( f ◦ ta ) = (Bα f ) ◦ ta .
(ii) Bα ( f ◦ τa ) = (Bα f ) ◦ τa .
(iii) Bα ( f ◦ ϕa ) = (Bα f ) ◦ ϕa .
Proof. By (3.7), we have

f ◦ ta (z) = f ◦ ta (z + w) dλα (w)
C

= f (a + z + w) dλα (w)
C
= f(a + z) = f◦ ta (z)
for any z ∈ C. This proves (i). Replacing a by −a in (i) leads to (ii).

Similarly, it follows from (3.7) that

f ◦ ϕa (z) = f ◦ ϕa (z + w) dλα (w)
C

= f (a − z − w) dλα (w)
C
= f(a − z) = f◦ ϕa (z).
This proves (iii).

For any positive integer n, we use Bnα f to denote the n-th iterate of the Berezin
transform of f , that is, we take the Berezin transform of f repeatedly n times to
obtain Bnα f .
Theorem 3.25. Suppose f ∈ L∞ (C) and n is a positive integer. Then
C f ∞
|Bnα f (z) − Bnα f (w)| ≤ √ |z − w| (3.13)
n

for all z and w in C, where C = 2 α /π .
Proof. Recall that the Berezin transform of f is

α
f (u)e−α |z−u| dA(u).
2
Bα f (z) =
π C
It follows that the difference
D = Bα f (z) − Bα f (w)
can be written as

α z + w −α |u−(z−w)/2|2
− e−α |u+(z−w)/2| dA(u).
2
f u+ e
π C 2
Let (z − w)/2 = reiθ with r ≥ 0. By the rotation invariance of the area measure,

α f ∞ −α |u−r|2 2
|D| ≤ e − e−α |u+r| dA(u)
π C

α f ∞
e−α (|u| |eα (u+ū)r − e−α (u+ū)r | dA(u).
2 +r2 )
=
π C
Write u = x + iy and dA(u) = dxdy. We obtain

α f ∞ ∞ −α y2 ∞
e−α (x +r ) |e−2rα x − e2rα x | dx
2 2
|D| ≤ e dy
π −∞ −∞
√ ∞ ∞
2 α f ∞ 2
e−y dy e−α (x +r ) e2rα x − e−2rα x dx
2 2
=
π −∞ 0
√ ∞
2 α f ∞
e−α (x−r) − e−α (x+r) dx
2 2
= √
π 0
√ ∞ ∞
2 α −α x2 −α x2
= √ f ∞ e dx − e dx
π −r r
√ r
2 α
e−α x dx
2
= √ f ∞
π −r
√ √
4r α 2 α
≤ √ f ∞ = √ f ∞ |z − w|.
π π
Thus, we have proved that

√
2 α
|Bα f (z) − Bα f (w)| ≤ √ f ∞ |z − w| (3.14)
π
for all f ∈ L∞ (C) and all z and w in C.

By Corollary 3.15, we have

α α
f (w)e− n |z−w| dA(w).
2
Bnα f (z) = B f (z) =
α
n πn C
This, along with a simple change of variables, shows that

√
Bnα f (z) = Bα g(z/ n),
√
where g(z) = f ( n z). Combining this with the estimate in (3.14), we obtain the
desired Lipschitz estimate in (3.13).

3.3 Fixed Points of the Berezin Transform 113
3.3 Fixed Points of the Berezin Transform
In the theory of Bergman spaces, it follows from a theorem of Ahern, Flores, and
Rudin that a function is fixed by the Berezin transform in that context if and only
if the function is harmonic, as long as the Berezin transform of the function is well
defined. No other assumption on the function is necessary. See [1].
Therefore, it is natural to ask if the fixed points of the Berezin transform in our
context here are exactly the harmonic functions as well. It turns out that the answer
is negative in general, but positive under certain conditions.
Proposition 3.26. Suppose f is a harmonic function on C satisfying condition (I1 ).
Then f = f .
Proof. If f is harmonic, then f ◦ tz is harmonic for every z. It follows from the mean
value theorem for harmonic functions that

f ◦ tz (0) = f ◦ tz (w) dλα (w).
C
This shows that f (z) = f(z) for every z ∈ C.

The following result gives a partial converse to the proposition above.
Proposition 3.27. If f ∈ L∞ (C), then the following conditions are equivalent:
(a) f = f .
(b) f is harmonic.
(c) f is constant.
Proof. Since f is bounded, the equivalence of (b) and (c) follows from the well-
known maximum modulus principle for harmonic functions. If f is constant, then
clearly f = f . If f = f , then f(n) = f for all positive integers n. By Theorem 3.25,
there exists a positive constant C such that
C
| f (z) − f (w)| ≤ √ |z − w|
n
for all z and w in C with z = w. Let n → ∞. We see that f must be constant.

Finally, in this section, we show by an example that there are more functions than
the harmonic ones that are fixed by the Berezin transform.
Lemma 3.28. For any complex ζ , let
∞
1
eζ t−t dt.
2
I(ζ ) = √
π −∞
We have I(ζ ) = eζ
2 /4
.
Proof. It is clear that I(ζ ) is an entire function of ζ . Differentiating under the

integral sign, we obtain
∞
1

teζ t−t dt
2
I (ζ ) = √
π −∞
∞
1 ζ ζ
eζ t−t dt + I(ζ )
2
= √ t−
π −∞ 2 2
ζ
= I(ζ ).
2
It follows that I(ζ ) = Ceζ /4 for some constant C and all ζ ∈ C. It is well known
2
that I(0) = 1. Thus, I(ζ ) = eζ /4 for all ζ ∈ C.

2

Now fix two complex constants a and b such that a2 + b2 = 8απ i and consider
the function
f (z) = eax+by , z = x + iy ∈ C,
which clearly satisfies condition (I1 ). A direct calculation shows that
Δ f = (a2 + b2) f = 8απ i f ,
so f is not harmonic. On the other hand,

f(z) = f (w + z) dλα (w)
C

= f (z) eau+bv dλα (w),
C
where w = u + iv. Separating the variables, we obtain
f(z) = f (z)I(a, α )I(b, α ),
where
∞
α
eζ t−α t dt.
2
I(ζ , α ) =
π −∞
A simple change of variables gives

√ √
f(z) = f (z)I(a/ α )I(b/ α ),
3.3 Fixed Points of the Berezin Transform 115
where I(ζ ) is the function considered in Lemma 3.28 above. An application of

Lemma 3.28 then gives
f(z) = f (z)e(a +b )/(4α ) = f (z).

2 2
This shows that the function f is fixed by the Berezin transform, but it is not
harmonic.
3.4 Fock–Carleson Measures 117
3.4 Fock–Carleson Measures

Theorem 3.29. Suppose μ is a positive Borel measure on C, 0 < p < ∞, and 0 <
r < ∞. Then the following conditions are equivalent:
(a) There exists a positive constant C such that

α α
| f (w)e− 2 |w| | p dμ (w) ≤ C | f (w)e− 2 |w| | p dA(w)
2 2
C C
for all entire functions f .

(b) There exists a positive constant C such that

pα
e− 2 |z−w|
2
dμ (w) ≤ C
C
for all z ∈ C.
(c) There exists a constant C > 0 such that μ (B(z, r)) ≤ C for all z ∈ C.
Proof. Fix a positive radius r and consider the lattice rZ2 in C. Let {zn } denote any
fixed arrangement of this lattice into a sequence. For any entire function f , we set

α
| f (w)e− 2 |w| | p dμ (w).
2
I( f ) =
C
Then

α
I( f ) ≤ ∑ | f (w)e− 2 |w| | p dμ (w).
2
n B(zn ,r)
By Lemma 2.32 and the triangle inequality, there exists a constant C1 > 0 such that

α α
| f (w)e− 2 |w| | p ≤ C1 | f (u)e− 2 |u| | p dA(u)
2 2
B(zn ,2r)
for all w ∈ B(zn , r). If condition (c) holds, then we can find a positive constant C2
(independent of f ) such that

α
I( f ) ≤ C2 ∑ | f (u)e− 2 |u| | p dA(u)
2
n B(zn ,2r)
for all entire functions f . It is clear that there exists a positive integer N such
that every point in the complex plane belongs to at most N of the disks B(zn , 2r).
Therefore,

α
| f (u)e− 2 |u| | p dA(u).
2
I( f ) ≤ C2 N
C
This shows that condition (c) implies condition (a).

To show that condition (a) implies condition (b), simply take f = kz and apply
Lemma 2.33.
Finally, if condition (b) holds, then

pα
e− 2 |z−w|
2
dμ (w) ≤ C
B(z,r)
for all z ∈ C. This clearly implies that
pα 2
μ (B(z, r)) ≤ Ce 2 r
for all z ∈ C.

It is interesting to notice that condition (c) is independent of p and α . It follows
that if condition (a) holds for some p > 0 and some α , then it holds for every p and
every α (with the constant C dependent on p and α ).
Similarly, condition (a) is independent of r. Therefore, if condition (c) holds for
some r > 0, then it holds for every r > 0 (with the constant C dependent on r).
From now on, we will call any positive Borel measure μ that satisfies any of
the equivalent conditions (a)–(c) above a Fock–Carleson measure. Similarly, we say
that a positive Borel measure μ on C is a vanishing Fock–Carleson measure if

α
| fn (z)e− 2 |z| | p dμ (z) = 0,
2
lim
n→∞ C
whenever { fn } is a bounded sequence in Fαp that converges to 0 uniformly on

compact subsets. We proceed to show that being a vanishing Fock–Carleson
measure is also independent of p and α .
Theorem 3.30. Suppose p > 0, α > 0, r > 0, and μ is a positive Borel measure on
C. Then the following conditions are equivalent:
(i) μ is a vanishing Fock–Carleson measure.
pα
e− 2 |z−w|
2
(ii) dμ (w) → 0 as z → ∞.
C
(iii) μ (B(z, r)) → 0 as z → ∞.
Proof. By the proof of Theorem 3.29, there exists a positive constant C (independent
of z) such that

pα
e− 2 |z−w|
2
μ (B(z, r)) ≤ C dμ (w)
C
for all z ∈ C. So condition (ii) implies (iii).

For any sequence zn → ∞, it is easy to see that the sequence of functions
eα z̄n w
fn (w) = kzn (w) = , w ∈ C,
eα |zn |
2 /2
satisfy fn p,α = 1 and fn (w) → 0 uniformly on compact sets. Therefore, condition

(i) implies (ii).
On the other hand, carefully examining the proof of Theorem 3.29, we see that
there is a positive constant C (independent of f ) such that

2 p
f (w)e−α |w| /2 dμ (w) (3.15)
C

2 p
≤ C ∑ μ (B(zk , r)) f (w)e−α |w| /2 dA(w),
k B(zk ,2r)
where {zk } is a fixed arrangement into a sequence of the lattice rZ2 . If condition (iii)
holds, then z → μ (B(z, r)) is a bounded function, and for any ε > 0, there exists a
positive integer N such that μ (B(zk , r)) < ε whenever k > N. Thus, for any bounded
sequence { fn } in Fαp that converges to 0 uniformly on compact sets, we can estimate
the sequence

2 p
In = fn (w)e−α |w| /2 dμ (w)
C
according to (3.15) as follows:
N
2 p
In ≤ C ∑ fn (w)e−α |w| /2 dA(w) (3.16)
k=1 B(zk ,2r)
∞
2 p
+Cε ∑ fn (w)e−α |w| /2 dA(w),
k=N+1 B(zk ,2r)
where C is a positive constant independent of n. Since fn (w) → 0 uniformly on

compact sets in C, we have
N
2 p
lim ∑ fn (w)e−α |w| /2 dA(w) = 0.
k=1 B(zk ,2r)
n→∞
Let n → ∞ in (3.16). We obtain

2 p
lim sup fn (w)e−α |w| /2 dμ (w)
n→∞ C

2 p
≤ Cε ∑ fn (w)e−α |w| /2 dA(w).
k=N+1 B(zk ,2r)
There is a positive integer m (depending on r only) such that every point in the
complex plane belongs to at most m of the disks D(zk , 2r). Therefore,
∞
2 p 2 p
∑ fn (w)e−α |w| /2 dA(w) ≤ m fn (w)e−α |w| /2 dA(w) ≤ C,
k=N+1 B(zk ,2r) C
where C is another positive constant independent of n (since { fn } is a bounded se-

quence in Fαp ). Therefore, we can find yet another positive constant C (independent
of n and ε ) such that

2 p
lim sup fn (w)e−α |w| /2 dμ (w) ≤ Cε .
n→∞ C
Since ε is arbitrary, we have

2 p
lim fn (w)e−α |w| /2 dμ (w) = 0.
n→∞ C
This shows that condition (iii) implies condition (i). The proof of the theorem is
complete.

Carefully examining the proof of Theorems 3.29 and 3.30 above, we obtain
the following characterization of Fock–Carleson and vanishing Fock–Carleson
measures.
Corollary 3.31. Suppose μ is a positive Borel measure on C, r > 0, and {zn } is
any arrangement into a sequence of the lattice rZ2 . Then
(a) μ is a Fock–Carleson measure if and only if { μ (B(zk , r))} is in l ∞ .
(b) μ is a vanishing Fock–Carleson measure if and only if the sequence
{ μ (B(zk , r))} is in c0 .
Here, l ∞ denotes the space of all bounded sequences, and c0 is the space of all
sequences tending to 0.
Let μ be a complex, regular Borel measure μ on the complex plane. Define

α 2 −α |w|2 α
e−α |z−w| dμ (w),
2
μ
(z) = |kz (w)| e dμ (w) =
π C π C
whenever these integrals converge. If d μ (z) = f (z)dA(z) and f satisfies condition

= f. Thus, we are going to call μ
(I1 ), it is clear that μ the Berezin transform of the
measure μ .
Taking p = 2 in Theorems 3.29 and 3.30, we see that a positive Borel measure
μ on C is a Fock–Carleson measure if and only if μ ∈ L∞ (C), and μ is a vanishing
Fock–Carleson measure if and only if μ ∈ C0 (C).
We also note that when the radius r is fixed, the function z → μ (B(z, r)) is a
constant multiple of the averaging function
μ (B(z, r))
μ
r (z) = .
π r2
Thus, conditions on the function z → μ (B(z, r)) can be replaced with the corre-
sponding conditions on the averaging function μ
r .
3.5 Functions of Bounded Mean Oscillation 123
3.5 Functions of Bounded Mean Oscillation
For any positive radius r and every exponent p ∈ [1, ∞), we define BMOrp to be the
space of locally area-integrable functions f on C such that
f BMOrp = sup MO p,r ( f )(z) < ∞,

z∈C
where
1
1 p
MO p,r ( f )(z) = |f −
fr (z)| p dA .
π r2 B(z,r)
Here,

1
fr (z) = 2 f dA
πr B(z,r)
is the mean (average) of f over the Euclidean disk B(z, r). Clearly, BMOrp is a linear
space.
When p = 2, it is easy to see that

1
MO22,r ( f )(z) = | f (u) − f (v)|2 dA(u) dA(v). (3.17)
2(π r2 )2 B(z,r) B(z,r)
It is also easy to check that
MO22,r ( f )(z) = |
f |2 r (z) − | fr (z)|2 . (3.18)
Lemma 3.32. Let 1 ≤ p < ∞, r > 0, and f be a locally area-integrable function on

C. Then f ∈ BMOrp if and only if there exists some C > 0 such that for any z ∈ C,
there is a complex constant cz with

1
| f (w) − cz | p dA(w) ≤ C. (3.19)
π r2 B(z,r)
Proof. If f ∈ BMOrp , then (3.19) holds with C = f BMO

p
p and cz = f r (z).
r
On the other hand, if (3.19) holds, then by the triangle inequality for the L p
integral,
1
1 p
MO p,r ( f )(z) = |f −
fr (z)| p dA
π r2 B(z,r)
1p
1
≤ | f − cz | p dA + | fr (z) − cz |.
π r2 B(z,r)
By Hölder’s inequality,
1
1 1
p
| fr (z) − cz | = 2 ( f − cz ) dA ≤ | f − c z | p
dA .
π r B(z,r) π r B(z,r)
2
It follows that MO p,r ( f )(z) ≤ 2C for all z ∈ C, so that f ∈ BMOrp .

For any r > 0, we consider the space BOr of continuous functions f on C such
that the function
ωr ( f )(z) = sup{| f (z) − f (w)| : w ∈ B(z, r)}
is bounded on C. We think of ωr ( f )(z) as the local oscillation of f at the point z.

Lemma 3.33. The space BOr is independent of r. Moreover, a continuous function
f on the complex plane belongs to BOr if and only if there exists a constant C > 0
such that
| f (z) − f (w)| ≤ C(|z − w| + 1) (3.20)

Proof. If f satisfies the condition in (3.20), then clearly f ∈ BOr .
To prove the other direction, assume that f ∈ BOr . Thus, there exists a positive
constant M such that
| f (u) − f (v)| ≤ M, (3.21)
whenever |u − v| ≤ r.
Let z and w be two arbitrary points in the complex plane. We are going to show
that (3.20) holds for some positive constant C that is independent of z and w.
If |z − w| ≤ r, then (3.20) holds with C = M. If |z − w| > r, we place points
z0 , . . . , zn on the line segment from z to w in such a way that z0 = z, zn = w, |zk −
zk+1 | = r for 0 ≤ k < n − 1, and |zn−1 − zn | ≤ r. By the triangle inequality and (3.21),
n−1
| f (z) − f (w)| ≤ ∑ | f (zk ) − f (zk+1 )| ≤ nM.
k=0
Since (n − 1)r ≤ |z − w| ≤ nr, we have
nr ≤ |z − w| + r ≤ max(1, 1/r)(|z − w| + 1).
With C = max(M, 1, 1/r), we obtain the desired estimate in (3.20).

Since BOr is actually independent of the radius r, we will write BO for BOr . The
initials in BO stand for bounded oscillation. It is clear that
f BO = sup{| f (z) − f (w)| : |z − w| ≤ 1}
defines a complete seminorm on BO.

We will make the connection between BMOrp and the weighted Gaussian
measures dλα with the help of Fock–Carleson measures. More specifically, for any
1 ≤ p < ∞ and r > 0, we use BArp to denote the space of Lebesgue measurable
functions f on C such that | f | p r (z) is bounded. By the characterization of Fock–
Carleson measures in Sect. 3.4, the space BArp is independent of r. Therefore, we
will write BA p for BArp . More specifically, a Lebesgue measurable function f on C
belongs to BArp if and only if
p
f BA p
p = sup | f | (z) < ∞,
z∈C
where |f | p is the Berezin transform of | f | p with respect to the Gaussian measure
dλα . Although the weight parameter α appears in the definition of the norm above,
the space BA p is independent of α .
The space BA p depends on p. In fact, if 1 ≤ p < q < ∞, then BAq ⊂ BA p and the
containment is strict.
We now describe the structure of BMOrp in terms of the relatively simple spaces
BO and BA p . Recall that ϕz (w) = z − w.
Theorem 3.34. Let α > 0, r > 0, and 1 ≤ p < ∞. Suppose f is a locally area-
integrable function on C. Then the following conditions are equivalent:
(a) f ∈ BMOrp .
(b) f ∈ BO + BA p .
(c) f satisfies condition (I1 ), and there exists a positive constant C such that

| f ◦ ϕz (w) − f(z)| p dλα (w) ≤ C (3.22)
C
for all z ∈ C.
(d) There exists a positive constant C such that for any z ∈ C, there is some complex
number cz with

| f ◦ ϕz (w) − cz | p dλα (w) ≤ C. (3.23)
C
p
Proof. Let f ∈ BMO2r and |z − w| ≤ r. We have
| fr (z) −
fr (w)| ≤ | fr (z) − f2r (z)| + | f2r (z) − fr (w)|

1
≤ | f (u) −
f2r (z)| dA(u)
π r2 B(z,r)

1
+ | f (u) −
f2r (z)| dA(u).
π r2 B(w,r)
Since B(z, r) and B(w, r) are both contained in B(z, 2r), it follows from Hölder’s
inequality that the two integral summands above are both bounded by a constant
that is independent of z and w. This proves that fr belongs to BOr = BO.
On the other hand, we can show that the function g = f − fr belongs to BA p
p p
whenever f ∈ BMO2r . In fact, it follows from (3.17) that f ∈ BMO2r implies that
p
f ∈ BMOr , and it follows from the triangle inequality for L p integrals that
1 1
p 1
p
|g| r (z) =
p | f (u) − fr (u)| dA(u)p
π r B(z,r)
2
1
1
p
≤ | f (u) − f r (z)| p
dA(u)
π r2 B(z,r)
1p
1 r (u) −
+ | f f r (z)| p
dA(u)
π r2 B(z,r)
≤ f BMOrp + ωr ( fr )(z).
Since fr ∈ BOr and f ∈ BMOrp , we have g ∈ BA p .

p
Thus, we have proved that f ∈ BMO2r implies
f = fr + ( f −
fr ) ∈ BO + BA p .
Since r is arbitrary, we conclude that BMOrp ⊂ BO + BA p , which proves that

condition (a) implies condition (b).
It is clear that every function in BO satisfies condition (I p ). Also, every function
in BA p satisfies condition (I p ). Therefore, condition (b) implies that f satisfies
condition (I p ). Since p ≥ 1, f also satisfies condition (I1 ). In particular, condition
(b) implies that the Berezin transform of f is well defined.
By the triangle inequality and Hölder’s inequality,
f ◦ ϕz − f(z)L p (dλα ) ≤ f ◦ ϕz L p (dλα ) + | f(z)| ≤ 2 |

f | p (z).
We see that condition (3.22) holds whenever f ∈ BA p . On the other hand, it follows
from Hölder’s inequality that

f ◦ ϕz − f(z)Lpp (dλα )= | f (z−w) − f(z)| p dλα (w)
C

≤ | f (z−w) − f (z−u)| p dλα (w)dλα (u).
C C
This together with Lemma 3.33 shows that for any f ∈ BO,

f ◦ ϕz − f(z)Lpp (dλα ) ≤ C p [|u − w| + 1]p dλα (w) dλα (u).
C C
The integral on the right-hand side above converges. Thus, condition (3.22) holds
for all f ∈ BO as well, and we have proved that condition (b) implies condition (c).
Mimicking the proof of Lemma 3.32, we easily obtain the equivalence of
conditions (c) and (d).
Finally, if condition (3.22) holds, we can find a positive constant C such that

C
| f (w) − f(z)| p dA(w)
π r2 B(z,r)

≤ | f (w) − f(z)| p |kz (w)|2 dλα (w)
C

= | f ◦ ϕz (w) − f(z)| p dλα (w).
C
This, along with Lemma 3.32, then shows that condition (c) implies condition (a).

As a consequence of Theorem 3.34, we see that the space BMOrp is independent
of r and the Berezin transform of every function in BMOrp is well defined. Thus, we
will write BMO p for BMOrp and define a complete seminorm on BMO p by
f BMO p = sup f ◦ ϕz − f(z)L p (dλα ) = sup f ◦ tz −

f (z)L p (dλα ) .
z∈C z∈C
One of the nice features of this seminorm is that it is invariant under the actions of
ta , τa , and ϕa .
The proof of Theorem 3.34 also shows that every function in BMO p satisfies
condition (I p ). In particular, BMO p ⊂ L p (C, dλα ).
Theorem 3.35. If 1 < p < ∞, then there exists a positive constant C = C(p, α ) such
that
| f(z) − f(w)| ≤ C f BMO p |z − w|
for all z and w in C and all f ∈ BMO p .

Proof. Fix any z ∈ C and fix any directional parameter θ . Consider the curve γ (t) =
z + eiθ t, which is traced out by a particle that starts at z, with unit speed, and in the
θ -direction. Recall that

α
f(γ (t)) = f (u)e−α |γ (t)−u| dA(u).
2
π C
Differentiating under the integral sign gives

d 2α 2
f (u)e−α |γ (t)−u| Re γ (t)(γ (t) − u) dA(u).
2
f (γ (t)) = −
dt π C
For any fixed t, the function

h(u) = Re γ (t)(γ (t) − u)
is harmonic, so it is fixed by the Berezin transform. It follows that

α
e−α |γ (t)−u| Re γ (t)(γ (t) − u) dA(u) =
2
h(γ (t)) = 0.
π C
Therefore, d f(γ (t))/dt is equal to

2α 2
( f (u) −
f (γ (t)))e−α |γ (t)−u| Re γ (t)(γ (t) − u) dA(u).
2
−
π C
Let q be the conjugate exponent, 1/p + 1/q = 1. Then by Hölder’s inequality,

|d f(γ (t))/dt| is less than or equal to
1
2α 2 p
| f (u) − f(γ (t))| p e−α |γ (t)−u| dA(u)
2
π C
times
1
q
q −α |γ (t)−u|2
|γ (t) − u| e dA(u) . (3.24)
C
The integral in (3.24) is, via a simple change of variables, equal to

|u|q e−α |u| dA(u),
2
which is clearly convergent. Therefore, there exists a positive constant C = C(α , p)

such that

d
f(γ (t)) ≤ CMO p ( f )(γ (t)) ≤ C f BMO p
dt
for all t, where
f BMO p = sup MO p ( f )(z) = sup f ◦ ϕz − f(z)L p (dλα ) .

z∈C z∈C
Integrating with respect to t, we obtain
| f(z) − f(w)| ≤ C f BMO p |z − w|

p
The following result gives another way to split the space BMO into the sum of
two simpler spaces: a space of “smooth” functions and a space of “small” functions.
Theorem 3.36. Suppose f ∈ BMO p and 1 ≤ p < ∞. Then f ∈ BO and f −
f ∈ BA p .
Proof. It is easy to see that there is a positive constant C such that

1
| f(z) −
fr (z)| ≤ | f (w) − f(z)| dA(w)
π r2 B(z,r)

≤C | f (w) − f(z)||kz (w)|2 dλα (w)
B(z,r)

≤C | f ◦ ϕz (w) −
f (z)| dλα (w)
C
≤ C f ◦ ϕz − f(z)L p (dλα ) ,
where the last step follows from Hölder’s inequality. This shows that f − fr is a
bounded function. Since a bounded continuous function belongs to both BO and
BA p , we have f−
fr ∈ BO ∩ BA p .
Write
f = ( f − fr ) − ( f−
f− fr ),
and recall from Theorem 3.34 that f −

fr is in BA p . We conclude that f − f belongs
p
to BA . Similarly, we can write
f = fr + ( f−
fr )
and infer that f ∈ BO.

Corollary 3.37. If 1 < p < ∞, then
BMO p = LIP + BA p ,
where LIP is the space of all Lipschitz functions on C. Moreover, a canonical

decomposition is given by f = f+ ( f − f).
The next result characterizes entire functions in BMO p .

Proposition 3.38. Suppose 1 ≤ p < ∞ and f is an entire function. Then f ∈ BMO p
if and only if f is a linear polynomial.
Proof. When f is entire, we have fr = f because of the mean value theorem. It
follows from Theorem 3.34 (and its proof) that f = fr ∈ BO whenever f ∈ BMO p .
Thus, there exists a positive constant C such that
| f (z) − f (w)| ≤ C(|z − w| + 1)
for all z and w. Let w = 0 and use Cauchy’s estimate. We conclude that f must be a
linear polynomial.
Conversely, if f is a linear polynomial, then f is Lipschitz in the Euclidean
metric. In particular, f ∈ BO, and so f ∈ BMO p .

Let VMOrp denote the space of locally area-integrable functions f such that
lim MO p,r ( f )(z) = 0.

z→∞
p p p p
It is clear that VMOr is a subspace of BMOr . Just like BMOr , the space VMOr is
p
also independent of r, and we will write VMO p for VMOr .
Similarly, we consider the space VOr consisting of continuous functions f such
that
lim ωr ( f )(z) = 0.
z→∞
It can be shown that VOr is independent of r, and we will write VO for VOr . The
initials in VO stand for “vanishing oscillation.”
We also consider the space VArp consisting of functions such that

1
lim | f (w)| p dA(w) = 0.
z→∞ π r2 B(z,r)
According to the characterizations of vanishing Fock–Carleson measures in

Sect. 3.4, the space VArp is independent of r and consists of functions f such
that |
f | p (z) → 0 as z → ∞. We will write VA p for VArp . The initials in VA p stand for
“vanishing average.” The following theorem describes the structure of VMO p .
Theorem 3.39. Suppose 1 ≤ p < ∞, r > 0, and f is locally area integrable. Then
the following conditions are equivalent:
(i) f ∈ VMO p = VMOrp .
(ii) MO p ( f )(z) → 0 as z → ∞.
(iii) f ∈ VO + VA p .
Moreover, there are two canonical decompositions for condition (iii) above:
f = f+ ( f − f), f = fr + ( f −

fr ).
We omit the proof.

Corollary 3.40. Suppose f is an entire function. Then f ∈ VMO p if and only if f
is constant.
3.6 Notes 133
3.6 Notes
The Berezin transform was introduced in [23] and then studied systematically in
[23–27] for a number of reproducing Hilbert spaces. It has become an indispensable
tool in the study of operators on function spaces, including Hankel operators,
Toeplitz operators, and composition operators. See [250] for applications of the
Berezin transform in the theory of Bergman spaces. In particular, the proofs of
Propositions 3.3–3.6 were adapted from the corresponding ones in [250].
In the setting of Fock spaces and when parametrized appropriately, the Berezin
transform is nothing but the heat transform. This connection with the heat equation
makes the Berezin transform on Fock spaces particularly useful. The semigroup
property of the heat transforms was first observed in [30].
The Lipschitz estimate for the Berezin transform of a bounded linear operator
on the Fock space is due to Coburn. See [54, 55]. Propositions 3.9–3.11 and
Corollary 3.12 are taken from [55], and these results will be needed in Chap. 6 when
we study Toeplitz operators on the Fock space.
Theorem 3.25, the Lipschitz estimate for the Berezin transform of a bounded
function, was first proved in [29]. Together with the semigroup property, this result
shows that the Berezin transform is a rapidly smoothing operation on bounded
functions, and consequently, a bounded function that is fixed by the Berezin
transform must be constant. On the other hand, there exist unbounded functions
fixed by the Berezin transform that are not harmonic. The example in Sect. 3.3 was
taken from [84]. This example shows the sharp contrast with the Bergman space
theory, where the fixed points of the Berezin transform are exactly the harmonic
functions; see [1].
The characterization of Fock–Carleson measures is analogous to the characteri-
zation of Carleson measures for Bergman spaces. The material in Sect. 3.4 is taken
from [132]. See [250] for the corresponding results in the Bergman space theory.
Note that the notion of Carleson measures was initially introduced in the Hardy
space setting, where a geometric characterization is much more difficult. See [76].
The notion of BMO and VMO using a fixed Euclidean radius was first introduced
in [32, 257]. This idea was then generalized to the setting of bounded symmetric
domains in [21] and to the case of strongly pseudoconvex domains in [149], with
the Euclidean metric replaced by the Bergman metric. The resulting spaces are
independent of the particular radius used, but the dependence on the exponent p
was observed and studied in [248] in the context of Bergman spaces on the unit ball.
The extension to the Fock space setting is straightforward.
The Lipschitz estimate for the Berezin transform of a function in BMO was first
proved in [21] in the context of Bergman spaces on bounded symmetric domains.
The extension to the Fock space, Theorem 3.35, was first carried out in [13].
3.7 Exercises 135
3.7 Exercises
√
1. Show that the Lipschitz constant 2 α in Corollary 3.8 is best possible.
2. Show that the spaces BMO p and VMO p are complete under the norm
f = f BMO p + | f(0)|.
3. Characterize the multipliers of the spaces BMO p and VMO p .

4. Show that the function |z| belongs to BMO p but the function |z|2 does not
belong to BMO p .
5. Show that the function √|z| belongs to VMO p .
6. Show that the function ei |z| belongs to VMO p .
7. Study the behavior of the Berezin transform of the function ln |z|, which is
harmonic everywhere except the origin.
8. If f ∈ L∞ (C), show that the sequence { f(n) } converges to a constant function
as n → ∞. Moreover, the convergence is uniform on any compact subset of C.
9. If f is locally L p -integrable and
lim f (z) = L
z→∞
exists, then f ∈ VMO p .

10. A function f is “eventually slowly varying” if, for any ε > 0, there exist positive
numbers R and δ such that | f (z) − f (w)| < ε whenever |z| > R, |w| > R, and
|z − w| < δ . Show that every eventually slowly varying function is in VMO p .
11. Characterize harmonic functions in BMO p .
12. Suppose α , β , and γ are positive parameters. Show that for 1 ≤ p ≤ ∞, we have
Qα Lβp ⊂ Lγp if and only if α 2 /γ ≤ 2α − β .
13. Show that the Berezin transform Bα is never bounded on Lβp , where α and β
are positive weight parameters.
14. If f ∈ BMO1 , show that Bα (| f |) − |Bα f | is bounded for α > 0.
15. Does the boundedness of Bα (| f |) − |Bα f | imply f ∈ BMO1 ?
16. Consider the previous two problems for 1 < p < ∞.
17. Show that Bα fr (z) = Bα /r2 f (rz), where fr (z) = f (rz).
18. Show that Bα is a bounded and self-adjoint operator on L2 (C, dA).
19. Show that BAq ⊂ BA p whenever 1 ≤ p ≤ q < ∞. Furthermore, the inclusion is
strict if p < q.
20. If f ∈ BMO p , then | f | ∈ BMO p . Similarly, if f ∈ VMO p , then | f | ∈ VMO p .
Chapter 4
Interpolating and Sampling Sequences
In this chapter, we characterize interpolating and sampling sequences for the Fock
spaces Fαp . The characterizations are based on a certain notion of uniform density
on the complex plane. So we will first spend some time discussing this geometric
notion of density which also has applications in other areas of analysis and physics.

DOI 10.1007/978-1-4419-8801-0 4,
138 4 Interpolating and Sampling Sequences
4.1 A Notion of Density 139
4.1 A Notion of Density
Let Z = {zn } be a sequence of distinct points in C. For any set S in C, we let

n(Z, S) = |Z ∩ S| denote the number of points in Z ∩ S. There are two families of sets
we are going to use in this chapter: Euclidean disks and squares. More specifically,
we will use
S = B(w, r) = {z ∈ C : |z − w| < r} ,
and
S = S(w, r) = {z ∈ C : |Re z − Rew| < r/2, |Im z − Imw| < r/2} .
The area of B(w, r) is π r2 , while the area of S(w, r) is r2 .

The lower and upper densities of Z are then defined as
n(Z, B(w, r))

D− (Z) = lim inf inf ,
r→∞ w∈C π r2
and
n(Z, B(w, r))
D+ (Z) = lim sup sup ,
r→∞ w∈C π r2
respectively.
The following result gives an alternative description of these densities in terms
of squares. Note that in the definition above and the proposition below, the quotients
n(Z, B(w, r))/(π r2 ) and n(Z, S(w, r))/r2 represent the average number of points
from Z per square unit in the disk B(w, r) and the square S(w, r), respectively.
Proposition 4.1. For any sequence Z of distinct points in C, let
− (Z) = lim inf inf n(Z, S(w, r)) ,

D
r→∞ w∈C r2
and
+ (Z) = lim sup sup n(Z, S(w, r)) .
D
r→∞ w∈C r2
− (Z) and D+ (Z) = D

Then we have D− (Z) = D + (Z).
Proof. Fix any positive number ε . It is clear that there exist a finite number of
disjoint open squares S(w j , r j ), 1 ≤ j ≤ N, in B(0, 1) such that

0 < π − r12 + · · · + rN2 < ε .
For any w ∈ C and r > 0, it is easy to see that z ∈ S(w + rw j , rr j ) if and only if
(z − w)/r ∈ S(w j , r j ). It follows that the squares S(w + rw j , rr j ) are disjoint and
contained in B(w, r). Thus,

n(Z, B(w, r)) ≥ n Z, ∪Nj=1 S(w + rw j , rr j )
N
= ∑ n(Z, S(w + rw j , rr j ))
j=1
N
n(Z, S(w + rw j , rr j ))
= ∑ (rr j )2
· (rr j )2 .
j=1
It follows that
n(Z, B(w, r)) N

n(Z, S(w + rw j , rr j )) r2j
π r2
≥ ∑ (rr j )2
·
π
j=1
n
n(Z, S(ζ , rr j )) r2j
≥ ∑ inf
(rr j )2
· .
π
j=1 ζ ∈C
Taking the infimum over w, we obtain
n(Z, B(w, r)) N

n(Z, S(w, rr j )) r2j
inf
w∈C π r2
≥ ∑ w∈C
inf
(rr j )2
· .
π
j=1
Letting r → ∞ then leads to

N r2j π − ε −
− (Z) ∑
D− (Z) ≥ D ≥ D (Z).
j=1 π π
Since ε is arbitrary, we must have D− (Z) ≥ D − (Z).

On the other hand, there exist a finite number of squares S(w j , r j ), 1 ≤ j ≤ N,
that cover the unit disk B(0, 1) and satisfy
0 < r12 + · · · + rN2 − π < ε .
For any w ∈ C and r > 0, we have

N
B(w, r) ⊂ S(w + rw j , rr j )
j=1
so that

n(Z, B(w, r)) ≤ n Z, ∪Nj=1 S(w + rw j , rr j )
N
≤ ∑ n(Z, S(w + rw j , rr j ))
j=1
N
n(Z, S(w + rw j , rr j ))
= ∑ (rr j )2
· (rr j )2 .
j=1
4.1 A Notion of Density 141
It follows that
n(Z, B(w, r)) N

n(Z, S(w + rw j )) r2j
π r2
≤ ∑ (rr j )2
·
π
j=1
N
n(Z, S(ζ , rr j )) r2j
≤ ∑ sup (rr j )2
· .
π
j=1 ζ ∈C
First, take the supremum over w ∈ C and then let r → ∞. We obtain

N r2j π + ε +
+ (Z) ∑
D+ (Z) ≤ D ≤ D (Z).
j=1 π π
Since ε is arbitrary, we must have D+ (Z) ≤ D + (Z).

In the previous two paragraphs, we tried to cover the unit disk by a finite number
of squares whose total area is arbitrarily close to the area of the unit disk. If we now
try to cover the unit square S(0, 1) by a finite number of disks whose total area is
arbitrarily close to the area of the unit square, then the same arguments show that
− (Z) ≥ D− (Z) and D
D + (Z) ≤ D+ (Z). This completes the proof of the proposition.

The following result shows that the upper and lower densities can also be defined
in terms of arbitrary sets of Lebesgue measure 1. Note that the Euclidean disk
B(w, r) is just a translation of a dilation of the unit disk |z| < 1.
Theorem 4.2. Let I be any subset of C of Lebesgue measure 1 whose boundary has
Lebesgue measure 0. Then we have
n(Z, w + rI)
r→∞ w∈C r2
and
n(Z, w + rI)
D+ (Z) = lim sup sup .
r→∞ w∈C r2
Proof. The proof is similar to that of Proposition 4.1. We will not need the full
strength of the theorem and will omit its proof here. We refer the interested reader
to [36] for details.

We conclude the section with an example for which we can explicitly compute
the uniform densities.
Proposition 4.3. For any lattice
Λ = {ω + mω1 + nω2 : m ∈ Z, n ∈ Z} ,
we have
1
D+ (Λ ) = D− (Λ ) = .
|Im (ω1 ω 2 )|
Proof. The fundamental region of the lattice Λ is congruent to the parallelogram

spanned by ω1 = a1 + ia2 and ω2 = b1 + ib2, whose area is

det a1 a2 = |a1 b2 − a2b1 | = |Im (ω1 ω 2 )|.
b1 b2
When r is very large, the number of points in Λ ∩ B(w, r) is roughly the area of
B(w, r) divided by the area of the fundamental region of Λ . It follows that
(π r2 )/|Im (ω1 ω 2 )| 1
D+ (Λ ) = D− (Λ ) = lim = .
r→∞ πr 2 |Im (ω1 ω 2 )|

As a special case, if r is any positive number, then the uniform densities of the
square lattice rZ2 are given by
D+ (rZ2 ) = D− (rZ2 ) = 1/r2 .

In particular, if r = π /α , then the uniform densities of the lattice Λα = π /α Z2
are given by
D+ (Λα ) = D− (Λα ) = α /π .
4.2 Separated Sequences 143
4.2 Separated Sequences
Let Z = {zn } be a sequence of distinct points in the complex plane. We say that Z is
separated if
δ (Z) = inf {|zn − zm | : n = m} > 0.
When Z is separated, the number δ = δ (Z) will be called the separation con-
stant of Z.
The next result is a necessary condition that the values of a function in Fαp taken
on a separated sequence must satisfy.
Proposition 4.4. Let Z = {zn } be a separated sequence and 0 < p < ∞. Then there
exists a positive constant C, independent of f , such that

∞ p
−α |zn |2 /2
∑ n
f (z )e p
≤ C f p,α
n=1
for all f ∈ Fαp .

Proof. Let δ = δ (Z) be the separation constant of Z. By Lemma 2.32, there exists
a positive constant C, independent of n and f , such that

| f (zn )e−α |zn | | f (z)e−α |z|
2 /2 2 /2
|p ≤ C | p dA(z)
B(zn ,r)
for all f ∈ Fαp and all n ≥ 1, where r = δ /2. By the definition of the separation
constant, the Euclidean disks B(zn , r) are all disjoint. Therefore,
∞ ∞
∑ | f (zn )e−α |zn | /2| p ≤ C ∑ | f (z)e−α |z|
2 2 /2
| p dA(z)
n=1 n=1 B(zn ,r)

| f (z)e−α |z|
2 /2
≤C | p dA(z)
C
2π C p
= f p,α .
pα
This proves the proposition.

Based on the proposition above, we now make the definition of interpolating
sequences for Fαp .
Let Z = {zn } denote a sequence of distinct points in the complex plane. We say
that Z is an interpolating sequence for Fαp , 0 < p < ∞, if for every sequence {vn } of
values satisfying
∞
−α |zk |2 /2 p
∑ k
v e < ∞, (4.1)
k=1
there exists a function f ∈ Fαp such that f (zk ) = vk for all k ≥ 1.

Similarly, we say that a sequence Z = {zn } of distinct points in C is an

interpolating sequence for Fα∞ if for every sequence {vn } of values satisfying
sup |vn |e−α |zn |

2 /2
< ∞, (4.2)
n≥1
there exists a function f ∈ Fα∞ such that f (zn ) = vn for all n ≥ 1.

Given any sequence Z = {zn } and any entire function f , we write
1/p
∞
α 2 p
f |Z p,α = ∑ f (zn )e− 2 |zn |
n=1
for 0 < p < ∞ and

α
f |Z∞,α = sup | f (zn )|e− 2 |zn | .
2
n≥1
The following result shows that if Z is an interpolating sequence for Fαp , then
interpolation can be performed in a stable way.
Lemma 4.5. Suppose 0 < p ≤ ∞ and Z = {zn } is an interpolating sequence for Fαp .
Then there exists a positive constant C with the following property: whenever {vn }
is a sequence such that {vn e−α |zn | /2 } ∈ l p there exists a function f ∈ Fαp such that
2
f (zn ) = vn for all n and

f p,α ≤ C f |Z p,α . (4.3)
α
Proof. Let X p denote the Banach space of sequences {vk } such that {vk e− 2 |zk | } ∈
2
l p . Let JZ denote the space of all functions f ∈ Fαp such that f (z) = 0 for all z ∈ Z.
It is clear that JZ is a closed subspace of Fαp . For any sequence v = {vk } ∈ X p , there
exists a function f ∈ Fαp such that f (zk ) = vk for all k ≥ 1. We define T (v) = f + JZ .
Then T is a well-defined linear mapping from X p into the quotient space Fαp /JZ . It is
easy to check that T has a closed graph in X p × (Fαp /JZ ). Therefore, by the closed-
graph theorem, the mapping T is continuous, which implies the desired estimate.

If Z is an interpolating sequence for Fαp , we are going to use N p (Z) = N p (Z, α )
to denote the smallest constant C satisfying the inequality in (4.3). We put N p (Z) =
N p (Z, α ) = ∞ when Z is not an interpolating sequence for Fαp . We also use the
convention that N p (0)
/ = 0.
We say that a sequence Z = {zn } of distinct points in C is a sampling sequence
for Fαp , 0 0 such that
∞ p
− α2 |zn |2
C−1 f pp,α ≤ ∑ f (zn )e p
≤ C f p,α (4.4)
n=1

Sampling for Fα∞ requires a slightly different treatment. More specifically, we say
that an arbitrary set Z in C is a sampling set for Fα∞ if there exists a constant C > 0
such that
α
f ∞,α ≤ C sup | f (z)|e− 2 |z|
2
(4.5)
z∈Z
for all f ∈ Fα∞ . When Z is a sequence, we use the term “sampling sequence” instead
of “sampling set.”
We use M p (Z) = M p (Z, α ) to denote the smallest constant C such that
f p,α ≤ C f |Z p,α
for all f ∈ Fαp . Thus, Z is a sampling set for Fα∞ if and only if M∞ (Z) < ∞, and it is a
sampling sequence for Fαp , 0 < p < ∞, if and only if M p (Z) < ∞ and f |Z p,α < ∞
We use the convention that the empty set is not a sampling set for Fαp , which
should be easy to conceive and accept. In particular, we are going to write
/ = ∞.
M∞ (0)
Recall that for any complex number a, the Weyl unitary operator Wa is defined by
α
Wa f (z) = eα āz− 2 |a| f (z − a).
2
Each Wa is a surjective isometry on Fαp . As a consequence of this translation

invariance, we immediately obtain
N p (Z + a) = N p (Z), M p (Z + a) = M p (Z), (4.6)
which allows us to translate our analysis around an arbitrary point to the origin 0.
Our next step is to show that every interpolating sequence for Fαp must be
separated, and every sampling sequence for Fαp must contain a separated sequence
that is still sampling. The following estimate will be needed for this purpose as well
as several other results.
Lemma 4.6. Suppose 0 0 such that
α 2 p
|S(ζ + z)| − |S(ζ )| p ≤ C|z| p − 2 |u|
e f (u) dA(u)
B(ζ ,3δ )
for all ζ ∈ C and all z with |z| ≤ δ .

Proof. For convenience, we write
α
fζ (w) = W−ζ f (w) = e−α ζ̄ w− 2 |ζ | f (ζ + w).
2
It is easy to see that

α
|S(ζ + z)| = e− 2 |z| | fζ (z)|,
2
|S(ζ )| = | fζ (0)|.
It follows that

|S(ζ + z)| − |S(ζ )| = e− α2 |z|2 | fζ (z)| − | fζ (0)|
α 2

= e− 2 |z| − 1 | fζ (z)| + | fζ (z)| − | fζ (0)|
α 2

≤ 1 − e− 2 |z| | fζ (z)| + | fζ (z) − fζ (0)|
α 2 α 2

= e 2 |z| − 1 e− 2 |z| fζ (z) + | fζ (z) − fζ (0)|
α 2 α 2

= e 2 |z| − 1 e− 2 |z| fζ (z) + |z|| fζ (w)|,
where the last step follows from the mean value theorem with some w satisfying
|w| < |z|.
By Lemma 2.32, there exists a constant C1 > 0 such that
α 2 p α 2 p
− 2 |z| − 2 |u|
e fζ (z) ≤ C1 e fζ (u) dA(u)
B(z,δ )
α 2 p
− 2 |u|
≤ C1 e fζ (u) dA(u)
B(0,2δ )
α 2 p
− 2 |u|
= C1 e f (u) dA(u).
B(ζ ,2δ )
The second inequality above follows from the triangle inequality, and the last
equality follows from a change of variables.
On the other hand, it follows from Cauchy’s integral formula that

1 fζ (u) du
fζ (w) = .
2π i |u−w|=δ (u − w)2
Consequently,
1
| fζ (w)| ≤ sup | fζ (u)| ≤ C2 sup | fζ (u)|e−α |u| /2 .
2
δ |u−w|=δ |u−w|=δ
Another application of Lemma 2.32, followed by the triangle inequality and a

change of variables, gives
α 2 p
− 2 |u|
| fζ (w)| p ≤ C3 e f (u) dA(u).
B(ζ ,3δ )
The desired result now follows from the triangle inequality
|u + v| p ≤ 2 p (|u| p + |v| p)
and the elementary inequality

α
0 < e 2 |z| − 1 ≤ C4 |z|2 ≤ C4 δ |z|,
2
|z| ≤ δ .
This completes the proof of the lemma.

Corollary 4.7. For 0 < p ≤ ∞, there is a positive constant C = C(α , p) such that

|S(z1 )| − |S(z2)| ≤ C|z1 − z2 | f p,α
for all f ∈ Fαp and all complex numbers z1 and z2 .

Proof. The case |z1 − z2 | ≤ 1 follows from the lemma above (and its proof, which
gives a version for p = ∞), while the case |z1 − z2 | > 1 is obvious.

Lemma 4.8. Suppose 0 1, we do not
do anything.
If |zn − zm | ≤ 1, we consider the sequence {ak }, where an = 1 and ak = 0 for
k = n. Since Z is an interpolating sequence for Fαp , there exists a function f ∈ Fαp
such that f (zk )e−α |zk | /2 = ak for all k ≥ 1 and
2
f p,α ≤ N p (Z) f |Z p,α = N p (Z).
With the notation S(z) = e−α |z| /2 f (z) from Lemma 4.6 and Corollary 4.7, we have
2

1 = |an | − |am | = |S(zn )| − |S(zm )| ≤ CN p (Z)|zn − zm |,
where C is a positive constant that only depends on α and p. This shows that the
sequence Z is separated.

We now proceed to show that every sampling sequence for Fαp must contain a
separated subsequence that is also a sampling sequence for Fαp . We break the proof
into two cases: 0 < p < ∞ and p = ∞.
Lemma 4.9. Suppose 0 < p < ∞ and Z = {zn } is any sequence of complex numbers.
Then the following two conditions are equivalent:
(a) There exists a positive constant C such that
∞
α 2 p
∑ f (zn )e− 2 |zn | ≤ C f pp,α
n=1

(b) The sequence Z is a union of finitely many separated sequences.
Proof. Condition (a) above simply says that the measure

∞
μ= ∑ δzn
n=1
is a Fock–Carleson measure for Fαp , where δz is the unit point mass at z. Therefore,
according to (an obvious variant of) Theorem 3.29, condition (a) is equivalent to the
existence of a positive integer N such that any square S ⊂ C of side length 1 contains
at most N points from Z, which is clearly equivalent to the condition that Z is the
union of finitely many separated sequences.

An obvious consequence of the above result is that every sampling sequence for
Fαp , where 0 < p < ∞, contains a separated subsequence. The following result shows
that this is true for p = ∞ as well and we can do more than that.
Lemma 4.10. If Z = {zn } is a sampling sequence for Fα∞ , then Z contains a
separated subsequence Z that is also a sampling sequence for Fα∞ .
Proof. Fix a sufficiently small positive number ε whose exact value will be specified
later. Let z1 = z1 , discard the terms in the sequence {zn } that are within ε of z1 , and
denote the remaining terms by {z11 , z12 , · · · } with the original order. Let z2 = z11 ,
discard the terms in the sequence {z1n } that are within ε of z2 , and denote by
{z21 , z22 , · · · } the remaining terms in the original order. Continuing this process,
infinitely many times if necessary, we obtain a subsequence Z = {zn } of Z which
clearly satisfies the condition |zi − zj | ≥ ε whenever i = j. In particular, Z is
separated. Furthermore, for any zk , either it was discarded during the process above,
in which case it is within ε of some point in the sequence Z , or it eventually gets
picked as a term in Z . Either way, we have d(zk , Z ) < ε so that

Z= Z ∩ B(z , ε ) . (4.7)
z ∈Z
Write Z = Z ∪ Z as a disjoint union. Clearly,

f |Z∞,α = max f |Z ∞,α , f |Z ∞,α ≤ f |Z ∞,α + f |Z ∞,α .
Given any w ∈ Z , it follows from (4.7) that there exists some z ∈ Z such that
|w − z| ≤ ε . By the triangle inequality and Corollary 4.7,

|S(w)| ≤ |S(z)| + |S(z)| − |S(w)|
≤ f |Z ∞,α + Cε f ∞,α ,
where C is a positive constant independent of ε and f . Therefore,
f |Z∞,α ≤ 2 f |Z ∞,α + Cε f ∞,α

for all f ∈ Fα∞ . Since Z is a sampling sequence for Fα∞ , there exists a positive constant
c such that c f ∞,α ≤ f |Z∞,α for all f ∈ Fα∞ . Thus,
(c − Cε ) f ∞,α ≤ 2 f |Z ∞,α
for all f ∈ Fα∞ . If the value of ε was chosen such that c − Cε > 0, then there is
another positive constant C > 0 such that C f ∞,α ≤ f |Z ∞,α for all f ∈ Fα∞ ,
which means that Z is sampling for Fα∞ .

We want to show that the lemma above holds for p < ∞ as well. But the proof is
more complicated.
Lemma 4.11. Suppose 0 < p < ∞ and Z = {zk } is a sampling sequence for Fαp .
Then Z contains a separated subsequence that is also a sampling sequence for Fαp .
Proof. By Lemma 4.9, we can write Z = Z1 ∪ Z2 ∪ · · · ∪ Zn as a disjoint union of
separated sequences. We prove the result by induction on n. If n = 1, there is nothing
to prove. Thus, we assume n > 1 and proceed to show that we can find a subsequence
Z of Z such that:
(a) Z is sampling for Fα .
p
(b) Z is the disjoint union of n − 1 separated sequences.

Let δ be the separation constant for Zn (so that |z − w| ≥ δ for all z and w in Zn
with z = w) and write Z = Z1 ∪ · · · ∪ Zn−1 . Fix any positive constant ε < δ /8 and
split Zn into two parts:

< ε , Γ = z ∈ Zn : d(z, Z)
Γ = z ∈ Zn : d(z, Z) ≥ε .
Let Z = Z ∪ Γ . Putting Γ together with Z1 , we have
Z = (Z1 ∪ Γ ) ∪ · · · ∪ Zn−1 ,
and each of the n − 1 sequences above is separated. We will show that Z is sampling
for Fαp when ε is sufficiently small.
Since Z = Z ∪ Γ , we will be done if Γ is empty. If Γ is not empty, we write
Γ = {ζk }. For each k, there exists a point ak ∈ Z such that |ζk − ak | < ε . For i = j,
we have
|ai − a j | = |(ai − ζi ) − (a j − ζ j ) + (ζ j − ζi )|
≥ |ζi − ζ j | − |(ai − ζi ) − (a j − ζ j )|
3
≥ δ − 2ε > δ .
4
In particular, the points in the sequence {ak } are distinct.

Since Z = Z ∪ Γ is sampling for Fα , there is a positive constant c such that
p
c f pp,α ≤ f |Z pp,α = f |Z pp,α + f |Γ pp,α

for all f ∈ Fαp . Using the notation S(z) = f (z)e−α |z|

2 /2
and the triangle inequality,
we have
f |Γ pp,α = ∑ |S(ζk )| p
k
= ∑ [|S(ζk )| − |S(ak )| + |S(ak )|] p

k
p
≤ 2 p ∑ |S(ζk )| − |S(ak )| + |S(ak )| p .
k
Since the ak ’s are distinct points from Z ⊂ Z , we have
∑ |S(ak )| p ≤ f |Z pp,α , f ∈ Fαp .

k
By Lemma 4.6, with δ /8 in place of δ , we can find a constant C > 0 that is

independent of ε and f such that
p
α 2 p
∑ |S(ζk )| − |S(ak )| ≤ Cε ∑
p
B(ak ,δ /2)
f (z)e− 2 |z| dA(z).
k k
Since the sequence {ak } is separated with separation constant at least 3δ /4, there is
another constant C > 0, independent of ε and f , such that
p
∑|S(ζk )| − |S(ak)| ≤ C ε p f pp,α
k
for all f ∈ Fαp . It follows that
c f pp,α ≤ 2 pC ε p f pp,α + (2 p + 1) f |Z pp,α
so that

c − 2 pC ε p f pp,α ≤ (2 p + 1) f |Z pp,α
for all f ∈ Fαp . If the value of ε was chosen such that c − 2 pC ε > 0, then the
sequence Z is sampling for Fαp . This completes the proof of the lemma.

4.3 Stability Under Weak Convergence 151
4.3 Stability Under Weak Convergence
In this section, we consider a notion of weak convergence for relatively closed

subsets in the complex plane and establish several results about sampling and
interpolation that are preserved under weak convergence.
We say that a set in the complex plane is relatively closed if its intersection with
any compact set is still compact. Given a nonempty and relatively closed subset A
of C, let
At = {z ∈ C : d(z, A) < t} , 0 < t < 1.
So At is the set of all points in C that are within distance t of the set A. If A and B
are two nonempty and relatively closed subsets of the complex plane, we define
[A, B] = inf {t : A ⊂ Bt , B ⊂ At }
and call it the Hausdorff distance between A and B. It can be verified that this is
indeed a metric. The assumption that A and B are relatively closed ensures that
[A, B] = 0 only when A = B.
Alternatively,
[A, B] = max (d ∗ (A, B), d ∗ (B, A)) ,
where
d ∗ (A, B) = sup d(z, B) = sup inf |z − w|
z∈A z∈A w∈B
is the asymmetric “distance” from A to B.

From the definition above, we see that [A, B] < ε if and only if the following two
conditions hold:
1. For any a ∈ A, there exists b ∈ B such that |a − b| < ε .
2. For any b ∈ B, there exists a ∈ A such that |a − b| < ε .
Suppose {An } and A are all nonempty and relatively closed subsets of the
complex plane. We say that {An } converges strongly to A if [An , A] → 0 as n → ∞.
We say that {An } converges weakly to A if {An ∩ F} converges strongly to A ∩ F for
every compact set F such that none of An ∩ F and A ∩ F is empty. Since [A, B] is a
distance, the limit of strong and weak convergence is unique.
To simplify notation and statements, we say that a sequence {An } of sequences
converges weakly to the empty set if we can write
An = {an1 , an2 , · · · , } , |an1 | ≤ |an2 | ≤ · · ·
for each n ≥ 1 and ank → ∞ as n → ∞ for each k ≥ 1.

In what follows, whenever we consider a sequence, we assume that it consists of
distinct points and has no finite accumulation point. In particular, such a sequence
is relatively closed in C and can be rearranged so that the modulus of its terms is
nondecreasing. We use the notation W (Z) to denote the collection of weak limits of
all the translates Z + z of Z. The set W (Z) will play a crucial role in our analysis.
We first prove a certain compactness property for uniformly separated sequences
in the complex plane.
Proposition 4.12. For each n ≥ 1, let Zn be a separated sequence. If δ =
infn δ (Zn ) > 0, then there exists a subsequence {Znk } and a separated sequence Z
(possibly empty) such that {Znk } converges weakly to Z.
Proof. We write Zn = {zn1 , zn2 , · · · } with |zn1 | ≤ |zn2 | ≤ · · · . If zn1 → ∞ as n → ∞,
then for every k, we have znk → ∞ as n → ∞. In this case, {Zn } converges weakly to
the empty set.
If zn1 → ∞ as n → ∞, we can find a subsequence {Zn j } such that zn j 1 → z1 as
j → ∞. Then either zn j 2 → ∞ as j → ∞, which implies that for every k ≥ 2, we have
zn j k → ∞ as j → ∞, or {Zn j } has a subsequence whose second components converge
to some z2 ∈ C. In the latter case, the process continues.
There are now two possibilities: either the process terminates after a finite
number, say N, of iterations, which produces a subsequence of {Zn } that converges
weakly to a finite sequence Z = {z1 , · · · , zN }, or the process never stops, which via
a diagonalization argument produces a subsequence of {Zn } that converges weakly
to an infinite sequence Z = {z1 , z2 , · · · }. The condition infn δ (Zn ) > 0 ensures that
the limit sequence Z is separated as well. This proves the desired result.

The following result gives an alternative description of weak convergence for
separated sequences.
Proposition 4.13. Suppose each Zn is a separated sequence with δ = infn δ (Zn ) >
0. Write Zn = {zn1 , zn2 , · · · } with |zn1 | ≤ |zn2 | ≤ · · · . Then {Zn } converges weakly to
Z if and only if one of the following is true:
(a) Z = 0/ is the empty set, and for every k ≥ 1 we have znk → ∞ as n → ∞.
(b) Z = {z1 , · · · , zN } is a finite set, znk → zk for every 1 ≤ k ≤ N, and znk → ∞ for
every k > N.
(c) Z = {z1 , z2 , · · · } is an infinite (separated) sequence and znk → zk for every k ≥ 1.
Proof. It is clear from the definition that any one of the above conditions implies that
{Zn } converges weakly to Z. The other implication follows from Proposition 4.12
and its proof, if we start out with an arbitrary subsequence of {Zn }. Here, we use the
fact that znk → zk (where zk is either finite or infinite) if and only if each subsequence
of {z1k , z2k , · · · } converges to zk .

We now prove that any weak limit of sampling sequences for Fα∞ remains a
sampling sequence for Fα∞ .
Proposition 4.14. Suppose {Zn } converges weakly to Z. Then
M∞ (Z) ≤ lim inf M∞ (Zn ),

n→∞
where M∞ (Z) denotes the Fα∞ sampling constant for Z.

/ we can write Zn = {znk } with |zn1 | ≤ |zn2 | ≤ · · · and have zn1 → ∞

Proof. If Z = 0,
as n → ∞, which implies that znk → ∞ as n → ∞ for every k. Choosing f = 1 in
α
f ∞,α ≤ M∞ (Zn ) sup e− 2 |znk | | f (znk )|
2
shows that M∞ (Zn ) → ∞. The desired result is then obvious.

Next, assume that Z is nonempty. Since M∞ (Z) is the smallest M such that
f ∞,α ≤ M f |Z∞,α ,
we can write
f ∞,α 1
M∞ (Z) = sup = sup .
f ∈Fα∞ f |Z∞,α f ∞,α =1 f |Z∞,α
It follows that the constant M = M∞ (Z) is given by
M −1 = inf f |Z∞,α .
f ∞,α =1
Thus, for any ε ∈ (0, 1), we can find a unit vector f ∈ Fα∞ such that
f |Z∞,α < M −1 + ε .
This is true even when M = ∞. Also, by translation invariance (namely, we can

translate Z and Zn simultaneously if necessary), we may assume that | f (0)| > 1 − ε .
By Corollary 4.7, there exists a positive number δ = Cε , where C > 0 is
independent of ε , such that
α 2
− 2 |w| α 2
e | f (w)| − e− 2 |z| | f (z)| < ε
whenever |w − z| < δ . Since {Zn } converges weakly to Z, there exists a positive

integer N such that

Zn ∩ B̄(0, ε −2 ), Z ∩ B̄(0, ε −2 ) < δ /2
whenever n > N, where B̄(0, r) is the closed disk with center 0 and radius r. Here, we
may assume that ε is small enough so that none of Zn ∩ B̄(0, ε −2 ) and Z ∩ B̄(0, ε −2 )
is empty.
Let a = 1 − (δ ε 2 /2) and assume that ε and δ are small enough so that a ∈
(0, 1). If n > N and w ∈ Zn ∩ B̄(0, ε −2 ), there exists some z ∈ Z ∩ B̄(0, ε −2 ) such
that |w − z| < δ /2. It follows from the triangle inequality that
δ δ
|aw − z| ≤ a|z − w| + (1 − a)|z| < + = δ.
2 2
Therefore,
α α α α
e− 2 |w| | f (aw)| = e− 2 |aw| | f (aw)|e− 2 (1−a )|w| ≤ e− 2 |aw| | f (aw)|
2 2 2 2 2
α 2
α 2 α 2
≤ e− 2 |aw| | f (aw)| − e− 2 |z| | f (z)| + e− 2 |z| | f (z)|
< ε + f |Z∞,α < M −1 + 2ε .
On the other hand, if |w| > ε −2 , then

α α α
e− 2 |w| | f (aw)| = e− 2 |aw| | f (aw)|e− 2 (1−a
2 2 2 )|w|2
α
≤ f ∞,α e− 2 (1−a)|w|
2
α −4 Cα
≤ e− 2 (1−a)ε = e − 4ε .
We may assume that ε is small enough so that

α
e− 2 |w| | f (aw)| ≤ e−(Cα )/(4ε ) < M −1 + 2ε
2
for all |w| > ε −2 . Combining this with the last estimate in the previous paragraph,
we conclude that the function g(z) = f (az) satisfies
g|Zn ∞,α < M −1 + 2ε , n > N.
Since | f (0)| > 1 − ε , we have
g∞,α ≥ |g(0)| = | f (0)| > 1 − ε .
It follows that
g∞,α 1−ε
M∞ (Zn ) ≥ ≥ −1
g|Zn ∞,α M + 2ε
for all n > N. Thus,

1−ε
lim inf M∞ (Zn ) ≥ .
n→∞ M −1 + 2ε
The desired result now follows by letting ε → 0.

As a consequence of the proposition above, we see that small perturbations of
a sampling sequence for Fα∞ remain sampling sequences for Fα∞ . More specifically,
we have the following.
Corollary 4.15. Suppose Z = {zn } is a sampling sequence for Fα∞ . There exists a
positive number δ such that any sequence W = {wn } satisfying |zn − wn | < δ , n ≥ 1,
is still a sampling sequence for Fα∞ .
The discussion above was about the behavior of the sampling constant M∞ (Z)
under weak convergence. The following result concerns the sampling constant
M p (Z) when p < ∞. Recall that for any separated sequence Z, we use
δ (Z) = inf {|z − w| : z ∈ Z, w ∈ Z, z = w}
to denote the separation constant of Z.

Proposition 4.16. For each n, let Zn be a separated sequence in C. If infn δ (Zn ) > 0,
then
M p (Z, α ) ≤ lim inf M p (Zn , α ), 0 < p < ∞,
n→∞
whenever Zn converges weakly to Z.

Proof. When Z = 0, / the desired result is proved just as in the case p = ∞. See the
proof of Lemma 4.14. So we assume Z = 0/ in the rest of the proof.
Let δ = infn δ (Zn ). It follows from Proposition 4.13 that Z is separated and
δ (Z) ≥ δ .
Given any ε > 0, we follow the same argument at the beginning of the proof of
Proposition 4.14 to find a unit vector f in Fαp such that
f |Z p,α ≤ M −1 + ε ,
where M = M p (Z, α ) (which may be infinite).

For any fixed and large enough radius R, we can find a positive integer N such that
[Zn ∩ B̄(0, R), Z ∩ B̄(0, R)] < min(δ /6, ε ), n > N.
Thus, for any n > N and z ∈ Zn ∩ B̄(0, R), we can find some w ∈ Z such that
δ
|z − w| < , |z − w| < ε .
6
Since Z is separated with separation constant at least δ , we see that different z
correspond to different w. By Lemma 4.6, there exists a positive constant C =
C(α , p, δ ) such that

α 2 α 2 p
| f (z)|e− 2 |z| − | f (w)|e− 2 |w| ≤ Cε p | f (u)e−α |u|
2 /2
| p dA(u).
B(w,δ /2)
If 0 < p ≤ 1, it follows from the triangle inequality that

α 2 p α 2 p α 2 α 2 p
f (z)e− 2 |z| ≤ f (w)e− 2 |w| + | f (z)|e− 2 |z| − | f (w)|e− 2 |w|

α 2 p α 2 p
≤ f (w)e− 2 |w| + Cε p f (u)e− 2 |u| dA(u).
B(w,δ /2)
Sum over all z ∈ Zn ∩ B̄(0, R), observe that different z correspond to different w, and
use the facts that f is a unit vector in Fαp and δ (Z) ≥ δ . We obtain
f |Zn ∩ B̄(0, R) pp,α ≤ f |Z pp,α + Cε p .
Since C is independent of R, letting R → ∞ gives
f |Zn pp,α ≤ f |Z pp,α + Cε p < (M −1 + ε ) p + Cε p
for all n > N. It follows that

1
1 p
M p (Zn , α ) ≥
(M −1 + ε ) p + Cε p
for all n > N, and so

1
1 p
lim inf M p (Zn , α ) ≥ .
n→∞ (M −1 + ε ) p + Cε p
Since ε is arbitrary, we must have
lim inf M p (Zn , α ) ≥ M = M p (Z, α ).

n→∞
If 1 ≤ p < ∞, we apply the version of the triangle inequality for p > 1 to get
⎡ ⎤1
p p
α
f |Zn ∩ B(0, R) p,α = ⎣ ∑ f (z)e− 2 |z| ⎦
2
z∈Zn ∩B(0,R)
⎡ ⎤1
p p
α
≤⎣ ∑ f (w)e− 2 |w| ⎦
2
w∈Z∩B(0,R)

1
p p
α
∑ f (u)e− 2 |u| dA(u)
2
+ Cε p
w∈Z B(w,δ /2)
≤ f |Z p,α + C1/pε ≤ M −1 + (1 + C1/p)ε .
Since C is independent of R, letting R → ∞ gives us
f |Zn p,α ≤ M −1 + (1 + C1/p)ε
for all n > N. It follows that

−1
M p (Zn ) ≥ M −1 + (1 + C1/p)ε , n > N,
so that
−1
lim inf M p (Zn , α ) ≥ M −1 + (1 + C1/p)ε .
n→∞
But ε is arbitrary and C is independent of ε , so we must have
lim inf M p (Zn , α ) ≥ M = M p (Z, α ).

n→∞
This completes the proof of the proposition.

Corollary 4.17. Suppose 0 < p < ∞ and Z is a separated sequence with separation
constant δ . If Z is sampling for Fαp and Z is another sequence such that [Z, Z ] is
sufficiently small, then Z is also a sampling sequence for Fαp .
Proof. This follows from Proposition 4.16.

Carefully examining the proof of Lemmas 4.14 and 4.16, we see that more can be
done. More specifically, if Z is separated, then there exists a constant C > 0 such that
M p (Z , α ) ≤ C[Z, Z ]M p (Z)
for sequences Z that are sufficiently close to Z. Here, the constant C only depends
on p and α .
This concludes the discussion about the stability of sampling sequences under
weak convergence. Next, we consider the stability of interpolating sequences under
weak convergence.
Proposition 4.18. Suppose {Zn } converges to Z weakly. Then
N p (Z, α ) ≤ lim inf N p (Zn , α )

n→∞
for all 0 < p ≤ ∞.

Proof. The case Z = 0/ is obvious. Also, by working with a subsequence if necessary,
we may assume that
lim inf N p (Zn ) = lim N p (Zn ) < ∞.

n→∞ n→∞
In particular, we may assume that

S = sup N p (Zn , α ) < ∞.
n
By the proof of Lemma 4.8, we have δ = infn δ (Zn ) > 0. Then it follows easily from
Proposition 4.13 that the sequence Z is also separated and its separation constant is
at least δ .
With the help of Proposition 4.13, we may also assume that
Zn = {zn1 , zn2 , · · · } , Z = {z1 , z2 , · · · } ,
with znk → zk , as n → ∞, for every appropriate k (depending on whether Z is finite

or infinite).
Fix a positive number ε and a sequence v = {vk } ∈ l p . If Z is a finite sequence of

length m, we assume that vk = 0 for k > m. For each n, there exists some function
fn ∈ Fαp such that
α
fn (znk )e− 2 |znk | = vk ,
2
k ≥ 1,
and
fn p,α ≤ N p (Zn ) fn |Zn p,α ≤ Svl p .
By a normal family argument, we may assume that
lim fn (z) = f (z)

n→∞
uniformly on compact subsets of the complex plane. By Fatou’s lemma, we have

p
f ∈ Fα with
f p,α ≤ lim inf fn p,α ≤ vl p lim inf N p (Zn ).
n→∞ n→∞
Furthermore, for any fixed zk ∈ Z, we have

α α
f (zk )e− 2 |zk | = lim fn (znk )e− 2 |znk | = vk .
2 2
n→∞
It follows that f |Z p,α = vl p so that
f p,α ≤ f |Z p,α lim inf N p (Zn ).

n→∞
This shows that

N p (Z) ≤ lim inf N p (Zn )
n→∞
and completes the proof of the proposition.

Corollary 4.19. Suppose 0 < p ≤ ∞ and Z is a separated sequence. If Z is an
interpolating sequence for Fαp , then there exists a positive constant σ such that Z is
interpolating for Fαp whenever [Z , Z] < σ .
Proof. This follows from Proposition 4.18.

4.4 A Modified Weierstrass σ -Function 159
4.4 A Modified Weierstrass σ -Function
A key tool in our proof of the sufficiency of the sampling and interpolating
conditions is a special, modified Weierstrass σ -function. Thus, we let Λα = {ωmn }
denote the square lattice in C that is defined by

ωmn = π /α (m + in),
where m and n run over all integers. Recall that the Weierstrass σ -function
associated to Λα is defined by

z z 1 z2
σα (z) = z ∏
1− exp + ,
m,n ωmn ωmn 2 ωmn
2
where the prime denotes the omission of the factor corresponding to m = n = 0. By

Proposition 1.19, σα (z) is an entire function with Λα as its zero set.
Also, recall that for any a ∈ C, the Weyl unitary operator Wa is defined by
α
Wa f (z) = eα az− 2 |a| f (z − a).
2
Proposition 4.20. The function σα is quasiperiodic in the sense that
Wωmn σα (z) = (−1)m+n+mn σα (z)
for all z and ωmn . Consequently, if

1 1
Rα = z = x + iy : |x| ≤ π /α , |y| ≤ π /α
2 2
is the fundamental region for Λα , then for any z ∈ C, there exists some w ∈ Rα
such that
α α
|σα (z)|e− 2 |z| = |σα (w)|e− 2 |w| .
2 2
Furthermore, there exists a positive constant c such that

α
|σα (z)|e− 2 |z| ≥ cd(z, Λα )
2
for all z ∈ C, where

d(z, Λα ) = min {|z − w| : w ∈ Λ }
is the Euclidean distance from z to Λα .

Proof. See Proposition 1.20 and Corollary 1.21.

The reciprocal density parameter α in Λα is critical for the Fock spaces Fαp . More
precisely, we will see that Λβ is interpolating for Fαp if and only if β < α ; and Λβ is
sampling for Fαp if and only if β > α . When β = α , Λβ is neither interpolating nor
sampling for Fαp , but is a set of uniqueness for Fαp ; see Lemma 5.7.
We will need to perturb the zeros of the Weierstrass σ -function σα (z). Let Z =
{zmn } be a sequence of distinct points in C. If there exists a constant Q > 0 (not
necessarily small!) such that |ωmn − zmn | ≤ Q for all ωmn ∈ Λα , then we say that Z
is uniformly close to Λα . For any sequence Z = {zmn } that is uniformly close to Λα ,
we define an associated function as follows:

z z 1 z2
g(z) = gZ (z) = (z − z00 ) ∏ 1 − exp + . (4.8)
m,n zmn zmn 2 ωmn 2
Here, we assume that z00 is the point of Z closest to 0. Note that both zmn and ωmn
appear in the formula above; it was not a misprint.
Lemma 4.21. Let Z be uniformly close to Λ = Λα and let g be its associated
function defined above. Then g is an entire function and the zero set of g is exactly
Z. Moreover, there exist positive constants C1 , C2 , and c such that
α
|g(z)|e− 2 |z| ≥ C1 e−c|z| log |z| d(z, Z)
2
(4.9)
and
α
|g(z)|e− 2 |z| ≤ C2 ec|z| log |z|
2
(4.10)
for all z ∈ C. Moreover,

α
|g (zmn )|e− 2 |zmn | ≥ C1 e−c|zmn | log |zmn |
2
(4.11)
for all m and n.

Proof. The convergence of the infinite product defining g and the determination
of the zero set of g are similar to the corresponding problems for the Weierstrass
product in Chap. 1. We leave the routine details to the reader.
We may write
α
e− 2 |z| σα (z)
2
α
e− 2 |z| g(z) =
2
d(z, Z)h(z),
d(z, Λ )
where the factor e−α |z|
2 /2
σα (z)/d(z, Λ ) is bounded below (see Proposition 4.20) and
g(z)d(z, Λ )
h(z) = .
σα (z)d(z, Z)
It is easy to see that h is continuous and nonvanishing on the complex plane. So
|h(z)| is bounded below on |z| ≤ 2Q. Here, Q is the constant that satisfies |zmn −
ωmn | ≤ Q for all (m, n). To show that h(z) is bounded below for |z| > 2Q, we rewrite
h(z) = h1 (z)h2 (z)h3 (z),

where

1 1
h1 (z) = exp z ∑
zmn
−
ωmn
,
|zmn |≤2|z|
d(z, Λ ) z − z00 1 − z/zmn

h2 (z) =
d(z, Z) z ∏ 1 − z/ωmn ,
|zmn |≤2|z|
and
(1 − z/zmn) exp(z/zmn )
h3 (z) = ∏ (1 − z/ωmn) exp(z/ωmn )
.
|zmn |>2|z|
Since Z is uniformly close to Λ , we have

1 1 C

zmn − ωmn ≤ |ωmn |2
for some constant C > 0 and all (m, n) = (0, 0). Using this and the elementary
estimates
N N
1
|ew | ≥ e−|w| , ∑∑ n 2 + m2
∼ log N, (4.12)
n=1 m=1
we can find positive constants C and c such that
|h1 (z)| ≥ Ce−c|z| log |z| , z ∈ C. (4.13)
Rewrite h2 (z) as

∏ [1 − (ωmn − zmn )/(ωmn − z)]
h2 (z) = ϕ (z) ,
∏ [1 − (ωmn − zmn )/ωmn ]
where
d(z, Λ ) z − z00 1 − z/zkl
ϕ (z) = ,
d(z, Z) z 1 − z/ωkl

ωkl is the point in Λ that is closest to z, and the finite product ∏ is taken over all
(m, n) such that
(m, n) = (0, 0), (m, n) = (k, l), |zmn | ≤ 2|z|.
It is clear that ϕ (z) is bounded below for |z| ≥ 2Q.

Since Q satisfies |zmn − ωmn | ≤ Q for all m and n, the condition |zmn | ≤ 2|z|
implies that
|ωmn | ≤ 2|z| + Q, |ωmn − z| ≤ 3|z| + Q.
It follows that

∏ 1 − ωmn − zmn ≤ ∏ 1 + Q
ωmn |ωmn |

Q
≤ ∏ 1+ : 0 < |ωmn | ≤ 2|z| + Q .
|ωmn |

To estimate the other product ∏ in h2 (z) above, we move a few additional factors
into ϕ (z) and further assume that |z − ωmn | > Q. Therefore, we can find a positive
constant C, independent of z, such that
∏ {(1 − Q/|ωmn − z|) : Q < |ωmn − z| ≤ 3|z| + Q}

|h2 (z)| ≥ C
∏{(1 + Q/|ωmn|) : 0 < |ωmn | ≤ 2|z| + Q}
for all z ∈ C. If we write z = w + ωkl , where |w| is a bounded function of z, then by

the translation invariance of Λ , we have
∏ {(1 − Q/|ωmn − w|) : Q < |ωmn − w| ≤ 3|z| + Q}

|h2 (z)| ≥ C
∏ {(1 + Q/|ωmn|) : 0 < |ωmn | ≤ 2|z| + Q}
for all z ∈ C. Take the logarithm of the above inequality, use the fact that log(1+x) ∼
x when x is small, and observe that

1
∑ |ωmn − w| : δ < |ωmn − w| < R ∼ R
as R → ∞ (which is easily obtained with the help of polar coordinates), we see that
there are positive constants c and C such that
|h2 (z)| ≥ Ce−c|z| , z ∈ C. (4.14)
To estimate h3 (z), observe that |zmn | > 2|z| implies
(1 − z/zmn ) exp(z/zmn )
1−
(1 − z/ωmn ) exp(z/ωmn )
∼ (1 − z/zmn ) exp(z/zmn ) − (1 − z/ωmn) exp(z/ωmn )
2
z2 z2 z
∼ 2 − 2 =O .
zmn ωmn ωmn
3
It follows that
1
log |h3 (z)| ≥ −C1 |z|2 ∑ | ω mn |
3
≥ −C2 |z|
|zmn |>2|z|
so that
|h3 (z)| ≥ Ce−c|z| , z ∈ C, (4.15)
for some positive constants c and C.

Inserting the estimates (4.13), (4.14), and (4.15) into h = h1 h2 h3 and then into the
function e−α |z| /2 g(z), we have proved the inequality in (4.9), which in turn gives
2
|g(z) − g(zmn )| − α |z|2 d(z, Z)

e 2 ≥ C1 e−c|z| log |z|
|z − zmn | |z − zmn |
for all z = zmn . Fix zmn , let z → zmn , and observe that d(z, Z) = |z − zmn | when z is
sufficiently close to zmn . We then obtain (4.11).
To prove (4.10), we write g = σα H, or
α α
e− 2 |z| g(z) = e− 2 |z| σ (z)H(z).
2 2
α
The quasiperiodicity of σα implies that the factor e− 2 |z| σα (z) is bounded. Rewrite
2
H = H1 H2 H3 , where H1 = h1 , H3 = h3 , and
z − z00 1 − z/zmn
H2 (z) =
z |z ∏ 1 − z/ωmn ,
mn |≤2|z|
and estimate the functions Hk the same way we did hk , the result is (4.10). This
completes the proof of the lemma.

Lemma 4.22. Let g be the function associated to Z = {zmn }. For any positive radius
R, there exists a positive constant C such that

g(z)

z − zmn ≤ C
for all (m, n) and all |z| ≤ R.

Proof. It is clear that

g(z) |g(z)| d(z, Z) |g(z)|

z − zmn = d(z, Z) |z − zmn | ≤ d(z, Z) .
The desired result then follows from the fact that the function g(z)/d(z, Z) is
continuous on the whole complex plane.

Lemma 4.23. Suppose Z is a sequence that is uniformly close to Λα . Then,
D+ (Z) = D− (Z) = α /π .
Proof. Suppose Z = {zmn }, Λα = {ωmn }, and |zmn − ωmn | ≤ Q for all m and n,
where Q is a positive constant. When r is much larger than Q, the number of points
in Z ∩ B(w, r) is roughly the same as the number of points in Λα ∩ B(w, r). More
precisely, it is easy to see that
n(Z, B(w, r))
lim =1
r→∞ n(Λα , B(w, r))
and the convergence is uniform in w ∈ C. This clearly gives the desired result.

The following result is usually referred to as a Lagrange-type interpolation
formula.
Proposition 4.24. Let Z = {zmn } be a separated sequence in C that is uniformly
close to Λβ and let g be the function associated to Z by (4.8). If α < β , then every
function f ∈ Fα∞ can be written as
f (zmn ) g(z)
f (z) = ∑
,
m,n (zmn ) z − zmn
g
where the series converges uniformly on compact subsets of C.

Proof. Since | f (zmn )| ≤ Ceα |zmn |
2 /2
, it follows from (4.11) that

f (zmn )
≤ C exp − 1 (β − α )|zmn |2 + c|zmn | log |zmn |
g (zmn ) 2
for all m and n. This, along with Lemma 4.22, shows that the series converges
uniformly on compact subsets of C.
To show that the series actually converges to f (z), we argue as follows. For each
sufficiently large r, it is easy to see that we can find a simple closed pass S = Sr
such that
d(S, Z) ≥ δ (Z)/2, d(S, 0) > r, |S| ≤ 8π r, (4.16)
where δ (Z) is the separation constant of Z. Let U be the region bounded by S. For
any z ∈ U − Z, we have by the calculus of residues that

1 f (ζ ) dζ f (z) f (zmn ) 1
= − ∑ .
2π i S (ζ − z)g(ζ ) g(z) zmn ∈U g (zmn ) z − zmn
By (4.9), with α replaced by β , (4.16), and the fact that

α
| f (ζ )|e− 2 |ζ | ≤ f ∞,α ,
2
ζ ∈ C,
we see that the integral on the left-hand side above tends to 0 as r → ∞. This proves
the desired expansion for f .

4.5 Sampling Sequences 165
4.5 Sampling Sequences
We say that a set Z in C is a set of uniqueness for Fαp if every function in Fαp that
vanishes on Z must be identically zero. Recall that a sequence Z is a zero set for Fαp
if there exists a function f ∈ Fαp whose zero set is exactly Z. Thus, a zero sequence
is not a set of uniqueness. But we cannot say that Z is a set of uniqueness if and only
if Z is not a zero set for Fαp . It is obvious that each sampling sequence for Fαp is a
set of uniqueness for Fαp . We use the convention that the empty set is not a set of
uniqueness for Fαp , which is again easy to conceive and accept.
Recall that W (Z) is the collection of weak limits of all the translates Z + z of Z.
Lemma 4.25. A separated sequence Z is sampling for Fα∞ if and only if every A ∈
W (Z) is a set of uniqueness (and hence nonempty) for Fα∞ .
Proof. First assume that Z is a sampling sequence. Let A ∈ W (Z) be the weak
limit of some sequence An = Z + ζn , ζn ∈ C. Although the set A may not be a
sequence, it follows from the proof of Proposition 4.14 and the translation invariance
of M∞ (Z) that
M∞ (A) ≤ lim inf M∞ (An ) = M∞ (Z) < ∞,
n→∞
where M∞ (A), just as in the case of sequences, is the smallest M such that
α 2

f ∞,α ≤ M sup | f (z)|e− 2 |z| : z ∈ A
for all f ∈ Fα∞ . So A is a sampling set for Fα∞ . In particular, A is a set of uniqueness
for Fα∞ .
Next, assume that Z is not sampling for Fα∞ . Then there exists a sequence { fn } of
unit vectors in Fα∞ such that fn |Z∞,α → 0 as n → ∞. For each n, we use continuity
to find some zn ∈ C such that
1
| fn (zn )|e−α |zn |
2 /2
= .
2
Let
α
gn (z) = fn (z + zn )e−α zn z− 2 |zn | .
2
Then for each n we have
gn∞,α = fn ∞,α = 1, |gn (0)| = 1/2.
Also,
lim gn |An ∞,α = lim fn |Z∞,α = 0.
n→∞ n→∞
By a normal family argument, we may assume that gn (z) → g(z) uniformly on

compact subsets of C. Clearly, g ∈ Fα∞ , g∞,α ≤ 1, and g(0) = 0. Let A be a weak
limit of the Fα∞ sampling sets An = Z − zn , possibly empty. The existence of such an
A follows from Proposition 4.12.
If A is empty, it is certainly not a set of uniqueness for Fα∞ . If A is not empty, we

fix any point a ∈ A. For any integer k, we can find a point ζk in some Ank such that
|a − ζk | < 1/k. By Corollary 4.7, there exists a positive constant C such that
α 2
− 2 |a| α
|gnk (a)| − e− 2 |ζk | |gnk (ζk )| ≤ C|a − ζk |
2
e
for all k. Let k → ∞ and use the inequality

α
e− 2 |ζk | |gnk (ζk )| ≤ gnk |Ank ∞,α .
2
We obtain g(a) = 0. So g vanishes on A but g(0) = 0. Thus, A is not a set of

uniqueness for Fα∞ . This completes the proof of the lemma.

Lemma 4.26. If M∞ (Z, α ) < ∞, then M∞ (Z, α + ε ) < ∞ for all sufficiently small
ε > 0.
Proof. By Lemma 4.10, Z contains a separated subsequence which is also sampling
for Fα∞ . By working with such a subsequence if necessary, we may assume that Z is
already separated.
Suppose M∞ (Z, α ) < ∞, but for a decreasing sequence of positive numbers εn
approaching 0, we have M∞ (Z, α + εn ) = ∞. We will obtain a contradiction.
For each n, we can find a unit vector fn in Fα∞+εn such that
fn |Z∞,α +εn < εn .
Using the intermediate value theorem for continuous functions, we can also find a
point ζn ∈ C such that
α + εn 1
| fn (ζn )|e− 2 |ζn |
2
= .
2
Let
α + εn
gn (z) = fn (z + ζn )e−(α +εn )ζ n z− 2 |ζn |
2
, n ≥ 1.
Then
1
gn ∞,α +εn = fn ∞,α +εn = 1, |gn (0)| = .
2
Note that
gn∞,α +ε1 ≤ gn ∞,α +εn = 1
for all n. With the help of a normal family argument and passing to a subsequence of
{gn } if necessary, we may assume that gn (z) → g(z) uniformly on compact subsets.
The limit function g is entire, and |g(0)| = 1/2. For any z ∈ C, we have
α α + εn 2
e− 2 |z| |g(z)| = lim e− 2 |z|
2
|gn (z)| ≤ lim gn ∞,α +εn = 1.
n→∞ n→∞
Thus g ∈ Fα∞ with g∞,α ≤ 1.

Let Zn = Z − ζn for every n. Then
gn |Zn ∞,α +εn = fn |Z∞,α +εn < εn .
Since Z is separated, we have infn δ (Zn ) > 0. By Proposition 4.12, {Zn } contains a
weakly convergent subsequence. Let A be the weak limit of some sequence {Znk }.
Then A ∈ W (Z).
If A is empty, it cannot be a set of uniqueness. Assume A = 0/ and fix some
point a ∈ A. For any positive integer j, there exists some point w j ∈ Znk j such that
|a − w j | < 1/ j. By Corollary 4.7, there exists a positive constant C such that

α + ε nk j 2 α + ε nk
− 2 |a| − 2 j |w j |2
e |gnk j (a)| − e |gnk j (w j )| < C|a − w j |

for all j. Letting j → ∞ leads to g(a) = 0. This shows that g ∈ Fα∞ , g(0) = 0, but g
vanishes on A. So A is not a set of uniqueness for Fα∞ . This contradicts Lemma 4.25
as we are assuming that Z is a sampling sequence for Fα∞ .

Lemma 4.27. For any fixed positive number r, the sequence {σk (r)} defined by
α r2
1
σk (r) = t k e−t dt
k! 0
is decreasing in k and tends to 0 as k → ∞.

Proof. It is well known that the incomplete gamma function
∞
Γ (a, z) = t a−1 e−t dt
z
has the property that

k
zj
Γ (k + 1, z) = k!e−z ∑ .
j=0 j!
It follows that
∞ ∞
1 k −t k −t
σk (r) = t e dt − t e dt
k! 0 α r2
1
= k! − Γ (k + 1, α r2 )
k!
k
(α r2 ) j
= 1 − e− α r ∑
2
,
j=0 j!
which is clearly decreasing in k and tends to 0 as k → ∞.

Lemma 4.28. If Z is a sampling sequence for Fα∞ , then D− (Z) > α /π .
Proof. By Lemma 4.10, Z contains a separated subsequence which is also sampling
for Fα∞ . Therefore, by working with such a subsequence if necessary, we may
assume that Z is already separated.
In view of Lemma 4.26, we just need to show that D− (Z) ≥ α /π . So let us
assume the contrary and write D− (Z) = α /π (1 + 2ε ) for some positive number ε
(the case D− (Z) = 0 can be handled similarly). We will show that this leads to a
contradiction.
Recall that
n(Z, B(w, r))
r→∞ w∈C π r2
where n(Z, B(w, r)) is the number of points in Z ∩ B(w, r). So the assumption
D− (Z) = α /π (1 + 2ε ) implies that there exist sequences {rn } and {wn } such that
rn → ∞ and
n(Z, B(wn , rn )) α
< , n ≥ 1.
2
rn 1+ε
Let
√ √
Rn = rn / 1 + ε Bn = B(0, rn ) = B(0, 1 + ε Rn ),
and
√
Nn = n(Z, B(wn , rn )) = n(Z, B(wn , 1 + ε Rn )).
Then Nn is the number of points in (Z − wn ) ∩ Bn and
α rn2
≤ Nn < α R2n .
1 + 2ε
In particular, Nn → ∞ as n → ∞.
To simplify the notation, we fix any n and write B = Bn , R = Rn , and N = Nn .
Let p = pn be “the” (unique up to a unimodular constant multiple) polynomial with
(Z − wn ) ∩ Bn as its zero set, normalized so that p2,α = 1.
We can write

N
αk k N
p(z) = ∑ ak fk (z), fk (z) = z, ∑ |ak |2 = 1.
k=0 k! k=0
It is easy to see that the functions { fk } are also orthogonal over the disk B:
√
fk (z) fm (z) dλα (z) = σk ( 1 + ε R)δk,m ,
B
where the constants σk are from Lemma 4.27. It follows from this and Lemma 4.27
that
N
B
|p(z)|2 dλα (z) = ∑ |ak |2 B
| fk (z)|2 dλα (z)
k=0
N √ N √
= ∑ |ak |2 σk ( 1 + ε R) ≥ ∑ |ak |2 σN ( 1 + ε R)
k=0 k=0
√
1 α (1+ε )R N −t
2
= σN ( 1 + ε R) = t e dt
N! 0
(1+ε )N (1+ε )N
1 1
≥ t N e−t dt ≥ t N e−t dt
N! 0 N! N
(1+ε )N
NN N N e−N
≥ e−t dt = (1 − e−ε N ).
N! N N!
This, together with Stirling’s formula

√
N! ∼ N N e−N N, N → ∞,
shows that there exists a constant C = C(α , ε ) > 0 (independent of N) such that

C C
|p(z)|2 dλα (z) ≥ √ ≥ √ .
B N αR
Since

α α 2 2
|p(z)|2 dλα (z) ≤ (1 + ε )R2 sup p(z)e− 2 |z| ,
B π z∈B
we can find another positive constant C = C(α , ε ) (independent of R) such that

α 2 3
p∞,α ≥ sup p(z)e− 2 |z| ≥ CR− 2 .
z∈B
On the other hand, for any z outside B and 0 ≤ k ≤ N, we can write
|z|2 = (1 + t)R2, t ≥ ε,
and deduce from

∞
(α R2 )k (α R2 ) j
∑ = eα R
2
≤
k! j=0 j!
that
αk
| fk (z)|2 e−α |z| = (1 + t)k R2k e−α (1+t)R
2 2
k!
(α R2 )k e−α R −α tR2 +k log(1+t)
2
= e
k!
≤ e−α tR
2 +k log(1+t)
≤ e−α tR
2 +N log(1+t)
≤ e−α tR
2 +α R2 log(1+t)
.
Since t ≥ ε , there exists another constant c = c(α , ε ) > 0 (independent of R) such

that
| fk (z)|2 e−α |z| ≤ e−2cR

2 2
for all 0 ≤ k ≤ N and z outside B. By the Cauchy–Schwarz inequality and the fact
that ∑Nk=0 |ak |2 = 1, we have
2
N

2 −α |z|2
= ∑ ak fk (z) e−α |z|
2
|p(z)| e
k=0
N N
∑ |ak |2 ∑ | fk (z)|2 e−α |z|
2
≤
k=0 k=0
≤ (N + 1)e−2cR ≤ (α R2 + 1)e−2cR .
2 2
for all z outside B. From this, we deduce that

p|Zn ∞,α = sup |p(z)|e−α |z| /2 : z ∈ Zn ∩ (C − B)
2

≤ α R2 + 1e−cR ,
2
where Zn = Z − wn .
Finally, if we set
α
gn (z) = eα wn z− 2 |wn | pn (z − wn ),
2
then
− 32
gn ∞,α = pn ∞,α ≥ CRn
and
$
α R2n + 1e−c|Rn |
2
gn |Z∞,α = pn |Zn ∞,α ≤
so that
gn |Z∞,α 5
≤ C Rn2 e−c|Rn |
2
gn ∞,α
for all n ≥ 1, where C and c are positive constants independent of n. Since Rn → ∞
as n → ∞, we conclude that
gn|Z∞,α
lim = 0.
n→∞ gn ∞,α
This contradicts with the assumption that Z is a sampling sequence for Fα∞ and
completes the proof of the lemma.

Lemma 4.29. Suppose 0 < p ≤ ∞ and Z is a sampling sequence for Fαp . Then Z is
a set of uniqueness for Fα∞ .
Proof. By Lemmas 4.10 and 4.11, we may assume that Z is separated.
The case p = ∞ is obvious. Suppose 0 < p < ∞, Z is sampling for Fαp , but Z is
not a set of uniqueness for Fα∞ . Then there exists a function f ∈ Fα∞ , not identically
zero, such that f vanishes on Z. Let g(z) = f (rz), where 0 < r < 1. Then g ∈ Fαp ,
g is not identically zero, and g vanishes on Z/r. This is impossible because by
Corollary 4.17, the sequence Z/r is sampling for Fαp when r is sufficiently close
to 1. Therefore, Z must be a set of uniqueness for Fα∞ .

Lemma 4.30. Suppose 0 α /π .
Proof. Again, by working with a subsequence of Z if necessary, we may assume
that Z is already separated.
Recall that W (Z) consists of all weak limits of translates of Z. Since every
translation of Z is also a sampling sequence for Fαp with the same separation
constant, it follows from Proposition 4.16 that every sequence in W (Z) is sampling
for Fαp as well. Combining this with Lemmas 4.25 and 4.29, we conclude that Z is a
sampling sequence for Fα∞ . Thus, D− (Z) > α /π by Lemma 4.28.

This completes the proof for the necessity of the sampling condition D− (Z) >
α /π for Fαp . We now proceed to prove the sufficiency. This will be accomplished
with the help of the Weierstrass σ -function and its variant g(z) discussed in
the previous section. The first step is to show that every sequence contains a
subsequence that is uniformly close to a square lattice Λγ and whose uniform lower
density changes very little.
Lemma 4.31. Suppose 0 < α < β and Z is a sequence with D− (Z) = β /π . There
exists a subsequence Z of Z such that Z is uniformly close to Λγ for some
α < γ < β.
Proof. Fix γ ∈ (α , β ) and choose ε > 0 such that γ + ε < β . The condition D− (Z) =
β /π implies that there exists a positive number r such that any square of side length
r contains at least (γ + ε )r2 /π points from Z.
We decompose C into the disjoint union of a sequence of squares (half open, half
closed) of side length r: C = ∪{Sk : k ≥ 1}. Since the area of each Sk is r2 and the
area of the fundamental region of Λγ is π /γ , each Sk contains r2 /(π /γ ) = γ r2 /π
points from Λγ (plus or minus a few points that can be neglected for our purpose).
But Sk contains at least (γ + ε )r2 /π points from Z. So for each k, we can choose
|Λγ ∩ Sk | points from Z to match those in Λγ ∩ Sk . We do this for each k, and the
result is a subsequence
√ of Z that is uniformly √ close to Λγ . More specifically, we
have |zmn − ωmn | ≤ 2 r for all m and n, where 2 r is the length of the diagonal of
each Sk .

We now prove that the condition D− (Z) > α /π is sufficient for a separated
sequence Z to be a sampling sequence of Fαp . For clarity, we break the proof into
three cases: 0 
α /π , then Z is a sampling sequence for Fαp .
Proof. Given a function f ∈ Fαp ⊂ Fα∞ we need to estimate the integral

α 2 p
I= f (z)e− 2 |z| dA(z)
C
from above. By Lemma 4.31, we may assume that Z is uniformly close to a square
lattice Λβ with β > α . Let Ω = Rα be the fundamental region for the square lattice
Λα = {ωmn } = {−ωmn }. Then by Lemma 1.13 and a change of variables, we have
p
− α2 |z|2
I=∑ e Wωkl f (z) dA(z),
k,l Ω
where Wωkl are the Weyl unitary operators defined in Sect. 2.6.
To estimate each summand on the right-hand side above, first observe that Z + ωkl
is uniformly close to Λβ as well, with a constant Q that is independent of k and l.
Thus, we can use Proposition 4.24 to write
Wωkl f (zmn + ωkl ) gωkl (z)

Wωkl f (z) = ∑ ,
m,n gωkl (zmn + ωkl ) z − zmn − ωkl

where gωkl is the Weierstrass σ -type function associated to the sequences Z + ωkl
and Λβ .
For ε = (β − α )/2, we can write
α
e−ε |zmn +ωkl | e− 2 |zmn | | f (zmn )|
2 2
|Wωkl f (zmn + ωkl )|
= β . (4.17)
|gωkl (zmn + ωkl )|

e− 2 |zmn +ωkl | |gωkl (zmn + ωkl )|
2
Let q = p/(p − 1) so that 1/p + 1/q = 1. Then by Hölder’s inequality and (4.11) in
Lemma 4.21, we see that |Wωkl f (z)| p is less than or equal to
α p

Ch(z) ∑ e− 2 |zmn | f (zmn ) e−ε |zmn +ωkl | +c|zmn +ωkl | log |zmn +ωkl | ,
2 2
m,n
where
q qp
gωkl (z)
h(z) = hkl (z) = ∑e −ε |zmn +ωkl |2
z − zmn − ω
kl

.
m,n
By Lemmas 1.12 and 4.22, the positive function h(z) is bounded on Ω with an upper
bound that is independent of k and l. In particular, the integral

pα 2
h(z)e− 2 |z| dA(z)
Ω
is dominated by a positive constant that is independent of k and l. Therefore, there

exist positive constants C and C such that
α p

I ≤ C ∑ ∑ e− 2 |zmn | f (zmn ) e−ε |zmn +ωkl | +c|zmn +ωkl | log |zmn +ωkl |
2 2
k,l m,n
α p

= C ∑ e− 2 |zmn | f (zmn ) ∑ e−ε |zmn +ωkl | +c|zmn +ωkl | log |zmn +ωkl |
2 2
m,n k,l

≤C f |Z pp,α ,
which is the desired estimate. Note that the last estimate above follows from
Lemma 1.12.

Lemma 4.33. Suppose 0 
α /π . Then Z is sampling for Fαp .
Proof. With notation from the proof of the previous lemma, we use the assumption
0 < p ≤ 1 to get

W f (z + ω ) p g (z) p
ωkl mn kl ωkl
|Wωkl f (z)| ≤ ∑
p
.

m,n g ωkl
(zmn + ω kl ) z − zmn − ωkl
Combining this with (4.11) and (4.17), we obtain positive constants C and c, both
independent of k and l, such that
p p
gωkl (z) − α |z |2
|Wωkl f (z)| ≤ C ∑
p e 2 mn f (zmn ) E(m, n, k, l),
z − zmn − ω
m,n kl
where
E(m, n, k, l) = e−pε |zmn +ωkl | mn +ωmn | log |zmn +ωkl |
2 +c|z
.
Integrate the above inequality over Ω with respect to e−pα |z| /2 dA(z) and notice that
2
Lemma 4.22 implies

p

gωkl (z) e− 2 |z|2 dA(z) ≤ C
pα
Ω
z − zmn − ω
kl
for some constant C > 0 that is independent of k and l. We obtain another constant
C > 0 such that
α p

I ≤ C ∑ ∑ e− 2 |zmn | f (zmn ) E(m, n, k, l)
2
k,l m,n
α p

= C ∑ e− 2 |zmn | f (zmn ) ∑ E(m, n, k, l)
2
m,n k,l
α p

≤ C ∑ e− 2 |zmn | f (zmn )
2
m,n

= C f |Z pp,α ,
which is the desired estimate.

Lemma 4.34. Any separated sequence Z with D− (Z) > α /π is a sampling
sequence for Fα∞ .
Proof. With notation from the proof of the previous two lemmas, we have
f ∞,α = sup Skl ,
k,l
where
α 2
Skl = sup e− 2 |z| |Wωkl f (z)| : z ∈ Ω .
To shorten the displays below, let
e(m, n, k, l) = e−ε |zmn +ωkl | mn +ωkl | log |zmn +ωkl |

2 +c|z
.
Then by (4.17), (4.11), and Lemmas 4.22 and 1.12, we have

α 2 gωkl (z) − α |z |2
Skl≤ C sup ∑ e− 2 |z| e 2 mn f (zmn ) e(m, n, k, l)
z − zmn − ω
z∈Ω m,n kl
≤ C f |Z∞,α ∑ e(m, n, k, l)
m,n
≤ C f |Z∞,α ,
which proves the desired result.

This completes the proof of the sufficiency of the condition D− (Z) > α /π for Z
to be a sampling sequence of Fαp . We summarize the main results of this section as
the following two theorems.
Theorem 4.35. A set Z is sampling for Fα∞ if and only if Z contains a separated
sequence Z such that D− (Z ) > α /π .
Theorem 4.36. Let Z be a sequence in C and 0 α /π .
Corollary 4.37. If Z is separated and 0 α /π .
4.6 Interpolating Sequences 177
4.6 Interpolating Sequences
In this section, we characterize interpolating sequences for Fαp by the condition

D+ (Z) < α /π . We begin with the sufficiency, which is still based on the modified
Weierstrass σ -function associated to a separated sequence that is uniformly close
to a square lattice. The first step is to show that every separated sequence can be
expanded to a sequence that is uniformly close to a square lattice and whose uniform
upper density increases very little.
Lemma 4.38. Let Z be a separated sequence in C with D+ (Z) = β /π and β < α .
We can expand Z to a separated sequence Z such that Z is uniformly close to a
square lattice Λγ with γ ∈ (β , α ).
Proof. Let γ ∈ (β , α ) and choose ε > 0 such that β < γ − ε . The condition D+ (Z) =
β /π implies that there is some large r such that any square of side length r contains
at most (γ − ε )r2 /π points from Z.
Just as in the proof of Lemma 4.31, we decompose the complex plane into the
disjoint union of squares (half open, half closed) of side length r: C = ∪Sk . Each
Sk contains at most (γ − ε )r2 /π points from Z. On the other hand, each Sk contains
r2 /(π /γ ) = γ r2 /π points from Λγ . Therefore, we can add a certain number of points
in each Sk to Z to match the number of points in Λγ ∩ Sk so that the expanded
sequence Z will be uniformly close to Λγ . It is easy to see that we can also do the
expansion in such a way that the new sequence Z remains separated.

Lemma 4.39. Suppose 0 < p ≤ ∞ and Z is a separated sequence. If D+ (Z) < α /π ,
then Z is interpolating for Fαp .
Proof. If we remove any number of points from an interpolating sequence for Fαp ,
what remains is still an interpolating sequence for Fαp : we just assign the value 0
to f (z) for those removed z. So by Lemma 4.38, we may as well assume that Z is
uniformly close to the square lattice Λβ = {ωmn } with D+ (Z) = β /π and β < α .
For any sequence {akl } of values for which
α

akl e− 2 |zkl | ∈ l p ,
2
we claim that the interpolation problem f (zkl ) = akl is solved explicitly by the
function
gmn (z − zmn )
f (z) = ∑ amn eα zmn z−α |zmn |
2
, (4.18)
m,n z − zmn
where gmn denotes the generalized Weierstrass σ -function associated with the
sequences Z − zmn and Λγ as given in (4.8) (it is easy to see that Z − zmn is uniformly
close to Λγ ). More specifically,
gmn (z − zmn )
z − zmn
is equal to

z − zmn z − zmn 1 (z − zmn )2
∏ 1 − zkl − zmn exp zkl − zmn + 2 ω 2 .
(k,l)=(m,n) kl
In particular,
gmn (zmn − zmn ) = gmn (0) = 1, gmn (zkl − zmn ) = 0,
for (k, l) = (m, n). Since

α α
α 2 2 gmn (z − zmn )
e− 2 |z| | f (z)| ≤ ∑ e− 2 |zmn | amn e− 2 |z−zmn | ,
2
m,n z − zmn
the series above can be written as

α α −β
− 2 |zmn |2 − 2 |z−zmn |2 − β2 |z−zmn |2 gmn (z − zmn )
∑ e amn e e z − zmn .
m,n
By (4.10), there exist positive constants C, C , and c such that

α
α 2
e− 2 |z| | f (z)| ≤ C ∑ e− 2 |zmn | amn e−δ |z−zmn | +c|z−zmn | log |z−zmn |
2 2
m,n
α δ

≤ C ∑ e− 2 |zmn | amn e− 2 |z−zmn |
2 2
m,n
for all z ∈ C, where δ = (α − β )/2. Since Z = {zmn } is uniformly close to the square
lattice Λβ = {ωmn }, we can find another positive constant C such that
α
α 2
e− 2 |z| | f (z)| ≤ C ∑ e− 2 |zmn | amn e−σ |z−ωmn |
2 2
(4.19)
m,n
α
for all z ∈ C, where σ = δ /4. Since the sequence {e− 2 |zmn | amn } is bounded, it
2
follows from (4.19) and Lemma 1.12 that the series in (4.18) converges absolutely
to an entire function f with f (zkl ) = akl for all (k, l).
It remains for us to show that the function f defined in (4.18) belongs to Fαp . Just
as in the previous section, we break the proof into three cases: 0 < p ≤ 1, 1 < p < ∞,
and p = ∞.
The case p = ∞ is the easiest. In fact, if the sequence e−α |zmn | /2 amn is bounded,
2
then by (4.19), there is a positive constant C such that

α
e− 2 |z| | f (z)| ≤ C ∑ e−σ |z−ωmn | .
2 2
m,n
This, along with Lemma 1.12, shows that f ∈ Fα∞ .

If 0 < p ≤ 1, it follows from (4.19) and Hölder’s inequality that

α 2 p α 2 p
f (z)e− 2 |z| ≤ C ∑ amn e− 2 |zmn | e−pσ |z−ωmn| .
2
m,n
Integrate term by term and use the translation invariance of the area measure. We
see that
p p
− α2 |z|2 − α2 |zmn |2
∑ amn e−pσ |z| dA(z).
2
e f (z) dA(z) ≤ C e
C m,n C
This shows that f ∈ Fαp whenever the series {amn e−α |zmn | /2 } is in l p .
2
The case 1 < p < ∞ follows from complex interpolation. In fact, examining the
arguments in the previous two paragraphs, we see that the linear operator
{cmn } → ∑ cmn e−σ |z−ωmn |

2
m,n
maps l ∞ to L∞ (C, dA) and l 1 to L1 (C, dA). Therefore, this operator maps l p to
L p (C, dA) for any 1 < p < ∞. This, along with (4.19), shows that f ∈ Fαp whenever
the sequence {amn e−α |zmn | /2 } belongs to l p .
2

The lemma above shows that the condition D+ (Z) < α /π is sufficient for a
separated sequence Z to be interpolating for Fαp . Next, we will prove that this density
condition is also necessary.
Lemma 4.40. Let 0 < p ≤ ∞. There is no sequence in C that is both sampling for
Fαp and interpolating for Fαp .
Proof. Assume the contrary and let Z be a sequence that is both sampling and
interpolating for Fαp . Then Z is separated and sampling for Fαp+ε for all sufficiently
small ε , because we have characterized sampling sequences for Fαp using the “open”
condition D− (Z) > α /π .
Fix a point ζ ∈ Z and use the assumption that Z is interpolating for Fαp to find
a function g ∈ Fαp such that g(ζ ) = 1 and g(z) = 0 for all z ∈ Z − {ζ }. Then, the
function f (z) = (z − ζ )g(z) is not identically zero, belongs to Fαp+ε , and vanishes
on Z. Thus, Z cannot possibly be sampling for Fαp+ε . This contradiction shows that
Z cannot be simultaneously sampling and interpolating for Fαp .

Lemma 4.41. Suppose 0 < p ≤ ∞ and Z is interpolating for Fαp . If Z is a set of
uniqueness for Fαp , then it must be a sampling sequence for Fαp .
Proof. Since Z is interpolating for Fαp , it must be separated by Lemma 4.8. Given
any function f ∈ Fαp , the sequence wn = f (zn ) has the property that {wn e−α |zn | /2 } ∈
2
l p . By the definition of N p (Z), there exists some function g ∈ Fαp such that g(zk ) = wk
for all k and g p,α ≤ N p (Z)g|Z p,α . Since Z is a set of uniqueness for Fαp and
f (zk ) = wk = g(zk ) for all k, we must have g = f , and so f p,α ≤ N p (Z) f |Z p,α
for all f ∈ Fαp . This says that Z is sampling for Fαp .

As consequences of the two lemmas above, we obtain the following corollaries:

Corollary 4.42. Let 0 < p ≤ ∞ and let Z be an interpolating sequence for Fαp . Then,
there exists a function f ∈ Fαp , not identically zero, such that f vanishes on Z.
Note that the above corollary does NOT say that every interpolating sequence for
Fαp is an Fαp -zero set because f may have additional zeros other than those in Z. In
fact, there exist examples of Fαp -interpolating sequences that are not Fαp -zero sets.
See Proposition 5.11.
Corollary 4.43. Let 0 < p ≤ ∞ and let Z be a sampling sequence for Fαp . For any
ζ ∈ Z, the sequence Z − {ζ } remains a sampling sequence for Fαp .
Proof. This is clear from the already-proved characterization of sampling sequences
for Fαp in terms of the lower density because deleting a single point from a sequence
does not alter the density of the sequence.
We give another proof that only relies on the fact that if Z is sampling for Fαp ,
then it is also sampling for Fαp+ε for sufficiently small ε .
So suppose Z is sampling for Fαp but Z = Z − {ζ } is not, where ζ ∈ Z. Without
loss of generality, we may also assume that Z is separated. Then, there exists a
sequence of unit vectors { fn } in Fαp such that fn |Z p,α → 0 as n → ∞. By a normal
family argument, we may as well assume that fn (z) → f (z) uniformly on compact
sets. By Fatou’s lemma, we have f ∈ Fαp . From fn |Z p,α → 0, we deduce that
f (z) = 0 for all z ∈ Z . Since
fn |Z p,α ≥ 1/M p (Z) > 0
for all n, we see that f (ζ ) = 0. The function (z − ζ ) f (z) is not identically zero,
vanishes on Z, and belongs to Fα +ε for any ε > 0. This contradicts the fact that Z is
p
p
a sampling sequence for Fα +ε .

Thus, sampling sequences for Fαp are stable under the following two operations:
deleting a finite number of points or adding any number of separated points from
outside the sequence.
Corollary 4.44. Let 0 < p ≤ ∞. If Z = {zn } is an interpolating sequence for Fαp ,
then so is Z ∪ {ζ } for any ζ ∈ Z.
Proof. By Corollary 4.42, there is a function g ∈ Fαp that is not identically zero but
vanishes on Z. By dividing out an appropriate power of z − ζ if necessary (which
preserves membership in Fαp ), we may assume that g(ζ ) = 0. Multiplying g by a
constant if necessary, we may further assume that g(ζ ) = 1.
Given a sequence {v} ∪ {vn } of values with {vn e−α |zn | /2 } ∈ l p , we can find a
2
p
function f ∈ Fα such that f (zn ) = vn for all n. The function
F(z) = f (z) + (v − f (ζ ))g(z)

belongs to Fαp and satisfies
F(ζ ) = v, F(zn ) = vn , n ≥ 1.
This shows that Z ∪ {ζ } is still an interpolating sequence for Fαp .

We see that interpolating sequences for Fαp
are stable under the following two
operations: deleting any number of points from the sequence or adding a finite
number of distinct points from outside the sequence.
A key tool for the rest of this section is the following quantity:
α
ρ p (z, Z) = sup | f (z)|e− 2 |z| ,
2
0 < p ≤ ∞,
f
where Z = {zn } and the supremum is taken over all unit vectors f in Fαp such that
f (zn ) = 0 for all n. We think of ρ p (z, Z) as some kind of distance from z to the
sequence Z. A normal family argument shows that the supremum in the definition
of ρ p (z, Z) is always attained.
By Corollary 2.8, we always have 0 ≤ ρ p (z, Z) ≤ 1. It is obvious that ρ p (z, Z) = 0
when z ∈ Z. We are going to show that ρ p (z, Z) = 0 only when z ∈ Z, provided that
Z is an interpolating sequence for Fαp .
Lemma 4.45. If Z is interpolating for Fα , where 0 0 when
p
z ∈ Z.
Proof. Actually, we only need to assume that Z is not a set of uniqueness (we
already know that every interpolating sequence for Fαp is not a set of uniqueness
for Fαp ). In fact, if f is any function in Fαp that is not identically zero and vanishes
on Z, then f cannot possibly have a zero at z of infinite order. Therefore, by dividing
out a finite and nonnegative power of w − z, which does not ruin membership in Fαp ,
we arrive at a function in Fαp that vanishes on Z but has a nonzero value at z.

The following result is a quantitative version of Corollary 4.44.
Lemma 4.46. Let Z = {z1 , z2 , · · · } and z0 ∈ Z. We have
1 + 2N p(Z)
N p (Z ∪ {z0 }) ≤
ρ p (z0 , Z)
for all 0 < p ≤ ∞.
Proof. We may assume that N p (Z) < ∞, that is, Z is an interpolating sequence for
Fαp . Given a sequence of values {v0 , v1 , v2 , · · · } with the l p norm of
α α α

v0 e− 2 |z0 | , v1 e− 2 |z1 | , v2 e− 2 |z2 | , · · ·
2 2 2
equal to 1, there is a function f ∈ Fαp such that f (zn ) = vn for all n ≥ 1 and
f p,α ≤ N p (Z) f |Z p,α ≤ N p (Z).

On the other hand, by Lemma 4.45, there exists a function f0 ∈ Fαp such that f0
vanishes on Z, f0 p,α ≤ 1, and
α
e− 2 |z0 | f0 (z0 ) = ρ p (z0 , Z).
2
Now the function
v0 − f (z0 ) α
f0 (z)e− 2 |z0 |
2
g(z) = f (z) +
ρ p (z0 , Z)
belongs to Fαp , solves the interpolation problem g(zn ) = vn for all n ≥ 0, and satisfies
|v0 − f (z0 )| − α |z0 |2

g p,α ≤ f p,α + e 2
ρ p (z0 , Z)
α α
|v0 |e− 2 |z0 | + | f (z0 )|e− 2 |z0 |
2 2
≤ N p (Z) +
ρ p (z0 , Z)
1 + f p,α
≤ N p (Z) +
ρ p (z0 , Z)
1 + N p(Z)
≤ N p (Z) +
ρ p (z0 , Z)
1 + 2N p(Z)
≤
ρ p (z0 , Z)
1 + 2N p(Z)
= g|(Z ∪ {z0 }) p,α .
ρ p (z0 , Z)

Lemma 4.47. Given positive constants δ0 , l0 , and α , there exists a positive
constant C = C(δ0 , l0 , α ) such that if N p (Z, α ) ≤ l0 and d(z, Z) ≥ δ0 , then ρ p (z, Z) ≥
C. Here, 0 < p ≤ ∞.
Proof. Let us assume the contrary, namely, there exists a sequence Zn of interpolat-
ing sets for Fαp and a sequence zn of points in C such that
N p (Zn , α ) ≤ l0 , d(zn , Zn ) ≥ δ0 , n ≥ 1,
and ρ p (zn , Zn ) → 0 as n → ∞.
By translation invariance, we may assume that each zn = 0. Going down to a
subsequence if necessary, we may also assume that Zn converges weakly to Z ,
where Z may be empty.
By Lemma 4.18, N p (Z , α ) ≤ l0 . Also, d(0, Zn ) ≥ δ0 shows that 0 is not in Z . By
Lemma 4.45, there exists a function f ∈ Fαp such that f vanishes on Z , f p,α ≤ 1,
and f (0) = r > 0. We may further assume that
α
lim f (z)e− 2 |z| = 0.
2
(4.20)
z→∞
In fact, the above condition is automatically satisfied for f ∈ Fαp when 0 < p < ∞.
If p = ∞, we modify the construction above as follows. Pick a complex number ζ
such that ζ ∈ Z and ζ = 0. Then Z ∪ {ζ } is still an interpolating sequence for Fαp .
Thus, there exists a function g ∈ Fαp such that g vanishes on Z ∪ {ζ } and g(0) = 0.
Then the function f (z) = g(z)/(z − ζ ) belongs to Fαp , vanishes on Z , satisfies the
condition in (4.20), and f (0) = 0.
Since {Zn } converges weakly to Z , the sequence εn = f |Zn p,α converges to 0
as n → ∞, which follows easily from (4.20). Now, choose gn ∈ Fαp with gn = f on
Zn and gn p,α ≤ l0 εn and define
f (z) − gn (z)
fn (z) = .
f p,α + l0 εn
For each n, it is clear that fn p,α ≤ 1 and fn = 0 on Zn . Since
|gn (0)| ≤ gn p,α ≤ l0 εn → 0
as n → ∞, we also have
r
ρ p (0, Zn ) ≥ | fn (0)| → > 0,
f p,α
which is a contradiction.

Lemma 4.48. Given positive constants l0 and α , there is a constant C = C(l0 , α ) >
0 such that if N p (Z, α ) ≤ l0 , then

log ρ p (z, Z) dA(z) ≥ −C|Q|2
Q
for every square Q with area |Q| ≥ 1.

Proof. By the proof of Lemma 4.8, there exists a point z0 ∈ Q and a positive constant
δ = δ (α , l0 ) such that d(z0 , Z) ≥ δ . By translation invariance, we may assume that
z0 = 0. It then follows from Lemma 4.47 that there is a function f with f p,α ≤ 1,
f |Z = 0, and | f (0)| ≥ σ , where σ = σ (α , l0 ) is another positive constant. Since
α
ρ p (z, Z) ≥ e− 2 |z| | f (z)|,
2
z ∈ C,
it follows from the subharmonicity of log| f (z)| that

2π
α 2 1
log | f (0)| ≤ r + log ρ p (reiθ , Z) dθ
2 2π 0

for all r ≥ 0. Multiply bothsides by r, integrate with respect to r from 0 to 2|Q|,
and observe that Q ⊂ B(0, 2|Q|) and ρ p ≤ 1. The desired result follows.

We can now prove the necessity of the condition D+ (Z) < α /π for Z to be an
interpolating sequence of Fαp .
Lemma 4.49. Suppose 0 2 and divide it
into N = [R] × [R] squares Q j , 1 ≤ j ≤ N, each of side length s = R/[R], where [R]
denotes the integer part of R. It is clear that 1 ≤ s ≤ 2.
Since Z is interpolating for Fαp , it is separated. Thus, for each j, we can find
some point z j ∈ Q j such that d(z j , Z) ≥ δ0 , where δ0 is a positive constant that only
depends on N p (Z) and α . Let Z j = Z ∪ {z j } and use Lemmas 4.46 and 4.47 to find a
positive constant l, independent of j, such that N p (Z j ) ≤ l for all j. By Lemma 4.48,
we can find a positive constant C = C(l, α ) such that

log ρ p (z, Z j ) dA(z) ≥ −C, 1 ≤ j ≤ N.
Qj
For any z ∈ Q j , we choose a function f such that f vanishes on Z j −z, f p,α ≤ 1,

and f (0)√= ρ p (0, Z j −z) = ρ p (z, Z j ). By Jensen’s formula applied to the disk |ζ | < r,
where 2 2 < r < R/2 (Jensen’s formula works for f (0) = 0, but the final estimate
below clearly holds for f (0) = 0 as well),
log ρ p (z, Z j ) = log | f (0)|

2π
dθ r r
≤ log | f (reiθ )| − ∑ log − log
0 2π ζ ∈Z,|z− ζ |<r
|z − ζ | |z − z j|
α r2 r r
≤ − ∑ log+ − log √
2 ζ ∈Z
|z − ζ | 2 2
α r2 r r
≤ − ∑ log+ − log √ ,
2 ζ ∈Z∩Q−
|z − ζ | 2 2
where Q− is the square of side length R− 2r inside Q sharing the same center with Q
and having sides parallel to the corresponding ones of Q. In other words, Q− consists
of those points whose distance to the complement of Q exceeds r. We integrate this
inequality with respect to area measure over Q j , use Lemma 4.48, and obtain

−C ≤ log ρ p (z, Z j ) dA(z)
Qj

α r2 r r
≤ |Q j | − ∑ log+ dA(z) − |Q j | log √ .
2 ζ ∈Z∩Q− Qj |z − ζ | 2 2
Summing over j, we obtain

α r2 2 r r
−CN 2 ≤ R − ∑ log+ dA(z) − R2 log √ .
2 ζ ∈Z∩Q− Q |z − ζ | 2 2
For any ζ ∈ Q− , the disk |z − ζ | < r is contained in Q so that

r r
log+ dA(z) = log dA(z)
Q |z − ζ | |z−ζ |<r |z − ζ |

r π r2
= log dA(z) = .
|z|<r |z| 2
Since N 2 ≤ R2 , it follows that

π r2 α r2 r
n(Z, Q− ) ≤ − log √ + C R2 ,
2 2 2 2
where n(Z, Q− ) denotes the number of points from Z contained in Q− . This can be
rewritten as

n(Z, Q− ) α 2 r 2C R2
≤ − log √ + . (4.21)
(R − 2r)2 π π r2 2 2 π r2 (R − 2r)2
Fix r and let R → ∞. Then by Proposition 4.1,
α 2 r 2C
D+ (Z) ≤ − log √ + 2 .
π π r2 2 2 πr
If r was chosen large enough so that

r
C − log √ < 0,
2 2
then D+ (Z) < α /π .

We summarize the main result of this section as follows:
Theorem 4.50. Suppose Z is a sequence in C and 0 < p ≤ ∞. Then Z is an
interpolating sequence for Fαp if and only if Z is separated and D+ (Z) < α /π .
Corollary 4.51. Suppose Z is a separated sequence in C and 0 < p ≤ ∞. Then Z is
interpolating for Fαp if and only if D+ (Z) < α /π .
4.7 Notes 187
4.7 Notes
The main results of this chapter are due to Seip and Wallsten, and our presentation
follows their papers [206] and [209] very closely. In turn, those two papers follow
Beurling’s 1977–1978 lectures on balayage and interpolation at the Mittag–Lefler
Institute very closely. In particular, the density notion introduced in Sect. 4.1 can be
found in Beurling’s lectures [36].
We chose to follow the more classical and original arguments of estimating
certain perturbations of the Weierstrass σ -function because this is more in line
with the traditional approaches to entire functions. But we point out that there are
now more modern and more powerful techniques for interpolation and sampling
problems that work in much more general settings. For example, many ideas used in
[203,208] to characterize interpolating and sampling sequences for Bergman spaces
can be adapted to work for Fock spaces as well.
See [205] for a complete description of interpolating and sampling sequences
for Bergman spaces on the unit disk. The books [78, 119, 203] contain more details
about the Bergman space results than Seip’s original papers. The interested reader
will find many additional papers in the bibliography about various interpolation and
sampling problems.
4.8 Exercises 189
4.8 Exercises
1. Suppose Z = {zn } is a sequence of interpolation for Fαp and {vn } is a sequence

of complex numbers such that {vn e−α |zn | /2 } ∈ l p . Show that the minimal
2
interpolation problem
% &
inf f p,α : f (zn ) = vn , n ≥ 1
has a unique solution.

2. If f ∈ Fαp for some 0 0, then for any complex number a, the
function g(z) = (z − a) f (z) belongs to Fβq for all 0 < q ≤ ∞ and β > α .
3. Prove Theorem 4.2.
4. Show that there exist two interpolating sequences for Fαp whose union is
sampling for Fαp .
5. If Z is not a set of uniqueness for Fαp , then ρ p (z, Z) = 0 if and only if z ∈ Z.
6. Show that for any ε > 0, there exists a positive constant C = C(ε , α , p) such
that

α 2 p α 2 p
f (z)e− 2 |z| ≤ C f (w)e− 2 |w| dA(w)
ε <|w−z|<2ε
for all z ∈ C.
7. Show that the incomplete gamma function has the property that
k
zj
Γ (k + 1, z) = k!e−z ∑
j=0 j!
for all k and z.

8. Show that
N N
1
∑∑ n 2 + m2
∼ log N
n=1 m=1
as N → ∞.
9. Show that
1 1
∑ (n2 + m2 )3/2
∼
N
n2 +m2 >N 2
as N → ∞.
10. Let δ be a positive number. Show that for any w ∈ C, we have

1
∑ |ωmn − w|
: δ < |ωmn − w| < R ∼ R
as R → ∞, where Λ = {ωmn } is any lattice.

11. Suppose Z is uniformly close to Λα . Show that
D+ (Z) = D− (Z) = α /π .
12. Justify the last step in the proof of Lemmas 4.32 and 4.34.
13. If Z is an interpolating sequences for Fαp , then any subset of Z is also an
interpolating sequence for Fαp .
α
14. Show that |σα (ωmn )|e− 2 |ωmn | is a positive constant independent of m and n,
2

where σα is the derivative of σα .
p
15. If Z is sampling for Fα , then adding any separated sequence to Z will create a
p
sampling sequence for Fα again.
16. If Z = {zn } is a sequence in C such that
% & 2
inf |z j − zk | : j = k > √ ,
α
then Z is an interpolating sequence for Fαp . See Tung [225]. √

17. If Z = {zn } is a sequence in C and there is a positive number ε < 1/ α such
that the disks B(zn , ε ) cover the whole complex plane, then Z is a sampling
sequence for Fαp .
18. Suppose Z = {zn } is separated and T is the operator from Fαp to l p defined by
α 2
T ( f ) = e− 2 |zn | f (zn ) . (4.22)
Show that:
(a) T is onto if and only if Z is interpolating for Fαp .
(b) T is bounded below if and only if Z is sampling for Fαp .
(c) T is one-to-one if and only if Z is a uniqueness set for Fαp .
Prove or disprove that T has closed range if and only if Z is either interpolating
or sampling for Fαp .
19. Suppose Z = {zn } is separated, 1 ≤ p < ∞, and 1/p + 1/q = 1. Then Z is an
interpolating sequence for Fαp if and only if there exists a positive constant c
such that
∞
q
∞
α
∑ ak e− 2 |z−zk | dA(z) ≥ c ∑ |ak |q
2
C k=1 k=1
for every sequence {ak } ∈ l q .

20. Suppose Z = {zn } is separated, 1 ≤ p < ∞, and 1/p + 1/q = 1. Then Z is a
sampling sequence for Fαp if and only if every function f ∈ Fαq has the form
∞
α
∑ ak eα zk z− 2 |zk |
2
f (z) =
k=1
for some {ak } ∈ l q .

4.8 Exercises 191
21. Let μ and ν be two positive measures. If A1 and A2 are two sets that are
measurable with respect to both μ and ν . Show that

ν (A1 ) ν (A2 ) ν (A1 ∪ A2 ) ν (A1 ) ν (A2 )
min , ≤ ≤ max , .
μ (A1 ) μ (A2 ) μ (A1 ∪ A2 ) μ (A1 ) μ (A2 )
22. Make precise the word “roughly” used in the proof of Proposition 4.3.
23. For a sequence Z = {zn } of distinct points in C, show that the following
(a) Z is sampling for Fα2 .
(b) Atomic decomposition holds on Z.
(c) The operator
∞
∑ f (zn )eα zzn −α |zn |
2
S f (z) = (4.23)
n=1
is bounded and invertible on Fα2 .

24. Show that the operator S defined in (4.23) is bounded on Fα2 if and only if Z is
the union of finitely many separated sequences.
25. Handle the case D− (Z) = 0 in the proof of Lemma 4.28.
26. Suppose Z = {zmn }, Λα = {ωmn }, and Λβ = {λmn }. If
|zmn − λmn | ≤ Q
for all (m, n), then there exists a positive constant Q = Q (α , β , Q) such that
for any (k, l), there exists some (k , l ) with the property that
|(zmn + ωkl ) − (λmn + λk l )| ≤ Q
for all (m, n).

27. Suppose Z = {zmn } is uniformly close to Λ = Λ (ω , ω1 , ω2 ) = {ωmn } with
|zmn − ωmn | ≤ Q for all (m, n). Show that for any ε > 0, there exists some
constant C = C(ε , Q, ω , ω1 , ω2 ) > 0 such that
∑ e−ε |zmn |
2
≤ C.
m,n
Hint: write |z|2 = |ω + (z − ω )|2 = |ω |2 |1 + (z − ω )/ω |2.

Chapter 5
Zero Sets for Fock Spaces
In this chapter, we study zero sets for the Fock spaces Fαp . Throughout this book, we
say that a sequence Z = {zn } ⊂ Ω is a zero set for a space X of analytic functions
in Ω if there exists a function f ∈ X, not identically zero, such that Z is exactly the
zero sequence of f , counting multiplicities.

DOI 10.1007/978-1-4419-8801-0 5,
194 5 Zero Sets for Fock Spaces
5.1 A Necessary Condition 195
5.1 A Necessary Condition
Recall from Theorem 2.12 that every function f ∈ Fαp is of order 2. Therefore, by
Hadamard’s factorization theorem, the zero sequence {zn } of f , with the origin
removed, must satisfy
∞
1
∑ |zn |3 < ∞.
n=1
In this section, we improve upon this estimate and obtain the following necessary
condition for a sequence {zn } to be a zero set for Fαp .
Theorem 5.1. Suppose 0 < p ≤ ∞ and {zn } is the zero sequence of a function f ∈
Fαp with f (0) = 0. Then√ there exist a positive constant c and a rearrangement of
{zn } such that |zn | ≥ c n for all n.
Proof. Without loss of generality, we may assume that f (0) = 1 and p = ∞. Let
{zn } denote the zero sequence of f , repeated according to multiplicity and arranged
so that 0 < |z1 | ≤ |z2 | ≤ |z3 | ≤ · · · .
Fix any positive radius r such that f has no zero on |z| = r and let n(r) denote
the number of zeros of f in |z| < r. By Jensen’s formula,
n(r) 2π
r 1
∑ log =
|zk | 2π 0
log | f (reiθ )| dθ .
k=1
Since f ∈ Fα∞ , we have

α 2
| f (reiθ )| ≤ f ∞,α e 2 r , 0 ≤ θ ≤ 2π , r > 0.
It follows that
n(r)
r α 2
∑ log |zk | ≤ 2
r + C,
k=1
where C = log f ∞,α . Rewrite the above inequality as
n(r)
r α
∏ |zk | ≤ exp 2
r2 + C
k=1
and observe that

n n(r)
r r
∏ |zk | ≤ ∏ |zk |
k=1 k=1
for any positive integer n (independent of r). Then

n
r α
∏ |zk | ≤ exp 2
r2 + C
k=1
for all positive integers n and all r > 0 such that f has no zero on |z| = r. Since
{|zk |} is nondecreasing, we have
rn α
≤ exp r 2
+ C ,
|zn |n 2
or

1 1 α 2 C
≤ exp r + , (5.1)
|zn | r 2n n
where n is any positive integer and r is any radius such that f has no zero on |z| = r.
There are only a countable number of radius r such that f has zeros on |z| = r.
Therefore,
√ for any positive integer n, we can choose a sequence {rk } such that rk →
n as k → ∞ and f has no zero on each |z| = rn . Combining this with (5.1), we
conclude that

1 1 α log f ∞,α
≤ √ exp + , n ≥ 1.
|zn | n 2 n
√
It is then clear that there is some positive constant c such that |zn | ≥ c n for all
n ≥ 1.

Note that the assumption f (0) = 0 is not a critical one. In fact, if f ∈ Fαp and it
has a zero of order m at the origin, then the function g defined by g(z) = f (z)/zm is
in Fαp and does not vanish at the origin.
Corollary 5.2. Suppose 0 2.

The function
sin(δ z2 )
f (z) =
δ z2
used in the proof of Theorem 5.4 shows that the estimate in Theorem 5.1 is best
possible. More specifically, we can find a positive constant C in this case such that
√ √
C−1 n ≤ |zn | ≤ C n
for all n ≥ 1.
5.2 A Sufficient Condition 197
5.2 A Sufficient Condition
The purpose of this section is to prove the following sufficient condition for zero
sequences of Fαp .
Theorem 5.3. Suppose that {zn } is a sequence of complex numbers such that
∞
1
∑ |zn |2 < ∞. (5.2)
n=1
Then {zn } is a zero set for Fαp , where 0 < p ≤ ∞.

Proof. Suppose that {zn } satisfies condition (5.2). We may also assume that the
sequence {zn } has been ordered in such a way that {|zn |} is nondecreasing. Consider
the Weierstrass product
∞
z
f (z) = ∏ E1 ,
n=1 zn
where E1 (z) = (1 − z)ez . By Theorem 1.6, f is entire, and {zn } is the zero sequence
of f . We will show that this function f belongs to all the Fock spaces Fαp , where
0 0.
If |z| < 1/2, we have
log |E1 (z)| = Re [log(1 − z) + z]

2
z z3 |z|4
= Re − − − − ···
2 3 4

2 1 |z| |z|2
≤ |z| + + + ···
2 3 4

1 1 1
≤ |z|2 1 + + 2 + · · ·
2 2 2
= |z|2 .
On the other hand, we have
|E1 (z)| ≤ (1 + |z|)e|z| , log |E1 (z)| ≤ |z| + log(1 + |z|), (5.3)
for all z. It follows that for any positive A, there exists a positive number R such that
log |E1 (z)| ≤ A|z|2 , |z| > R.

On the annulus 1/2 ≤ |z| ≤ R, the function |z|2 log |E1 (z)| is continuous except at
z = 1, where it tends to −∞. Hence, there is a constant B such that
1
log |E1 (z)| ≤ B|z|2 , ≤ |z| ≤ R.
2
Combining the estimates from the last three paragraphs, we conclude that
log |E1 (z)| ≤ M|z|2 , z ∈ C,
where M = max(1, A, B).

Given any positive ε , we can find a positive integer N such that
∞
1 ε
∑ < .
n=N+1 |zn |
2 2M
From this, we deduce that

∞ ∞
2
z ε
∑ log |E1 (z/zn )| ≤ M ∑ zn ≤ 2 |z|2
n=N+1 n=N+1
for all z ∈ C. Using (5.3) again, we can find some r1 > 0 such that
ε 2
log |E1 (z)| ≤ |z| , |z| > r1 ,
2S
where
N
1
S= ∑ |zn |2 .
n=1
Set r2 = r1 |zN |. Then |z| > r2 implies that |z/zn | > r1 for 1 ≤ n ≤ N. It follows that
N
ε
∑ log |E1 (z/zn )| ≤ 2 |z|2 , |z| > r2 .
n=1
Therefore,
∞
log | f (z)| = ∑ log |E1(z/zn )| < ε |z|2
n=1
for all |z| > r2 , or | f (z)| < eε |z| for all |z| > r2 . Since ε is arbitrary, we see that
2
f ∈ Fαp for all α > 0 and 0 < p ≤ ∞.

Note that the proof above can easily be adapted to show that the function P(z) f (z)
belongs to Fαp for any polynomial P(z). Therefore, if {zn } satisfies (5.2), then
{zn } ∪ F is also a zero set for Fαp , where F is any finite set. It is permitted to
have the origin contained in F.
5.3 Pathological Properties 199
5.3 Pathological Properties
In this section, we present examples to show certain pathological properties of zero

sequences of Fock spaces. More specifically, we will show that:
(i) The union of two zero sequences for Fαp is not necessarily a zero sequence for
Fαp again.
(ii) A subsequence of a zero sequence for Fαp is not necessarily a zero sequence for
Fαp again.
(iii) If α = β , then the spaces Fαp and Fβq have different zero sequences.
(iv) An interpolating sequence for Fαp is not necessarily a zero sequence for Fαp .
Theorem 5.4. Suppose α > 0 and 0 < p ≤ ∞. There exist two zero sequences for
Fαp whose union is no longer a zero sequence for Fαp .
Proof. Fix δ ∈ (πα /8, α /2) and consider the sequence

Z = ekπ i/2 nπ /δ : k = 0, 1, 2, 3; n = 1, 2, 3, · · · .
It is easy to see that Z is the zero sequence of the entire function
sin(δ z2 )
f (z) = .
δ z2
Converting the sine function above to complex exponential functions and using
the assumption that δ < α /2, we easily check that f ∈ Fαp . Therefore, Z is a zero
sequence for Fαp .
Let Z = {eπ i/4 z : z ∈ Z} be a rotation of the sequence Z above. Then Z is also
an Fαp zero sequence. Clearly, Z and Z are disjoint. We now arrange Z ∪ Z into a
single sequence {zn } such that
|z1 | ≤ |z2 | ≤ |z3 | ≤ · · · .
If {zn } is a zero sequence for Fαp ⊂ Fα∞ , it follows from the proof of Theorem 5.1
that there exists a positive constant C such that
n
r α 2
∏ |zk | ≤ Ce 2 r
k=1
for all n ≥ 1 and r > 0. Square both sides, replace n by 8n, and integrate from 0
to ∞ with respect to the measure re−β r , where β > α . We obtain another positive
2
constant C such that
(8n)! 8n 1
∏ |zk |2 ≤ C
β 8n k=1
for all n ≥ 1. It is easy to see that this reduces to

δ 8n (8n)!
≤ C, n ≥ 1.
πβ (n!)8
By Stirling’s formula, there exists yet another positive constant C, independent of n,
such that
√
8δ 8n n
≤C
πβ n4
for all n ≥ 1. This clearly implies that 8δ ≤ πβ . Since β can be arbitrarily close to
α , we have δ ≤ πα /8, which is a contradiction. This shows that {zn } is not an Fαp
zero set and completes the proof of the theorem.

Theorem 5.5. Let α > 0 and 0 < p ≤ ∞. There exists an Fαp zero sequence {zn }
and a subsequence {znk } which is not an Fαp zero sequence.
Proof. Fix a positive constant δ such that δ < α /2 and consider the following entire
function:
eiδ z − 1
2
f (z) = .
iδ z2
It is easy to check that f ∈ Fαp . Thus, its zero set

' ( ' (
2nπ 2nπ
± : n = 1, 2, 3, · · · ∪ ±i : n = 1, 2, 3, · · ·
δ δ
is an Fαp zero sequence. Let {zn } denote the subsequence consisting of real elements
in the above set. We proceed to show that {zn } is not an Fαp zero set.
Again, aiming to arrive at a contradiction later, we assume that g is a function in
Fαp that vanishes precisely on {zn }. It is clear that ρ1 (g) = m(g) = 2; see Sect. 1.1
for definitions and properties of these numbers. By Theorem 1.10, we always have
ρ (g) ≥ ρ1 (g), so g must be of order greater than or equal to 2. Combining this with
Theorem 2.12, we conclude that g must be of order 2. By Lindelöf’s theorem (see
Theorem 1.11), the function g must be of maximum (infinite) type since the sums
1
S(r) = ∑ z2
∼ log r, r > 1,
|zn |≤r n
are clearly unbounded. By Theorem 2.12 again, the function g cannot possibly be in
p p
Fα . This contradiction shows that {zn } is not an Fα zero set.

We now consider zero sets for different Fock spaces. The Weierstrass σ -functions
play a significant role here.
Recall that for any positive α ,

π
Λα = ωmn = (m + in) : m ∈ Z, n ∈ Z
α
is the square lattice in the complex plane with fundamental region

1 π 1 π
Ωα = z = x + iy : |x| < , |y| < .
2 α 2 α
The Weierstrass σ -function associated to Λα is the following infinite product:

z z 1 z2
σα (z) = z ∏ 1 − exp + ,
m,n ωmn ωmn 2 ωmn
2
where the product is taken over all integers m and n with ωmn = 0.
Lemma 5.6. Let 0 < α1 < α < α2 < ∞. We have:
(a) σα ∈ Fαp2 for all 0 < p ≤ ∞.
(b) σα ∈ Fαp1 for any 0 < p ≤ ∞.
(c) σα ∈ Fα∞ .
(d) σα ∈ fα∞ , and so σα ∈ Fαp for any 0 < p < ∞.
Proof. It follows from the quasiperiodicity of σα that if z = ωmn + w and w ∈ Ωα ,

then
α α
|σα (z)|e− 2 |z| = |σα (w)|e− 2 |w| .
2 2
(5.4)
Since the function |σα (w)|e−α |w| /2 is bounded on the relatively compact set Ωα ,
2
there exists a positive constant C such that

α
|σα (z)| ≤ Ce 2 |z| ,
2
z ∈ C.
This clearly implies that σα ∈ Fα∞ and σα ∈ Fαp2 for all 0 < p ≤ ∞.
If S is any compact set contained in the fundamental region of Λα , then there
exists a positive constant δ such that
α
|σα (w)|e− 2 |w| ≥ δ ,
2
w ∈ S.
This together with (5.4) shows that

α
|σα (z)|e− 2 |z| ≥ δ ,
2
z ∈ S + ωmn ,
for all (m, n). This clearly shows that σα ∈ fα∞ . Since Fαp ⊂ fα∞ for 0 < p < ∞, we
have σα ∈ Fαp for any 0 < p < ∞. Also, Fαp1 ⊂ fα∞ for all 0 < p ≤ ∞. So σα ∈ Fαp1
for all 0 < p ≤ ∞.

Lemma 5.7. Suppose 0 < p < ∞ and f ∈ Fαp . If f (z) = 0 for all z ∈ Λα , then f is
identically zero.
Proof. By the Weierstrass factorization theorem, we can write f = hσα , where h is
an entire function. In view of the quasiperiodicity of σα , we have

α 2 p α 2 p
f (z)e− 2 |z| dA(z) = ∑ |h(z + ωmn )| p σα (z)e− 2 |z| dA(z),
C m,n Ωα
where Ωα is the fundamental region of σα . Let D be any small disk centered at 0

and contained in 12 Ωα . Then by Corollary 1.21, there exists a positive constant C
such that

α 2 p
f (z)e− 2 |z| dA(z) ≥ C ∑ |h(z + ωmn )| p dA(z).
C m,n Ωα −D
Since the function z → |h(z + ωmn )| p is subharmonic, there exists a positive constant
δ (independent of (m, n)) such that

|h(z + ωmn )| p dA(z) ≥ δ |h(z + ωmn )| p dA(z)
Ωα −D Ωα
for all (m, n). It follows that there is another positive constant C such that

α 2 p
f (z)e− 2 |z| dA(z) ≥ C |h(z)| p dA(z).
C C
This is impossible unless h is identically zero.

p ≤ ∞, 0 < q ≤ ∞, and α1 =
α2 . Then Fαp1
q
Theorem 5.8. Suppose 0 < and Fα2 have
different zero sets.
Proof. Without loss of generality, let us assume that α1 < α < α2 . By Lemma 5.6,
the Weierstrass function σα belongs to Fαq2 , so its zero sequence Λα is a zero set for
Fαq2 . On the other hand, if f ∈ Fαp1 ⊂ Fα2 and f vanishes on Λα , then it follows from
Lemma 5.7 that f is identically zero. Therefore, Λα cannot possibly be a zero set
for Fαp1 .

The remaining question for us now is this: do Fαp and Fαq have different zero sets
whenever p = q? As of this writing, there is no complete answer, but it is easy to
produce examples of such pairs that do not have the same zero sets. The simplest
example is Z = Λα , which is a zero set for Fα∞ , but not a zero set for any Fαp when
0 2 because
the function f (z) = σα (z)/z belongs to Fαp if and only if p > 2. However, this
sequence Z is not a zero set for Fα2 . To see this, suppose f is a function in Fα2 ,
not identically zero, such that f vanishes on Z. By Weierstrass factorization, we
have f (z) = [σα (z)/z]g(z) for some entire function g that is not identically zero.
Mimicking the proof of Lemma 5.7, we can show that

g(z) 2

|z|>1
z dA(z) < ∞.
It follows from polar coordinates and the Taylor expansion of g that this is
impossible unless g is identically zero. This actually shows that Z = Λα − {0} is
a uniqueness set for Fα2 . In the above arguments, the point 0 can be replaced by any
other point in Λα .
On the other hand, if Z is the resulting sequence when two points a and b are
removed from Λα , then the function
σα (z)
f (z) =
(z − a)(z − b)
belongs to Fα2 and has Z as its zero sequence. Therefore, Z is a zero set for Fα2 .
Consequently, it is possible to go from a uniqueness set to a zero set by removing
just one point. Equivalently, it is possible to add just a single point to a zero set of
Fα2 so that the resulting sequence becomes a uniqueness set for Fα2 . This shows how
delicate the problem of characterizing zero sets for Fαp is.
We can also show by an example that it is generally very difficult to distinguish
q
between zero sets for Fαp and Fα . More specifically, for any positive integer N with
q
N p > 2, if Z is an Fα zero set and if N points {z1 , · · · , zN } are removed from Z, then
the remaining sequence Z is an Fαp zero set. To see this, let Z be the zero sequence of
a function f ∈ Fα , not identically zero, then Z is the zero sequence of the function
q
f (z)
g(z) = ,
(z − z1 ) · · · (z − zN )
which is easily seen to be in Fαp . Therefore, zero sets for Fαp and Fαq may be different,
but they are not too much different.
Let Z be a zero sequence for Fαp and let IZ denote the set of functions f in Fαp
such that f vanishes on Z. In the classical theories of Hardy and Bergman spaces,
the space IZ is always infinite dimensional. This is no longer true for Fock spaces.
Theorem 5.9. For any 0 < p ≤ ∞ and k ∈ {1, 2, · · · } ∪ {∞}, there exists a zero set
Z for Fαp such that dim(IZ ) = k.
Proof. The case k = ∞ is trivial; any finite sequence Z will work. So we assume that
k is a positive integer in the rest of the proof.
We first consider the case p = ∞ and k > 1. In this case, we consider Z = Λα −

{a1 , · · · , ak−1 }, where a1 , · · · , ak−1 are (any) distinct points in Λα and
σα (z)
f (z) = .
(z − a1 ) · · · (z − ak−1 )
It follows from Corollary 1.21 that f ∈ Fα∞ and Z is exactly the zero sequence of f .
Furthermore, if h is a polynomial of degree less than or equal to k − 1, then the
function f (z)h(z) is still in Fα∞ .
On the other hand, if F is any function in Fα∞ that vanishes on Z, then we can
write
σα (z)g(z)
F(z) = f (z)g(z) = ,
(z − a1 ) · · · (z − ak−1 )
where g is an entire function. For any positive integer n, let Cn be the boundary
of
the square centered at 0 with horizontal and vertical side length (2n + 1) π /α . It
is clear that

d(Cn , Λα ) ≥ π /α /2, n ≥ 1.
So there exists a positive constant C such that

α
|σα (z)|e− 2 |z| ≥ C,
2
z ∈ Cn , n ≥ 1.
This together with the assumption that F ∈ Fα∞ implies that there exists another
positive constant C such that
|g(z)| ≤ C|z − a1| · · · |z − ak−1| (5.5)
for all z ∈ Cn and n ≥ 1. By Cauchy’s integral estimates, the function g must be a

polynomial of degree at most k − 1.
Therefore, when p = ∞, k > 1, and Z = Λα − {a1 , · · · , ak−1 }, we have shown that
a function F ∈ Fα∞ vanishes on Z if and only if
σα (z)h(z)
F(z) = ,
(z − a1 ) · · · (z − ak−1 )
where h is a polynomial of degree less than or equal to k − 1. This shows that

dim(IZ ) = k.
When p = ∞ and k = 1, we simply take Z = Λα . The arguments above can be
simplified to show that a function F ∈ Fα∞ vanishes on Z if and only if F = cσα for
some constant c.
Next, we assume that 0 2, or equivalently,

σ (z)e− α2 |z|2 p
α
dA(z) < ∞. (5.6)
|z|>1 zN
Remove any N + k − 1 points {a1 , · · · , aN+k−1 } from Λα and denote the remaining
sequence by Z. Then Z is the zero sequence of the function
σα (z)
,
(z − a1 ) · · · (z − aN+k−1 )
which belongs to Fαp in view of (5.6). In fact, if g is any polynomial of degree less
than or equal to k − 1, then it follows from (5.6) that g times the above function
belongs to IZ .
Conversely, if f is any function in Fαp that vanishes on Z, then we can write
σα (z)g(z)
f (z) = ,
(z − a1 ) · · · (z − aN+k−1 )
where g is an entire function. Since Fαp ⊂ Fα∞ , it follows from (5.5) and Cauchy’s
integral estimates that g is a polynomial with degree less than or equal to N + k − 1.
If the degree of g is j > k − 1, then
g(z) 1
∼ , z → ∞.
(z − a1 ) · · · (z − aN+k−1 ) zN+k−1− j
This together with f ∈ Fαp shows that (5.6) still holds when N is replaced by N + k −
1 − j, which contradicts our minimality assumption on N. Thus, j ≤ k − 1, which
shows that IZ is k dimensional.

The following result describes the structure of IZ when it is finite dimensional.
Theorem 5.10. Suppose Z is a zero set for Fαp and dim(IZ ) = k is a positive integer.
Then there exists a function g ∈ IZ such that IZ = gPk−1 , where Pk−1 is the set of all
polynomials of degree less than or equal to k − 1.
Proof. First, observe that if dim(IZ ) = k < ∞, then Z = Z ∪ {a1 , · · · , ak } is a
uniqueness set for Fαp for all {a1 , · · · , ak }. Here, the union in Z should be understood
in the sense of zero sequences, where multiplicities are taken into account. In fact,
if there exists a function f ∈ Fαp , not identically zero, such that f vanishes on Z ,
then the functions
f (z) f (z)
f (z), , ··· ,
z − a1 z − ak
all belong to Fαp and vanish on Z. Here again, if zeros of higher multiplicity
are involved, then some obvious adjustments should be made. It is clear that the
functions listed above are linearly independent, so the dimension of IZ is at least
k + 1, a contradiction.
Next, observe that if dim(IZ ) > m, then Z = Z ∪ {a1 , · · · , am } is not a uniqueness
set for Fαp for any collection {a1 , · · · , am }. To see this, pick any m + 1 linearly
independent functions f1 , · · · , fm+1 from IZ , let
f = c1 f1 + · · · + cm+1 fm+1 ,
and consider the system of linear equations
c1 f1 (a j ) + · · · + cm+1 fm+1 (a j ) = 0, 1 ≤ j ≤ m.
Once again, obvious adjustments should be made when there are zeros of higher
multiplicity. The homogeneous system above has m equations but m + 1 unknowns,
so it always has nonzero solutions c j , 1 ≤ j ≤ m + 1. With such a choice of c j , the
function f is not identically zero but vanishes on Z , so Z is not a uniqueness set.
It follows that if 1 ≤ j < k and Z = Z ∪ {a1 , · · · , a j }, then Z is not a uniqueness
set for Fα . We can actually show that Z is a zero set for Fα . In fact, if f is a function
p p
p
in Fα , not identically zero, such that f vanishes on Z (but not necessarily exactly on
Z ), then the conclusion of the previous paragraph shows that the number of zeros
of f in addition to those in Z cannot exceed k − j. If these additional zeros a are
divided out of f by the appropriate powers of z − a, the resulting function is still in
Fα and vanishes exactly on Z . Thus, Z is a zero set for Fα .
p p
Fix a function g ∈ IZ that has exactly Z as its zero set. If f is any function in IZ ,
not identically zero, then just as in the previous paragraph, we can show that the
zeros of f must be of the form Z = Z ∪ {a1 , · · · , a j }, where j ≤ k − 1. Thus, we
can factor f as follows: f = gPeh , where P ∈ Pk−1 and h is entire. It is clear that
dividing a polynomial out of f , whenever the division is possible, always results in
p
a function in Fα . Therefore, the function geh belongs to IK as well. It follows that
the function ge − g = g(eh − 1) belongs to IZ . If h is not constant, then by Picard’s
h
theorem, eh − 1 has infinitely many zeros, so geh − g is a function in IZ that has

infinitely many zeros in addition to those in Z, a contradiction. This shows that h is
constant and IZ ⊂ gPk−1 . A count of dimension then gives IZ = gPk−1 .

In the classical theories of Hardy and Bergman spaces, every interpolating
sequence is necessarily a zero sequence. We now show that this is not true for Fock
spaces.
Proposition 5.11. There exists an interpolating sequence for Fαp that is not a zero
set for Fαp .
√
Proof. Fix some δ > 2/ α . For any positive integer k, let Zk denote the set of
k + 1 points evenly spaced in the first quadrant on the circle |z| = kδ , including the
end-points kδ and kδ i. Let
∞

Z= Zk = {z1 , z2 , · · · , zn , · · · }.
k=1
Since the distance between any two neighboring points in Zk is

π
2kδ sin > δ,
4k
√
the sequence Z is separated with a separation constant greater than 2/ α . This
p
implies that Z is an interpolating sequence for Fα ; see Exercise 16 in Chap. 4.
If Z is the zero sequence of some function f ∈ Fαp , then by Theorem 5.1, the order
ρ of f is less than or equal to 2. On the other hand, for the sequence Z, we have
m = ρ1 = 2; see Sect. 1.1 for the definition of these constants. By Theorem 1.10, we
have ρ ≥ m = 2. Thus, ρ = 2, and Lindelöf’s theorem (Theorem 1.11) applies.
For r ∈ (mδ , (m + 1)δ ), we have
m k
1 1
S(r) = ∑ = ∑ ∑ e−iπ j/k
k=1 (kδ )
2 2
|zk |<r zk j=0
m
1 1 + e−iπ /k m
1 cos(π /2k)
= ∑ (kδ )2 1 − e−iπ /k ∑ (kδ )2 sin(π /2k)
=
k=1 k=1
2i m 1 2i 2i
∼− ∑ k ∼ − πδ 2 log m ∼ − πδ 2 log r
πδ 2 k=1
as r → ∞. This shows that S(r) is not bounded in r. By Lindelöf’s theorem, f has

infinite type. This contradicts with Theorem 2.12, which asserts that f must have
type less than or equal to α /2 when f is of order 2. Therefore, Z cannot be a zero
sequence for Fαp .

5.4 Notes 209
5.4 Notes
Theorem 5.1, the necessary condition for zero sets of Fock spaces, was obtained in
[249]. Theorem 5.3, the sufficient condition for zero sets of Fock spaces, is classical
and follows from the general theory of entire functions. The proof of Theorem 5.3
here is basically from [67].
The results in Sect. 5.3 were mostly from [249, 258]. The motivation for [249]
was Horowitz’s study of zero sets for Bergman spaces; see [127–129]. The
most intriguing results concerning zero sequences for Fock spaces are probably
Theorems 5.9 and 5.10, which were proved in [258]. One interesting problem that
remains open is the following: if p = q, do Fαp and Fαq always have different zero
sequences?
Lemma 5.7 shows that Λα is a set of uniqueness for Fαp when 0 < p < ∞. This
result as well as its proof are from [209]. Proposition 5.11, which is a little surprising
when compared to the corresponding questions in the Hardy and Bergman space
settings, is from [225].
5.5 Exercises 211
5.5 Exercises
1. We say that an entire function f (z) belongs to the Nevanlinna–Fock class Fα∗ if

log+ | f (z)| dλα (z) < ∞.
C
Show that the zero sequence {zn } of any function f in Fα∗ with f (0) = 0 satisfies
the following condition:
∞
e−α |zn |
2
∑ |zn |2 < ∞.
n=1
2. Let a be a nonzero complex number. Solve the extremal problem
sup{Re f (0) : f 2,α ≤ 1, f (a) = 0}.
3. Suppose Z is a zero set for Fαp and k is a positive integer. Show that the following
(a) dim(IZ ) ≤ k.
(b) Z ∪ {a1 , · · · , ak } is a uniqueness set for Fαp for all {a1 , · · · , ak }.
(c) Z ∪ {a1 , · · · , ak } is a uniqueness set for Fαp for some {a1 , · · · , ak }.
4. Suppose Z is a zero set for Fαp and k is a positive integer. Show that the following
(a) dim(IZ ) = k.
(b) For any {a1 , · · · , ak }, the sequence Z ∪ {a1, · · · , ak−1 } is not a uniqueness
set for Fαp but Z ∪ {a1, · · · , ak } is.
(c) For some {a1 , · · · , ak−1 }, the sequence Z ∪ {a1 , · · · , ak−1 } is not a unique-
ness set for Fαp , but for some {b1 , · · · , bk }, the sequence Z ∪ {b1 , · · · , bk } is
a uniqueness set for Fαp .
5. If Z is a zero set for Fαp , then the sequence remains a zero set for Fαp after any
finite number of points are removed from it.
6. Suppose 0 0, provided that |zn | = 0, 1. Show that this is false in general if ε = 0.
8. Suppose {zn } is the zero sequence of a function f ∈ Fαp , where f (0) = 1, 0 <
p < ∞, and {|zn |} is nondecreasing. Show that
C αe 2
n np
1
∏ |zn | p ≤ √n n
f pp,α
k=1
for all n ≥ 1, where C is a positive constant independent of n and f .

q
9. Suppose 0 2. Show that if any N points are removed from Z, the remaining sequence
becomes a zero set for Fαp .
10. Let Z be a zero set for Fα2 with 0 ∈ Z. Show that there is no function GZ ∈ Fα2
such that GZ (0) > 0, GZ 2,α = 1, Z(GZ ) = Z, and f /GZ 2,α ≤ f 2,α for
all f ∈ Fα2 with f |Z = 0. See [119] for information about the corresponding
problem in the Bergman space setting.
11. Suppose f ∈ Fαp has order 2 and type α /2. Then f must have infinitely many
zeros. See [22].
12. Show that the function
∞
1
f (z) = ∑ n√n! zn
n=1
belongs to F12 but the function z f (z) is no longer in F12 . See [22].
13. If Z is a zero set for Fαp and dim(IZ ) < ∞, then every function in IZ has order 2
and type α /2.
14. If Z is a zero set for Fαp and dim(IZ ) < ∞, then any two functions in IZ whose
zeros are exactly those in Z can only differ by a constant multiple. Thus, there
is essentially just one function that vanishes exactly on Z.
Chapter 6
Toeplitz Operators
There is a rich history of Toeplitz operators, especially those on the Hardy space. In
particular, Toeplitz operators on the Hardy space provide ample examples of shifts,
isometries, and Fredholm operators. They also provide motivating examples in index
theory and the theory of invariant subspaces.
In this chapter, we study Toeplitz operators on the Fock space Fα2 . Problems
considered include boundedness, compactness, and membership in the Schatten
classes. The approach here is more closely related to the theory of Toeplitz operators
on the Bergman space that was developed over the past thirty years or so.

DOI 10.1007/978-1-4419-8801-0 6,
214 6 Toeplitz Operators
6.1 Trace Formulas 215
6.1 Trace Formulas
Recall that for any fixed weight parameter α , the orthogonal projection
P : L2α → Fα2
is an integral operator,

P f (z) = K(z, w) f (w) dλα (w),
C
where
K(z, w) = eα zw
is the reproducing kernel of the Hilbert space Fα2 .

Given ϕ ∈ L∞ (C), we define a linear operator Tϕ : Fα2 → Fα2 by
Tϕ ( f ) = P(ϕ f ), f ∈ Fα2 .
We call Tϕ the Toeplitz operator on Fα2 with symbol ϕ . It is clear that Tϕ is bounded
with Tϕ ≤ ϕ ∞ .
Proposition 6.1. For any complex numbers a and b, and for any bounded functions
ϕ and ψ , we have:
(i) Taϕ +bψ = aTϕ + bTψ .
(ii) Tϕ = Tϕ∗ .
(iii) Tϕ ≥ 0 if ϕ ≥ 0.
Proof. These follow easily from the definitions. We omit the routine details.

One of the main differences between Toeplitz operators on the Fock space and
those on Hardy and Bergman type spaces is the lack of bounded analytic and
harmonic symbols in the Fock space setting. In fact, by the maximum modulus
principle, if an analytic or harmonic function on C is bounded, it has to be a constant.
By the integral representation for the orthogonal projection P, we can write

Tϕ ( f )(z) = K(z, w) f (w)ϕ (w) dλα (w)
C

α
K(z, w) f (w)e−α |w| ϕ (w) dA(w).
2
=
π C
This motivates us to define Toeplitz operators on Fα2 with much more general
symbols.
If μ is a Borel measure on C, we define the Toeplitz operator Tμ as follows:

α
K(z, w) f (w)e−α |w| dμ (w),
2
Tμ ( f )(z) = z ∈ C.
π C
Note that Tϕ is very loosely defined here, because it is not clear when the integrals
above will converge, even if the measure μ is finite, as the kernel function K(z, w)
is unbounded for any fixed z = 0.
To make things a little more precise, we say that a complex Borel measure μ on
C satisfies condition (M) if

|K(z, w)|e−α |w| d|μ |(w) < ∞
2
(6.1)
C
for all z ∈ C. Because of the exponential form of the reproducing kernel, it is clear
that the above is equivalent to

|K(z, w)|2 e−α |w| d|μ |(w) < ∞
2
(6.2)
C
for all z ∈ C. When d μ (z) = ϕ (z) dA(z), the measure μ satisfies condition (M) if
and only if the function ϕ satisfies condition (I1 ). See Sect. 3.2 for the definition of
condition (I p ).
If μ satisfies condition (M), then the Toeplitz operator Tμ above is well defined
on a dense subset of Fα2 . In fact, if
N
f (w) = ∑ ck K(w, ak )
k=1
is any finite linear combination of kernel functions in Fα2 , then it follows from
condition (M) and the Cauchy–Schwarz inequality that Tμ ( f ) is well defined. Recall
from Lemma 2.11 that the set of all finite linear combinations of kernel functions is
dense in Fα2 .
If μ satisfies condition (M), the Berezin transform of μ (see Sect. 3.4) is well
defined:

α α
|kz (w)|2 e−α |w| dμ (w) = e−α |z−w| dμ (w),
2 2
μ
(z) =
π C π C
where
α
K(z, z) = eα wz− 2 |z|
2
kz (w) = K(w, z)/
are the normalized reproducing kernels of Fα2 .

If μ is positive or if the Toeplitz operator Tμ happens to be a bounded operator
on Fα2 , then it is easy to see that
μ
(z) = Tμ kz , kz α , z ∈ C.
When dμ (z) = ϕ (z) dA(z), we get back to Tϕ and ϕ.

In the rest of this section, we focus on the case of trace-class Toeplitz operators.
We will obtain several trace formulas related to Toeplitz operators. These trace
formulas will then be used in the next two sections to study bounded and compact
Toeplitz operators.
The definition of the Berezin transform ϕ and the Toeplitz operator Tϕ requires
that the function ϕ satisfy condition (I1 ). But in the study of Toeplitz operators, we
often need to require that ϕ satisfy condition (I2 ), which is slightly stronger than
condition (I1 ).
If the Toeplitz operator Tϕ is bounded on Fα2 , we have

Tϕ f (z) = f (w)ϕ (w)K(z, w) dλα (w),
C
and it is easy to check that

KTϕ (w, z) = ϕ (u)K(u, z)K(w, u) dλα (u). (6.3)
C
See Sect. 3.1 for the definition of KS (w, z) for any bounded linear operator S on Fα2 .
If we further assume that ϕ satisfies condition (I2 ), then it is also easy to check that
the uniqueness of KTϕ implies
ϕ K( · , z) − KTϕ ( · , z) ⊥ Fα2 (6.4)

for all z ∈ C.
Theorem 6.2. Suppose ϕ is Lebesgue measurable on C and S is a bounded linear
operator on Fα2 . If
(1) ϕ satisfies condition (I2 ),
(2) Tϕ is bounded on Fα2 ,
(3) )Tϕ S) is trace class,
(4) C C |ϕ (z)||K(w, z)||KS (w, z)| dλα (w) dλα (z) < ∞,
then we have

α dA(z).
tr (Tϕ S) = ϕ (z)KS (z, z) dλα (z) = ϕ (z)S(z) (6.5)
C π C
Proof. By hypothesis (1), each function ϕ K( · , z) is in L2α , and by (6.4), we can

write
KTϕ ( · , z) = ϕ K( · , z) − H( · , z),
where H( · , z) ⊥ Fα2 . By Corollary 3.12,

tr (Tϕ S) = dλα (w) KTϕ (z, w)KS (w, z) dλα (z)
C C

= dλα (w) [ϕ (z)K(z, w) − H(z, w)]KS∗ (z, w) dλα (z)
C C

= dλα (w) ϕ (z)K(z, w)KS∗ (z, w) dλα (z)
C C

= dλα (w) ϕ (z)K(w, z)KS (w, z) dλα (z).
C C
Hypothesis (4) allows the application of Fubini’s theorem, which, along with the
reproducing property in Fα2 and parts (3) and (8) of Proposition 3.9, leads to the
desired trace formulas.

Taking S to be the identity operator, we obtain the following trace formula for
Toeplitz operators on the Fock space.
Corollary 6.3 Suppose ϕ satisfies condition (I2 ). If Tϕ is in the trace class and
ϕ ∈ L1 (C, dA), then

α
tr (Tϕ ) = ϕ (z)K(z, z) dλα (z) = ϕ (z) dA(z). (6.6)
C π C
Proof. When S is the identity operator, we have KS (z, w) = K(z, w), so the condition

|ϕ (z)||K(w, z)||KS (w, z)| dλα (w) dλα (z) < ∞
C C
becomes

α
|ϕ (z)|K(z, z) dλα (z) = |ϕ (z)| dA(z) < ∞.
C π C

Note that there exist symbol functions ϕ such that Tϕ is in the trace class but
ϕ ∈ L1 (C, dA). See Exercise 10.
Corollary 6.4 Suppose ϕ is bounded and compactly supported in C. Then for any
bounded linear operator S on Fα2 , the operator Tϕ S is trace class and

α dA(z).
tr (Tϕ S) = ϕ (z)KS (z, z) dλα (z) = ϕ (z)S(z) (6.7)
C π C
Proof. It is easy to see that hypotheses (1)–(3) of Theorem 6.2 are satisfied. To
check hypothesis (4) of Theorem 6.2, we write

I= |ϕ (z)| dλα (z) |KS (w, z)||K(w, z)| dλα (w).
C C
From the definition KS (w, z) = S∗ K( · , z)(w), we deduce that

|KS (w, z)|2 dλα (w) = S∗K( · , z)22,α .
C
It follows from this and the Cauchy–Schwarz inequality that

|KS (w, z)||K(w, z)| dλα (w) ≤ S∗ Kz 2,α Kz 2,α ≤ SK(z, z).
C
Thus,

α S
I ≤ S |ϕ (z)|K(z, z) dλα (z) = |ϕ | dA < ∞,
C π C
as ϕ is bounded and compactly supported.

As a consequence of Corollary 6.4, we show that every trace-class operator on
Fα2 can be approximated by trace-class Toeplitz operators in the trace norm, and
every compact operator on Fα2 can be approximated by compact Toeplitz operators
in norm.
Theorem 6.5. Let C denote the set of all Toeplitz operators Tϕ , where ϕ is
continuous and has compact support in C. Then:
(1) C is trace-norm dense in the trace class T of Fα2 .
(2) C is norm dense in the space K of all compact operators on Fα2 .
Proof. Let L denote the space of all bounded linear operators on Fα2 . Then, it is well
known that T ∗ = L and K∗ = T, with the duality pairing given by S, T = tr (ST ).
To prove (2), assume that C is not norm dense in K. By the Hahn–Banach
theorem, there must be a nonzero operator S in T such that
Tϕ , S = 0, Tϕ ∈ C.
By Corollary 6.4,

α dA(z)
0 = Tϕ , S = tr (Tϕ S) = ϕ (z)S(z)
π C
for all continuous functions ϕ with compact support in C. This implies that S = 0.
So S = 0, a contradiction which proves (2).
The proof for (1) is similar, and we omit the details here.

6.2 The Bargmann Transform 221
6.2 The Bargmann Transform
The connection between Toeplitz operators on the Fock space and pseudodifferential
operators on L2 (R, dx) is established by the Bargmann transform, and the most
elementary way to understand the Bargmann transform is via the classical Hermite
polynomials.
Recall that for any nonnegative integer n, the nth Hermite polynomial Hn (x) is
defined by
2 d
n
Hn (x) = (−1)n ex n e−x .
2
dx
The first five Hermite polynomials are given by
H0 (x) = 1,
H1 (x) = 2x,
H2 (x) = 4x2 − 2,
H3 (x) = 8x3 − 12x,
H4 (x) = 16x4 − 48x2 + 12.
In general, it is easy to check that each Hn has degree n and

Hn (x) = 2xHn−1 (x) − Hn−1 (x), n ≥ 1,
which can be used to compute Hn inductively. In particular, the leading term of

Hn (x) is (2x)n .
Lemma 6.6. For nonnegative integers m and n, let

Hn (x)Hm (x)e−x dx.
2
Imn =
R
√
Then Imn = 0 for m = n and Inn = 2n n! π .
Proof. For any polynomial f , we use integration by parts n times to get

dn −x2
Hn (x) f (x)e−x dx = (−1)n
2
f (x) e dx
R R dxn

f (n) (x)e−x dx.
2
=
R
If m < n and f = Hm , then f (n) ≡ 0 and so Imn = 0.

If f = Hn , then f (x) = (2x)n + · · ·, and so f (n) ≡ 2n n!. It follows that

√
e−x dx = 2n n! π .
2
Inn = 2n n!
R

Theorem 6.7. For any nonnegative integer n, let

1
2α 4 1 √
e−α x Hn ( 2α x).
2
hn (x) = √
π n
2 n!
Then {hn } is an orthonormal basis of L2 (R, dx).

Proof. It follows from a change of variables and Lemma 6.6 that {hn } is an
orthonormal set. In particular, for any positive integer N, the functions
h0 (x), h1 (x), · · · , hN (x),
are linearly independent. It follows that the polynomials

√ √ √
H0 ( 2α x), H1 ( 2α x), · · · , HN ( 2α x) (6.8)
are linearly independent in the vector space of all polynomials of degree less than
or equal to N. A dimensionality argument then shows every polynomial of degree
less than or equal to N can be written as a linear combination of the polynomials in
(6.8). Therefore, the condition

f (x)hn (x) dx = 0, n ≥ 0,
R
implies that

f (x)P(x)e−α x dx = 0
2
for all polynomials P, which, according to Lemma 3.16, implies that f = 0 almost
everywhere. Thus, the set {hn } is complete in L2 (R, dx).

We now define the Bargmann transform. Let f be a function on R satisfying the
condition that f (x)e|tx|−π x is integrable with respect to dx for any real t. Then for
2
any positive parameter α , we can define an analytic function Bα f by

1
2α 4 2 − α z2
Bα f (z) = f (x)e2α xz−α x 2 dx, z ∈ C. (6.9)
π R
This will be called the (parametrized) Bargmann transform of f .

Theorem 6.8. For any positive α , the Bargmann transform is an isometry from
L2 (R, dx) onto Fα2 .
Proof. It suffices for us to show that for any nonnegative integer n, we have Bα hn =
en , where

αn n
en (z) = z .
n!
To this end, first observe that if u = x − z, where x is fixed, then d/du = −d/dz. It
follows that

dn −(x−z)2 n d
n
−u2 −u2
= e−x Hn (x).
2
n
e = (−1) n
e = e Hn (u)
dz z=0 du u=x u=x
Therefore, by Taylor’s formula,

∞
zn
e−(x−z) = ∑ e−x Hn (x)
2 2
.
n=0 n!
√
Replace x by 2α x and replace z by α /2 z. Then
1 ∞
2α 4 2 − α z2
π
e2α xz−α x 2 = ∑ ek (z)hk (x).
k=0
Multiply both sides by hn (x) and integrate over the real line. The desired result
Bα hn = en then follows from the fact that {hk } is orthonormal in L2 (R, dx).

Proposition 6.9. The inverse of the Bargmann transform is given by
1
2α 4 2 − α z2
[B−1
α f ](x) = f (z)e2α xz−α x 2 dλα (z), (6.10)
π C
where f ∈ Fα2 .
Proof. Fix any polynomial f ∈ Fα2 and any function g ∈ L2 (R, dx) that is compactly
supported. Since
Bα : L2 (R, dx) → Fα2 ⊂ L2α
is an isometry, we have
B−1 −1
α f , gL2 (R) = Bα Bα f , Bα gα = f , Bα gα
1
2α 4 2 − α z2
= f (z) dλα (z) g(x)e2α xz−α x 2 dx
π C R
1
2α 4 2 − α z2
= g(x) dx f (z)e2α xz−α x 2 dλα (z)
π R C
= F, gL2 (R) ,
where
1
2α 4 2 − α z2
F(x) = f (z)e2α xz−α x 2 dλα (z).
π C
This proves the desired formula for B−1 2

α f , as the polynomials are dense in Fα and
the compactly supported functions in L (R, dx) are dense there.
2

Proposition 6.10. Let a = r + is ∈ C and ka be the normalized reproducing kernel
of Fα2 at point a. Then
1
2α 4
[B−1 e−2α i(rD+sX) e−α x ,
2
α ka ](x) = (6.11)
π
where e−2α i(rD+sX) is the pseudodifferential operator defined in (1.21).
Proof. Let c = (2α /π )1/4. By Proposition 6.9 and the reproducing property in Fα2 ,

2 − α z2
[B−1
α ka ](x) = c e2α xz−α x 2 ka (z) dλα (z)
C

α 2 −α x2 α 2
= ce− 2 |a| e2α xz− 2 z eα az dλα (z)
C
α 2 −α x2 +2α xa− α a2
= ce− 2 |a| 2
α 2 +s2 )−α x2 +2α x(r−is)− α (r−is)2

= ce− 2 (r 2
= ce−2α ixs+α irs−α x

2 +2α xr−α r2
.
On the other hand, by (1.21),
e2α i(−rD−sX) e−α x = e−2α isx+α irs−α (x−r) = e−2α isx+α irs−α x
2 2 2 +2α xr−α r2
.

Lemma 6.11. We have

e−2π izx−π x dx = e−π z
2 2
(6.12)
R
for all complex numbers z.
Proof. Recall that

1
2α 4
e− α x
2
h0 (x) =
π
is the first vector in the orthonormal basis {hn} of L2 (R, dx). By Theorem 6.8 and
its proof, Bα (h0 ) = e0 = 1, or equivalently,

2α α 2
e2α xz−2α x dx = e 2 z .
2
π R

Replacing z by −i 2π /α z and changing x to π /(2α )x, we obtain the desired
identity in (6.12).

The above lemma simply states that the Fourier transform of e−π x is e−π z . In
2 2
what follows, the equivalent form (obtained by suitable changes of variables)

−iα xz− α2 x2 2 π − α z2
e dx = e 2 (6.13)
R α
will be more convenient for us to use.

Theorem 6.12. Suppose σ (w) = σ (u, v), with w = v + iu, is a symbol function and
σ (D, X) is the Weyl pseudodifferential operator on L2 (R, dx) with symbol σ . Let
T = Bα σ (D, X)B−1
α on Fα . Then T (z) = B2α σ (z) for all z ∈ C.
2
Proof. Recall that

σ (D, X) = σ (p, q)e2π i(pD+qX) dp dq
R R
α 2 α α 2α i(pD+qX)
= σ p, q e dp dq.
π R R π π
T(z)= Bα σ (D, X)B−1

α kz , kz α
= σ (D, X)B−1 −1
α kz , Bα kz L2 (R)
α 2 α α
= σ p, q e2α i(pD+qX) B−1 −1
α kz , Bα kz L2 (R) dpdq.
π R R π π
To simplify notation, let us write
ρ (p, q) = e2α i(pD+qX)
for real p and q, and proceed to compute the integral
I = e2α i(pD+qX) B−1 −1

α kz , Bα kz L2 (R) .
Let z = r + is. By Proposition 6.10, Lemma 1.28, and the fact that each ρ (−r, −s)
is a unitary operator on L2 (R, dx), we have
I = c2 ρ (p, q)ρ (−r, −s)e−α x , ρ (−r, −s)e−α x L2 (R)

2 2
= c2 e2α i(−ps+qr)ρ (−r, −s)ρ (p, q)e−α x , ρ (−r, −s)e−α x L2 (R)
2 2
= c2 e2α i(−ps+qr)ρ (p, q)e−α x , e−α x L2 (R) ,

2 2

where c2 = 2α /π .
By (1.21) and the change of variables x → x − (p/2),

2 2α i(−ps+qr)
e2α iqx+α ipq−α (x+p)
2−α x2
I=c e dx
R

p 2
−α (x− 2p )2
= c2 e2α i(−ps+qr) e2α iqx−α (x+ 2 ) dx
R

α 2
= c2 e2α i(−ps+qr)− 2 p e2α iqx−2α x dx.
2
It follows from another change of variables and Lemma 6.11 that

√α
2α iqx−2α x2 π
e−2π i 2π qx−π x dx
2
e dx =
R 2α R

π −π · α q2 π − α q2
= e 2 π = e 2 .
2α 2α
Therefore,
α 2 α α 2α i(−ps+qr)− α (p2 +q2 )
T(z) = σ p, q e 2 dp dq
π R R π π

π2
σ (p, q)e2π i(−ps+qr)− 2α (p
2 +q2 )
= dp dq.
R R
We rewrite T(z) as

π2
e2π i(−ps+qr)− 2α (p σ (u, v)e−2π i(pu+qv) du dv.
2 +q2 )
dp dq
R R R R
Interchanging the order of integration above, we see that T(z) is equal to

π 2 2 ]+[−2π iq(−r+v)− π 2 q2 ]
σ (u, v) du dv e[−2π ip(s+u)− 2α p 2α dp dq.
R R R R
Evaluate the inner integrals using Lemma 6.11 again. We obtain

2α
T(z) = σ (u, v)e−2α [(s+u)
2 +(−r+v)2 ]
du dv.
π R R
Since z = r + is and w = v + iu, we can rewrite the above formula as

2α
T(z) = σ (w)e−2α |w−z| dA(w) = B2α σ (z).
2
π C
This completes the proof of the theorem.

The rest of this section is devoted to showing that every Toeplitz operator on Fα2
is unitarily equivalent to an anti-Wick pseudodifferential operator on L2 (R, dx).
We begin with the unbounded operator A of differentiation on Fα2 together with

its adjoint. Thus,
1
A f (z) = f (z), A∗ f (z) = z f (z). (6.14)
α
We show that, via the Bargmann transform Bα , these operators are unitarily
equivalent to certain familiar operators on L2 (R, dx).
Lemma 6.13. For any positive α , we have
B−1
α ABα = X + iD = Z, B−1 ∗ ∗
α A Bα = X − iD = Z , (6.15)
where X, D, and Z are the (unbounded) operators on L2 (R, dx) defined in Sect. 1.4.
Proof. Let Cc (R) denote the space of continuous functions on R having compact
support. Then Cc (R) is dense in L2 (R, dx). Given f ∈ Cc (R), we differentiate
1
2α 4 2 − α z2
Bα f (z) = e2α xz−α x 2 f (x) dx
π R
under the integral sign to obtain

1
2α 4 2 − α z2
ABα f (z) = (2x − z)e2α xz−α x 2 f (x) dx. (6.16)
π R
This gives
ABα f = 2Bα X f − A∗ Bα f ,
and hence
B−1 −1 ∗
α ABα + Bα A Bα = 2X. (6.17)
On the other hand, we can rewrite (6.16) as

1
1 2α 4 d 2α xz−α x2 − α z2
ABα f (z) = − f (x) e 2 dx + A∗Bα f (z).
α π R dx
Apply integration by parts to the integral above. We obtain
ABα f = 2iBα D f + A∗ Bα f ,
and hence
B−1 −1 ∗
α ABα − Bα A Bα = 2iD. (6.18)
Solving for B−1 −1 ∗

α ABα and Bα A Bα from (6.17) and (6.18), we obtain the desired
results.

We now establish the relationship between anti-Wick pseudodifferential

operators on L2 (R, dx) and Toeplitz operators on Fα2 .
Theorem 6.14. Let
σ (z) = σ (z, z) = ∑ cnm zn zm
be real analytic and

σ (Z, Z ∗ ) = ∑ cnm Z n Z ∗m
be the anti-Wick pseudodifferential operator on L2 (R, dx). We have
Bα σ (Z, Z ∗ )B−1
α = Tϕ , (6.19)
where Tϕ is the Toeplitz operator on Fα2 with symbol ϕ (z) = σ (z, z) = σ (z).
Proof. By Lemma 6.13, we have
α = ∑ cnm A A .
Bα σ (Z, Z ∗ )B−1 n ∗m
Thus, for f ∈ Fα2 , we have

n n
1 ∂
α f (z) = ∑ cnm
Bα σ (Z, Z ∗ )B−1 (zm f (z)).
α ∂ zn
If f has the property that the function zm f (z) is also in Fα2 (all polynomials, which
are dense in Fα2 , clearly have this property), then we can write

zm f (z) = wm f (w)eα zw dλα (w).
C
Differentiating under the integral sign n times, we obtain

∂n m
(z f (z)) = α n wn wm f (w)eα zw dλα (w).
∂ zn C
Therefore,

Bα σ (Z, Z ∗ )B−1
α f (z) =
C
∑ cnm wn wm f (w)eα zw dλα (w)

= ϕ (w) f (w)eα zw dλα (w)
C
= Tϕ f (z).
This proves the desired relation.

6.3 Boundedness 229
6.3 Boundedness
In this section, we obtain necessary and sufficient conditions for the Toeplitz
operator Tϕ to be bounded on Fα2 . These conditions are based on the Berezin
transform or the heat transform at particular time points.
The main results of the section can be summarized as follows:
(a) If Tϕ is bounded on Fα2 , then Bβ ϕ is bounded for all β ∈ (0, 2α ).
(b) If Bβ ϕ is bounded for some β > 2α , then Tϕ is bounded on Fα2 .
(c) If ϕ ≥ 0, then Tϕ is bounded on Fα2 if and only if Bα ϕ is bounded if and only if
ϕr is bounded, where r is any fixed radius. Here,

1
ϕr (z) = ϕ (w) dA(w)
π r2 B(z,r)
is the averaging function of ϕ with respect to area measure.

(d) If ϕ ∈ BMO1 , then Tϕ is bounded on Fα2 if and only if T|ϕ | is bounded on Fα2 if
and only if Bα ϕ is bounded.
The proof of (c) uses the characterizations of Fock–Carleson measures and
is almost straightforward. The result in (d) follows from (c) and the translation
invariant characterization of BMO1 .
The proof of (a) depends on some general trace estimates and the semigroup
property of the weighted Berezin transforms. The proof of (b) requires certain
estimates from the theory of pseudodifferential operators.
We now get down to the details.
Recall that the standard orthonormal basis for Fα2 is given by

αn n
en (z) = z , n = 0, 1, 2, 3, · · · .
n!
For any nonnegative integer n, let Pn denote the rank-one projection from Fα2 onto
the one-dimensional subspace generated by en . Thus,
Pn f = f , en en , n ≥ 0, f ∈ Fα2 .
It follows from (3.3), the definition of KS (w, z), that
KPn (z, w) = en (z)en (w), n ≥ 0.
For any parameter t ∈ (−1, 1), we consider the operator

∞
T (t) = (1 − t) ∑ t n Pn , (6.20)
n=0
with the usual convention that T (0) = P0 . It is clear that the series above converges
in the norm topology of Fα2 . By property (7) of Proposition 3.9, we have
∞
KT (t) (z, w) = (1 − t) ∑ t n KPn (z, w)
n=0
∞
= (1 − t) ∑ t n en (z)en (w)
n=0
= (1 − t)eα tzw ,
and the series converges uniformly on compact subsets of C × C.

Let S1 denote the norm in the trace class S1 . Since each Pn is a positive trace-
class operator with Pn S1 = tr (Pn ) = 1, the series in (6.20) also converges in S1
with
∞
1−t
T (t) S1 ≤ (1 − t) ∑ |t|n PnS1 = , (6.21)
n=0 1 − |t|
and
∞
tr (T (t) ) = (1 − t) ∑ t n tr (Pn ) = 1. (6.22)
n=0
Recall that for each a ∈ C, we have the Weyl unitary operator Wa on Fα2 , a
weighted translation operator, defined by
α
Wa f (z) = f (z − a)ka (z) = eα za− 2 |a| f (z − a).
2
We always have
Wa∗ = W−a , Wa Tϕ Wa∗ = Tϕ ◦τa , Wa∗ Tϕ Wa = Tϕ ◦ta ,
where τa (z) = z − a and ta (z) = z + a. The translation invariance of the parametrized

Berezin transform also gives
Bβ (ϕ ◦ τa ) = (Bβ ϕ ) ◦ τa , Bβ (ϕ ◦ ta ) = (Bβ ϕ ) ◦ ta.
Now for every t ∈ (−1, 1) and every a ∈ C, we consider the operator

(t)
Ta = Wa T (t)Wa∗ .
(t) (t)
Thus, T0 = T (t) , each Ta is still in the trace class, and it follows from the well-
known trace identity tr (AB) = tr (BA) that
(t)
tr (Ta ) = tr [T (t)Wa∗Wa ] = tr (T (t) ) = 1
for all t ∈ (−1, 1) and a ∈ C.

6.3 Boundedness 231
Theorem 6.15. Suppose ϕ satisfies condition (I2 ) and Tϕ is bounded on Fα2 . Then
(t)
tr (Tϕ Ta ) = Bβ ϕ (a) (6.23)
√
for all −1 < t < 2 − 1, where β = α (1 − t).
Proof. We first prove the result for a = 0. The problem is reduced to checking
hypothesis (4) of Theorem 6.2. In fact, it would then follow from (6.5) that

tr Tϕ T (t) = ϕ (z)KT (t) (z, z) dλα (z)
C

ϕ (z)eα t|z| dλα (z)
2
= (1 − t)
C

α (1 − t)
ϕ (z)e−α (1−t)|z| dA(z)
2
=
π C
= Bβ ϕ (0).
Thus, we need to estimate the integral

I(t) = |ϕ (z)| dλα (z) |K(z, w)||KT (t) (z, w)| dλα (w)
C C

= (1 − t) |ϕ (z)| dλα (z) |eα (1+t)zw | dλα (w)
C C

α (1+t)2
|z|2
= (1 − t) |ϕ (z)|e 4 dλα (z)
C

α (1 − t) (1+t)2
|ϕ (z)|eα [ 4 −1]|z|
2
= dA(z)
π C

α (1 − t) α
|ϕ (z)e− 2 |z| |e−δ (t)|z| dA(z),
2 2
=
π C
where

(1 + t)2 α α
δ (t) = α 1 − − = (1 − 2t − t 2 ).
4 2 4
By the Cauchy–Schwarz inequality, I < ∞ whenever δ (t) >√0. It is elementary that

for t ∈ (−1, 1), we have δ (t) > 0 if and only if −1 < t < 2 − 1. This proves the
desired result for a = 0.
In general, note that Tϕ is bounded if and only if Tϕ ◦ta is bounded. Thus,
(t)
tr (Tϕ Ta ) = tr (Tϕ Wa T (t)Wa∗ ) = tr (Wa∗ Tϕ Wa T (t) )
= tr (Tϕ ◦ta T (t) ) = Bβ (ϕ ◦ ta )(0)
= Bβ ϕ (a),

As a consequence of the theorem above, we obtain the following necessary

condition for a Toeplitz operator to be bounded on Fα2 , one of the main results of
this section.
Theorem 6.16. Suppose ϕ satisfies condition (I2 ) and Tϕ is bounded on Fα2 . Then
Bβ ϕ is bounded for all β with 0 < β < 2α .
√
Proof. Let β = α (1√− t) with −1 < t < 1. The condition −1 < t < 2 − 1 is
equivalent to α (2 − 2) < β < 2α . Also, according to the trace-norm estimate in
(6.21), we have
(t) 1−t
Ta S1 = Wa T (t)Wa∗ S1 = T (t) S1 ≤ .
1 − |t|
Combining this with Theorem 6.15, we obtain
(t) (t) 1−t

|Bβ ϕ (a)| = |tr (Tϕ Ta )| ≤ Tϕ Ta S1 ≤ Tϕ
1 − |t|
for all a ∈ C. This shows that
1−t
Bβ ϕ ∞ ≤ Tϕ < ∞ (6.24)
1 − |t|
√
whenever α (2 − 2)√< β < 2α .
If 0 < β ≤ α (2 − 2) < α , we can find a positive number γ such that
1 1 1
= + .
β γ α
By Theorem 3.13, the semigroup property of the heat transform Ht , we have H1/β =
H1/γ H1/α . In terms of the parametrized Berezin transforms, we have Bβ = Bγ Bα . By
what was proved in the previous paragraph, or directly from Bα ϕ (a) = Tϕ ka , ka α ,
the boundedness of Tϕ on Fα2 implies Bα ϕ ∞ ≤ Tϕ . Since Bγ is a contraction on
L∞ , we have
Bβ ϕ ∞ = Bγ Bα ϕ ∞ ≤ Bα ϕ ∞ ≤ Tϕ .

Our next goal is to show that if Bβ ϕ is bounded for some β > 2α , then
Tϕ is bounded on Fα2 . This is accomplished with the help of the theory of
pseudodifferential operators.
Theorem 6.17. Suppose g satisfies condition (I2 ) and σ (D, X) is the pseudodiffer-
ential operator on L2 (R, dx) with symbol
σ (ζ , x) = σ (z) = B2α g(z), z = x + iζ .
Then Tg = Bα σ (D, X)B−1

α and B2α σ (z) = Bα g(z).
6.3 Boundedness 233
Proof. Let T = Bα σ (D, X)B−1

α . By Theorem 6.12, we have
T(z) = B2α σ (z) = B2α B2α g(z).
By the semigroup property (Corollary 3.15), we have
B2α B2α g = Bα g = Tg .
It follows that the operators T and Tg have the same Berezin symbol. Since the
mapping S → S is one-to-one, we conclude that T = Tg .

Theorem 6.18. Let g be a symbol function on C that satisfies condition (I2 ). If there
exists some β ∈ (2α , ∞) such that Bβ g ∈ L∞ (C), then Tg is bounded on Fα2 .
Proof. Let σ (z) = B2α g(z). In view of Theorem 6.17, the Toeplitz operator Tg on
Fα2 is unitarily equivalent to the pseudodifferential operator σ (D, X) on L2 (R, dx).
We proceed to show that the pseudodifferential operator σ (D, X) is bounded.
Let γ be the positive number satisfying
1 1 1
= + .
2α β γ
By the semigroup property of the parametrized Berezin transforms, we have
σ (z) = B2α g(z) = Bγ Bβ g(z).
Let ϕ (z) = Bβ g(z). Then ϕ is in L∞ (C), and

γ
ϕ (w)e−γ |z−w| dA(w).
2
σ (z) =
π C
Differentiating under the integral sign, we see that for any nonnegative integers n
and m, we have

∂ n+m σ
hmn (z − w, z − w)ϕ (w)e−γ |z−w| dA(w),
2
(z) =
∂ zn ∂ zm C
where hmn is a polynomial of degree m + n. Thus, for all z ∈ C, we have

n+m
∂ σ
(z) ≤ ϕ ∞ |hmn (z − w, z − w)|e−γ |z−w| dA(w)
2
∂ zn ∂ zm C

|hmn (u, u)|e−γ |u| dA(u) < ∞.
2
= ϕ ∞
C
This shows that ∂ m+n σ /∂ zn ∂ zm is bounded on C for all nonnegative integer n

and m. By Theorem 1.24, the pseudodifferential operator σ (D, X) is bounded on
L2 (R, dx).

When the symbol function ϕ is nonnegative, we have the following characteriza-
tion for boundedness.
Theorem 6.19. Suppose ϕ ≥ 0 satisfies condition (I1 ). Then the following condi-
tions are equivalent:
(a) Tϕ is bounded on Fα2 .
(b) ϕ = Bα ϕ ∈ L∞ (C).
(c) Bβ ϕ ∈ L∞ (C), where β is any fixed positive weight parameter.
(d) ϕr ∈ L∞ (C), where r is any fixed positive radius.
Proof. The equivalences of (a), (b), and (d) follow from the characterization of
Fock–Carleson measures in Sect. 3.4. In fact, when ϕ is nonnegative, we have

Tϕ f , f α = | f |2 ϕ dλα .
C
The densely defined positive operator Tϕ is bounded if and only if there exists a
constant C > 0 such that
Tϕ f , f α ≤ C f 22,α , f ∈ Fα2 ,
which is the same as

| f |2 ϕ dλα ≤ C | f |2 dλα , f ∈ Fα2 .
C C
This condition simply means that the measure dμ (z) = ϕ (z) dA(z) is Fock–Carleson.
The equivalence of (b) and (c) follows from Theorem 3.23.

As a consequence of the above theorem, we obtain the following characterization
of bounded Toeplitz operators on Fα2 induced by symbols from BMO1 .
Theorem 6.20. Suppose ϕ ∈ BMO1 . Then the following conditions are equiva-
lent:
(a) Tϕ is bounded on Fα2 .
(b) ϕ = Bα ϕ ∈ L∞ (C).
(c) Bβ ϕ ∈ L∞ (C), where β is any fixed positive weight parameter.
(d) ϕr ∈ L∞ (C), where r is any fixed positive radius.
Proof. By (3.22) of Theorem 3.34, there exists a constant C > 0 such that
ϕ ◦ ϕz − ϕ
(z)L1 (dλα ) ≤ C
6.3 Boundedness 235
for all z ∈ C, where ϕz (w) = z − w. By the triangle inequality, we also have
ϕ ◦ ϕz L1 (dλα ) − |ϕ(z)| ≤ C
for all z ∈ C, which is the same as
|*
ϕ | − |ϕ| ∈ L∞ (C).
Therefore, ϕ ∈ L∞ (C) if and only if |* ϕ | ∈ L∞ (C). It follows from this and the
characterization of bounded Toeplitz operators with nonnegative symbols (see
Theorem 6.19) that the condition ϕ ∈ L∞ (C) implies that T|ϕ | is bounded on Fα2 .
Let ϕ = f + ig, where f and g are the real and imaginary parts of ϕ , respectively.
Since | f | ≤ |ϕ | and |g| ≤ |ϕ |, and nonnegative symbols induce positive operators,
we see that the boundedness of T|ϕ | implies that both T| f | and T|g| are bounded on
Fα2 .
Since f is real-valued, we can write f = f + − f − , where
f + = max( f , 0), f − = max(0, − f ),
are the positive and negative parts of f , respectively. It follows from 0 ≤ f + ≤ | f |

and 0 ≤ f − ≤ | f | that T f + and T f − are both bounded on Fα2 . Thus, T f = T f + − T f − is
bounded. Similarly, Tg is bounded. This shows that the condition ϕ ∈ L∞ (C) implies
the boundedness of Tϕ on Fα2 . Since the inverse implication is obvious, we have
proved the equivalence of (a) and (b).
Recall from the proof of Theorem 3.36 that Bβ ϕ − ϕr is bounded when
ϕ ∈ BMO1 . This shows that conditions (b), (c), and (d) are equivalent whenever
ϕ ∈ BMO1 . This completes the proof of the theorem.

6.4 Compactness 237
6.4 Compactness
In this section, we discuss the compactness of Toeplitz operators on Fα2 . The main
results are parallel to those in the previous section. All conditions in this section are
in terms of membership in the space C0 (C) which consists of continuous functions
f on C such that f (z) → 0 as z → ∞. In several approximation arguments, we will
also need the space Cc (C), consisting of continuous functions f on C with compact
support. It is clear that Cc (C) is dense in C0 (C) in the supremum norm of C0 (C).
Theorem 6.21. Suppose ϕ satisfies condition (I2 ) and Tϕ is compact on Fα2 . Then
Bβ ϕ ∈ C0 (C) for all β ∈ (0, 2α ).
Proof. Recall from Theorem 6.16 and its proof that, for any β ∈ (0, 2α ), there exists
a positive constant C = C(β ) such that Bβ f ∞ ≤ CT f whenever T f is bounded
on Fα2 .
If Tϕ is compact on Fα2 , then by Theorem 6.5, there exists a sequence { fn } of
functions in Cc (C) such that
1
Tϕ − T fn ≤ , n ≥ 1.
n
Therefore,
1
Bβ ϕ − Bβ fn ∞ ≤ CTϕ − T fn <
n
for all n ≥ 1. Each fn has compact support, so Bβ fn ∈ C0 (C). Since C0 (C) is closed
in the supremum norm, we conclude that Bβ ϕ is in C0 (C) as well.

Theorem 6.22. Suppose g is a symbol function that satisfies condition (I2 ). If there
exists some β ∈ (2α , ∞) such that Bβ g ∈ C0 (C), then Tg is compact on Fα2 .
Proof. As in the proof of Theorem 6.18, the Toeplitz operator Tg on Fα2 is unitarily
equivalent to the pseudodifferential operator σ (D, X) on L2 (R, dx), where σ (z) =
B2α g(z). Furthermore, it follows from Theorem 6.18 that Tg and σ (D, X) are both
bounded operators with

γ
ϕ (w)e−γ |z−w| dA(w),
2
σ (z) = Bγ ϕ (z) =
π C
where ϕ (z) = Bβ g(z). For any pair of nonnegative integers m and n, there is a
polynomial hmn (z, z) such that

∂ m+n σ
hmn (z − w, z − w)ϕ (w)e−γ |z−w| dA(w).
2
(z) = (6.25)
∂ zm ∂ zn C
The integral transform T defined by

hmn (z − w, z − w) f (w)e−γ |z−w| dA(w)
2
T f (z) =
C
is bounded on L∞ (C). See the proof of Theorem 6.18. If f is compactly supported,

say on |z| ≤ R, then

hmn (z − w, z − w) f (w)e−γ |z−w| dA(w)
2
T f (z) =
|w|≤R

hmn (w, w) f (z − w)e−γ |w| dA(w),
2
=
|w−z|≤R
and so

|hmn (w, w)|e−γ |w| dA(w).
2
|T f (z)| ≤ f ∞
|w−z|≤R
The convergence of the integral

|hmn (w, w)|e−γ |w| dA(w)
2
clearly implies that T f (z) → 0 as z → ∞. Thus, T maps Cc (C) into C0 (C). Since
Cc (C) is dense in C0 (C) in the norm topology of L∞ (C), we infer from the
boundedness of T : L∞ (C) → L∞ (C) that T maps C0 (C) into C0 (C). This, along with
(6.25), shows that ∂ m+n σ /∂ zm ∂ zm is in C0 (C) for any pair of nonnegative integers
m and n. By Theorem 1.25, the pseudodifferential operator σ (D, X) is compact on
L2 (R, dx), and hence the Toeplitz operator Tg is compact on Fα2 .

Theorem 6.23. Suppose ϕ is nonnegative and satisfies condition (I1 ). Then, the
following conditions are equivalent:
(a) Tϕ is compact on Fα2 .
(b) ϕ ∈ C0 (C).
(c) Bβ ϕ ∈ C0 (C), where β is any fixed positive weight parameter.
(d) ϕr ∈ C0 (C), where r is any fixed positive radius.
Proof. The equivalence of (a), (b), and (d) follow from the characterization of
vanishing Fock–Carleson measures in Sect. 3.4. See the proof of Theorem 6.19 for
the connection to Fock–Carleson measures. The equivalence of (b) and (c) follows
from Theorem 3.23.

The rest of this section is devoted to the compactness of Toeplitz operators with
symbols in BMO1 .
Lemma 6.24. Suppose f ∈ BMO1 and f = Bα f is bounded. Then
T f Kz = Kz [P( f ◦ ϕz )] ◦ ϕz (6.26)
for all z ∈ C, where P : L2α → Fα2 is the orthogonal projection, T f is the Toeplitz
operator on Fα2 , Kz is the reproducing kernel of Fα2 , and ϕz (w) = z − w.
6.4 Compactness 239
Proof. Since BMO1 and the Berezin transform are both translation invariant, we see
that for any z ∈ C, we have
f ◦ ϕz ∈ BMO1 , Bα ( f ◦ ϕz ) ∈ L∞ (C).
In particular, each side of (6.26) is well defined.

By the definition of Toeplitz operators and a change of variables,

T f Kz (w) = P( f Kz )(w) = f (u)Kz (u)Kw (u) dλα (u)
C

= f (ϕz (u))Kz (ϕz (u))Kw (ϕz (u))|kz (u)|2 dλα (u)
C

= f (ϕz (u))eα (zw−uw+zu) dλα (u).
C
On the other hand,

Kz (w)[P( f ◦ ϕz )](ϕz (w)) = eα zw f (ϕz (u))eα (z−w)u dλα (u)
C

= f (ϕz (u))eα (zw+zu−wu) dλα (u).
C
This proves the desired identity.

Lemma 6.25. Suppose f ∈ BMO and f is bounded. Then there exists a positive
1
constant C such that

sup |P( f ◦ ϕz )(w)| ≤ Ceα |w|
2 /4
(6.27)
z∈C
for all w ∈ C.
Proof. Recall from the proof of Theorem 6.20 that if f ∈ BMO1 and f is bounded,
then |*
f | is bounded as well. By translation invariance of BMO1 and the Berezin
transform, there exists a positive constant C such that
Bα (| f ◦ ϕz |)(w) ≤ C, z, w ∈ C.
By Theorem 3.29, there exists another positive constant C (independent of z) such

that

|g(u)|| f ◦ ϕz (u)| dλα (u) ≤ C |g(u)| dλα (u)
C C
for all entire functions g. In particular,

|P( f ◦ ϕz )(w)| = f ◦ ϕz (u)e α wu
dλα (u)
C

≤ | f ◦ ϕz (u)||eα wu | dλα (u)
C

≤C |eα wu | dλα (u)
C
α
= Ce 4 |w| .
2

Lemma 6.26. Suppose f ∈ BMO1 and f = Bα f ∈ C0 (C). Then:
(a) For any a ∈ C, we have P( f ◦ ϕz )(a) = T f ◦ϕz 1(a) → 0 as z → ∞.
(b) T f ◦ϕz 1 → 0 weakly in Fα2 as z → ∞.
Proof. By Theorem 6.20 and the fact that T f ◦ϕz = Uz T f Uz , where Uz f = f ◦ ϕz kz is

a self-adjoint unitary operator, there exists a constant C > 0 such that T f ◦ϕz ≤ C
for all z ∈ C. In particular, T f ◦ϕz 1 ≤ C for all z ∈ C. Since
T f ◦ϕz 1(a) = T f ◦ϕz 1, Ka
and the set of all finite linear combinations of kernel functions is dense in Fα2 , we
see that (a) and (b) are actually equivalent.
To prove part (b), it suffices to show that
lim T f ◦ϕz 1, un = 0 (6.28)

z→∞
for every nonnegative integer n because the set of polynomials is dense in Fα2 .
Fix a nonnegative integer n and a point a ∈ C. Observe that
f(ϕz (a)) = f
◦ ϕz (a) = e−α |a| T f ◦ϕz Ka , Ka ,
2
where
∞
αk k k
Ka (u) = eα ua = ∑ ua.
k=1 k!
It follows that
∞
α k+ j + ,
f(ϕz (a)) = e−α |a| ∑
2
T f ◦ϕz uk , u j ak a j .
k, j=0 k! j!
Thus, for any positive radius r, the integral

f(ϕz (u))un eα |u| dA(u)
2
Ir (z) =
|u|<r
6.4 Compactness 241
can be written as
∞
α k+ j + ,
Ir (z) = ∑ k! j!
T f ◦ϕz uk , u j
|v|<r
vk+n v j dA(v)
k, j=0
∞
α 2k+n
= ∑ k!(k + n)! Tf ◦ϕz uk , uk+n |v|<r
|v|2(k+n) dA(v)
k=0
∞
α 2k+n
=π ∑ k!(k + n + 1)! Tf ◦ϕz uk , uk+n r2(k+n+1)
k=0

αn - .
= πr 2(n+1)
T f ◦ϕz 1, u + Σr,n (z) ,
n
(n + 1)!
where
∞
α 2k+n
Σr,n (z) = ∑ k!(k + n + 1)! Tf ◦ϕz uk , un+k r2k .
k=1
As z → ∞, we have f(ϕz (u)) → 0 for every u ∈ C. By the dominated convergence

theorem,

f(ϕz (u))un eα |u| dA(u) = 0
2
lim Ir (z) = lim
z→∞ z→∞ |u|<r
for any r > 0. It follows that

αn
lim T f ◦ϕz 1, u + Σr,n(z) = 0
n
(6.29)
z→∞ (n + 1)!
for any fixed r > 0. Since T f ◦ϕz ≤ C for all z ∈ C, where C is independent of z,
we see that
∞
α 2k+n uk uk+n 2k
|Σr,n (z)| ≤ C ∑ r
k=1 k!(k + n + 1)!

∞
α 2k+n k!(k + n)! 2k
=C∑ r
k=1 k!(k + n + 1)! α 2k+n
∞
(α r2 )k
∑
n
≤ Cα 2
k=1 k!

= C α 2 eα r − 1
n 2
for all r > 0, n ≥ 0, and z ∈ C. Given any ε > 0, choose a small enough positive
radius r such that
2
C α 2 eα r − 1 < ε .
n
Then by (6.29), we have

(n + 1)!
lim sup |T f ◦ϕz 1, un ≤ ε.
z→∞ αn
This proves (6.28) and completes the proof of the lemma.

We can now characterize the compactness of Toeplitz operators with symbols in
BMO1 in terms of the Berezin transform.
Theorem 6.27. If f ∈ BMO1 , then T f is compact on Fα2 if and only if f ∈ C0 (C).
Proof. It suffices to show that the condition f ∈ C0 (C) implies the compactness of
T f on Fα2 . The other implication is obvious.
So let us assume that f ∈ BMO1 and f ∈ C0 (C). We will actually prove that the
operator
T f : Fα2 → L2α
is compact, which clearly implies the desired compactness of T f : Fα2 → Fα2 .
For any positive radius R, we consider the operator
T fR = MχR T f : Fα2 → L2α ,
where χR is the characteristic function of the open ball |z| < R and MχR is the
operator of multiplication on L2α by χR . It follows from the boundedness of T f and a
simple normal family argument that each T fR is compact. Thus, the compactness of
T f will follow if we can show that
lim T fR − T f Fα2 →L2α = 0. (6.30)

R→∞
Given g ∈ Fα2 , we have
(T f − T fR )g(z) = (1 − χR)T f g(z)

= (1 − χR(z))T f g, Kz α
= (1 − χR(z))g, T f Kz α

= g(u)(1 − χR(z))T f Kz (u) dλα (u).
C
Thus, T f − T fR is an integral operator with kernel
K Rf (z, u) = (1 − χR(z))T f Kz (u).
By Schur’s test (Lemma 2.14), whenever there exists a positive function h on C such
that

|K Rf (z, u)|h(z) dλα (z) ≤ C1 h(u), u ∈ C, (6.31)
C
6.4 Compactness 243
and

|K Rf (z, u)|h(u) dλα (u) ≤ C2 h(z), z ∈ C, (6.32)
C
we then have
T f − T fR 2F 2 →L2 ≤ C1C2 . (6.33)
α α
We will arrive at constants C1 and C2 such that the product C1C2 tends to 0 as R → ∞,
which then implies the compactness of T f .
α 2
Let h(z) = K(z, z) = e 2 |z| and consider the integrals

I(z) = |K Rf (z, u)|h(u) dλα (u), z ∈ C,
C
from (6.32). It is clear that I(z) = 0 for |z| < R. For |z| ≥ R, we have

I(z) = |T f Kz (u)| K(u, u) dλα (u),
C
which by Lemma 6.24 can be written as

I(z) = |Kz (u)||P( f ◦ ϕz )(ϕz (u))| K(u, u) dλα (u).
C
Making the change of variables u → ϕz (u) and simplifying the result, we get
α
α
|P( f ◦ ϕz )(u)| eα (z−u)z e− 2 |z−u| dA(u).
2
I(z) =
π C
Fix p ∈ (1, ∞) and σ ∈ (α /4, α /2). Let 1/p + 1/q = 1. By Hölder’s inequality,

α α

|P( f ◦ ϕz (u))|e−σ |u| eσ |u| |eα (z−u)z |e− 2 |z−u| dA(u)
2 2 2
I(z) =
π C
1
α p −pσ |u|2
p
≤ |P( f ◦ ϕz (u))| e dA(u)
π C
1
qσ |u|2 qα (z−u)z − q2α |z−u|2
q
× e |e |e dA(u) .
C
The second integral above can be written as

qα 2 qα
qα 2 2 qα qα 2
e 2 |z| +qσ |u| e− 2 |z| + 2 (z−u)z+ 2 (z−u)z− 2 |z−u| dA(u),
C
which is equal to

qα 2 2 qα qα 2 α
2 |z| +qσ |u| − 2 |u| 2 |z| e−q( 2 −σ )|u| dA(u).
2 2
e dA(u) = e
C C
On the other hand, it follows from Lemma 6.25 that for all z ∈ C, we have
α
|P( f ◦ ϕz )(u)| p e−pσ |u| ≤ Ce−p(σ − 4 )|u| ,
2 2
with the function on the right-hand side above integrable with respect to dA. This,
along with Lemma 6.26 and the dominated convergence theorem, shows that the
constants
1p
p −pσ |u|2
C1,R = sup |P( f ◦ ϕz (u))| e dA(u)
|z|≥R C
tend to 0 as R → ∞. Therefore, we can find constants C1,R such that C1,R → 0 as

R → ∞ and I(z) ≤ C1,R h(z) for all z ∈ C. This proves the desired estimate in (6.32).
The integrals

J(u) = |K Rf (z, u)|h(z) dλα (z)
C
from (6.31) are slightly easier to estimate. In fact, by Lemma 6.24 and a change of
variables,

J(u) ≤ |T f Kz (u)| K(z, z) dλα (z)
C

α α 2
= |Kz (u)||P( f ◦ ϕz )(ϕz (u))|e− 2 |z| dA(z)
π C

α α
|Kz+u (u)||P( f ◦ ϕz+u)(z)|e− 2 |z+u| dA(z)
2
=
π C

α α |u|2 α 2
= e2 |P( f ◦ ϕz+u )(z)|e− 2 |z| dA(z).
π C
By Lemma 6.25, there is a positive constant C such that

α α 2 − α |z|2 α α
J(u) ≤ Ce 2 |u| e 4 |z| dA(z) = Ce 2 |u| e− 4 |z| dA(z).
2 2 2
2
C C
This proves the desired estimate in (6.31) and completes the proof of the theorem.

Corollary 6.28 Let f ∈ BMO1 , α > 0, and β > 0. Then Bα f ∈ C0 (C) if and only
if Bβ f ∈ C0 (C).
Proof. Without loss of generality, assume that 0 < α < β . If Bβ f ∈ C0 (C), then
by Proposition 3.21, Bα f ∈ C0 (C). We do not need the assumption that f ∈ BMO1
here.
If Bα f ∈ C0 (C), then by Theorem 6.27, T f is compact on Fα2 , which, according
to Theorem 6.21, implies that Bγ f ∈ C0 (C) for all 0 < γ < 2α . Repeat this process
a certain number of times, we will then get Bβ f ∈ C0 (C).

6.5 Toeplitz Operators in Schatten Classes 245
6.5 Toeplitz Operators in Schatten Classes
For μ ≥ 0, we are going to determine when the Toeplitz operator Tμ on Fα2 belongs
to the Schatten class S p . The case when p ≥ 1 is relatively easy and will be taken up
first.
Recall that for any bounded linear operator T on Fα2 we define the Berezin
transform T by
T(z) = T kz , kz , z ∈ C,
where kz are the normalized reproducing kernels in Fα2 . If T is positive on Fα2 , then

α
tr (T ) = T(z) dA(z).
π C
See Proposition 3.3. In particular, T is in the trace-class S1 if and only if the integral
above converges. As a consequence, we obtain the following trace formula for
Toeplitz operators on the Fock space.
Proposition 6.29. Suppose μ is a positive Borel measure on C and satisfies
condition (M). Then Tμ is in the trace-class S1 if and only if μ is finite on C.
Moreover, tr (Tμ ) = (α /π )μ (C).
Proof. Since all integrands below are nonnegative, we use Fubini’s theorem to
obtain

α
tr (Tμ ) = μ
(z) dA(z)
π C

α
eα |z| dλα (z) |eα z̄w |2 e−α (|z| +|w| ) dμ (w)
2 2 2
=
π C C

α
e−α |w| dμ (w) |eα z̄w |2 dλα (z)
2
=
π C C

α α
= dμ (w) = μ (C).
π C π
This also shows that tr (Tμ ) < ∞ if and only if μ (C) < ∞.

Lemma 6.30. If p ≥ 1 and ϕ ∈ L p (C, dA), then Tϕ ∈ S p .
Proof. If ϕ ∈ L p (C, dA), then ϕ ◦ ta ∈ L p (C, dA) by a simple change of variables.
It follows that ϕ ◦ ta ∈ L p (C, dλα ) for every a ∈ C. Thus, ϕ satisfies condition (I p).
Since p ≥ 1, ϕ also satisfies condition (I1 ) so that Tϕ is densely defined on Fα2 .
The rest is proved in exactly the same way that Proposition 7.11 in [250] was
proved.

Lemma 6.31. Suppose r > 0, μ is a positive Borel measure on C, and
μ (B(z, r))
μ
r (z) = , z ∈ C.
π r2
If μ
r is in L p (C, dA) for some 0 < p < ∞, then μ satisfies condition (M), and the
Toeplitz operators Tμ and Tμr are both bounded on Fα2 . Moreover, there exists a
positive constant C (independent of μ ) such that Tμ ≤ CTμr .
Proof. Let

C= μ (B(z, r)) p dA(z) < ∞.
C
For any a ∈ C, we have

μ (B(z, r)) p dA(z) ≤ C.
B(a,r/2)
When z ∈ B(a, r/2), we have B(a, r/2) ⊂ B(z, r) by the triangle inequality. It follows
that μ (B(z, r)) ≥ μ (B(a, r/2)), and so
π r2
μ (B(a, r/2)) p ≤ C, a ∈ C.
4
This shows that the function a → μ (B(a, r/2)) is bounded. By Theorem 3.29 (with
p = 2 there), the measure μ satisfies condition (M), and the Toeplitz operator Tμ is
bounded on Fα2 , which in turn implies that the function z → μ (B(z, r)) is bounded.
Thus, Tμr is bounded on Fα2 as well.
Given f ∈ Fα2 , we use Fubini’s theorem to obtain

π r2 Tμr f , f = π r2 | f (z)|2 μ
r (z) dλα (z)
C

= | f (z)|2 μ (B(z, r)) dλα (z)
C

= | f (z)|2 dλα (z) χB(z,r) (w) dμ (w)
C C

= dμ (w) | f (z)|2 χB(w,r) (z) dλα (z)
C C

α
| f (z)e−α |z|
2 /2
= dμ (w) |2 dA(z).
π C B(w,r)
Combining the above identity with Lemma 2.32, we obtain a positive constant C
such that

| f (w)|2 e−α |w| dμ (w) = Tμ f , f .
2
CTμr f , f ≥
C

Note that the condition μ
r ∈ L p (C, dA) for 0 < p < ∞ implies that

lim μ (B(z, r)) p dA(z) = 0.
a→∞ B(a,r/2)
Refining the arguments in the above proof then shows that μ (B(a, r/2)) → 0 as
a → ∞, which implies that Tμ is compact on Fα2 and μ r ∈ C0 (C).
For the remainder of this section, we let {an } denote any fixed arrangement of
the square lattice rZ2 into a sequence. We are now ready to characterize positive
Toeplitz operators in S p when p ≥ 1.
Theorem 6.32. Suppose μ ≥ 0, r > 0, and p ≥ 1. If μ satisfies condition (M), then
(a) The operator Tμ is in the Schatten class S p .
(b) The function μ
(z) is in L p (C, dA).
(c) The function μ (B(z, r)) is in L p (C, dA).
(d) The sequence { μ (B(an, r))} is in l p .
Proof. That (a) implies (b) follows from Proposition 3.5. The elementary inequality
μ
r (z) ≤ C μ (z) (see the proof of Theorem 3.29) shows that condition (b) implies (c).
If the averaging function μ r (z), which differs from μ (B(z, r) by a constant, is in
L p (C, dA), then it follows from Lemma 6.30 that Tμr is in S p . Combining this with
Lemma 6.31, we conclude that Tμ is in S p . This proves that (c) implies (a). Hence,
conditions (a), (b), and (c) are equivalent.
To prove that condition (d) is equivalent to the other conditions, we first assume
that condition (b) holds, which implies that the function μ (B(z, 2r)) is in L p (C, dA).
Choose a positive integer m such that each point in the complex plane belongs to at
most m of the disks B(an , r). Then
∞
m
C
μ (B(z, 2r)) p dA(z) ≥ ∑ μ (B(z, 2r)) p dA(z).
n=1 B(an ,r)
For each z ∈ B(an , r), we deduce from the triangle inequality that
μ (B(z, 2r)) ≥ μ (B(an , r)).
Therefore, ∞
m
C
μ (B(z, 2r)) p dA(z) ≥ π r2 ∑ μ (B(an , r)) p .
n=1
This shows that condition (b) implies (d).

To finish the proof, we assume that condition (d) holds, that is,
∞
∑ μ (B(an, r)) p < ∞.
n=1
It is easy to see that we also have

∞
∑ μ (B(zn , r)) p < ∞,
n=1
where {zn } is any arrangement of the lattice (r/2)Z2 . In fact, for each point zk that
is not in the lattice {an }, the disk B(zk , r) is covered by six adjacent disks B(ak , r).
Therefore,
∞
C
μ (B(z, r/2)) p dA(z) ≤ ∑ μ (B(z, r/2)) p dA(z)
n=1 B(zn ,r/2)
∞
≤ ∑ μ (B(zn , r)) p dA(z)
n=1 B(zn ,r/2)
π r2 ∞
= ∑ μ (B(zn , r)) p < ∞.
4 n=1
This shows that condition (d) implies (c), as the equivalence of (c) to (b) implies that
if condition (c) holds for one positive radius, then it will hold for any other positive
radius. This completes the proof of the theorem.

Specializing to the case when
α
dμ (z) = ϕ (z) dA(z),
π
we obtain the following corollary concerning Toeplitz operators induced by non-
negative functions.
Corollary 6.33 Suppose ϕ ≥ 0, p ≥ 1, and r > 0. If ϕ satisfies condition (I1 ), then
(a) The Toeplitz operator Tϕ belongs to S p .
(b) The Berezin transform ϕ belongs to L p (C, dA).
(c) The averaging function

1
π r2 B(z,r)
belongs to L p (C, dA).

(d) The averaging sequence {ϕr (an )} belongs to l p .
We now turn our attention to the case 0 0, and 0 < p ≤ 1. If μ satisfies condition (M),
then the following conditions are equivalent:
(a) The function μ
(z) is in L p (C, dA).
(b) The function μ (B(z, r)) is in L p (C, dA).
(c) The sequence { μ (B(an, r))} is in l p .
Proof. We begin with the inequality

∞
α α
e−α |z−w| dμ (w) ≤ ∑ e−α |z−w| dμ (w).
2 2
μ
(z) =
π C π n=1 B(an ,r)
For w ∈ B(an , r), we have
|z − w|2 ≥ (|z − an| − |an − w|)2 ≥ |z − an|2 − 2r|z − an|.
It follows that
∞
α
∑ e−α |z−an| +2α r|z−an| μ (B(an, r)).
2
μ
(z) ≤
π n=1
Since 0 < p ≤ 1, Hölder’s inequality gives

α p ∞
∑ e−pα |z−an| +2prα |z−an| μ (B(an , r)) p .
2
μ
(z) p ≤
π n=1

α p ∞
∑ μ (B(an, r)) p e−pα |z−an|
2 +2prα |z−a
n|
μ
(z) p dA(z) ≤ dA(z).
C π n=1 C
By an obvious change of variables, the integral above equals

e−pα |z|
2 +2prα |z|
dA(z),
C
which is easily seen to be convergent. Thus, the condition { μ (B(an , r))} ∈ l p implies
μ
∈ L p (C, dA).
On the other hand, there exists a positive integer m such that every point in the
complex plane belongs to at most m of the disks B(an , r). Thus,
∞
m
C
μ
(z) p dA(z) ≥ ∑ μ
(z) p dA(z).
n=1 B(an ,r)
For any z ∈ B(an , r), we have

α α
e−α |z−w| dμ (w) ≥ e−α |z−w| dμ (w)
2 2
μ
(z) =
π C π B(an ,r)
α
≥ e−4α r μ (B(an , r)).
2
π
It follows that
∞
(z) p dA(z) ≥ α r2 e−4pα r ∑ μ (B(an, r)) p .
2
m μ
C n=1
Thus, μ ∈ L p (C, dA) implies that { μ (B(an , r))} ∈ l p , which proves the equivalence
of conditions (a) and (c).
That condition (a) implies (b) follows from the inequality μ (B(z, r)) ≤ C μ (z)
observed in the proof of Theorem 3.29.
To prove that condition (b) implies (c), we assume that the function μ (B(z, r))
is in L p (C, dA). Consider the lattice (r/2)Z2 and arrange it into a sequence {zn }.
There exists a positive integer m such that every point in the complex plane belongs
to at most m of the disks B(zn , r/2). Therefore,
∞
m
C
μ (B(z, r)) p dA(z) ≥ ∑ μ (B(z, r)) p dA(z).
n=1 B(zn ,r/2)
For each z ∈ B(zn , r/2), the triangle inequality gives us that
μ (B(z, r)) ≥ μ (B(zn , r/2)).
Thus,

π r2 ∞
m
C
μ (B(z, r)) p dA(z) ≥ ∑ μ (B(zn , r/2)) p.
4 n=1
By the equivalence of conditions (a) and (c), the function μ belongs to
L p (C, dA). Applying the equivalence of (a) and (c) once more, we conclude that
{ μ (B(an , r))} ∈ l p . This completes the proof of the lemma.

Lemma 6.35. Suppose μ ≥ 0, 0 < p ≤ 1, and μ satisfies condition (M). If the
function μ
belongs to L p (C, dA), then the operator Tμ belongs to S p .
Proof. Since μ
belongs to L p (C, dA) and μ dominates μ r , Lemma 6.31 shows that
*
Tμ is bounded. Thus, Tμ = μ
and the desired result follows from Proposition 3.6.

We will need the following lemma, which can be found as Proposition 1.29 in
[250].
Lemma 6.36. If 0 < p ≤ 2, then for any orthonormal basis {en } of a separable
Hilbert space H and any compact operator T on H, we have
∞ ∞
p
T S p ≤ ∑ ∑ |Ten , ek | p .
n=1 k=1
We are now ready to characterize Toeplitz operators Tμ in S p when 0 0, 0 < p ≤ 1, and μ satisfies condition (M).

Then the following conditions are equivalent:
(a) Tμ belongs to the Schatten class S p .
(b) μ
(c) μ
r belongs to L p (C, dA).
(d) {μ
r (an )} belongs to l p .
Proof. The equivalence of (b), (c), and (d) was proved in Lemma 6.34. That
condition (b) implies condition (a) was proved in Lemma 6.35. Therefore, to finish
the proof, we will show that condition (a) implies (d).
To this end, fix some large R with R > 2r and use Lemma 1.14 to partition {an }
into N sublattices such that the Euclidean distance between any two points in each
sublattice is at least R. Let {ζn } be such a sublattice and let
∞
ν= ∑ μ χn ,
n=1
where χn is the characteristic function of B(ζn , r). Since Tμ ∈ S p and μ ≥ ν , we

have Tν ≤ Tμ , and so Tν ∈ S p with Tν S p ≤ Tμ S p .
Let {en } be an orthonormal basis for Fα2 and define a linear operator A on Fα2
by Aen = kζn , n ≥ 1, where kζ is the normalized reproducing kernel of Fα2 at ζ .
By the proof of Theorem 2.34, the operator A is bounded. Let T = A∗ Tν A. Then
T S p ≤ A2Tμ S p .
We split the operator T as T = D + E, where D is the diagonal operator defined
on Fα2 by
∞
Df = ∑ Ten , en f , en en ,
n=1
and E = T − D. Since 0 < p ≤ 1, it follows from the triangle inequality that
T Spp ≥ DSpp − ESpp . (6.34)
Also, D is a positive diagonal operator, so

∞ ∞
DSpp = ∑ Ten , en p = ∑ Tν kζn , kζn p (6.35)
n=1 n=1
α p ∞ p
=
π ∑ C
e −α |z−ζn |2
dν (z)
n=1
α p ∞ p
∑ e−α |z−ζn | dν (z)
2
≥
π n=1 B(ζn ,r)
∞
≥ C1 ∑ ν (B(ζn , r)) p .
n=1
On the other hand, by Lemma 6.36, we have

∞ ∞
ESpp ≤ ∑ ∑ |Een, ek | p = ∑ |Tν kζn , kζk | p
n=1 k=1 n=k
α p p

=
π ∑ C kζn (z)kζk (z)e
−α |z|2
dν (z) . (6.36)
n=k
A straightforward calculation shows that

2 α |z−ζn |2 α |z−ζk |2
kζn (z)kζk (z)e−α |z| = e− 2 e− 2 ,
so (6.36) gives us
α p p
α |z−ζn |2 α |z−ζk |2
ESpp ≤
π ∑ C
e− 2 e− 2 dν (z) . (6.37)
n=k
If n = k, then |ζn − ζk | ≥ R. Thus, for |z − ζn | ≤ R2 , the triangle inequality gives us

|z − ζk | ≥ R2 . Therefore, for each z ∈ C, at least one of |z − ζn | ≥ R2 and |z − ζk | ≥ R2
must hold. From this, we deduce that
α |z−ζn |2 α |z−ζk |2 α R2 α |z−ζn |2 α |z−ζk |2

e− 2 e− 2 ≤ e− 16 e− 4 e− 4 .
Plugging this into (6.37), we obtain

α p p
pα R2 α |z−ζn |2 α |z−ζk |2
ESpp ≤
π
e− 16
∑ C
e− 4 e− 4 dν (z) . (6.38)
n=k
Since the measure ν is supported on ∪ j B(ζ j , r), we have

∞
α 2 − α |z−ζ |2 α 2 − α |z−ζ |2
e− 4 |z−ζn | 4 k dν (z) = ∑ e− 4 |z−ζn | 4 k dμ (z)
C j=1 B(ζ j ,r)
∞ α α
∑ e− 4 |z∗ −ζn | − 4 |z∗ −ζk |
2 2
= μ (B(ζ j , r)).
j=1
The last step above follows from the mean value theorem with
z∗ = z∗ (n, k, j) ∈ B(ζ j , r).
Since 0 < p ≤ 1, it follows from Hölder’s inequality that

p ∞
α 2 − α |z−ζ |2 pα pα
e− 4 |z−ζn | ∑ μ (B(ζ j , r)) p e− 4 |z∗ −ζn | − 4 |z∗ −ζk | ,
2 2
4 k dν (z) ≤
C j=1
and so
α p pα
∞ ∞ pα pα
e− 16 R ∑ ∑ μ (B(ζ j , r)) p e− 4 |z∗ −ζn | − 4 |z∗ −ζk |
2 2 2
ESpp ≤
π n,k=1 j=1
α p pα
∞ ∞ pα 2 pα
e− 16 R ∑ μ (B(ζ j , r)) p ∑ e− 4 |z∗ −ζn | − 4 |z∗ −ζk |
2 2
= .
π j=1 n,k=1
If n = j, then |ζ j − ζn | ≥ R > 2r, so by the triangle inequality,

r 1
|z∗ − ζn | ≥ |ζ j − ζn | − r = |ζ j − ζn | 1 − > |ζ j − ζn |.
|ζ j − ζn | 2
This holds trivially for n = j as well. Thus,

α p pα
∞ ∞ pα 2 − pα |ζ −ζ |2
e− 16 R ∑ μ (B(ζ j , r)) p ∑ e− 16 |ζ j −ζn |
2
ESpp ≤ 16 j k
π j=1 n,k=1
2
α p ∞ ∞
− p16α R2 − p16α |ζ j −ζn |2
=
π
e ∑ μ (B(ζ j , r)) ∑ e p
j=1 n=1
2
α p ∞ ∞
− p16α R2 − p16α |ζ j −an |2
≤
π
e ∑ μ (B(ζ j , r)) ∑ e p
j=1 n=1
2
α p ∞ ∞
− p16α R2 − p16α |an |2
=
π
e ∑ μ (B(ζ j , r)) p ∑ e .
j=1 n=1
The last series above is clearly convergent. So we can find a positive constant C2 ,
independent of R, such that
pα
∞
ESpp ≤ C2 e− 16 R ∑ μ (B(ζ j , r)) p .
2
j=1
Going back to (6.34) and (6.35), we deduce that

pα 2
∞
T Spp ≥ DSpp − ESpp ≥ C1 − C2 e− 16 R ∑ μ (B(ζ j , r)) p .
j=1
Since C1 and C2 do not depend on R, setting R > 0 large enough gives us

∞
∑ μ (B(ζ j , r)) p ≤ C3 Tμ Spp ,
j=1
where C3 is another positive constant. Since this holds for each of the N subse-
quences of {an }, we obtain
∞
∑ μ (B(an, r)) p ≤ C3 NTμ Spp (6.39)
n=1
for all positive Borel measures μ such that

∞
∑ μ (B(an, r)) p < ∞.
n=1
Finally, an easy approximation argument shows that (6.39) holds for all positive
Borel measures μ with Tμ ∈ S p . This proves that condition (a) implies (d), and thus
completes the proof of Theorem 6.37.

Again, specializing to the case when
α
dμ (z) = ϕ (z) dA(z),
π
we obtain the following corollary concerning Toeplitz operators induced by non-
negative functions:
Corollary 6.38 Suppose ϕ ≥ 0, 0 0, and ϕ satisfies condition (I1 ).
Then the following conditions are equivalent:
(a) The Toeplitz operator Tϕ belongs to S p .
(b) The Berezin transform ϕ belongs to L p (C, dA).
(c) The averaging function

1
π r2 B(z,r)

(d) The sequence {ϕr (an )} belongs to l p .
6.6 Finite Rank Toeplitz Operators 255
6.6 Finite Rank Toeplitz Operators
In this section, we consider the following problem: when does a Toeplitz operator
Tμ have finite rank on the Fock space Fα2 ? It turns out the problem is pretty tricky.
If μ has compact support in C, we will be able to determine exactly when Tμ
has finite rank. But on the other hand, we will also construct a radial function ϕ ,
not identically zero, such that Tϕ = 0 in a natural way on the Fock space. This is
something unique for the Fock space setting. In particular, in the Fock space setting,
the Berezin transform ϕ → ϕ is not one-to-one if no additional assumptions are
made about ϕ .
Let n be a positive integer and denote by P(Cn ) the algebra of all holomorphic
polynomials on Cn . For any tuple k = (k1 , · · · , kn ) of nonnegative integers, we write
zk = zk11 · · · zknn , |k| = k1 + · · · + kn .
These are the monomials in P(Cn ).

Given a permutation σ on {1, · · · , n}, we write
σ (z) = (zσ (1) , · · · , zσ (n) ), z = (z1 , · · · , zn ) ∈ Cn .
A function f : Cn → C is called symmetric if f (σ (z)) = f (z) for all z ∈ Cn and all

permutations σ on {1, · · · , n}. We say that f : Cn → C is antisymmetric if f (σ (z)) =
sgn(σ ) f (z) for all z ∈ Cn and all permutations σ on {1, · · · , n}.
A set U ⊂ Cn is called permutation-invariant if σ (z) ∈ U for all z ∈ U
and all permutations σ on {1, · · · , n}. Obviously, the notions of symmetric and
antisymmetric functions can also be defined on any permutation-invariant subset
of Cn . In particular, if R is any positive radius, we let CS (R) denote the space of all
symmetric, complex-valued, and continuous functions f on the closed ball B(0, R)
in Cn .
For any complex-valued function f on a permutation-invariant subset U of Cn ,
we can define two functions, called the symmetrization and antisymmetrization of
f , respectively, as follows:
1
n! ∑
fs (z) = f (σ (z)), z ∈ U,
σ
and
1
n! ∑
fa (z) = sgn(σ ) f (σ (z)), z ∈ U,
σ
where the sums are taken over all permutations on {1, · · · , n}.
Let Ps (Cn ) denote the subspace of P(Cn ) consisting of all symmetric poly-
nomials. Similarly, let Pa (Cn ) denote the subspace of P(Cn ) consisting of all
antisymmetric polynomials.
Let P∗ (Cn ) denote the vector space of all conjugate linear functionals on P(Cn ).
If μ is a finite complex Borel measure with compact support in C, then the Toeplitz
operator Tμ is well defined on the dense set P(C) in Fα2 . Furthermore, for any f ∈
P(C), we have Tμ ( f ) ∈ P∗ (C) in the sense that

α
f (z)g(z)e−α |z| dμ (z),
2
Tμ ( f )(g) = g ∈ P(C).
π C
Therefore, when restricted to polynomials, we can think of the Toeplitz operator Tμ

as a mapping from P(C) to P∗ (C). If Tμ : Fα2 → Fα2 has finite rank, then so does
Tμ : P(C) → P∗ (C).
Lemma 6.39. Suppose μ is a finite complex Borel measure on C with compact
support. If Tμ has rank less than n, then
⎛ ⎞
μ ( f1 g1 ) · · · μ ( fn g1 )
⎜ .. .. .. ⎟
det ⎝ . . . ⎠=0 (6.40)
μ ( f1 gn ) · · · μ ( fn gn )
for all complex polynomials fk and gk in P(C). Here,

α
f (z)g(z)e−α |z| dμ (z).
2
μ ( f g) = Tμ ( f )(g) =
π C
Proof. Given one-variable polynomials f1 , · · · , fn , the functionals Tμ ( f1 ), · · · , Tμ ( fn )

are linearly dependent because Tμ has rank less than n. So there are coefficients
c1 , · · · , cn , not all 0, such that
c1 Tμ ( f1 ) + · · · + cn Tμ ( fn ) = 0. (6.41)
If {g1 , · · · , gn } is another collection of polynomials of one complex variable, we

take the inner product of gk with both sides of (6.41) to obtain
⎛ ⎞⎛ ⎞ ⎛ ⎞
μ ( f1 g1 ) · · · μ ( fn g1 ) c1 0
⎜ .. .. .. ⎟ ⎜ .. ⎟ ⎜ .. ⎟
⎝ . . . =
⎠⎝ . ⎠ ⎝ . ⎠.
μ ( f1 gn ) · · · μ ( fn gn ) cn 0
Since the ck ’s are not all 0, we see that the determinant of the matrix above must
be 0.

Lemma 6.40. Suppose μ is a finite complex Borel measure on C with compact
support. If Tμ has rank less than n and
dμn (z1 , · · · , zn ) = e−α (|z1 |

2 +···+|z |2 )
n
dμ (z1 ) · · · d μ (zn )
is the product measure on Cn , then

f g d μn = 0 (6.42)
Cn
for all polynomials f ∈ P(Cn ) and all antisymmetric polynomials g ∈ P(Cn ).

Proof. Since the determinant is linear in each column, we can rephrase (6.40) as
follows:

f1 (z1 ) · · · fn (zn )Δ (g1 , · · · , gn )(z) d μn (z) = 0, (6.43)
Cn
where z = (z1 , · · · , zn ) and

⎛ ⎞
g1 (z1 ) · · · g1 (zn )
⎜ .. ⎟.
Δ (g1 , · · · , gn )(z) = det ⎝ ... ..
. . ⎠
gn (z1 ) · · · gn (zn )
Inserting monomials fk into (6.43) and then taking finite linear combinations, we
see that (6.43) remains valid if the product f1 (z1 ) · · · fn (zn ) is replaced by any
polynomial f ∈ P(Cn ). In other words,

f (z)Δ (g1 , · · · , gn )(z) d μn (z) = 0 (6.44)
Cn
for all f ∈ P(Cn ) and gk ∈ P(C), 1 ≤ k ≤ n.

If each gk is a monomial in P(C), then the function Δ (g1 , · · · , gn )(z) is an
antisymmetric polynomial in P(Cn ). On the other hand, it follows from the
elementary identities
1
n! ∑
[g1 (z1 ) · · · gn (zn )]a = (sgnσ )g1 (zσ (1) ) · · · gn (zσ (n) )
σ
1
= Δ (g1 , · · · , gn )(z)
n!
that any antisymmetric polynomial in P(Cn ) is a finite linear combination of

functions of the form Δ (g1 , · · · , gn )(z). This proves the desired result.

Lemma 6.41. Let K be a permutation invariant compact set in Cn , let Φs denote
the algebra consisting of all finite linear combinations of functions of the form ψϕ ,
where ψ and ϕ are symmetric polynomials in P(Cn ), and let Cs (K) denote the space
of symmetric continuous functions on K. Then Φs is dense in Cs (K) in the sense of
uniform convergence.
Proof. It is clear that Φs is an algebra that contains the constant functions and is
closed under complex conjugation. If it also separated points in K, the desired result
would then follow from the Stone–Weierstrass approximation theorem. But it is

easy to see that Φs does not separate points in K. In fact, if z ∈ K and w = σ (z) =
(zσ (1) , · · · , zσ (n) ), where σ is a permutation not equal to the identity, then z = w but
f (z) = f (w) for all f ∈ Φs .
To overcome this obstacle, we define an equivalence relation ∼ on K as follows:
z ∼ w if and only if w = σ (z) for some permutation σ . Let K = K/∼ be the quotient
space equipped with the standard quotient topology. It is clear that every function
in Cs (K) induces a function in C(K ), the space of complex-valued continuous
functions on the compact Hausdorff space K , and conversely, every function in
C(K ) can be lifted to a function in Cs (K). Also, it is easy to see that Φs separates
points in K . In fact, if the cosets of z = (z1 , · · · , zn ) and w = (w1 , · · · , wn ) are two
different points in K (in other words, if w is not a permutation of z), then the two
one-variable polynomials
n n
p(u) = ∏ (u − zk ), q(u) = ∏ (u − wk ),
k=1 k=1
either have different zeros or they have the same zeros with different multiplicities.
It follows that at least one Taylor coefficient of p differs from the corresponding
coefficient of q. Thus, there exists an elementary symmetric polynomial whose
values at z and w are different.
We can now apply the Stone–Weierstrass approximation theorem to conclude
that every function in Cs (K) can be uniformly approximated by a sequence of
functions in Φs .

Theorem 6.42. Suppose μ is a compactly supported finite complex Borel measure
on C such that the rank of Tμ is less than n, where n is a positive integer. Then μ is
supported on less than n points in C.
Proof. Recall that for z = (z1 , · · · , zn ),
⎛ ⎞
1 1 ··· 1
⎜ z1 z2 · · · zn ⎟
⎜ ⎟
V (z) = det ⎜ .. .. .. .. ⎟ = ∏(zi − z j )
⎝ . . . . ⎠ i> j
zn−1
1 z2 · · · zn−1
n−1
n
is called the Vandermonde determinant, which is an antisymmetric polynomial in

P(Cn ).
Fix a compact set E ⊂ C that contains the support of μ . Suppose the support of
μ contains n distinct points a1 , · · · , an . We will obtain a contradiction. To this end,
we choose a one-variable polynomial p ∈ P(C) such that p(ai ) = p(a j ) for all i = j
and consider the multiple-variable polynomial
Vp (z1 , · · · , zn ) = V (p(z1 ), · · · , p(zn )).
The choice of p ensures that Vp(a1 , · · · , an ) = 0.
It is easy to see that Vp is an antisymmetric polynomial in P(Cn ). Since the

product of a symmetric function and an antisymmetric function is antisymmetric,
an application of Lemma 6.40 to the functions ψ = ψ1Vp and ϕ = ϕ1Vp , where both
ψ1 and ϕ1 are symmetric polynomials in P(Cn ), shows that

F|Vp |2 dμn = 0 (6.45)
Cn
for all F ∈ Φs . Since μn is supported on the permutation invariant compact set E n =

E × · · · × E, it follows from Lemma 6.41 that (6.45) holds for all F ∈ Cs (E n ).
The measure |Vp |2 dμn is permutation invariant, which implies that

F|Vp |2 dμn = Fs |Vp |2 dμn
Cn Cn
for all F ∈ C(E n ), where Fs is the symmetrization of F. Thus, (6.45) holds for all
F ∈ C(E n ). Consequently, |Vp|2 dμn is the zero measure so that the support of μn is
contained in the zero variety of Vp . Since a = (a1 , · · · , an ) is contained in the support
of μn , we must have Vp(a1 , · · · , an ) = 0, which is a contradiction. This shows that μ
is supported on less than n distinct points in C.

Corollary 6.43 Let ϕ be a compactly supported and locally integrable function on
C. Then the Toeplitz operator Tϕ on Fα2 has finite rank if and only if ϕ = 0.
In the rest of this section, we present an example to show that it is necessary to
assume that the measure μ in Theorem 6.42 and ϕ in Corollary 6.43 are compactly
supported. These results will be false without this assumption. To better understand
the intricacy of the problem, we note that if ϕ is bounded, then it follows easily from
the integral representation of the projection Pα and Fubini’s theorem that

Tϕ f , g = ϕ (z) f (z)g(z) dλα (z)
C
for all polynomials f and g. A limit argument then shows that the above also holds
for all functions f and g in Fα2 .
Proposition 6.44. There exists a radial function ϕ , not identically zero, such that
Tϕ = 0 on Fα2 in the sense that

ϕ (z) f (z)g(z) dλα (z) = 0
C
for all polynomials f and g.

Proof. We start with two constants ρ and c satisfying

πi
c = exp (2 − ρ ) , 0 < ρ < 1.
2
Let z±ρ denote the branches given by

π 3π
z±ρ = |z|±ρ e±iρθ , θ∈ − , .
2 2
Define a function f on the closed upper half-plane by f (0) = 0 and

f (z) = exp cz−ρ + czρ , Im (z) ≥ 0, z = 0.
Obviously, f is analytic in the upper half-plane.

For θ ∈ [0, π ], we have
π πρ πρ πρ π
− <− ≤− + ρθ ≤ < .
2 2 2 2 2
Thus, for z = |z|eiθ with θ ∈ [0, π ] and |z| > 0, we have
πρ πρ
0 < cos < cos − + ρθ ≤ 1, (6.46)
2 2
and
πi πi

f (z) = exp |z|−ρ e− 2 (2−ρ )−ρθ i + |z|ρ e 2 (2−ρ )+ρθ i
πρ
= exp −(|z|−ρ + |z|ρ ) cos − + ρθ
2
πρ
+ i(|z|−ρ − |z|ρ ) sin − + ρθ .
2
In particular,
πρ
| f (z)| = exp −(|z|−ρ + |z|ρ ) cos − + ρθ
2
for z = |z|eiθ with θ ∈ [0, π ] and |z| > 0. This together with (6.46) shows that
lim f (z) = 0 = f (0), lim f (z) = 0,

z→0 z→∞
where z is restricted to the closed upper half-plane, so f is continuous on the closed

upper half-plane. Similarly, we can show that
lim f (k) (z) = 0, lim f (k) (z) = 0, (6.47)

z→0 z→∞
where k is any nonnegative integer and z is restricted to the closed upper half-plane.
By the formula for | f (z)|, the restriction of f to the real line belongs to L1 (R, dx).
In particular, the Fourier transform of f is well defined. Let g be the function eα x
2
times the Fourier transform of f , namely,

∞
−α x2
g(x)e = f (t)e−2π itx dt, −∞ < x < ∞.
−∞
Since f is analytic in the upper half-plane and continuous on the closed upper half-
plane, it follows from (6.47) and contour integration around the semicircle |z| = R,
Im (z) ≥ 0, that g(x) = 0 whenever x ∈ (−∞, 0]. So the function g is supported on
(0, ∞).
By the Fourier inversion formula, we can write
∞ ∞
−α t 2 +2π itx
g(t)e−α t
2 +2π itx
f (x) = g(t)e dt = dt
−∞ 0
for −∞ < x < ∞. Differentiating under the integral sign, we obtain

∞
(k)
g(t)t k e−α t dt,
2
f (0) = (2π i) k
k = 0, 1, 2, 3, · · · .
0
Since all derivatives of f vanish at the origin, we have

∞
g(t)t k e−α t dt = 0,
2
k = 0, 1, 2, 3, · · · .
0
Set ϕ (z) = g(|z|). Then ϕ is a radial function, so

ϕ (z)zk zm dλα (z) = 0
C
whenever k = m. On the other hand,

∞
g(r)r2k+1 e−α r dr = 0
2
ϕ (z)zk zk dλα (z) = 2α
C 0
for all k ≥ 0. This shows that

ϕ (z) f (z)g(z) dλα (z) = 0
C
for all polynomials f and g.

6.7 Notes 263
6.7 Notes
The systematic study of Toeplitz operators on the Fock space started in [28, 29],
where several important techniques were introduced that remain useful up to today.
For example, the use of the Berezin transform in function theoretic operator theory
began in [28].
The material in Sect. 6.1 is mostly from [30]. The Bargmann transform between
the Fock space Fα2 and L2 (R, dx) has been a well-known and very useful tool in
analysis. Our presentation in Sect. 6.2 follows Folland’s book [92] closely. Theo-
rems 6.12 and 6.14 are well known in the theory of pseudodifferential operators.
The idea of using the operators T (t) to study trace-class properties of Toeplitz
operators first appeared in [30]. Theorems 6.15–6.18, as well as their compactness
counterparts in Sect. 6.4, are all from [30]. The characterization of bounded and
compact Toeplitz operators with nonnegative symbols is very similar to the Bergman
space setting, and details are worked out in [132].
For Toeplitz operators with bounded symbols, the characterization of compact-
ness in terms of the Berezin transform is also analogous to the Bergman space
setting, which was first obtained by Axler and Zheng in [6] and later generalized
to BMO symbols by Zorborska in [259]. Our presentation here follows [15, 61]
closely.
When 1 ≤ p < ∞, the characterization of Toeplitz operators in the Schatten class
S p of the Fock space Fα2 is relatively easy and follows the Bergman space theory
very closely. However, if 0 < p < 1, there is a critical difference between the Fock
and Bergman space theories. More specifically, in the Bergman space theory, there
is a cutoff point when Schatten class Toeplitz operators are characterized using the
Berezin transform, while the cutoff disappears in the Fock space setting. The proof
of Theorem 6.37 here is simpler than the one first constructed in [132].
Theorem 6.42, the characterization of finite-rank Toeplitz operators induced by
compactly supported measures, is due to Luecking [153]. The proof in [153] is
purely algebraic and works in several different contexts, including Toeplitz opera-
tors on the Bergman space of various domains. The example in Proposition 6.44
was constructed in [105]. Note that Proposition 6.44 does not contradict with
Proposition 3.17 because the function in Proposition 6.44 is far worse than the
functions permitted in Proposition 3.17.
6.8 Exercises 265
6.8 Exercises
1. Let μ be a positive Borel measure on C satisfying condition (M). Then the

following conditions are equivalent:
(a) μ is a vanishing Fock–Carleson measure.
(b) μ − μR → 0 as R → ∞, where μR is the truncation of μ on the disk
B(0, R).
(c) There exists a sequence of finite Borel measures μn , each with compact
support, such that μ − μn → 0 as n → ∞.
2. Suppose p > 1. Show that there exists ϕ ≥ 0 such that Tϕ ∈ S p but ϕ ∈
L p (C, dA).
3. Suppose 0 < p < 1. Show that there exists ϕ ≥ 0 such that ϕ ∈ L p (C, dA) but
Tϕ ∈ S p .
4. Suppose
ϕ (z) = e( 5 + 5 i)|z| .
1 2 2
Show that the Toeplitz operator Tϕ is unitary on the Fock space F12 (α = 1)
and the Berezin transform ϕ vanishes at ∞ and belongs to L p (C, dA) for all
0 < p < ∞.
5. Recall that for any z ∈ C, we have the self-adjoint unitary operator Uz defined
by Uz f (w) = f (z − w)kz (w). Show that if Tϕ is bounded, then

Uz Tϕ Uz dλα (z) = Tψ ,
C
where ψ (w) = ϕ(−w) and the integral converges in the strong operator
topology.
6. If Tϕ is bounded, show that

Wz Tϕ Wz∗ dλα (z) = Tϕ .
C
7. Show that there exist functions ϕ such that ϕ ∈ L∞ (C) but Tϕ is not bounded
on Fα2 .
8. Show that there exist functions ϕ such that ϕ(z) → 0 as z → ∞ but Tϕ is not
compact on Fα2 .
9. Suppose ϕ is radial, that is, ϕ (z) = ϕ (|z|) for all z ∈ C. If ϕ satisfies condition
(I1 ), show that the densely defined Toeplitz operator Tϕ is diagonal with
respect to the standard basis of Fα2 . Characterize boundedness, compactness,
and membership in the Schatten classes for such Toeplitz operators in terms of
the moments of ϕ .
2 )
10. Suppose ϕ (z) = ei|z| . Show that Tϕ is in the trace class, but C |ϕ | dA = ∞.
11. If ϕ is bounded and compactly supported, then Tϕ belongs to S p for all

0 < p < ∞.
12. Show that the set of bounded Toeplitz operators on Fα2 is not norm-dense in the
space of all bounded linear operators on Fα2 . See [30].
13. Show that there exists no positive constant C such that B2α ϕ ∞ ≤ CTϕ for
all ϕ . See [30].
14. Let Tϕα denote the Toeplitz operator defined on Fα2 using the orthogonal
projection Pα : L2α → Fα2 . Show that
α /r2
Tϕαr f (z) = Tϕ f1/r (rz)
for all polynomials f .

15. Show that the operator
Tϕαr : Fα2 → Fα2
is unitarily equivalent to the operator
α /r2
Tϕ : Fα2/r2 → Fα2/r2 .
16. Suppose 1 ≤ p < ∞ and Bβ ϕ ∈ L p (C, dA) for some β > 2α . Then the Toeplitz
operator Tϕ : Fα2 → Fα2 belongs to the Schatten class S p . See [30] and [61].
2
17. Let c be a complex constant and ϕ (z) = ec|z| . Show that Tϕ is bounded on Fα2
∞
if and only if B2α ϕ ∈ L (C), Tϕ is compact on Fα2 if and only if B2α ϕ ∈ C0 (C),
and Tϕ belongs to the Schatten class S p if and only if B2α ϕ ∈ L p (C, dA). See
[30].
18. Define T : Fα2 → Fα2 by T f (z) = f (−z). Show that Tϕ − T ≥ 1 for any
bounded Toeplitz operator Tϕ on Fα2 . See [30].
19. Suppose T is a finite sum of finite products of Toeplitz operators on Fα2 induced
by bounded symbols. Show that T is compact on Fα2 if and only if T ∈ C0 (C).
20. Suppose ϕ (z) = | f (z)|e−σ |z| , where f is entire and σ > 0. Show that Tϕ is
2
bounded on Fα2 if and only if ϕ ∈ L∞ (C), Tϕ is compact on Fα2 if and only if

ϕ ∈ C0 (C), and Tϕ belongs to the Schatten class S p if and only if ϕ ∈ L p (C, dA).
21. Show that Hn (x) = 2xHn−1 (x) − Hn−1 (x) for all n ≥ 1.
Chapter 7
Small Hankel Operators
In this chapter, we study small Hankel operators on the Fock space Fα2 . Problems
considered in the chapter include boundedness, compactness, and membership in
the Schatten class S p . We will also determine when a small Hankel operator has
finite rank.

DOI 10.1007/978-1-4419-8801-0 7,
268 7 Small Hankel Operators
7.1 Small Hankel Operators 269
7.1 Small Hankel Operators
Recall that
P : L2α → Fα2
is the orthogonal projection. Let

2
Fα = f : f ∈ Fα2
2
and use P to denote the orthogonal projection from L2α onto F α .
Suppose ϕ is a function on C that satisfies condition (I1 ). Using the integral
representation for P (and hence P) we can define an operator hϕ on a dense subset
of Fα2 by

hϕ f (z) = P(ϕ f )(z) = K(w, z)ϕ (w) f (w) dλα (w).
C
In fact, as in the definition of Toeplitz operators, the assumption that ϕ satisfy

condition (I1 ) ensures that hϕ f is well defined whenever
n
f (z) = ∑ ck K(z, ak )
k=1
is a finite linear combination of reproducing kernels. The set of all such f is a dense
subspace of Fα2 .
The operator hϕ is traditionally called the small (or little) Hankel operator with
symbol ϕ . We say that hϕ is bounded on Fα2 if there exists a constant C > 0 such
that hϕ ( f )α ≤ C f α whenever f is a finite linear combination of reproducing
kernels. In this case, the domain of hϕ can be extended to the whole space Fα2 .
7.2 Boundedness and Compactness 271
7.2 Boundedness and Compactness
In this section, we determine when the small Hankel operator hϕ is bounded or

compact on the Fock space Fα2 . We will focus on the case when ϕ belongs to L2α . In
this case, we can further assume that ϕ is conjugate analytic. In fact, if ϕ ∈ L2α , then
ϕ satisfies condition (I1 ), and it is easy to check that hϕ = hP(ϕ ) , with P(ϕ ) ∈ Fα2 .
Theorem 7.1. Suppose ϕ ∈ Fα2 . Then, hϕ is bounded on Fα2 if and only if ϕ ∈ Fα∞/2 ,
that is, there exists a constant C > 0 such that
|ϕ (z)| ≤ Ceα |z|

2 /4
, z ∈ C.
Moreover, we always have
1
ϕ Fα∞/2 ≤ hϕ ≤ ϕ Fα∞/2 .
2
Proof. First, suppose that hϕ is bounded on Fα2 . Then there exists a positive constant
C such that
|hϕ f , g| ≤ hϕ f g, f , g ∈ Fα2 ,
or

f (w)g(w)ϕ (w) dλα (w) ≤ hϕ f g, f , g ∈ Fα2 .
C
Let f = g = kz be the normalized reproducing kernels in Fα2 . Then

kz2 (w)ϕ (w) dλα (w) ≤ hϕ , z ∈ C. (7.1)
C
Rewrite this as

−α |z|2

e Ce
(2α z)w
ϕ (w) dλα (w) ≤ hϕ , z ∈ C.
By the reproducing property in Fα2 , the integral above equals ϕ (2z), so
e−α |z| |ϕ (2z)| ≤ hϕ ,

2
z ∈ C.
Replacing z by z/2 shows that ϕ ∈ Fα∞/2 and ϕ Fα∞/2 ≤ hϕ .

Next, we suppose that ϕ ∈ Fα∞/2 so that the function
ψ (w) = 2ϕ (2w)e−α |w|

2
is bounded on C with ψ ∞ = 2ϕ Fα∞/2 . According to the reproducing formula

in F22α ,
z
ϕ (z) = ϕ 2 · = e2α (z/2)wϕ (2w) dλ2α (w)
2 C

= eα zw ϕ (2w) dλ2α (w) = Pα (ψ )(z).
C
Therefore, if f and g are polynomials (which are dense in Fα2 ), then

hϕ f , g = f g ϕ dλα = f g, Pα (ψ ) = f g, ψ = f g ψ dλα .
C C
Thus, by Hölder’s inequality,

|hϕ f , g| ≤ ψ ∞ | f g| dλα ≤ 2ϕ Fα∞/2 f g.
C
This shows that the small Hankel operator hϕ is bounded, and we have the norm
estimate hϕ ≤ 2ϕ Fα∞/2 .

Theorem 7.2. Suppose ϕ ∈ Fα2 . Then hϕ is compact on Fα2 if and only if f ∈ fα∞/2 ,
that is,
lim e−α |z|
2 /4
ϕ (z) = 0. (7.2)
z→∞
Proof. First, assume that ϕ is an entire function that satisfies condition (7.2). Then
there exists a sequence of polynomials {pk } such that
lim pk − ϕ Fα∞/2 = 0.

k→∞
By Theorem 7.1, we have hϕ − h pk → 0 as k → ∞. It is easy to see that each h pk

has finite rank and hence is compact. So hϕ is compact.
On the other hand, if hϕ is compact, then it follows from the proof of Theorem 7.1
that
lim e−α |z| ϕ (2z) = 0,
2
z→∞
because kz → 0 weakly in Fα2 as z → ∞. Replacing z by z/2 shows that condition

(7.2) must hold.

Corollary 7.3. Suppose f is an entire function. Then f = Pα (g) for some g ∈ L∞ (C)
if and only if f ∈ Fα∞/2 . Similarly, f = Pα (g) for some g ∈ C0 (C) if and only if
f ∈ fα∞/2 .
Proof. If f = Pα (g) for some g ∈ L∞ (C), then h f = hg is bounded, so by

Theorem 7.1, f ∈ Fα∞/2 .
If f = Pα (g) for some g ∈ C0 (C), then we can approximate g in L∞ (C) by
a sequence {gk } of functions with compact support in C. Each hgk is obviously
compact and
h f − hgk = hg−gk ≤ g − gk ∞ → 0
as k → ∞. It follows that h f is compact. By Theorem 7.2, f ∈ fα∞/2 .

On the other hand, if we define g(z) = 2 f (2z)e−α |z| , it follows from the proof
2
of Theorem 7.1 that f = Pα (g). If f ∈ Fα∞/2 , then g ∈ L∞ (C). Similarly, if f ∈ fα∞/2 ,

then g is in C0 (C). This completes the proof of the corollary.

7.3 Membership in Schatten Classes 275
7.3 Membership in Schatten Classes
Our next goal is to characterize small Hankel operators induced by entire functions
that belong to the Schatten classes S p . As usual, the cases 1 ≤ p < ∞ and 0 < p < 1
require different treatments. More specifically, we use complex interpolation for the
case 1 ≤ p < ∞, and we use atomic decomposition for the case 0 < p < 1.
Theorem 7.4. Suppose 1 ≤ p ≤ ∞, β = α /2, and ϕ is an entire function satisfying
condition (I1 ). Then hϕ is in the Schatten class S p if and only if ϕ ∈ Fβp .
Proof. By Theorem 7.1, the mapping F : Fβ1 + Fβ∞ → S∞ defined by F(ϕ ) = hϕ is

bounded (and conjugate linear) because Fβ1 is continuously contained in Fβ∞ .
If ϕ ∈ Fβ1 , then it follows from the reproducing formula in Fβ2 that

ϕ (z) = eβ zw ϕ (w) dλβ (w).
C
β
If we write Kw (z) = eβ zw for the reproducing kernel in Fβ2 , then it follows from
Fubini’s theorem that for polynomials f and g we have

hϕ f , g = f (z)g(z)ϕ (z) dλα (z)
C

β
= ϕ (w) dλβ (w) f (z)g(z)Kw (z) dλα (z)
C C

= ϕ (w) h β f , g dλβ (w).
C Kw
In the sense of Banach space valued integrals, we can rewrite the above as

hϕ = ϕ (w) h β dλβ (w). (7.3)
C Kw
It is easy to see that each h β is an operator of rank one, so by Theorem 7.1,

Kw
≤ 2Kwβ Fβ∞ = 2eβ |w|

2 /2
h β S1 = h β .
Kw Kw
Therefore, it follows from (7.3) that

2β β 2
|ϕ (w)|eβ |w|
2 /2
hϕ S1 ≤ 2 dλβ (w) = ϕ (w)e− 2 |w| dA(w).
C π C
This shows that hϕ belongs to the trace-class S1 whenever ϕ is in Fβ1 . On the other
hand, we have already shown in the previous section that hϕ is in S∞ whenever
ϕ ∈ Fβ∞ . An application of complex interpolation then shows that, for 1 ≤ p ≤ ∞,
the small Hankel operator hϕ is in the Schatten class S p whenever ϕ ∈ Fβp .
On the other hand, if the small Hankel operator hϕ belongs to the Schatten class
S p , where 1 ≤ p < ∞, then according to Proposition 3.5 and its proof, the function
Φ (z) = hϕ kz , kz
is in L p (C, dA), where kz are the normalized reproducing kernels of Fα2 . We compute
that

Φ (z) = ϕ (w)kz2 (w) dλα (w)
C

= e−α |z| ϕ (w)e2α zw dλα (w)
2
= e−α |z| ϕ (2z).

2
Obviously, the condition that
e−α |z| ϕ (2z) ∈ L p (C, dA)

2
is equivalent to the condition that
e−α |z|
2 /4
ϕ (z) ∈ L p (C, dA),
which in turn is equivalent to ϕ ∈ Fβp . This completes the proof of the theorem.

Note that if ϕ ∈ Fβ1 , we can also use atomic decomposition to prove that the
operator hϕ is in S1 . See the first part of the proof of the next theorem.
Theorem 7.5. Suppose 0 < p < 1, β = α /2, and ϕ is an entire function satisfying
condition (I1 ). Then hϕ is in the Schatten class S p if and only if ϕ ∈ Fβp .
Proof. First, assume that ϕ ∈ Fβ . By Theorem 2.34, we can write

p
∞
ϕ (z) = ∑ ck ϕk (z),
k=1
where {ck } ∈ l p and

β 2 +β z z
ϕk (z) = e− 2 |zk | k , k ≥ 1.
We may also assume that the sequence {zk } is dense enough to be a sampling
sequence for Fβp . Moreover, there is a constant C > 0, independent of ϕ , such that
∞
∑ |ck | p ≤ Cϕ pp,β .
k=1
It follows that
p
∞ ∞

hϕ Spp = ∑ ck hϕ k ≤ ∑ |ck | p hϕ k Spp .
k=1 k=1
Sp
Each operator hϕ k is a rank-one operator. In fact, if we use K α and K β to denote the

reproducing kernels of Fα2 and Fβ2 , respectively, then for any f ∈ Fα2 , we have
hϕ k f (z) = P(ϕ k f )(z) = ϕ k f , Kzα α

+ , + ,
= e−β |zk | /2 f Kzα , Kzβk = e−β |zk | /2 f Kzα , Kβαzk /α
2 2
α α
−β |zk |2 /2
=e f (β zk /α )Kzα (β zk /α ) = f (zk /2)ϕ k (z)
+ ,
= f , Kzαk /2 ϕ k (z).
α
Therefore,
hϕ k S p = hϕ k ≤ Kzαk /2 2,α ϕk 2,α = 1,
and so
∞
hϕ Spp ≤ ∑ |ck | p ≤ Cϕ pp,β .
k=1
On the other hand, if hϕ is in S p , we are going to show that ϕ ∈ Fβp . To this end,
we fix a square lattice Z = {zk } in C such that atomic decomposition holds on Z
for both Fβp and Fα2 . We also assume that 2Z is a sampling sequence for Fβp . Fix a
sufficiently large R and use Lemma 1.14 to decompose Z = Z1 ∪ · · · ∪ ZN into N
square lattices such that for each 1 ≤ k ≤ N and each pair {w1 , w2 } of distinct points
in Zk , we have |w1 − w2 | > R.
Fix an orthonormal basis {ek } for Fα2 and define an operator A on Fα2 as follows:

∞ ∞ α
∑ ck ek ∑ ck eα zzk − 2 |zk | .
2
A (z) =
k=1 k=1
By the atomic decomposition for Fα2 , the operator A is bounded and onto. Clearly,
we have A = A1 + · · · + AN , where

∞
α
∑ ck ek ∑ ck eα zzk − 2 |zk |
2
Aj (z) =
k=1 zk ∈Z j
for 1 ≤ j ≤ N. Each operator A j is also bounded on Fα2 .

We also consider the companion operators
B j : Fα2 → Fα2
defined by
B j( f ) = A j f , f ∈ Fα2 , 1 ≤ j ≤ N.
Since hϕ is in S p , so is the operator T = T1 + · · · + TN , where
T j = B∗j hϕ A j , 1 ≤ j ≤ N.
Write T = D + E, where D is diagonal with respect to the basis {ek } and satisfies
Dek , ek α = Tek , ek α for all k ≥ 1. If we write
α
fk (z) = eα zzk − 2 |zk | ,
2
then
∞ ∞
DSpp = ∑ |Dek , ek | p = ∑ |Tek , ek | p
k=1 k=1
∞ ∞ p

∑ |hϕ fk , f k | p = ∑ ϕ (2zk )e−α |zk |
2
=
k=1 k=1
≥ Cϕ pp,β ,
where C is a positive constant independent of ϕ . Note that the last inequality above
follows from the assumption that {2zk } is a sampling sequence for Fβp .
On the other hand, since 0 < p < 1, it follows from Lemma 6.36 that
ES p ≤ ∑ |Eek , el | p = ∑ |Tek , el | p

p
k,l k=l
N
= ∑ ∑ |hϕ A j ek , A j el | p.
j=1 k=l
Since
hϕ A j ek , A j el = 0
unless both zk and zl are in Z j , we see that
N
ESpp ≤ ∑∑ |hϕ fk , f l | p : k = l, zk ∈ Z j , zl ∈ Z j .
j=1
If ϕ is already in Fβp , we can write
∞
ϕ (z) = ∑ ci ϕi (z),
i=1
where
β
ϕi (z) = eβ zzi − 2 |zi |
2
and
∞
∑ |ci | p ≤ Cϕ βp ,p.
i=1
Here, C is a positive constant independent of ϕ . By Hölder’s inequality,
∞ N
ESpp ≤ ∑ |ci | p ∑ ∑ |hϕ i fk , f l | p : k = l, zk ∈ Z j , zl ∈ Z j .
i=1 j=1
|hϕ i fk , f l α | = e−β |zk −(zi /2)|

2 −β |z
l −(zi /2)|
2
.
Therefore,
∞
ESpp ≤ ∑ |ci | p ∑ e−pβ |zk −(zi /2)|
2 −pβ |z
l −(zi /2)|
2
.
i=1 |zk −zl |≥R
If 2δ is the separation constant for the sequence Z, then by Lemma 2.32, there
exists a positive constant C = C(δ , α , p) such that

−pβ [|zk − 2i |2 +|zl − 2i |2 ] e−pβ [|z|2 +|w|2 ] dA(z)dA(w).
z z
e ≤C
z z
B zk − 2i ,δ ×B zl − 2i ,δ
If (k, l) = (k , l ), then
zi zi zi zi
B zk − , δ × B zl − , δ ∩ B zk − , δ × B zl − , δ = 0.
/
2 2 2 2
Also,
zi zi
B zk − , δ × B zl − , δ ⊂ (z, w) ∈ C2 : |z − w| ≥ R − 2δ .
2 2
It follows that there exists a positive constant C, independent of large R, such that
∑ e−pβ |zk −(zi /2)|

2 −pβ |z
l −(zi /2)|
2
|zk −zl |≥R

e−pβ (|z|
2 +|w|2 )
C dA(z) dA(w).
|z−w|≥R−2δ
The above double integral tends to 0 as R → ∞. This, along with the fact that

DSpp ≤ 2 p T Spp + ESpp ,
shows that we can find a positive constant σ such that
σ ϕ p,β ≤ hϕ S p , (7.4)
where ϕ ∈ Fβp and σ is independent of ϕ .

The inequality in (7.4) was proved under the assumption that ϕ is already in Fβp .
The general case then follows from an easy approximation argument. In fact, if ϕ
is any entire function such that hϕ is in S p , then by Theorem 7.1, ϕ must be in Fβ∞ .
We consider the functions ϕr , 0 < r < 1, defined by ϕr (z) = ϕ (rz). Each ϕr ∈ Fβp ,
so by (7.4),
σ ϕr p,β ≤ hϕ r S p ≤ hϕ S p , 0 < r < 1.
Let r → 1. We obtain
σ ϕ p,β ≤ hϕ S p < ∞.

7.4 Finite Rank Small Hankel Operators 281
7.4 Finite Rank Small Hankel Operators
In this section, we characterize small Hankel operators on Fα2 whose range is finite
dimensional. Such operators are called finite rank operators.
We begin with an example. Suppose ϕ (z) = K(a, z) for some point a ∈ C. Then
for any function f ∈ Fα2 , we have

hϕ ( f )(z) = P(ϕ f )(z) = K(w, z)K(a, w) f (w) dλα (w)
C
= f (a)K(a, z).
So, in this case, the range of hϕ is the one-dimensional subspace spanned by the
function z → K(a, z). More generally, if
N
∂k
ϕ (z) = ∑ ck ∂ ak K(a, z)
k=0
for some point a ∈ C and some nonnegative integer N, then

N
∂k
hϕ ( f )(z) = ∑ ck ∂ ak C
K(w, z)K(a, w) f (w) dλα (w)
k=0
N
∂k
= ∑ ck ∂ ak [ f (a)K(a, z)] ,
k=0
which shows that hϕ is a finite rank operator whose range is spanned by the
following functions of z:
∂k
K(a, z), 0 ≤ k ≤ N.
∂ ak
We are going to show that these are essentially all the finite rank small Hankel
operators on Fα2 . But we first need the following elementary result from algebra.
Lemma 7.6. Let P(C) denote the ring of all complex polynomials of the variable z.
If J is an ideal in P(C) containing at least one nonzero polynomial, then there
are a finite number of complex numbers ak , 1 ≤ k ≤ N, and for each k, there
exists a nonnegative integer Nk , such that J consists of all polynomials ϕ with
the property that
ϕ (i) (ak ) = 0, 1 ≤ k ≤ N, 0 ≤ i ≤ Nk .
Proof. By a well-known fact in abstract algebra (see [146] for example), every ideal
J = (0) of P(C) is generated by a polynomial, that is, there exists a polynomial q
such that J = {pq : p ∈ P(C)}. If a1 , · · · , aN are the zeros of q, and each zero ak has
multiplicity 1 + Nk , then J has the desired form.

Theorem 7.7. A bounded small Hankel operator has finite rank if and only if it can
be written as hϕ , where
N Nk
ϕ (z) = ∑ ∑ cki ϕki (z). (7.5)
k=1 i=0
Here, ϕki (z) denotes the function
∂i
K(a, z)
∂ ai
evaluated at the point a = ak .
Proof. We have already proved that hϕ has finite rank if ϕ is given by (7.5).
To prove the other direction, we write the small Hankel operator as hϕ , where ϕ
is conjugate analytic. If hϕ has finite rank, then the restriction of hϕ on P(C) also
has finite rank. Consider the kernel of hϕ on P(C):

J= f ∈ P(C) : hϕ ( f ) = 0 .
It is easy to check that J is an ideal in P(C). In fact, if hϕ ( f ) = 0, then ϕ f , g = 0 for

all polynomials g (which are dense in Fα2 ). If p is any polynomial, then ϕ f , pg = 0
for all polynomials g. This can be rewritten as ϕ p f , g = 0 for all polynomials g,
which shows that hϕ (p f ) = 0 as well.
By Lemma 7.6, there exist points ak ∈ C, 1 ≤ k ≤ N, and for each k, there exists
a nonnegative integer Nk , such that

J = f ∈ P(C) : f (i) (ak ) = 0, 1 ≤ k ≤ N, 0 ≤ i ≤ Nk .
In other words, J is the intersection of the kernels of finitely many linear functionals
on P(C).
Let g = ϕ ∈ Fα2 . Then the linear functional on P(C) defined by
f → f , g = hϕ ( f ), 1
vanishes on J. Combining this with the conclusion from the previous paragraph, we
can find constants cki such that
5 6
N Nk N Nk
∂i
f , g = ∑ ∑ cki f (i)
(ak ) = f , ∑ ∑ cki i K(·, ak )
∂a
k=1 i=0 k=1 i=0
for all polynomials f . This shows that
N Nk
∂i
ϕ (z) = g(z) = ∑ ∑ cki ∂ ai K(ak , z),
k=1 i=0

7.5 Notes 283
7.5 Notes
Small Hankel operators on the Fock space were first studied in [138], where the
boundedness, compactness, and membership in Schatten classes S p for 1 ≤ p < ∞
were characterized. The case when 0 < p < 1 was taken up and settled in [231]. Our
presentation here follows [138] and [231] very closely.
7.6 Exercises 285
7.6 Exercises
1. For a symbol function ϕ , define a conjugate linear operator

hϕ : Fα2 → Fα2
by
hϕ ( f ) = P(ϕ f ). Show that hϕ is bounded if and only if
hϕ is bounded, hϕ is

compact if and only if hϕ is compact, and hϕ is in the Schatten class S p if and
only if
hϕ is in the Schatten class S p .
2. Suppose ϕ is an entire function. Define a bilinear form
Φ : Fα2 × Fα2 → C
by

Φ ( f , g) = ϕ f , gα = ϕ f g dλα .
C
Show that hϕ is bounded on Fα2 if and only if there exists a constant C > 0 such
that |Φ ( f , g)| ≤ C f 2,α g2,α for all f and g in Fα2 .
3. Formulate conditions for compactness and membership in Schatten classes for
hϕ on Fα2 in terms of the bilinear form Φ in the previous problem, where ϕ is
any entire function.
4. Suppose ϕ ∈ L2α . Show that hϕ = 0 if and only if ϕ ⊥ Fα2 .
5. Consider the integral transform

Vϕ (z) = hϕ kz , kz α = ϕ (w)kz (w)2 dλα (w).
C
Show that hϕ is bounded if and only if Vϕ is bounded, hϕ is compact if and

only if Vϕ ∈ C0 (C), and hϕ belongs to the Schatten class S p if and only if
Vϕ ∈ L p (C, dA).
6. If ϕ is entire, show that
Vϕ (z) = e−α |z| ϕ (2z)

2
for all z ∈ C.
7. If ϕ ∈ L p (C, dλα ) for some 1 < p < ∞, then ϕ satisfies condition (I1 ). In
particular, every function in Fα2 satisfies condition (I1 ).
8. Show that if ϕ satisfies condition (I1 ) with respect to the weight parameter
β = 3α /4, then Pα (ϕ ) satisfies condition (I1 ) with respect to the weight
parameter α .
9. Show that Theorems 7.1 and 7.2 remain valid with the weaker assumption that
ϕ is entire and satisfies condition (I1 ).
10. Verify directly that hϕ has finite rank when ϕ is a polynomial.
11. Show that hϕr S p ≤ hϕ S p for all 0 < r < 1.
Chapter 8
Hankel Operators
In this chapter, we study (big) Hankel operators Hϕ on the Fock space Fα2 .
Problems considered include, again, boundedness, compactness, and membership
in the Schatten classes. There are basically two theories here: one concerns the
simultaneous size estimates for both Hϕ and Hϕ , and one concerns the size estimates
for the single operator Hϕ . The former is similar to the situations in the more
classical Hardy and Bergman space settings, while the latter is unique to the Fock
space setting.

DOI 10.1007/978-1-4419-8801-0 8,
288 8 Hankel Operators
8.1 Boundedness and Compactness
Suppose ϕ ∈ L∞ (C). We can then define an operator Hϕ on Fα2 by
Hϕ ( f ) = (I − P)(ϕ f ),
where I is the identity operator on L2α and
P : L2α → Fα2
is the orthogonal projection. It is obvious that Hϕ is a bounded linear operator from

Fα2 into L2α Fα2 and Hϕ ≤ ϕ ∞ . We call Hϕ the (big) Hankel operator with
symbol ϕ . By the integral representation of the projection P, we have
Hϕ ( f )(z) = ϕ (z) f (z) − P(ϕ f )(z)

= (ϕ (z) − ϕ (w))K(z, w) f (w) dλα (w)
C
for all f ∈ Fα2 and z ∈ C.

Using this integral representation for Hankel operators with bounded symbols,
we can extend the definition of Hϕ to the case in which ϕ is not necessarily
bounded. In particular, if ϕ satisfies condition (I1 ), then Hϕ ( f ) will always be
defined whenever f is a finite linear combination of reproducing kernels in Fα2 .
A natural question arises: for which symbol functions ϕ is the Hankel operator Hϕ
bounded?
In this section, we answer the above question when ϕ is real-valued. Equiva-
lently, we characterize those symbol functions ϕ such that both Hϕ and Hϕ are
bounded. A similar characterization will be given for the simultaneous compactness
of Hϕ and Hϕ .
We begin with Hankel operators induced by symbol functions that are Lipschitz
in the Euclidean metric.
Lemma 8.1. If there exists a positive constant C such that
|ϕ (z) − ϕ (w)| ≤ C|z − w|
all complex numbers z and w. Then ϕ satisfies condition (I1 ) and Hϕ ≤
for
2π /α C.
Proof. It is easy to check that any Lipschitz function satisfies condition (I1 ) and
hence induces a well-defined Hankel operator. To estimate the norm of Hϕ , consider
the integrals

I(z) = |z − w||K(z, w)|K(w, w)|1/2 dλα (w), z ∈ C.
C
By a change of variables,

α α 2
I(z) = |z − w| eα zw− 2 |w| dA(w)
π C

α α |z|2 α
|z − w|e− 2 |z−w| dA(w)
2
= e 2
π C

α α 2 α
= e 2 |z| |w|e− 2 |w| dA(w)
2
π C

2π α |z|2
= e2 .
α
Thus,

2π
|ϕ (z) − ϕ (w)||K(z, w)|K(w, w) 1/2
dλα (w) ≤ CK(z, z)1/2
C α
for all z ∈ C. The desired norm estimate then follows from Lemma 2.14 and the
integral representation of the Hankel operator Hϕ .

Recall that for any a ∈ C, we define a unitary operator Ua on L2α by Ua f =
f ◦ ϕa ka , where ϕa (z) = a − z and ka is the normalized reproducing kernel of Fα2 at
a. It is easy to check that Ua2 = I, so Ua∗ = Ua−1 = Ua . Since Ua leaves the Fock space
Fα2 invariant, we have Ua P = PUa .
Lemma 8.2. Suppose f satisfies condition (I2 ). Then the operators T f and H f are
both densely defined on Fα2 . Moreover, we have
T f kz = Uz P( f ◦ ϕz ) = P( f ◦ ϕz ) ◦ ϕzkz (8.1)
and
H f kz = Uz (I − P)( f ◦ ϕz ) = [ f − P( f ◦ ϕz ) ◦ ϕz ]kz (8.2)
for all z ∈ C.
Proof. Since each Uz commutes with the projection P, we have
T f kz = P( f kz ) = PUz ( f ◦ ϕz ) = Uz P( f ◦ ϕz ).
This proves the desired results.

Proposition 8.3. Suppose f satisfies condition (I2 ). Then

max H f kz , H f kz ≤ MO( f )(z) ≤ H f kz + H f kz
$
for all z ∈ C, where MO( f ) = |
f |2 − | f|2 .
Proof. Since each kz is a unit vector, it follows from the Cauchy–Schwarz inequality
that
MO( f )2 (z) = f kz 2 − | f kz , kz |2

= f kz 2 − |P( f kz ), kz |2
≥ f kz 2 − P( f kz )2
= (I − P)( f kz )2 = H f kz 2 .
Replacing f by f , we also have MO( f )(z) ≥ H f kz . Thus,

MO( f )(z) ≥ max H f kz , H f kz .
On the other hand, it follows from Lemma 8.2 that
H f kz = Uz (I − P)( f ◦ ϕz ) = (I − P)( f ◦ ϕz )

= f ◦ ϕz − P( f ◦ ϕz ).
Similarly, we have
H f kz = f ◦ ϕz − P( f ◦ ϕz ) = f ◦ ϕz − P( f ◦ ϕz ).
Since f(z) = P( f ◦ ϕz )(0) and Pg(z) = g(0) whenever g ∈ Fα2 , we have
MO( f )(z) = f ◦ ϕz − P( f ◦ ϕz )(0)

≤ f ◦ ϕz − P( f ◦ ϕz ) + P( f ◦ ϕz ) − P( f ◦ ϕz )(0))
= H f kz + P( f ◦ ϕz ) − P( f ◦ ϕz )(0)
= H f kz + P[ f ◦ ϕz − P( f ◦ ϕz )]
≤ H f kz + f ◦ ϕz − P( f ◦ ϕz )
= H f kz + f ◦ ϕz − P( f ◦ ϕz )
= H f kz + H f kz .
This completes the proof of the proposition.

We can now prove the main result of this section.
Theorem 8.4. Suppose ϕ satisfies condition (I2 ). Then the following two conditions
are equivalent:
1. Both Hϕ and Hϕ are bounded on Fα2 .
2. The function ϕ belongs to BMO2 .
Proof. First, assume that ϕ ∈ BMO2 . Then, by Corollary 3.37, we can write ϕ =
ϕ1 + ϕ2 , where the function ϕ1 satisfies the Lipschitz estimate
|ϕ1 (z) − ϕ1 (w)| ≤ C|z − w|
and the Toeplitz operator T|ϕ2 |2 is bounded. By Lemma 8.1, the Hankel operator Hϕ1
is bounded on Fα2 . On the other hand, it follows from the identity
Hϕ∗2 Hϕ2 = T|ϕ2 |2 − Tϕ 2 Tϕ2
and the boundedness of T|ϕ2 |2 that Hϕ2 is also bounded on Fα2 . Therefore, Hϕ is
bounded. Since BMO2 is closed under complex conjugation, the assumption ϕ ∈
BMO2 implies that ϕ is also in BMO2 so that Hϕ is bounded on Fα2 as well.
Next, assume that both Hϕ and Hϕ are bounded on Fα2 . Then, it follows from the
inequality (see Proposition 8.3)
1/2
MO(ϕ )(z) = | ϕ |2 (z) − |ϕ(z)|2 ≤ Hϕ kz + Hϕ kz
that the function ϕ is in BMO2 .

A companion result for the simultaneous compactness of Hϕ and Hϕ is the
following:
Theorem 8.5. Suppose ϕ satisfies condition (I2 ). Then the following two conditions
are equivalent:
1. Both Hϕ and Hϕ are compact on Fα2 .
2. The function ϕ belongs to VMO2 .
Proof. If ϕ ∈ VMO2 , then ϕ − ϕr BMO2 → 0 as r → ∞, where ϕr is ϕ times the

characteristic function of the Euclidean ball B(0, r). It is easy to see that both Hϕr
and Hϕ r are compact on Fα2 . Since
Hϕ − Hϕr + Hϕ − Hϕ r ∼ ϕ − ϕr BMO2 ,
we conclude that both Hϕ and Hϕ can be approximated by compact operators in the

norm topology and so must be compact themselves.
Conversely, if Hϕ and Hϕ are both compact on Fα2 , then it follows from the second
inequality in Proposition 8.3 that ϕ is in VMO2 , as the normalized reproducing
kernels kz tend to 0 weakly in Fα2 .

Corollary 8.6. If ϕ is entire, then Hϕ is bounded if and only if ϕ is a linear
polynomial and Hϕ is compact if and only if ϕ is constant.
8.2 Compact Hankel Operators with Bounded Symbols 293
8.2 Compact Hankel Operators with Bounded Symbols
The purpose of this section is to show that for bounded symbol functions ϕ , the
Hankel operator Hϕ is compact on Fα2 if and only if Hϕ is compact on Fα2 . This
striking result probably reflects the lack of bounded analytic functions (except the
constants) in the complex plane, as the direct analogs for Hankel operators on the
more classical Hardy and Bergman spaces are false.
Lemma 8.7. If f ∈ L∞ (C), then |P f (z)| ≤ f ∞ eα |z|
2 /4
for all z ∈ C.
Proof. This follows directly from Corollary 2.5.

Lemma 8.8. Suppose F(w, z) is a nonnegative measurable function on C × C with

the property that there is a constant B > 0 such that
α
F(w, z) ≤ Be 4 |z| ,
2
z, w ∈ C. (8.3)
Then there exists another positive constant C such that
1
4
α zw+ 12 α |z|2 1 α |w|2
F(w, ϕw (z))|e | dλα (z) ≤ Ce 2 F(w, z) dλα (z)
2
C C
for all w ∈ C.
Proof. We make an obvious change of variables to rewrite the integral on the left-
hand side as

α α |w|2 α
F(w, u) e− 2 |u| dA(u).
2
e2
π C
Denote the integral above by I, apply Hölder’s inequality with exponents 4 and 4/3,
and use the assumption in (8.3). We obtain

3α |u|2 α
F(w, u)e− e− 8 |u| dA(u)
2
I= 8
C
1 3
4 4
4 − 32α |u|2 − α6 |u|2
≤ F(w, u) e dA(u) e dA(u)
C C
1
α 4
F(w, u)2 e−α |u| dA(u)
2
≤C
π C
1
4
=C F(w, z)2 dλα (z) .
C

Lemma 8.9. Suppose f ∈ L∞ (C). For any z ∈ C, we have

1 1
|P( f ◦ ϕw )(ϕw (z))||Kw (z)|Kw (w) 2 dλα (w) ≤ 4 f ∞ Kz (z) 2 ,
C
and

1 1
| f (z) − P( f ◦ ϕw )(ϕw (z))|Kw (z)|Kw (w) 2 dλα (w) ≤ 6 f ∞ Kz (z) 2 .
C
Proof. It follows from (8.1) that
|P( f ◦ ϕw )(ϕw (z))||Kw (z)| = |P( f Kw )(z)|

= f (u)Kw (u)K(z, u) dλα (u)
C

≤ f ∞ |K(u, w)||K(z, u)| dλα (u)
C
α
= f ∞ e 4 |z+w| .
2
Thus the integral

|P( f ◦ ϕw )(ϕw (z))||eα zw+ 2 α |w| | dλα (w)
1 2
I=
C
satisfies the following estimates:

α 2 + α |w|2
I ≤ f ∞ e 4 |z+w| 2 dλα (w)
C

α α 2 α4 zw 2 − α4 |w|2
= f ∞ e 4 |z| e e dA(w)
π C

α α4 zw 2 α
= 4 f ∞ e 4 |z|
2
e dλ 4 (w)
C
α |z|2 α |z|2 α
= 4 f ∞ e 2 |z| .
2
= 4 f ∞ e 4 e 4
This proves the first estimate. The second estimate follows from the triangle
inequality, the first estimate, and the fact that

1 1
|Kw (z)|Kw (w) 2 dλα (w) = 2Kz (z) 2 .

C
Theorem 8.10. Suppose f ∈ L∞ (C). Then

(a) T f is compact if and only if P( f ◦ ϕa ) → 0 as a → ∞.
(b) H f is compact if and only if f ◦ ϕa − P( f ◦ ϕa ) → 0 as a → ∞.
Proof. The proof of part (a) is exactly the same as the proof of part (b). The only
difference is in the projection that is used in the definitions of Toeplitz operators and
Hankel operators: T f = PM f and H f = (I − P)M f . Therefore, we use Q to denote
either P or I − P in the rest of the proof.
Since ka → 0 weakly in Fα2 as a → ∞, the compactness of
QM f : Fα2 → Q(L2α )
implies that QM f (ka ) → 0 in norm as a → ∞. By Lemma 8.2,
QM f (ka )2 = Q( f ◦ ϕa )2 , a ∈ C.
Thus the compactness of QM f implies Q( f ◦ ϕa ) → 0 as a → ∞.

Next, we assume that Q( f ◦ ϕa ) → 0 as a → ∞ and proceed to show that the
operator QM f is compact. Obviously, it is equivalent for us to show that the operator
(QM f )∗ : Q(L2α ) → Fα2 ⊂ L2α
is compact.
Given h ∈ Q(L2α ) and w ∈ C, we use Lemma 8.2 to write
(QM f )∗ h(w) = (QM f )∗ h, Kw = h, QM f Kw

= h, Q( f ◦ ϕw ) ◦ ϕw Kw

= h(z)Q( f ◦ ϕw )(ϕw (z))Kw (z) dλα (z).
C
For each positive number R, define an operator
SR : Q(L2α ) → L2α
by
SR h(w) = χR (w)(QM f )∗ h(w), w ∈ C,
where χR is the characteristic function of the ball {u ∈ C : |u| ≤ R}.

By Fubini’s theorem and a change of variables,

χR (w)|Q( f ◦ ϕw )(ϕw (z))|2 |Kw (z)|2 dλα (z)dλα (w)
C C

= Kw (w)Q( f ◦ ϕw )2 dλα (w)
|w|≤R

α
= QM f kw 2 dA(w)
π |w|≤R
≤ α R2 QM f 2 < ∞.
It follows that the operator SR is Hilbert–Schmidt. In particular, SR is compact.

We write

(QM f )∗ − SR g(w) = H(w, z)g(z) dλα (z), g ∈ Q(L2α ),
C
where
H(w, z) = (1 − χR(w))Q( f ◦ ϕw )(ϕw (z))Kw (z).
∗
test to obtain an estimate on the norm of (QM f ) −SR .
We are going to apply Schur’s
To this end, we let h(z) = K(z, z). It follows from Lemma 8.9 that

|H(w, z)|h(w) dλα (w) ≤ 6 f ∞ h(z)
C
for all z ∈ C. On the other hand, if we write
F(w, z) = (1 − χR(w))|Q( f ◦ ϕw )(z)|,
then by Lemma 8.7,

α
F(w, z) ≤ 2 f ∞ e 4 |z| ,
2
so we can apply Lemma 8.8. In fact, since
|H(w, z)| = F(w, ϕw (z))|Kw (z)|,
an application of Lemma 8.8 tells us that there exists a positive constant C,

depending on f only, such that
1
4
|H(w, z)|h(z) dλα (z) ≤ Ch(w) F(w, z)2 dλα (z)
C C
1
= Ch(w)(1 − χR(w))Q( f ◦ ϕw ) 2 .
By Schur’s test, there exists a positive constant C such that

(QM f )∗ − SR ≤ C sup Q( f ◦ ϕw )1/4 : |w| > R .
This shows that the condition

lim Q( f ◦ ϕa ) = 0
a→∞
implies that
lim (QM f )∗ − SR = 0.
R→∞
In other words, (QM f )∗ can be approximated in norm by compact operators, and so

it must be compact as well. This completes the proof of the theorem.

Lemma 8.11. For any f ∈ L∞ (C), there exists a positive constant C such that
f◦ ϕa − P( f ◦ ϕa ) ≤ C f ◦ ϕa − P( f ◦ ϕa ) 4
1
for all a ∈ C.
Proof. It follows from Corollary 2.5 that
α
| f(w) − P f (w)| ≤ 2 f ∞ e 4 |w|
2
(8.4)
for all w ∈ C. Since the Berezin transform fixes entire functions, we have

f(w) − P f (w) = ( f (z) − P f (z))|kw (z)|2 dλα (z)
C
so that

| f(w) − P f (w)| ≤ e−α |w|
2
| f (z) − P f (z)||Kw (z)|2 dλα (z) (8.5)
C
for all w ∈ C. By (8.4),

α
f− P f 2 = | f(w) − P f (w)|2 e−α |w| dA(w)
2
π C

2α
f ∞ | f(w) − P f (w)|e− 4 α |w| dA(w).
3 2
≤
π C
Using (8.5), Fubini’s theorem, and Corollary 2.5, we arrive at

8
f− P f 2 = f ∞ | f (z) − P f (z)|e 7 α |z| dλα (z)
4 2
7 C

8α
f ∞ | f (z) − P f (z)|e− 8 α |z| e− 56 α |z| dA(z).
3 2 3 2
=
7π C
Applying Hölder’s inequality (with exponents 4 and 4/3) and Lemma 8.7, we obtain
1
4
f− P f 2 ≤ C1 f ∞ | f (z) − P f (z)| e4 − 23 α |z|2
dA(z)
C
1
3 4
≤ C2 f ∞2 | f (z) − P f (z)|2 dλα (z)
C
3
= C2 f ∞2 f − P f 1/2 .
This shows that

3
f− P f ≤ C f ∞4 f − P f 1/4,
where the constant C is independent of f . Replacing f by f ◦ ϕa and using the

translation invariance of the Berezin transform, we obtain the desired estimate.

Lemma 8.12. If f ∈ L∞ (C) and H f is compact, then both H f and T f − f are compact.
Proof. By Theorem 8.10,
lim f ◦ ϕa − P( f ◦ ϕa ) = 0,
a→∞
which, according to Lemma 8.11, implies that
lim f◦ ϕa − P( f ◦ ϕa ) = 0.
a→∞
Since the projection P is bounded on L2α , we also have
lim P( f◦ ϕa ) − P( f ◦ ϕa ) = 0.

a→∞
By part (a) of Theorem 8.10, the Toeplitz operator T f − f is compact. Since
f◦ ϕa − P( f◦ ϕa ) ≤ f◦ ϕa − P( f ◦ ϕa ) + P( f ◦ ϕa ) − P( f◦ ϕa ),
we see that
lim f◦ ϕa − P( f◦ ϕa ) = 0,
a→∞
which, in view of part (b) of Theorem 8.10, shows that H f is compact.

Theorem 8.13. Suppose f ∈ L∞ (C). Then H f is compact if and only if H f is
compact.
Proof. Let g = f and assume that Hg is compact. By Theorem 8.10,
lim g ◦ ϕa − P(g ◦ ϕa) = 0.

a→∞
Combining this with Lemma 8.11, we see that
g ◦ ϕa − P(g ◦ ϕa) = 0,
lim
a→∞
and so by the triangle inequality,
lim g ◦ ϕa − g ◦ ϕa = 0.
a→∞
Since complex conjugation commutes with the Berezin transform, we also have
lim f ◦ ϕa −
f ◦ ϕa = 0,
a→∞
which implies that H f − f is compact. Using Lemma 8.12 and iteration, we conclude
that H f − f(m) is compact for every positive integer m.
On the other hand, Theorem 3.25 shows that f ∈ L∞ (C) implies
C
| f(m) (z) − f(m) (w)| ≤ √ |z − w|,
m
which, along with Lemma 8.1, shows that H f(m) → 0 as m → ∞. This, combined
with the fact that each H f − f(m) is compact, shows that H f is compact.

8.3 Membership in Schatten Classes
In this section, we characterize when the Hankel operators H f and H f belong to the
Schatten class S p simultaneously. Throughout the section, we fix a positive radius r
and write
1
MOr ( f )(z) = |
f |2 r (z) − | fr (z)|2 ,
2
and
1
MO( f )(z) = |
f |2 (z) − | f(z)|2 .
2
Lemma 8.14. Let 2 ≤ p < ∞. If H f and H f are both in the Schatten class S p , then
MO( f ) ∈ L p (C, dA).
Proof. If H f is in S p , then (H ∗f H f ) p/2 is in the trace class S1 , so by Proposition 3.3,

(H ∗f H f ) p/2 kz , kz dA(z) < ∞,
C
where kz are the normalized reproducing kernels of Fα2 . By Lemma 3.4,

H ∗f H f kz , kz p/2 dA(z) < ∞,
C
or

H f kz p dA(z) < ∞.
C
Similarly, if H f is in S p , then

H f kz p dA(z) < ∞.
C
The desired result then follows from Proposition 8.3.

Lemma 8.15. Let 0 < p ≤ 2. If MO( f ) ∈ L p (C, dA), then both H f and H f are in
the Schatten class S p .
Proof. By Proposition 8.3, the condition MO( f ) ∈ L p (C, dA) implies that the func-
tion z → H f kz is in L p (C, dA). This, along with Proposition 3.3 and Lemma 3.4,
shows that
α
∗
tr (H f H f ) p/2
= (H ∗f H f ) p/2 kz , kz dA(z)
π C

α
≤ H ∗ H f kz , kz p/2 dA(z)
π C f

α
= H f kz p dA(z) < ∞.
π C
Therefore, H f ∈ S p . Since the condition MO( f ) ∈ L p (C, dA) is closed under

complex conjugation, we also have H f ∈ S p.

Lemma 8.16. Suppose 2 ≤ p < ∞ and T is the integral operator defined by

T f (z) = G(z, w)K(z, w) f (w) dλα (w),
C
where G is a measurable function on C × C and K(z, w) is the reproducing kernel

of Fα2 . If

|G(z, w)| p |K(z, w)|2 dλα (z) dλα (w) < ∞,
C C
then T is in the Schatten class S p of L2α .

Proof. The case p = 2 follows from the classical characterization of Hilbert–
Schmidt integral operators on L2 spaces; see [113]. If G ∈ L∞ (C × C), then T is
dominated by the bounded operator Qα considered in Sect. 2.2, so the operator
T is bounded on L2α as well. The case 2 < p < ∞ then follows from complex
interpolation.

Lemma 8.17. Let 1 ≤ p < ∞. There exists a positive constant C = C p such that

1 + |z| p−1
| f(z) − f(0)| p dλα (z) ≤ C [MO( f )(z)] p dλα (z) (8.6)
C C |z|
for all f .
Proof. Recall from the proof of Theorem 3.35 that there exists a positive constant
C = C(α ) such that

d
f(tz/|z|) ≤ CMO( f )(tz/|z|)
dt
for all t ≥ 0 and z ∈ C − {0}. Thus,

|z|
d

| f (z) − f (0)| = f (tz/|z|) dt
0 dt
|z|
≤C MO( f )(tz/|z|) dt
0
1
= C|z| MO( f )(tz) dt.
0
Since p ≥ 1, an application of Hölder’s inequality gives

1
| f(z) − f(0)| p ≤ C p |z| p MO( f ) p (tz) dt.
0
This, along with Fubini’s theorem, shows that the integral

I= | f(z) −
f (0)| p dλα (z)
C
satisfies
1
I ≤ Cp |z| p dλα (z) MO( f ) p (tz) dt
C 0
1
= Cp dt |z| p MO( f ) p (tz) dλα (z)
0 C
1
= C |z| p e−α |z| MO( f ) p (tz) dA(z)
2
dt
0 C
1
dt

|z| p e−α |z|
2 /t 2
=C MO( f ) p (z) dA(z)
0 t 2+p C
1
= C t −(2+p)e−α |z|
2 /t 2
|z| p MO( f ) p (z) dA(z) dt
C 0
∞
= C t p e− α t
2 |z|2
|z| p MO( f ) p (z) dA(z) dt
C 1
∞
MO( f ) p (z)
= C t p e−α t dt,
2
dA(z)
C |z| |z|
where C = Cα /π . By L’Höpital’s rule,

∞
t p e−α t dt
2
|z| 1
lim = .
|z|→∞ |z| p−1 e−α |z|
2
2α
It follows that there exists another constant C > 0, independent of z, such that
∞
t p e−α t dt ≤ C 1 + |z| p−1 e−α |z|
2 2
|z|
for all z ∈ C. This proves the desired estimate.

Lemma 8.18. Suppose 2 ≤ p < ∞ and MO( f ) ∈ L p (C, dA). Then both H f and H f
are in the Schatten class S p .
Proof. First, consider the integral

I= | f(z) − f(w)| p |K(z, w)|2 dλα (z) dλα (w).
C C
By Fubini’s theorem and the change of variables formula, we have

α
I= dA(z) | f(z) − f (w)| p |kz (w)|2 dλα (w)
π C C

α
= dA(z) | f(z) − f (z − w)| p dλα (w)
π C C

α
= dA(z) | f◦ ϕz (0) − f ◦ ϕz (w)| p dλα (w),
π C C
where ϕz (w) = z − w. By Lemma 8.17 and the invariance of the Berezin transform
under the action of ϕz , there exists a positive constant C, independent of f , such that

I ≤C dA(z) ϕ (w)MO( f ◦ ϕz ) p (w) dλα (w)
C C

=C dA(z) ϕ (w)MO( f ) p (ϕz (w)) dλα (w),
C C
where ϕ (w) = (1 + |w| p−1 )/|w|. Changing variables again and applying Fubini’s
theorem, we obtain

I ≤C dA(z) ϕ (ϕz (w))MO( f ) p (w)|kz (w)|2 dλα (w)
C C

=C MO( f ) p (w) dA(w) ϕ (ϕz (w))|kw (z)|2 dλα (z)
C C

=C MO( f ) (w) dA(w)
p
ϕ (u) dλα (u).
C C
It is clear that the integral

α 1 + |z| p−1 −α |z|2
ϕ (u) dλα (u) = e dA(z)
C π C |z|
converges. It follows that I < ∞, and by Lemma 8.16, the Hankel operator H f
belongs to S p .
Next, we consider the function g = f − f. By the triangle inequality,
1 1
2
2 (z) 2 =
|g| | f (w) − f(w)|2 |kz (w)|2 dλα (w)
C
1
2
≤
| f (w) − f (z)| |kz (w)| dλα (w)
2 2
C
1
2
+ | f(z) − f(w)|2 |kz (w)|2 dλα (w)
C
1
2
= MO( f )(z) + | f
◦ ϕz (0) − f
◦ ϕz (w)|2 dλα (w) .
C
By assumption, the first term above is in L p (C, dA). The second term is also in
L p (C, dA). In fact, since p ≥ 2, an application of Hölder’s inequality gives
p
2

| f ◦ ϕz (w) − f ◦ ϕz (0)| dλα (w) dA(z)
2
C C

≤ dA(z) | f
◦ ϕz (w) − f
◦ ϕz (0)| p dλα (w)
C C

≤C MO( f ) p (w) dA(w).
C
The last inequality above was proved in the previous paragraph. We conclude that
√2 2 belongs to
the function |g| belongs to L p (C, dA). In other words, the function |g|
L (C, dA). By Corollary 6.33, the Toeplitz operator T|g|2 belongs to the Schatten
p/2
class S p/2. Since

Hg∗ Hg = T|g|2 − Tg Tg ≤ T|g|2 ,
the operator Hg∗ Hg belongs to the Schatten class S p/2 . This shows that Hg belongs
to S p , and consequently, H f = H f + H f − f belongs to S p . The condition MO( f ) ∈
L p (C, dA) is closed under complex conjugation, so we must have H f ∈ S p as well.

Recall that Z denotes the additive integer group and
Z2 = {n + im : n, m ∈ Z}
is the lattice of integers in the complex plane. Throughout this section, we fix a
positive integer N and consider the finer lattice

1 2 n + im
Z = : n, m ∈ Z .
N N
We also consider the following two special squares in the complex plane:

1 1
SN = x + iy : 0 ≤ x < , 0 ≤ y < ,
N N
and

1 2 1 2
QN = x + iy : − ≤ x < , − ≤ y < .
N N N N
If f is a Lebesgue measurable function on the complex plane, we write

JN ( f ) = | f (u) − f (v)|2 dA(u) dA(v).
QN QN
If E is a measurable set in C with 0 < A(E) < ∞ and f is integrable on E, we use

1
fE = f dA
A(E) E
to denote the average (mean) of f over the set E.

Lemma 8.19. Suppose f is locally square integrable and ν ∈ Z2 /N. Then

4N 4 |γ (ν )|
SN
| f ◦ tν − fSN |2 dA ≤ N 2 +
9 ∑ JN ( f ◦ ta),
a∈γ (ν )
where ta (z) = z + a is the translation by a and γ (ν ) is the canonical path in Z2 /N

from ν to 0 (see Sect. 1.2).
Proof. The case ν = 0 is trivial. If ν = 0, we write
γ (ν ) = {a0 , a1 , . . . , al }
in the order in which γ (ν ) is defined, where l + 1 = |γ (ν )| is the length of the path

γ (ν ). It is clear that
(SN + a j−1) ∪ (SN + a j ) ⊂ QN + a j−1, 1 ≤ j ≤ l.
We will estimate the integral

I= | f ◦ tν − fSN |2 dA
SN
using the elementary inequality
|z1 + · · · + zk |2 ≤ k(|z1 |2 + · · · + |zk |2 )
along with several natural “telescoping” decompositions.

We begin with the estimate

I= | f ◦ tal − ( f ◦ ta0 )SN |2 dA
SN

≤2 | f ◦ tal − ( f ◦ tal )SN |2 + |( f ◦ tal )SN − ( f ◦ ta0 )SN |2 dA.
SN

2 | f ◦ tal − ( f ◦ tal )SN |2 dA
SN

1
= | f ◦ tal (u) − f ◦ tal (v)|2 dA(u) dA(v)
A(SN ) SN SN
≤ N 2 JN ( f ◦ tal ).
On the other hand,

2 |( f ◦ tal )SN − ( f ◦ ta0 )SN |2 dA
SN
l
≤ 2l ∑ |( f ◦ ta j )SN − ( f ◦ ta j−1 )SN |2 dA
j=1 SN
l
≤ 4l ∑ |( f ◦ ta j )SN − ( f ◦ ta j−1 )QN |2
j=1 SN

+ |( f ◦ ta j−1 )QN − ( f ◦ ta j−1 )SN |2 dA.
Thus the quantity

D = |( f ◦ ta j )SN − ( f ◦ ta j−1 )QN |2
can be estimated as follows:

2
1

D= [ f ◦ ta j − ( f ◦ ta j−1 )QN ] dA
A(SN ) S N

≤ N2 | f − ( f ◦ ta j−1 )QN |2 dA
SN +a j

≤ N2 | f − ( f ◦ ta j−1 )QN |2 dA
QN +a j−1

= N2 | f ◦ ta j−1 − ( f ◦ ta j−1 )QN |2 dA
QN
N4
= JN ( f ◦ ta j−1 ).
18
Similarly,
N4
|( f ◦ ta j−1 )QN − ( f ◦ ta j−1 )SN |2 ≤ JN ( f ◦ ta j−1 ).
18
Therefore,
l
4lN 4
2
SN
|( f ◦ tal )SN − ( f ◦ ta0 )SN |2 dA ≤
9 ∑ JN ( f ◦ ta j−1 ).
j=1

Lemma 8.20. Suppose f satisfies condition (I2 ). There exists a positive constant
C = CN (depending on N) such that
∑ ∑ e−α |ν |
2 /3
sup MO( f )2 (z) ≤ C JN ( f ◦ ta ).
z∈SN ν ∈Z2 /N a∈γ (ν )
Proof. For any constant c, we have

| f ◦ tz − c|2 dλα = |
f |2 (z) − c f(z) − c f(z) + |c|2
C
= |
f |2 (z) − | f(z)|2 + | f(z) − c|2
≥ |
f |2 (z) − | f(z)|2 .
Thus, for any z ∈ C, we have

MO( f ) (z) ≤2
| f ◦ tz − fSN |2 dλα
C

= ∑ | f (w + z) − fSN |2 dλα (w)
ν ∈Z2 /N SN +ν +z

α
∑ | f (w) − fSN |2 e−α |w−z| dA(w)
2
=
π ν ∈Z2 /N SN +ν

α
∑ | f ◦ tν (w) − fSN |2 e−α |w−z+ν | dA(w).
2
=
π ν ∈Z2 /N SN
For w and z in SN , we have
|w − z + ν |2 ≥ |ν |2 + |w − z|2 − 2|w − z||ν |

≥ |ν |2 /2 − |w − z|2
≥ |ν |2 /2 − N −2.
It follows from this and Lemma 8.19 that

α α2 α
e N ∑ e− 2 |ν |
2
MO( f )2 (z) ≤ | f ◦ tν − fSN |2 dA
π ν ∈Z2 /N SN

α α2 α 4N 4 |γ (ν )|
e N ∑ e− 2 |ν | N 2 + ∑ JN ( f ◦ ta).
2
≤
π ν ∈Z2 /N
9 a∈γ (ν )
Since the length of γ (ν ) is comparable to |ν |, it is clear that we can find a constant

C = CN such that

α α2 4N 4 |γ (ν )| − α |ν |2 α
≤ CN e− 3 |ν |
2
e N N +
2
e 2
π 9
for all ν . This proves the desired result.

Lemma 8.21. Suppose f satisfies condition (I2 ). If 0 < p ≤ 2, then there exists a
positive constant C = CN , depending on N and p but not on f , such that

∑
p
[MO( f )(z)] p dA(z) ≤ CN [JN ( f ◦ tb )] 2 .
C
b∈Z2 /N
Proof. Let us consider the integral

I= [MO( f )(z)] p dA(z).
C
It is clear that

Z2
C= SN + u : u ∈ ,
N
and this is a disjoint union. It follows that

I= ∑ SN +u
[MO( f )(z)] p dA(z)
u∈Z2 /N
1
≤ ∑ sup{MO( f ) p (z) : z ∈ SN + u}
N 2 u∈Z 2 /N
1
= ∑ sup {MO( f ) p (u + z) : z ∈ SN }
N 2 u∈Z 2 /N
1
= ∑ sup {MO( f ◦ tu) p (z) : z ∈ SN } .
N 2 u∈Z 2 /N
Since 0 < p ≤ 2, it follows from Lemma 8.20 and Hölder’s inequality that
pα
∑ ∑
p
e− 6 |ν |
2
sup MO( f ◦ tu ) p (z) ≤ CN [JN ( f ◦ tu ◦ ta )] 2 .
z∈SN ν ∈Z2 /N a∈γ (ν )
Since tu ◦ ta = tu+a , we have
CN pα
∑ ∑ ∑
p
e− 6 |ν |
2
I≤ [JN ( f ◦ tu+a)] 2
N2 u∈Z2 /N ν ∈Z2 /N a∈γ (ν )
CN pα
∑ ∑ ∑
p
e− 6 |ν |
2
= [JN ( f ◦ tu+a)] 2
N2 ν ∈Z2 /N a∈γ (ν ) u∈Z2 /N
CN pα
∑ ∑
p
|γ (ν )|e− 6 |ν |
2
= [JN ( f ◦ tb )] 2 ,
N2 ν ∈Z2 /N b∈Z2 /N
where |γ (ν )| is the length of the path γ (ν ). Again, since |γ (ν )| is comparable to |ν |,

the series
pα
∑ |γ (ν )|e− 6 |ν |
2
ν ∈Z2 /N
converges. This proves the desired estimate.

Lemma 8.22. There exist a positive integer N and a positive constant CN such that
1 2
Iν ( f ) ≥ CN JN ( f ◦ tν ), ν∈ Z ,
N
for all locally square integrable f , where Iν ( f ) denotes the integral

2
α zw− α2 |w|2 −iIm (αν w)

( f (z) − f (w))e dA(w) e−α |z| dA(z).
2

+ν Q
QN N +ν
Proof. We can write Iν ( f ) as

2
α
( f (z) − f (w))e− 2 |z−w|
2 +iα Im (zw−ν w)
dA(w) dA(z),

+ν Q
QN N +ν
which, after a simultaneous change of variables and some simplifications, becomes

2
− α2 |z−w|2 +iα Im (zw)

dA(w) dA(z).
QN
Q ( f ◦ tν (z) − f ◦ tν (w))e
N
Fix any δ ∈ (0, 1/4) and choose a positive integer N such that
α 2 +iα Im (zw)
e− 2 |z−w| = 1 + γz,w, |γz,w | < δ ,
for all (z, w) ∈ QN × QN . To compress the expressions below, we write γ = γz,w .

Then, for any z ∈ QN , we deduce from the triangle inequality that the quantity
2

( f ◦ tν (z) − f ◦ tν )(1 + γ ) dA
Q
N
is greater than or equal to

2

Q ( f ◦ tν (z) − f ◦ tν ) dA − Q ( f ◦ tν (z) − f ◦ tν )γ dA ,
N N
which is greater than or equal to

2

Q ( f ◦ tν (z) − f ◦ tν ) dA
N
minus

2 ( f ◦ tν (z) − f ◦ tν ) dA ( f ◦ tν (z) − f ◦ tν )γ dA ,
QN QN
which is greater than or equal to

2 2

( f ◦ tν (z) − f ◦ tν ) dA − 2 δ | f ◦ tν (z) − f ◦ tν | dA .
Q Q
N N
It follows that
2

Iν ( f ) ≥
QN
Q ( f ◦ tν (z) − f ◦ tν (w)) dA(w) dA(z)
N
2
−2δ | f ◦ tν (z) − f ◦ tν (w)| dA(w) dA(z).
QN QN
The first integral above can be written as [9/(2N 2 )]JN ( f ◦ tν ), and according to
the Cauchy–Schwarz inequality, the second integral above is less than or equal to
(9/N 2 )JN ( f ◦ tν ). We conclude that

9 1
Iν ( f ) ≥ − 2δ JN ( f ◦ tν ).
N2 2
This completes the proof of the lemma.

In the remainder of this section, we fix a positive integer N such that Lemma 8.22
holds. We will need to decompose the lattice Z2 /N into more sparse sublattices. To
this end, we fix another positive integer M whose magnitude will be specified later.
For any j = ( j1 , j2 ), where each jk ∈ {1, 2, . . ., M}, we let
ν ν
1 2
ΛM j = ν = , : νk = jk mod M, k = 1, 2 .
N N
It is clear that
Z2 M
= ΛM
j ,
N j , j =1
1 2
the sublattices ΛM
j are disjoint, and the distance between any two points in the same
ΛMj is at least M/N.
Lemma 8.23. Suppose 0 < p < ∞ and f satisfies condition (I2 ). Then the Hankel
operators H f and H f both belong to the Schatten class S p if and only if the
commutator [M f , P] = M f P − PM f belongs to the Schatten class S p .
Proof. It is easy to see that
[M f , P] = [M f , P]P + [M f , P](I − P) = H f − H ∗f .
So the simultaneous membership of H f and H f in S p implies that [M f , P] is in S p .

To prove the other direction, note that
[M f , P]P = (M f P − PM f )P = M f P − PM f P = (I − P)M f P.
So the Hankel operator H f : Fα2 → L2α is just the restriction of [M f , P] on the space
Fα2 . It follows that the membership of [M f , P] in S p implies the membership of H f
in S p . But the condition [M f , P] ∈ S p implies [M f , P] ∈ S p , so [M f , P] ∈ S p implies
that both H f and H f are in S p .

Lemma 8.24. For any 2 ≤ p < ∞, there exists a positive constant C (depending on
N but independent of f ) such that
[M f , P]Spp ≤ C ∑ JN ( f ◦ tν ) p/2

ν ∈Z2 /N
for all f ∈ L2local (C, dA).

Proof. If f ∈ L2local (C, dA), then
MχE [M f , P]MχE ∈ S2 ⊂ S p , 2 ≤ p < ∞.

Here E is any bounded Borel set in C. Therefore, it suffices to show that there exists
a positive constant C, independent of f and E, such that
MχE [M f , P]MχE Spp ≤ C ∑ JN ( f ◦ tu ) p/2 (8.7)

u∈Z2 /N
for all bounded E and f ∈ L2local (C, dA).

Fix a bounded Borel set E and let F be any finite set in Z2 such that

E⊂
(SN + u) =: E.
u∈F
Since ST SS p ≤ ST S p S for all bounded operators S and all T ∈ S p , and since
MχE MχE = MχE , it suffices to estimate the S p norm of the operator:
Y= ∑ MχSN +u [M f , P]MχS
N +u
= ∑ Yv ,
u,u ∈F v∈Z2 /N
where
Yv = ∑ χF×F (u, u + v)MχSN +u [M f , P]MχSN +u+v .
u∈Z2 /N
For any given v ∈ Z2 /N, the family

% &
χSN +u+v f : f ∈ L2α , u ∈ Z2 /N
of subspaces are pairwise orthogonal in L2α . Since T S p ≤ T S2 when p ≥ 2, we

have
p
Yv S p = ∑ χF×F (u, u + v)MχSN +u [M f , P]MχSN +u+v Spp
u∈Z2 /N
≤ ∑ MχSN +u [M f , P]MχSN +u+v Sp2 . (8.8)

u∈Z2 /N
Since [M f , P] has ( f (z) − f (w))eα zw as its kernel function, we have
MχSN +u [M f , P]MχSN +u+v 2S2

= | f (z) − f (w)|2 |eα zw |2 dλα (z) dλα (w)
SN +u SN +u+v
α 2
| f (z) − f (w)|2 e−α |z−w| dA(z) dA(w)
2
=
π SN +u SN +u+v

≤ δ (v) | f ◦ tu (z) − f ◦ tu (w)|2 dA(z) dA(w), (8.9)
SN SN +v
where tu is the translation by u and

δ (v) = exp −α inf |(w − z) + v| .
2
w,z∈SN
It follows from the inequalities

1
|(w − z) + v|2 ≥ |v|2 + |w − z|2 − 2|w − z||v| ≥ |v|2 − |w − z|2
2
that there exists a positive constant B such that
α
δ (v) ≤ Be− 2 |v| ,
2
v ∈ Z2 /N.
Because A(SN ) = 1/N 2 , we have for any g ∈ L2local (C, dA) that

|g(z) − g(w)|2 dA(z) dA(w)
SN SN +v

≤2 |g(z) − gSN |2 + |gSN − g ◦ tv(w)|2 dA(z) dA(w)
SN SN

2 2
= |g − gSN |2 dA + |g ◦ tv − gSN |2 dA.
N2 SN N2 SN
It follows from the identity

1 1
|g − gQN |2 dA = |g − gSN |2 dA + |gSN − gQN |2
A(SN ) SN A(SN ) SN
that

1
|g − gSN |2 dA ≤ |g − gQN |2 dA ≤ JN (g).
SN SN 2
Applying Lemma 8.19 to the integral

|g ◦ tv − gSN |2 dA,
SN
we obtain

|g(z) − g(w)|2 dA(z) dA(w)
SN SN +v

1 4 4
≤ 2 JN (g) + N + N |γ (v)| ∑ JN (g ◦ ta)
2
N 9 a∈γ (v)

1 4
≤ N 2 + 2 + N 4 |γ (v)| ∑ JN (g ◦ ta ),
N 9 a∈γ (v)
where γ (v) is the discrete path in Z2 /N from 0 to v (see Sect. 1.2). Let g = f ◦ tu in
the above estimate and use (8.9). We see that
MχSN +u [M f , P]MχSN +u+v 2S2

− α2 |v|2 1 4 4
Be N + 2 + N |γ (v)| ∑ JN ( f ◦ tu ◦ ta ).
2
N 9 a∈γ (v)
Since p/2 ≥ 1, it follows from Hölder’s inequality that
∑
p
MχSN +u [M f , P]MχSN +u+v Sp2 ≤ h(v) [JN ( f ◦ tu ◦ ta )] 2 ,
a∈γ (v)
where
α 2 2
p 1 4
p + p−2
2 2
h(v) = Be− 2 |v| N 2 + 2 + N 4 |γ (v)| .
N 9
Combining this with (8.8), we obtain
∑ ∑
p
Yv Spp ≤ h(v) [JN ( f ◦ tu ◦ ta )] 2
u∈Z2 /N a∈γ (v)
∑ ∑
p
= h(v) [JN ( f ◦ tb )] 2 . (8.10)
u∈Z2 /N b∈γ (v)+u
For any b ∈ Z2 /N, we have b ∈ γ (v) + u if and only if −u ∈ γ (v) − b. Thus,

{u ∈ Z2 /N : b ∈ γ (v) + u}| = |γ (v) − b| = |γ (v)| ≤ 1 + |γ (v) .
Therefore,
∑ ∑
p
[JN ( f ◦ tb )] 2
u∈Z2 /N b∈γ (v)+u
p % &
= ∑ [JN ( f ◦ tb )] 2 | u ∈ Z2 /N : b ∈ γ (v) + u |
b∈Z2 /N
∑
p
= (1 + |γ (v)|) [JN ( f ◦ tb )] 2 .
b∈Z2 /N
A substitution of this in (8.10) gives us
∑
p
Yv Spp ≤ h(v)(1 + |γ (v)|) [JN ( f ◦ tb )] 2 .
b∈Z2 /N
Consequently,
Y S p ≤ ∑ Yv S p
v∈Z2 /N
⎡ ⎤1
p
∑ ∑
1 p
≤ [h(v)(1 + |γ (v)|)] ⎣ p [JN ( f ◦ tb )] ⎦ .
2
v∈Z2 /N b∈Z2 /N
From Lemma 1.12, the definition of h(v), and the elementary inequality |γ (v)| ≤
2|v|, we see that the constant
∑
1
C= [h(v)(1 + |γ (v)|)] p
v∈Z2 /N
is finite. With this constant C, the inequality in (8.7) holds for any bounded Borel
set E ⊂ C.

Lemma 8.25. Suppose 0 < p < 2 and f satisfies condition (I2 ). If both H f and
H f are in the Schatten class S p , then MO( f ) ∈ L p (C, dA). Moreover, there exists a
positive constant C, independent of f , such that

[MO( f )(z)] p dA(z) ≤ C H f Spp + H f Spp .
C
Proof. For any
j = ( j1 , j2 ) ∈ {1, 2, . . . , M} × {1, 2, . . ., M},
we fix an orthonormal basis {eν : ν ∈ ΛM

j } for Lα and define two sequences {hν }
2
and {ζν } in L2α as follows:
hν (w) = eα |w| e−α iIm (ν w) χQN +ν (w),

2 /2
ν ∈ ΛM
j ,
and
χQN +ν (z)[M f , P]hν (z)
ζν (z) = , ν ∈ ΛM
j .
χQN +ν [M f , P]hν
We also define two operators A j and B j on L2α as follows:
A j eν = ζν , B j eν = h ν , ν ∈ ΛM
j .
It is easy to check that both A j and B j extend to bounded linear operators on L2α .
In fact, since each hν is supported on QN + ν and different QN + ν are disjoint, we
have
⎛ ⎞2 2

B j ⎝ ∑ cν eν ⎠ = ∑ cν hν (w) dλα (w)

ν ∈ΛM
j
C
ν ∈ΛM
j

= ∑M |cν |2
QN +ν
|hν (w)|2 dλα (w)
ν ∈Λ j

α
= ∑M |cν |2
QN +ν π
dA(w)
ν ∈Λ j
9α
π N2 ν∑
= |cν |2 .
∈ΛN j
√ √
This shows that B j ≤ (3 α )/(N π ). A similar argument shows that A j ≤ 1.
Let W j = A∗j [M f , P]B j for each j. Then,
W j S p ≤ A j [M f , P]S p B j .
Since there are M 2 such j’s, we obtain

p
3 α
∑ W j Spp ≤M 2
N π
[M f , P]Spp
j
p
6 α p p
≤M 2
H f S p + H f S p .
N π
Here, we used the first identity in the proof of Lemma 8.23 and the fact that, for any
positive p and any Schatten class operators S and T , we always have
S + T Spp ≤ 2 p (SSpp + T Spp ). (8.11)
Fix a very large natural number R and consider the truncation ZR of the lattice
Z2 /N:
% &
ZR = ν = (ν1 , ν2 ) ∈ Z2 /N : |νk | ≤ R, k = 1, 2 .
For any j, we set Z j = ZR ∩ ΛM

j and denote by PZ j the orthogonal projection from
L2α onto the subspace spanned by {eν : ν ∈ Z j }. It is clear that
PZ j W j PZ j g = ∑ g, eν W j eν , eν eν .

ν ,ν ∈Z j
We are going to decompose PZ j W j PZ j into a “diagonal” part and an “off-diagonal”

part. More specifically, we define an operator D j by
D jg = ∑ g, eν W j eν , eν eν

ν ∈Z j
and set
E j = PZ j W j PZ j − D j .
Both D j and E j are finite rank operators, so they both belong to the Schatten class
S p . Also, it follows from (8.11) that
2 p W j Spp ≥ 2 p PZ j W j PZ j Spp ≥ D j Spp − 2 pE j Spp .
Since D j is diagonal, we have
D j Spp = ∑ |A∗j [M f , P]B j eν , eν | p

ν ∈Z j
= ∑ χQN +ν [M f , P]hν p
ν ∈Z j
p
2
= ∑ QN +ν
|(M f P − PM f )hν | dλα 2
.
ν ∈Z j
Note that
(M f P − PM f )hν (z) = f (z)Phν (z) − P( f hν )(z)

= ( f (z) − f (w))eα zw hν (w) dλα (w)
C

α α
( f (z) − f (w))eα zw− 2 |w|
2 −α iIm (ν w)
= dA(w).
π QN +ν
An application of Lemma 8.22 then produces a positive constant CN such that
∑ [JN ( f ◦ tν )] 2 .
p
D j Spp ≥ CN
ν ∈Z j
Next, we will obtain an upper bound for E j S p , which is much more involved
than the previous estimates. We begin with the following well-known fact from
operator theory: if 0 < p ≤ 2 and T is a compact operator on a separable Hilbert
space H, then
T Spp ≤ ∑ |Ten , em | p
n,m
for any orthonormal basis {en } of H. See Lemma 6.36. Thus,
E j Spp ≤ ∑ M |E j eν , eν | p
ν ,ν ∈Λ j
= ∑ |E j eν , eν | p
ν ,ν ∈Z j ,ν =ν

[M f , P]hν , χQN +ν [M f , P]hν p
= ∑
χ QN +ν [M f , P]h ν


χQN +ν [M f , P]hν , χQN +ν [M f , P]hν p
= ∑
χ [M f , P]h
QN +ν ν

≤ ∑ χQN +ν [M f , P]hν p .
Write χQN +ν [M f , P]hν p as
2 p
α zw− α2 |w|2 −α iIm (ν w)
2
( f (z) − f (w))e dA(w) dλα (z)

+ν Q
QN N +ν
p
and apply the Cauchy–Schwarz inequality in the inner integral. We see that E j S p
is less than or equal to [(3α )/(N π )] p times
p
2
∑ QN +ν QN +ν
| f (z) − f (w)| e 2 −α |z−w|2
dA(w) dA(z) .

1
|z − w| ≥ (M − 3)
N
whenever z ∈ QN + ν and w ∈ QN + ν (without loss of generality, we may assume

that M > 3). Thus E j Spp is less than or equal to the constant
p
3α pα
( M−3
N )
2
e− 2
Nπ
times the infinite sum

p
α 2
∑ f (w)|2 e− 2 |z−w| dA(w) dA(z)
2
| f (z) − .
ν ,ν ∈Z j ,ν =ν QN +ν QN +ν
Making the simultaneous change of variables
z → z + ν , w → w + ν ,
and estimating the resulting exponential function with the help of the triangle
inequality, we obtain a positive constant CN such that
pα
( M−3 pα | ν − ν | 2 p
N )
2
E j Spp ≤ CN e− 2
∑ e− 5 I(ν , ν ) 2 ,
where

I(ν , ν ) = | f ◦ tν (z) − f ◦ tν (w)|2 dA(w) dA(z).
QN QN
We enumerate the points in the path γ (ν , ν ) ⊂ ZR as {a0 , . . . , al } in such a way

that a0 = ν , al = ν , and
(SN + ak−1 ) ∪ (SN + ak ) ⊂ QN + ak−1 , 1 ≤ k ≤ l,
where l + 1 is the length of the path γ (ν , ν ). By the triangle inequality,
| f ◦ tν (z) − f ◦ tν (w)| ≤ | f ◦ tν (z) − ( f ◦ tν )QN |

+|( f ◦ tν )QN − f ◦ tν (w)|
l
+ ∑ |( f ◦ tak−1 )QN − ( f ◦ tak )QN |.
k=1
By Cauchy–Schwarz, the integrand | f ◦ tν (z) − f ◦ tν (w)|2 in I(ν , ν ) is less than or

equal to (l + 2) times
| f ◦ tν (z) − ( f ◦ tν )QN |2 + |( f ◦ tν )QN − f ◦ tν (w)|2

l
+ ∑ |( f ◦ tak−1 )QN − ( f ◦ tak )QN |2 .
k=1
Therefore, if we also assume N ≥ 3, the double integral I(ν , ν ) is less than or equal
to 9(l + 2)/N 2 times

| f ◦ tν (z) − ( f ◦ tν )QN |2 dA(z)
QN

+ | f ◦ tν (w) − ( f ◦ tν )QN |2 dA(w)
QN
l
+ ∑ |( f ◦ tak−1 )QN − ( f ◦ tak )QN |2 .
k=1
Since 0 < p/2 < 1, I(ν , ν ) p/2 is less than or equal to [9(l + 2)/N 2 ] p/2 times
p
2
| f ◦ tν − ( f ◦ tν )QN | dA 2
(8.12)
QN
p
2
+ | f ◦ tν − ( f ◦ tν )QN | dA 2
(8.13)
QN
2
l p
+ ∑ |( f ◦ tak−1 )QN − ( f ◦ tak )QN | 2

. (8.14)
k=1
It follows that E j Spp is less than or equal to
p
− p4α ( M−3 9(l + 2) 2
N )
2
CN e
N2
times
p
2
− α5 |ν −ν |2
∑ e
QN
| f ◦ t − ( f ◦ t )QN | dA
ν ν
2
(8.15)
p
α 2
∑ e− 5 |ν −ν |
2
+ | f ◦ tν − ( f ◦ tν )QN |2 dA (8.16)
QN
p
l 2
− α5 |ν −ν |2
+ ∑ e ∑ |( f ◦ tak−1 )QN −( f ◦ tak )QN |2 . (8.17)
ν ,ν ∈Z j ,ν =ν k=1
Since l is comparable to |ν − ν |, we can find another constant CN such that

p pα pα
[9(l + 1)/N 2 ] 2 e− 5 |ν −ν | ≤ CN e− 6 |ν −ν |
2 2
.
So the quantity in (8.15) is dominated by (up to a multiplicative constant that only

depends on N)
p
− p4α ( M−3 − p6α |ν −ν |2
2
N )
2
e ∑ e
QN
| f ◦ tν − ( f ◦ tν )QN | dA 2
,
which is equal to
pα
( M−3
N )
2
∑
p
Ce− 4 [JN ( f ◦ tν )] 2 ,

ν ∈Z j
where
2p
N2 pα 2
C=
18 ∑ e− 6 |ν −ν |
ν ∈Z j
p
N2 2 pα 2
≤
18 ∑ e− 6 |ν −ν |
ν ∈Z2 /N
2p
N2 pα
∑ e− 6 |ν |
2
= .
18 ν ∈Z2 /N
By symmetry, we get exactly the same estimate for the quantity in (8.16).
Since 0 < p/2 < 1, we can apply Hölder’s inequality in (8.17) and reduce our
estimate to the following quantity:
pα
l pα
( M−3
N )
2 2
S j = e− 4
∑ ∑ e− 6 |ν −ν | |( f ◦ tak−1 )QN − ( f ◦ tak )QN | p .
ν ,ν ∈Z j ,ν =ν k=1

Just like the computation we performed in the proof of Lemma 8.19, we have
|( f ◦ tak−1 )QN −( f ◦ tak )QN | p = | fQN +ak−1 − fQN +ak | p

≤ 2 p | fQN +ak−1 − fSN +ak | p + | fSN +ak − fQN +ak | p
p p
≤ CN JN ( f ◦ tak−1 ) 2 + JN ( f ◦ tak ) 2 .
Thus,
pα 2
∑ ∑
p
S j ≤ CN e− 6 |ν −ν | [JN ( f ◦ tu )] 2
ν ,ν ∈Z j ,ν =ν u∈γ (ν ,ν )
pα 2
∑ ∑ [JN ( f ◦ tu)] 2 χγ (ν ,ν ) (u)
p
= CN e− 6 |ν −ν |
ν ,ν ∈Z j ,ν =ν u∈Z j
pα 2
∑ [JN ( f ◦ tu)]
p
= CN 2
∑ e− 6 |ν −ν | χγ (ν ,ν ) (u).
u∈Z j ν ,ν ∈Z j ,ν =ν
By Lemma 1.15, there exists a constant C > 0, independent of u and R, such that
pα 2
∑ e− 6 |ν −ν | χγ (ν ,ν ) (u) ≤ C
for all u ∈ Z. Therefore,

pα
( M−3
N )
2
∑ [JN ( f ◦ tu)] 2
p
E j Spp ≤ CN e− 4
u∈Z j
for all j, where CN is yet another constant that depends on N only. Combining this
with the earlier lower estimate for D j S p , we see that there exist two constants CN1
and CN2 , which are both independent of M and R, such that

2 2 − p4α ( M−3 )
2
∑ [JN ( f ◦ tu)] 2 .
p
M 2
H f Spp + H f Spp ≥ CN − CN M e
1 N
u∈Z j
If we pick M such that
pα
( M−3
N ) > 0,
2
CN1 − CN2 M 2 e− 4
then we obtain a constant C > 0, independent of f and R, such that
∑ [JN ( f ◦ tu)] 2
p
p p
H f S p + H f S p ≥ C
u∈Z j
for all j ∈ {1, 2, . . . , M} × {1, 2, . . . , M}. Summing over all such j, we obtain a
constant C > 0, independent of the truncating constant R, such that
∑
p
H f Spp + H f Spp ≥ C [JN ( f ◦ tu)] 2 .
u∈ZR
Let R → ∞. We obtain
∑
p
H f Spp + H f Spp ≥ C [JN ( f ◦ tu )] 2 .
u∈Z2 /N
This, along with Lemma 8.21, completes the proof of Lemma 8.26.

Theorem 8.26. Suppose 0 0, N is any positive integer, and f satisfies
condition (I2 ). Then the following conditions are equivalent:
(a) The operators H f and H f both belong to the Schatten class S p .
(b) The function
MO( f )(z) = [|

f |2 (z) − | f(z)|2 ]1/2
is in L p (C, dA).
(c) The function
MOr ( f )(z) = [|

f |2 r (z) − | fr (z)|2 ]1/2
is in L p (C, dA).
(d) The sequence

1

[JN ( f ◦ tν )] 2 : ν ∈ Z2 /N
belongs to l p .
Proof. That (a) implies (b) follows from Lemmas 8.14 and 8.25. Lemmas 8.15 and
8.18 show that condition (b) implies (a). So (a) and (b) are equivalent.
By the double integral representations for MO( f ) and MOr ( f ), it is easy for us
to find a positive constant C = C(α , r) such that
MOr ( f )(z) ≤ CMO( f )(z), z ∈ C,
which shows that (b) implies (c).

To show that (c) implies (d), we fix any positive r and choose a sufficiently large
positive integer N such that
QN + ν ⊂ B(ζ , r), ν ∈ Z2 /N, ζ ∈ SN + ν . (8.18)
This is possible because of the triangle inequality for the Euclidean metric.
Consider the function:
1
2
Fr (z) = | f (u) − f (v)|2 dA(u) dA(v) .
B(z,r) B(z,r)
Since MOr ( f ) and Fr differ only by a multiplicative constant, condition (c) implies
that Fr ∈ L p (C, dA).
Let

I= Fr (z) p dA(z).
C
Since the complex plane is the disjoint union of SN + ν , ν ∈ Z2 /N, it follows from
the mean value theorem and (8.18) that

1
I= ∑ Fr (z) p dA(z) = ∑ Fr (ζν ) p
N 2 ν ∈Z
ν ∈Z2 /N SN +ν 2 /N
p
1 2
= 2 ∑ | f (u) − f (v)| dA(u) dA(v)
2
N ν ∈Z2 /N B(ζν ,r) B(ζν ,r)
p
1 2
≥ 2 ∑ | f (u) − f (v)| dA(u) dA(v)
2
N ν ∈Z2 /N QN +ν QN +ν
1
∑ [JN ( f ◦ tν )] 2 .
p
=
N 2 ν ∈Z 2 /N
Thus, condition (c) implies (d).

When 0 < p ≤ 2, Lemma 8.21 shows that condition (d) implies (b). When 2 ≤
p < ∞, Lemmas 8.23 and 8.24 show that condition (d) implies (a). Since (a) and (b)
are already equivalent, we see that condition (d) implies (a) for all 0 < p < ∞. This

8.4 Notes 327
8.4 Notes
The study of Hankel operators on the Fock space goes back to [28] at least, where the
compactness was studied for Hankel operators induced by bounded symbols. This
compactness problem is equivalent to the symbol calculus for Toeplitz operators
with bounded symbols modulo compact operators.
The introduction of BMO (and VMO) defined with a fixed radius into the study
of Hankel and Toeplitz operators was first made in [257] in the context of Bergman
spaces in the unit disk. The extension to Fock spaces was first carried out in [32].
One of the unique features of the Fock space theory is the following: when f is
bounded, the Hankel operator H f is compact on Fα2 if and only if H f is compact. This
result is due to Berger and Coburn [28,29], and it is not true for Hankel operators on
the Bergman space or the Hardy space. A partial explanation for this difference is
probably the lack of bounded analytic or harmonic functions on the entire complex
plane.
The material in Sect. 8.3 concerning membership of the Hankel operators H f in
Schatten classes is mostly from [131, 242]. Again, there is a key difference between
the Fock and Bergman theories. In the Bergman space setting, there is a cutoff point
when the invariant mean oscillation MO( f ) is used to describe the membership of
H f and H f in S p , while in the Fock space setting, this cutoff point disappears because
of the exponential decay of the Fock kernel e−α |z| .
2
8.5 Exercises 329
8.5 Exercises
1. Show that on the space Fα2 , we have Wa = eiTψ for any a ∈ C, where ψ (z) =
2Im (az).
2. Show that H f and H f both belong to the Schatten S p if and only if the sequence
{MOr ( f )(ν ) : ν ∈ Z2 /N} belongs to l p , where r > 0 and N is any positive integer.
3. For f ∈ L∞ (C), show that H f is Hilbert–Schmidt if and only if H f is Hilbert–
Schmidt. See [12].
4. Show that Theorems 8.4 and 8.5 remain valid with the weaker assumption that
ϕ ∈ L2α .
5. Show that Hϕ∗ Hϕ = T|ϕ |2 − Tϕ Tϕ .
6. If f kz 2 ≤ C as for all z ∈ C, show that H f and H f are both bounded. Similarly,
if f kz → 0 as z → ∞, then H f and H f are both compact.
α
7. Show that | f (z) − P f (z)| ≤ 2 f ∞ e 4 |z| for almost all z ∈ C and f ∈ L∞ (C).
2
8. Define and study Hankel operators on the Fock space Fαp when 1 ≤ p ≤ ∞.
References
1. P. Ahern, M. Flores, W. Rudin, An invariant volume-mean-value property. J. Funct. Anal.

111, 380–397 (1993)
2. A. Alexandrov, G. Rozenblum, Finite rank Toeplitz operators: some extensions of
D. Luecking’s theorem. J. Funct. Anal. 256, 2291–2303 (2009)
3. N. Aronszajn, Theory of reproducing kernels. Trans. Amer. Math. Soc. 68, 337–404 (1950)
4. G. Ascensi, Y. Lyubarskii, K. Seip, Phase space distribution of Gabor expansions. Appl.
Comput. Harmon. Anal. 26, 277–282 (2009)
5. S. Axler, The Bergman space, the Bloch space, and commutators of multiplication operators.
Duke Math. J. 53, 315–332 (1986)
6. S. Axler, D. Zheng, Compact operators via the Berezin transform. Indiana Univ. Math. J. 47,
387–400 (1998)
7. H. Bacry, A. Grossmann, J. Zak, Proof of the completeness of lattice states in the kq-
representation. Phy. Rev. B12, 1118–1120 (1975)
8. V. Bargmann, On a Hilbert space of analytic functions and an associated integral
transform I. Comm. Pure Appl. Math. 14, 187–214 (1961)
9. V. Bargmann, On a Hilbert space of analytic functions and an associated integral
transform II. Comm. Pure. Appl. Math. 20, 1–101 (1967)
10. V. Bargmann, Remarks on a Hilbert space of analytic functions, Proc. N.A.S. 48, 199–204
(1962)
11. V. Bargmann, P. Butera, L. Girardello, J. Klauder, On the completeness of coherent states.
Rep. Math. Phys. 2, 221–228 (1971)
12. W. Bauer, Hilbert–Schmidt Hankel operators on the Segal–Bargmann space. Proc. Amer.
Math. Soc. 132, 2989–2998 (2004)
13. W. Bauer, Mean oscillation and Hankel operators on the Segal–Bargmann space. Integr.
Equat. Operat. Theor. 52, 1–15 (2005)
14. W. Bauer, Berezin–Toeplitz quantization and composition formulas. J. Funct. Anal. 256,
3107–3142 (2009)
15. W. Bauer, L. Coburn, J. Isralowitz, Heat flow, BMO, and the compactness of Toeplitz
operators. J. Funct. Anal. 259, 57–78 (2010)
16. W. Bauer, K. Furutani, Compact operators and the pluriharmonic Berezin transform. Int.
J. Math. 19, 645–669 (2008)
17. W. Bauer, K. Furutani, Hilbert–Schmidt Hankel operators and berezin iteration. Tokyo
J. Math. 31, 293–319 (2008)
18. W. Bauer, H. Issa, Commuting toeplitz operators with quasi-homogeneous symbols on the
Segal–Bargmann space, J. Math. Anal. Appl. 386, 213–235 (2012)
19. W. Bauer, T. Le, Algebraic properties and the finite rank problem for toeplitz operators on the
Segal–Bargmann space. J. Funct. Anal. 261, 2617–2640 (2011)

DOI 10.1007/978-1-4419-8801-0, © Springer Science+Business Media New York 2012
332 References
20. W. Bauer, Y.J. Lee, Commuting toeplitz operators on the Segal–Bargmann space. J. Funct.
Anal. 260, 460–489 (2011)
21. D. Bekolle, C. Berger, L. Coburn, K. Zhu, BMO in the Bergman metric on bounded symmetric
domains. J. Funct. Anal. 93, 310–350 (1990)
22. C. Bénéteau, B. Carswell, S. Kouchedian, Extremal problems in the Fock space. Comput.
Methods Funct. Theory 10, 189–206 (2010)
23. F.A. Berezin, Covariant and contravariant symbols of operators. Math. USSR-Izv. 6,
1117–1151 (1972)
24. F.A. Berezin, Quantization. Math. USSR-Izv. 8, 1109–1163 (1974)
25. F.A. Berezin, Quantization in complex symmetric spaces. Math. USSR-Izv. 9, 341–379
(1975)
26. F.A. Berezin, General concept of quantization. Comm. Math. Phys. 40, 153–174 (1975)
27. F.A. Berezin, The relation between co- and contra-variant symbols of operators on classical
complex symmetric spaces. Soviet Math. Dokl. 19, 786–789 (1978)
28. C. Berger, L. Coburn, Toeplitz operators and quantum mechanics. J. Funct. Anal. 68, 273–299
(1986)
29. C. Berger, L. Coburn, Toeplitz operators on the Segal–Bargmann space. Trans. Amer. Math.
Soc. 301, 813–829 (1987)
30. C. Berger, L. Coburn, Heat flow and Berezin–Toeplitz estimates. Amer. J. Math. 116, 563–590
(1994)
31. C. Berger, L. Coburn, A symbol calculus for Toeplitz operators. Proc. Nat. Acad. Sci. USA
83, 3072–3073 (1986)
32. C. Berger, L. Coburn, K. Zhu, Toeplitz operators and function theory in n-dimensions.
Springer Lect. Notes Math. 1256, 28–35 (1987)
33. J. Bergh, J. Löfstrom, Interpolation Spaces. Grundlehren Math. Wiss. 223, Springer, (1976)
34. S. Bergman, The Kernel Function and Conformal Mapping, Math. Surveys V, (American
Mathematical Society, Providence, RI, 1950)
35. B. Berndtsson, J. Ortega-Cerdá, On interpolation and sampling in Hilbert spaces of analytic
functions. J. Reine Angew. Math. 464, 109–128 (1995)
36. A. Beurling, The Collected Works of Arne Beurling, vol. 2, (Harmonic Analysis, Boston,
1989)
37. O. Blasco, A. Galbis, On taylor coefficients of entire functions integrable against exponential
weights. Math. Nachr. 223, 5–21 (2001)
38. R. Boas, Entire Functions, (Academic Press, New York, 1954)
39. H. Bommier-Hato, H. Youssfi, Hankel operators on weighted Fock space. Integr. Equ. Oper.
Theory 59, 1–17 (2007)
40. H. Bommier-Hato, H. Youssfi, Hankel operators and the Stieltjes moment problem. J. Funct.
Anal. 258, 978–998 (2010)
41. A. Borichev, R. Dhuez, K. Kellay, Sampling and interpolation in large Bergman and Fock
spaces. J. Funct. Anal. 242, 563–606 (2007)
42. A. Borichev, Y. Lyubarskii, Riesz bases of reproducing kernels in Fock-type spaces. (English
summary) J. Inst. Math. Jussieu 9, 449–461 (2010)
43. O. Bratteli, D. Robinson, Operator Algebras and Quantum Statistical Mechanics, I, (Springer,
New York, 1979)
44. O. Bratteli, D. Robinson, Operator Algebras and Quantum Statistical Mechanics, II,
(Springer, New York, 1981)
45. S. Brekke, K. Seip, Density theorems for sampling and interpolation in the Bargmann–Fock
spaces III. Math. Scand. 73, 112–126 (1993)
46. B. Carswell, B. MacCluer, A. Schuster, Composition operators on the Fock space. Acta. Sci.
Math. (Szeged) 69, 871–887 (2003)
47. J. Cheeger, M. Gromov, M. Taylor, Finite propagation speed, kernel estimates for functions
of the Laplace operator. J. Differ. Geom. 17, 15–53 (1982)
48. X. Chen, K. Guo, Analytic hilbert spaces over the complex plane. J. Math. Anal. Appl. 268,
684–700 (2002)
References 333
49. X. Chen, S. Hou, A Beurling type theorem for the Fock space. Proc. Amer. Math. Soc. 131,
2791–2795 (2003)
50. H. Cho, B. Choe, H. Koo, Linear combinations of composition operators on the Fock–Sobolev
spaces, preprint, (2011)
51. H. Cho, K. Zhu, Fock–Sobolev spaces and their Carleson measures, to appear in J. Funct.
Anal.
52. B. Choe, On higher dimensional Luecking’s theorem. J. Math. Soc. Japan 61, 213–224 (2009)
53. B. Choe, K. Izuchi, H. Koo, Linear sums of two composition operators on the Fock space.
J. Math. Anal. Appl. 369, 112–119 (2010)
54. L. Coburn, A sharp berezin lipschitz estimate. Proc. Amer. Math. Soc. 135, 1163–1168 (2007)
55. L. Coburn, A Lipschitz estimate for Berezin’s operator calculus. Proc. Amer. Math. Soc. 133,
127–131 (2005)
56. L. Coburn, The Bargmann isometry and Gabor–Daubechies wavelet localization operators, in
Systems, Approximation, Singular Integral Operators, and Related Topics, (Bordeaux, 2000),
169–178; Oper. Theory Adv. Appl. 129, (Birkhauser, Basel, 2001)
57. L. Coburn, On the Berezin–Toeplitz calculus. Proc. Amer. Math. Soc. 129, 3331–3338 (2001)
58. L. Coburn, The measure algebra of the Heisenberg group. J. Funct. Anal. 161, 509–525 (1999)
59. L. Coburn, Berezin–Toeplitz quantization, in Algebraic Methods in Operator Theory,
(Birkhauser, Boston, 1994) pp. 101–108
60. L. Coburn, Deformation estimates for the Berezin–Toeplitz quantization. Comm. Math. Phys.
149, 415–424 (1992)
61. L. Coburn, J. Isralowitz, B. Li, Toeplitz operators with BMO symbols on the Segal–Bargmann
space. Trans. Amer. Math. Soc. 363, 3015–3030 (2011)
62. L. Coburn, B. Li, Directional derivative estimates for Berezin’s operator calculus. Proc. Amer.
Math. Soc. 136, 641–649 (2008)
63. L. Coburn, J. Xia, Toeplitz algebras and Rieffel deformation. Comm. Math. Phys. 168, 23–38
(1995)
64. M. Christ, On the ∂ equation in weighted L2 norms in C1 . J. Geom. Anal. 1, 193–230 (1991)
65. R. Coifman, R. Rochberg, Representation theorems for holomorphic and harmonic functions
in L p . Astérisque 77, 11–66 (1980)
66. R. Coifman, R. Rochberg, G. Weiss, Factorization theorems for Hardy spaces in several
complex variables. Ann. Math. 103, 611–635 (1976)
67. J. Conway, Functions of One Complex Variable, (Springer, New York, 1973)
68. I. Daubechies, Time-frequency localization operators–a geometric phase space approach.
IEEE Trans. Inform. Th. 34, 605–612 (1988)
69. I. Daubechies, The wavelet transform, time-frequency localization, and signal analysis. IEEE
Trans. Inform. Th. 36, 961–1005 (1990)
70. I. Daubechies, Ten Lectures on Wavelets, (Society for Industrial and Applied Mathematics,
Philladephia, 1992)
71. I. Daubechies, A. Grossmann, Frames in the Bargmann space of entire functions. Comm.
Pure. Appl. Math. 41, 151–164 (1988)
72. I. Daubechies, A. Grossmann, Y. Meyer, Painless nonorthogonal expansions. J. Math. Phys.
27, 1271–1283 (1986)
73. I. Daubechies, T. Paul, Time-frequency localization operators– a geometric phase space
approach II. Inverse Probl. 4, 661–680 (1988)
74. M. Dostanić, K. Zhu, Integral operators induced by the Fock kernel. Integr. Equat. Operat.
Theor. 60, 217–236 (2008)
75. R.J. Duffin, A.C. Schaeffer, A class of nonharmonic Fourier series. Trans. Amer. Math. Soc.
72, 341–366 (1952)
76. P. Duren, Theory of H p Spaces, 2nd edn, (Dover Publications, New York, 2000)
77. P. Duren, B. Romberg, A. Shields, Linear functionals on H p spaces with 0 < p < 1. J. Reine
Angew. Math. 238, 32–60 (1969)
78. P. Duren, A. Schuster, Bergman Spaces, (American Mathematical Society, Providence, RI,
2004)
334 References
79. P. Duren, A. Schuster, D. Vukotić, On Uniformly Discrete Sets in the Unit Disk, in
Quadrature Domains and Applications, Oper. Theory Adv. Appl. 156, (Birkhaüser, Basel,
2005) pp. 131–150
80. P. Duren, G. Taylor, Mean growth and coefficients of H p functions. Illinois J. Math. 14,
419–423 (1970)
81. A.E. Dzhrbashyan, Integral representation and continuous projections in certain spaces of
harmonic functions, Mat. Sb. 121, 259–271 (1983); (Russian). English translation: Math.
USSR-Sb 49, 255–267 (1984)
82. M. Englis, Toeplitz Operators on Bergman-Type Spaces, Ph.D. thesis, (MU CSAV, Prague,
1991)
83. M. Englis, Some density theorems for Toeplitz operators on Bergman spaces. Czechoslovak
Math. J. 40, 491–502 (1990)
84. M. Englis, Functions invariant under the Berezin transform. J. Funct. Anal. 121, 233–254
(1994)
85. M. Englis, Compact Toeplitz operators via the Berezin transform on bounded symmetric
domains. Integr. Equat. Operat. Theor. 33, 326–355 (1999)
86. M. Englis, Berezin transform on the harmonic Fock space. J. Math. Anal. Appl. 367, 75–97
(2010)
87. M. Englis, Toeplitz operators and localization operators. Trans. Amer. Math. Soc. 361,
1039–1052 (2009)
88. H.G. Feichtinger, On a new segal algebra. Monatsh. f. Math. 92, 269–289 (1981)
89. E. Fischer, Über algebraische Modulsysteme und lineare homogene partielle Differentialgle-
ichungen mit konstanten Koeffizienten. J. Reine Angew. Math. 141, 48–81 (1911)
90. V. Fock, Verallgemeinerung und Lösung der Diracschen statistischen Gleichung. Z. Physik
49, 339–357 (1928)
91. V. Fock, Konfigurationsraum und zweite Quantelung. Z. Phys. 75, 622–647 (1932)
92. G. Folland, Harmonic Analysis in Phase Space, Ann. Math. Studies 122, (Princeton
University Press, Princeton, NJ, 1989)
93. G. Folland, Fourier Analysis and its Applications, Brooks/Cole Publishing Company, (Pacific
Grove, California, 1992)
94. F. Forelli, W. Rudin, Projections on spaces of holomorphic functions in balls. Indiana Univ.
Math. J. 24, 593–602 (1974)
95. O. Furdui, Norm calculations of composition operators on Fock spaces. Acta Sci. Math.
(Szeged) 74, 281–288 (2008)
96. O. Furdui, On a class of integral operators. Integr. Equat. Operat. Theor. 60, 469–483 (2008)
97. D. Gabor, Theory of communication. J. Inst. Elect. Eng. 93, 429–457 (1946)
98. D. Garling, P. Wojtaszczyk, Some Bargmann spaces of analytic functions. Lect. Notes Pure
Appl. Math. 172, 123–138 (1995)
99. H.F. Gautrin, Toeplitz operators in Bargmann spaces. Integr. Equat. Operat. Theor. 11,
173–185 (1988)
100. I.C. Gohberg, M.G. Krein, Introduction to the Theory of Linear Nonselfadjoint Operators,
(Nauka, Moscow, 1965); (Russian). English translation: Trans. Math. Monographs 18,
(American Mathematical Society, Providence, RI, 1969)
101. A. Grishin, A. Russakovskii, Free interpolation by entire functions. J. Sov. Math. 48, 267–275
(1990)
102. K. Grochenig, H. Razafinjatovo, On Landau’s necessary conditions for sampling and
interpolation of band-limited functions. J. London Math. Soc. 54, 557–565 (1996)
103. K. Grochenig, D. Walnut, A Riesz basis for the Bargmann–Fock space related to sampling
and interpolation. Ark. Math. 30, 283–295 (1992)
104. A. Grossmann, J. Morlet, Decomposition of hardy functions into square integrable wavelets
of constant shape. SIAM J. Math. Anal. 15, 723–736 (1984)
105. S.M. Grudsky, N.L. Vasilevski, Toeplitz operators on the Fock space: radial component
effects. Integr. Equat. Operat. Theor. 44, 10–37 (2002)
106. W. Gryc, T. Kemp, Duality in Segal-Bargmann spaces. J. Funct. Anal. 261, 1591–1623 (2011)
References 335
107. V. Guillemin, Toeplitz operators in n-dimensions. Integr. Equat. Operat. Theor. 7, 145–205
(1984)
108. K. Guo, Quasi-invariant subspaces generated by polynomials with nonzero leading terms.
Studia Math. 164, 231–241 (2004)
109. K. Guo, Homogeneous quasi-invariant subspaces of the Fock space. J. Aust. Math. Soc. 75,
399–407 (2003)
110. K. Guo, K. Izuchi, Composition operators on Fock type spaces. Acta Sci. Math. (Szeged) 74,
807–828 (2008)
111. K. Guo, D. Zheng, Invariant subspaces, quasi-invariant subspaces, and Hankel operators.
J. Funct. Anal. 187, 308–342 (2001)
112. B. Hall, W. Lewkeeratiyutkul, Holomorphic sobolev spaces and the generalized Segal–
Bargmann transform. J. Funct. Anal. 217, 192–220 (2004)
113. P. Halmos, V. Sunder, Bounded Integral Operator on L2 Spaces, (Springer, Berlin, 1978)
114. G. Hardy, J. Littlewood, Some new properties of fourier constants. Math. Ann. 97, 159–209
(1926)
115. G. Hardy, J. Littlewood, Some properties of conjugate functions. J. Reine Angew. Math. 167,
405–423 (1931)
116. W.W. Hastings, A carleson measure theorem for Bergman spaces. Proc. Amer. Math. Soc. 52,
237–241 (1975)
117. W. Hayman, On a conjecture of korenblum. Anal. (Munich) 19, 195–205 (1999)
118. H. Hedenmalm, An invariant subspace of the Bergman space having the co-dimension 2
property. J. Reine Angew. Math. 443, 1–9 (1993)
119. H. Hedenmalm, B. Korenblum, K. Zhu, Theory of Bergman Spaces, (Springer, New York,
2000)
120. H. Hedenmalm, S. Richter, K. Seip, Interpolating sequences and invariant subspaces of given
index in the Bergman space. J. Reine Angew. Math. 477, 13–30 (1996)
121. H. Hedenmalm, K. Zhu, On the failure of optimal factorization for certain weighted Bergman
spaces. Complex Variables Theor. Appl. 19, 165–176 (1992)
122. A. Hinkkanen, On a maximum principle in Bergman space. J. Anal. Math. 79, 335–344 (1999)
123. F. Holland, R. Rochberg, Bergman kernels and Hankel forms on generalized Fock spaces.
Contemp. Math. 232, 189–200 (1999)
124. F. Holland, R. Rochberg, Bergman kernel asymptotics for generalized Fock spaces. J. Anal.
Math. 83, 207–242 (2001)
125. L. Hörmander, The Analysis of Linear Partial Differential Operators, IV, (Springer, Berlin,
1985)
126. L. Hörmander, An Introduction to Complex Analysis in Several Complex Variables, 3rd edn.
(Van Nostrand, Princeton, NJ, 1990)
127. C. Horowitz, Zeros of functions in the Bergman spaces, Ph. D. thesis, (University of
Michigan, Ann Arbor, 1974)
128. C. Horowitz, Zeros of functions in the Bergman spaces. Duke Math. J. 41, 693–710 (1974)
129. C. Horowitz, Factorization theorems for functions in the Bergman spaces. Duke Math. J. 44,
201–213 (1977)
130. R. Howe, Quantum mechanics and partial differential equations. J. Funct. Anal. 38, 188–254
(1980)
131. J. Isralowitz, Schatten p class Hankel operators on the Segal–Bargmann space H 2 (Cn , dμ )
for 0 < p < 1. J. Operat. Theor. 66, 145–160 (2011)
132. J. Isralowitz, K. Zhu, Toeplitz operators on the Fock space. Integr. Equat. Operat. Theor. 66,
593–611 (2010)
133. K. Izuchi, Cyclic vectors in the Fock space over the complex plane. Proc. Amer. Math. Soc.
133, 3627–3630 (2005)
134. K. Izuchi, K. Izuchi, Polynomials having leading terms over C2 in the Fock space. J. Funct.
Anal. 225, 439–479 (2005)
135. J. Janas, Unbounded Toeplitz operators in the Segal–Bargmann space. Studia Math. 99, 8799
(1991)
336 References
136. J. Janas, J. Stochel, Unbounded Toeplitz operators in the Segal–Bargmann space II. J. Funct.
Anal. 126, 418-447 (1994)
137. S. Janson, P. Jones, Interpolation between H p spaces: the complex method. J. Funct. Anal.
48, 58–80 (1982)
138. S. Janson, J. Peetre, R. Rochberg, Hankel forms and the Fock space. Revista Mat. Ibero-Amer.
3, 61–138 (1987)
139. J.R. Klauder, B.S. Skagerstam, Coherent States–Applications in Physics and Mathematical
Physics, (World Scientific Press, Singapore, 1985)
140. W. Knirsch, G. Schneider, Continuity and Schatten–von Neumann p-class membership of
Hankel operators with anti-holomorphic symbols on generalized Fock spaces. J. Math. Anal.
Appl. 320, 403–414 (2006)
141. B. Korenblum, A maximum principle for the Bergman space. Publ. Mat. 35, 479–486 (1991)
142. S. Krantz, Function Theory of Several Complex Variables, 2nd edn. (American Mathematical
Society, Providence, RI, 2001)
143. O. Kures, K. Zhu, A class of integral operators on the unit ball of Cn . Integr. Equ. Oper.
Theory 56, 71–82 (2006)
144. H. Landau, Necessary density conditions for sampling and interpolation of certain entire
functions. Acta Math. 117, 37–52 (1967)
145. H. Landau, Sampling, data transmission, and the Nyquist rate. Proc. IEEE 55, 1701–1706
(1967)
146. S. Lang, Algebra, Graduate Texts in Mathematics, 211, (Springer, New York, 2002)
147. B. Levin, Lectures on Entire Functions, Transl. Math. Monographs 150, (American Mathe-
matical Society, Providence, RI, 1996)
148. B. Levin, Y. Lyubarskii, Interpolation by special classes of entire functions and related
expansions in exponential series, Izv. Akad. Nauk SSSR Ser. Mat. 39, 657–702 (1975);
(Russian). English translation: Math. USSR-Izv 9, 621–662 (1975)
149. H. Li, BMO, VMO, and Hankel operators on the Bergman space of strongly pseudo-convex
domains. J. Funct. Anal. 106, 375–408 (1992)
150. N. Lindholm, Sampling in weighted L p spaces of entire functions in Cn and estimates of the
Bergman kernel. J. Funct. Anal. 182, 390–426 (2001)
151. D. Luecking, Forward and reverse Carleson inequalities for functions in Bergman spaces and
their derivatives. Amer. J. Math. 107, 85–111 (1985)
152. D. Luecking, Trace ideal criteria for Toeplitz operators. J. Funct. Anal. 73, 345–368 (1987)
153. D. Luecking, Finite rank Toeplitz operators on the Bergman space. Proc. Amer. Math. Soc.
136, 1717–1723 (2008)
154. W. Lusky, On the Fourier series of unbounded harmonic functions. J. London Math. Soc. 61,
568–580 (2000)
155. Y. Lyubarskii, Frames in the Bargmann space of entire functions. Adv. Soviet Math. 429,
107–113 (1992)
156. T. MacGregor, K. Zhu, Coefficient multipliers between the Bergman and Hardy spaces.
Mathematika 42, 413–426 (1995)
157. M. Morris, Understanding Quantum Physics, (Prentice Hall, New Jersey, 1990)
158. Y. Lyubarskii, K. Seip, Sampling and interpolation of entire functions and exponential
systems in convex domains. Ark. Mat. 32, 157–193 (1994)
159. Y. Lyubarskii, K. Seip, Complete interpolating sequences for Paley–Wiener spaces and
Muckenhoupt’s A p -condition. Rev. Mat. Iberoamer. 13, 361–376 (1997)
160. N. Marco, X. Massaneda, J. Ortega-Cerdá, Interpolating and sampling sequences for entire
functions. Geom. Funct. Anal. 13, 862–914 (2003)
161. X. Massaneda, P. Thomas, Interpolating sequences for Bargmann–Fock spaces in Cn . Indag.
Math. 11, 115–127 (2000)
162. A. Nakamura, F. Ohya, H. Watanabe, On some properties of functions in weighted Bergman
spaces. Proc. Fac. Sci. Tokai Univ. 15, 33–44 (1979)
163. D. Newman, H. Shapiro, Certain Hilbert spaces of entire functions. Bull. Amer. Math. Sco.
72, 971–977 (1966)
References 337
164. D. Newman, H. Shapiro, Fischer Spaces of Entire Functions, in 1968 Entire Functions and
Related Parts of Analysis, (American Mathematical Society, Providence, RI, 1966)
165. D. Newman, H. Shaprio, A Hilbert Space of Entire Functions Related to the Operational
Calculus, unpublished manuscript, (1964, 1971)
166. J. von Neumann, Foundations of Quantum Mechanics, (Princeton University Press, Princeton,
1955)
167. V.L. Oleinik, Carleson measures and the heat equation. J. Math. Sci. 101, 3133–3188 (2000)
168. J. Ortega-Cerdá, K. Seip, Beurling type density theorems for sampling and interpolation in
weighted L p spaces of entire functions. J. Anal. Math. 75, 247–266 (1998)
169. J. Ortega-Cerdá, K. Seip, Multipliers for entire functions and an interpolation problem of
Beurling. J. Funct. Anal. 161, 400–415 (1999)
170. J. Ortega-Cerdá, K. Seip, Fourier frames. Ann. Math. 155, 789–806 (2002)
171. J. Peetre, Paracommutators and Minimal Spaces, in Operators and Function Theory ed. S.C.
Power (Reidel, Dordrecht, 1985) pp. 163–224
172. J. Peetre, Invariant function spaces and Hankel operators–a rapid survey. Exposition. Math.
5, 5–16 (1987)
173. V.V. Peller, Hankel operators of the class S p and their applications (rational approximation,
Gaussian processes, the problem of majorizing operators). Mat. Sb. 113, 538–581 (1980);
(Russian). English translation: Math. USSR-Sb. 41, 443–479 (1980)
174. V.V. Peller, Hankel Operators and Their Applications, (Springer, New York, 2003)
175. L. Peng, Paracommutators of Schatten–Von Neumann class S p , 0 < p < 1. Math Scand. 61,
68–92 (1987)
176. A.M. Perelomov, On the completeness of a system of coherent states. Theor. Math. Phys. 6,
156–164 (1971)
177. A.M. Perelomov, Generalized Coherent States and Their Applications, (Springer, Berlin,
1986)
178. S. Pichorides, On the best values of the constants in the theorems of M. Riesz, Zygmund, and
Kolmogorov. Studia Math. 44, 165–179 (1972)
179. J. Pool, Mathematical aspects of the Weyl correspondence. J. Math. Phys. 7, 66–76 (1966)
180. S.C. Power, Hankel operators on Hilbert space. Bull. London Math. Soc. 12, 422–442 (1980)
181. S.C. Power, Hankel operators on Hilbert space. Res. Notes Math. 64, 87 (1982)
182. S.C. Power, Finite rank multivariable Hankel forms. Lin. Algebra Appl. 48, 237–244 (1982)
183. D. Quillen, On the representation of Hermitian forms as sums of squares. Invent. Math. 5,
237–242 (1968)
184. E. Ramirez de Arellano, N. Vasilevski, Toeplitz operators on the Fock space with presymbols
discontinuous on a thick set. Math. Nachr. 180, 299–315 (1996)
185. E. Ramirez de Arellano, N. Vasilevski, Bargmann projection, three-valued functions, and
corresponding Toeplitz operators. Contemp. Math. 212, 185–196 (1998)
186. R. Rochberg, Trace ideal criteria for Hankel operators and commutators. Indiana Univ. Math.
J. 31, 913–925 (1982)
187. R. Rochberg, Decomposition Theorems for Bergman Spaces and Their Applications, in
Operators and Function Theory ed. by S.C. Power (Reidel, Dorddrecht, 1985) pp. 225–277
188. C. Rondeaux, Classes de Schatten d’opérateurs pseudo-différentiels. Ann. Sci. École Norm.
Sup. 4, 67–81 (1984)
189. G. Rozenblum, N. Shirokov, Finite rank Bergman–Toeplitz and Bargmann–Toeplitz operators
in many dimensions. Complex Anal. Oper. Theor. 4, 767-775 (2010)
190. W. Rudin, Functional Analysis, 2nd edn. (McGraw–Hill, New York, 1991)
191. W. Rudin, Function Theory in the Unit Ball of Cn , (Springer, New York, 1980)
192. Sangadji, K. Stroethoff, Compact Toeplitz operators on generalized Fock spaces. Acta Sci.
Math. (Szeged) 64, 657–669 (1998)
193. A. Schuster, On Seip’s description of sampling sequences for Bergman spaces. Complex
Variables 42, 347–367 (2000)
194. A. Schuster, The maximum principle for the Bergman space and the Möbius pseudodistance
for the annulus. Proc. Amer. Math. Soc. 134, 3525–3530 (2006)
338 References
195. A. Schuster, K. Seip, A Carleson type condition for interpolation in Bergman spaces. J. Reine
Angew. Math. 497, 223–233 (1998)
196. A. Schuster, K. Seip, Weak conditions for interpolation in holomorphic spaces. Publ. Mat. 44,
277–293 (2000)
197. A. Schuster, D. Varolin, Sampling sequences for Bergman spaces, 0 < p < 1. Complex
Variables 47, 243–253 (2002)
198. I.E. Segal, Lectures at the Summer Seminar on Applied Mathematics, (Boulder, Colorado,
1960)
199. I.E. Segal, Mathematical Problems of Relativistic Physics, (American Mathematical Society,
Providence, RI, 1963)
200. I.E. Segal, The Complex Wave Representation of the Free Boson Field, in Topics in
Functional Analysis: Essays dedicated to M.G. Krein on the occasion of his 70th birthday,
ed. by I. Gohberg, M. Kac (Academic Press, New York, 1978) pp. 321–343
201. K. Seip, Reproducing formulas and double orthogonality in Bargmann and Bergman spaces.
SIAM J. Math. Anal. 22, 856–876 (1991)
202. K. Seip, Regular sets of sampling and interpolation for weighted Bergman spaces. Proc. Amer.
Math. Soc. 117, 213–220 (1993)
203. K. Seip, Interpolation and Sampling in Spaces of Analytic Functions, (American
Mathematical Society, Providence, RI, 2004)
204. K. Seip, Density theorems for sampling and interpolation in the Bargmann–Fock space
I. J. Reine Angew. Math. 429, 91–106 (1992)
205. K. Seip, Beurling type density theorems in the unit disk. Invent. Math. 113, 21–39 (1993)
206. K. Seip, Density theorems for sampling and interpolation in the Bargmann–Fock space. Bull.
Amer. Math. Soc. 26, 322–328 (1992)
207. K. Seip, On Korenblum’s density condition for the zero sequences of A−α . J. Analyse Math.
67, 307–322 (1995)
208. K. Seip, Interpolating and sampling in small Bergman spaces, to appear in Collect. Math.
209. K. Seip, R. Wallstén, Density theorems for sampling and interpolation in the Bargmann–Fock
space II. J. Reine Angew. Math. 429, 107–113 (1992)
210. K. Seip, H. Youssfi, Hankel operators on Fock spaces and related Bergman kernel estimates.
preprint, (2010)
211. S. Semmes, Trace ideal criteria for Hankel operators and applications to Besov spaces. Integr.
Equat. Operat. Theor. 7, 241–281 (1984)
212. M. Shubin, Pseudodifferential Operators and Spectral Theory, (Springer, Berlin, 1987)
213. B. Simon, Trace Ideals and Their Applications, London Math. Soc. Lecture Notes Series 35,
(Cambridge University Press, Cambridge, 1979)
214. P. Sjögren, Un contre-exemple pour le noyau reproduisant de la mesure gaussienne dans
le plan complexe, Seminaire Paul Krée (Equations aux dérivées partienlles en dimension
infinite) 1975/76, Paris
215. E. Stein, Interpolation of linear operators. Trans. Amer. Math. Soc. 83, 482–492 (1956)
216. E. Stein, G. Weiss, Interpolation of operators with change of measures. Trans. Amer. Math.
Soc. 87, 159–172 (1958)
217. R. Strichartz, L p contractive projections and the heat semigroup for differential forms.
J. Funct. Anal. 65, 348–357 (1986)
218. K. Stroethoff, Hankel and Toeplitz operators on the Fock space. Mich. Math. J. 39, 3–16
(1992)
219. R. Supper, Zeros of functions of finite order. J. Inequal. Appl. 7, 49–60 (2002)
220. M. Tatari, S. Vaezpour, A. Pishinian, On some properties of Fock space Fα2 by frame theory.
Int. J. Contemp. Math. Sci. 5, 1107–1114 (2010)
221. D. Timotin, Cp estimates for certain kernels: the case 0 < p < 1, J. Funct. Anal. 72, 368–380
(1987)
222. H. Triebel, Interpolation Theory, Function Spaces, and Differential Operators, (VEB, Berlin,
1977)
223. J.Y. Tung, Fock spaces, Ph.D. thesis, (University of Michigan, Ann Arbor, 2005)
References 339
224. J.Y. Tung, Taylor coefficients of functions in Fock spaces. J. Math. Anal. Appl. 318, 397–409
(2006)
225. J.Y. Tung, Zero sets and interpolating sets in Fock spaces. Proc. Amer. Math. Soc. 134,
259–263 (2005)
226. J.Y. Tung, On Taylor coefficients and multipliers in Fock spaces. Contemp. Math. 454,
135–147 (2008)
227. G. Valiron, Sur la formule d’interpolation de Lagrange. Bull. Sci. Math. 49, 181–192 (1925)
228. N. Vasilevski, V. Kisil, E. Ramirez, R. Trujilo, Toeplitz operators with discontinuous
presymbols in the Fock space. Dokl. Math. 52, 345–347 (1995)
229. D. Vukotić, A sharp estimate for A p functions in Cn . Proc. Amer. Math. Soc. 117, 753–756
(1993)
230. D. Vukotić, On the coefficient multipliers of Bergman spaces. J. London Math. Soc. 50,
341–348 (1994)
231. R. Wallsten, The S p -criterion for Hankel forms on the Fock space, 0 < p < 1. Math. Scand.
64, 123–132 (1989)
232. C. Wang, Some results on Korenblum’s maximum principle J. Math. Anal. Appl. 373,
393–398 (2011)
233. C. Wang, Domination in the Bergman space and Korenblum’s constant. Integr. Equat. Operat.
Theor. 61, 423–432 (2008)
234. C. Wang, Behavior of the constant in Korenblum’s maximum principle. Math. Nachr. 281,
447–454 (2008)
235. C. Wang, On a maximum principle for Bergman spaces with small exponents. Integr. Equat.
Operat. Theor. 59, 597–601 (2007)
236. C. Wang, On Korenblum’s maximum principle. Proc. Amer. Math. Soc. 134, 2061–2066
(2006)
237. C. Wang, An upper bound on Korenblum’s maximum principle. Integr. Equat. Operat. Theor.
49, 561–563 (2004)
238. C. Wang, On Korenblum’s constant. J. Math. Anal. Appl. 296, 262–264 (2004)
239. C. Wang, Refining the constant in a maximum principle for the Bergman space. Proc. Amer.
Math. Soc. 132, 853–855 (2004)
240. E.T. Whittaker, On the functions which are represented by the expansions of interpolation
theory. Proc. R. Soc. Edinburgh 35, 181–194 (1915)
241. E.T. Whittaker, G.N. Watson, A Course of Modern Analysis, 4th edn. (Cambridge University
Press, Cambridge, 1996)
242. J. Xia, D. Zheng, Standard deviation and Schatten class Hankel operators on the Segal–
Bargmann space. Indiana Univ. Math. J. 53, 1381–1399 (2004)
243. J. Xia, D. Zheng, Two-variable Berezin transform and Toeplitz operators on the Fock space,
preprint
244. R.M. Young, An Introduction to Nonharmonic Fourier Series, (Academic Press, New York,
1980)
245. K. Zhu, Positive Toeplitz operators on weighted Bergman spaces of bounded symmetric
domains. J. Operat. Theor. 20, 329–357 (1988)
246. K. Zhu, A Forelli–Rudin type theorem. Complex Variables 16, 107–113 (1991)
247. K. Zhu, Schatten class Hankel operators on the Bergman space of the unit ball. Amer. J. Math.
113, 147–167 (1991)
248. K. Zhu, BMO and Hankel operators on Bergman spaces. Pacific J. Math. 155, 377–397 (1992)
249. K. Zhu, Zeros of functions in Fock spaces. Complex Variables 21, 87–98 (1993)
250. K. Zhu, Operator Theory in Function Spaces, 2nd edn. (American Mathematical Society,
Providence, RI, 2007)
251. K. Zhu, Spaces of Holomorphic Functions in the Unit Ball, (Springer, New York, 2005)
252. K. Zhu, Interpolating and recapturing in reproducing Hilbert spaces. Bull. Hong Kong Math.
Soc. 1, 21–33 (1997)
253. K. Zhu, Evaluation operators on the Bergman space. Math Proc. Cambridge Philos. Soc. 117,
513–523 (1995)
340 References
254. K. Zhu, Interpolating sequences for the Bergman space. Michigan Math. J. 41, 73–86 (1994)
255. K. Zhu, Invariance of Fock spaces under the action of the Heisenberg group. Bull. Sci. Math.
135, 467–474 (2011)
256. K. Zhu, Duality of Bloch spaces and norm convergence of Taylor series. Michigan Math. J.
38, 89–101 (1991)
257. K. Zhu, VMO, ESV, and Toeplitz operators on the Bergman space. Trans. Amer. Math. Soc.
302, 617–646 (1987)
258. K. Zhu, Maximal zero sequences for Fock spaces, preprint, (2011)
259. N. Zorboska, Toeplitz operators with BMO symbols and the Berezin transform. Int. J. Math.
Math. Sci. 46, 2929–2945 (2003)
Index
Symbols S(w, r), square centered at w with side length r,

B(a, r), Euclidean disk, 63 139
BA p , functions of bounded averages, 125 S1 , trace class, 24
BArp , functions of bounded averages, 125 S2 , Hilbert–Schmidt class, 24
BO, functions of bounded oscillation, 124 S p , Schatten class, 24
BOr , functions of bounded oscillation, 124 Tμ , Toeplitz operator on Fα2 , 216
Bα f , Berezin transform of f , 101 Tϕ , Toeplitz operator on Fα2 , 215
C0 (C), space of continuous functions vanishing Ua , weighted translation operator, 76
at ∞, 23 VA p , functions of vanishing averages, 130
Cc (C), space of continuous functions with VArp , functions of vanishing averages, 130
compact support, 23 VO, functions of vanishing oscillation, 130
D, diffferential operator, 19 VOr , functions of vanishing oscillation, 130
D+ (Z), upper (uniform) density, 139 W (Z), weak limits of translates of Z, 165
D− (Z), lower (uniform) density, 139 Wa , Weyl operator, 76
En (z), elementary factor, 4 X, multiplication operator, 19
Fαp , Fock space, 36 Z, the operator X + iD, 19
H f , Hankel operator on Fα2 , 287 Z ∗ , the operator X − iD, 19
Hn (x), Hermite polynomials, 221 [A, B], Hausdorff distance between two sets,
Ht , heat transform, 101 151
I, identity operator, 19 [D, X], commutator, 20
K(z, w), reproducing kernel in Fα2 , 34 [X,Y ]θ , complex interpolation space, 59
KH (z, w), reproducing kernel for H, 78 Γ (a, z), incomplete gamma function, 167
KS (z, w), kernel function induced by S, 99 H, Heisenberg group, 25
Kα (z, w), reproducing kernel in Fα2 , 34 Hn , Heisenberg group, 25
Kw , reproducing kernel in Fα2 , 34 Λ , lattice, 9
Lαp , the space L p (C, d λ pα /2 ), 36 Λ (ω , ω1 , ω2 ), lattice, 9
MO( f )(z), invariant mean oscillation of f at z, Λα , square lattice, 16
127, 290 f |Z p,α , sequence norm, 144
MO p,r ( f )(z), mean oscillation of f on B(z, r), Z, integer group, 9
123 Z2 , integer lattice, 9
M p (Z), stable sampling constant, 145 BMO, bounded mean oscillation, 123
M p (Z, α ), stable sampling constant, 145 BMO p , bounded mean oscillation, 123
Np (Z), stable interpolation constant, 144 BMOrp , bounded mean oscillation, 123
Np (Z, α ), stable interpolation constant, 144 χS , characteristic function, 11
Pα , orthogonal projection from L2 (C, d λα ) δ (Z), separation constant, 143
onto Fα2 , 34 δ (x), δ function, 22
Qα , integral operator, 43 δz , point mass at z, 148

DOI 10.1007/978-1-4419-8801-0, © Springer Science+Business Media New York 2012
342 Index
γ (z), discrete path between 0 and z, 11 C

γ (z, w), discrete path between z and w, 11 Calderón–Vaillancourt theorem, 23
σ̂ (p, q), Fourier transform, 22 canonical decomposition, 129
λα , Gaussian measure, 33 Carleson measure, 117, 148
f , gα , inner product in L2 (C, d λα ), 33 closed-graph theorem, 144
ωr ( f )(z), oscillation of f over B(z, r), 124 commutator, 312
ωmn , lattice points, 9 complex interpolation, 59
ρ p (z, Z), certain “distance” from z to Z, 181 Condition (I1 ), 101
σ (D, X), pseudodifferential operator, 19 Condition (I2 ), 101
τa , translation by −a, 75 Condition (Ip ), 101
tr (T ), trace of T , 96 Condition (M), 216
ϕa , ϕa (z) = a − z, 75 congruent parallelogram, 10
VMO, vanishing mean oscillation, 130
VMO p , vanishing mean oscillation, 130
VMOrp , vanishing mean oscillation, 130 D

f r (z), mean of f over B(z, r), 123 decomposition, 10, 64
μr , averaging function of μ , 246 density, 139
T, Berezin transform of T on Fα2 , 95 diagonal operator, 251, 278, 318

f , Berezin transform of f , 101 diagonalization argument, 152
μ (z), Berezin transform of μ or Tμ , 216 diagonalization process, 152
d(z, S), Euclidean distance from z to S, 18 dilation, 23
f α∞ , Fock space, 39 dilation operator, 36
f r , dilation of f by r, 23 discriminant, 48
h f , small Hankel operator on Fα2 , 269 dominated convergence theorem, 39
hn (x), Hermite functions, 222 double pole, 13
kz , normalized reproducing kernel, 35 doubly periodic, 13
n(Z, S), number of points in Z ∩ S, 139 dual space, 53
rZ2 , square lattice, 11 duality, 53
ta , translation by a, 75
Bα , (parametrized) Bargmann transform, 222
B−1 α , inverse Bargmann transform, 223
E
eigenvalue, 98
eigenvector, 98
A embedding, 56
anti-Wick correspondence, 20 equivalence relation, 258
anti-Wick pseudodifferential operator, 226 even function, 14
antisymmetric function, 255 extremal function, 38
antisymmetric polynomial, 255
antisymmetrization, 255
arithmetic mean, 60 F
atomic decomposition, 63, 277 Fatou’s lemma, 39
finite genus, 6
finite order, 6
B finite rank, 5
Bargmann isometry, 221 finite rank Hankel operator, 281
Bargmann transform, 221 finite rank operator, 255
Berezin symbol, 93 finite rank Toeplitz operator, 255
Berezin transform, 93 fixed points of the Berezin transform, 113
Berezin transform of functions, 101 Fock projection, 61
Berezin transform of operators, 93 Fock spaces, 33
Bergman space, 4, 57, 293 Fock–Carleson measure, 117
big Hankel operator, 287 Fourier inversion formula, 22
bounded mean oscillation, 123 Fourier transform, 20
bounded oscillation, 125 fundamental region, 9, 64, 142, 202
Index 343
G Lindelöf’s theorem, 7, 200

Gaussian measure, 33 Liouville’s theorem, 3
Gaussian weights, 87 Lipschitz, 99
genus, 5 Lipschitz estimate, 99
geometric mean, 60 Lipschitz functions, 289
local oscillation, 124
lower density, 139
H
Hadamard factorization, 6
Hankel operator, 287 M
Hardy space, 4, 57, 293 maximal invariant Fock space, 77
Hausdorff distance, 151 maximum modulus principle, 7
heat equation, 102 maximum order, 41
heat transform, 102, 229 maximum principle for Fock spaces, 81
Heisenberg group, 25, 76 maximum type, 6, 41
Hermite polynomials, 221 mean oscillation, 123
Hilbert–Schmidt class, 96 mean value theorem, 3
Hilbert–Schmidt integral operator, 302 minimal invariant Fock space, 77
modified Weierstrass σ -function, 159
I
ideal, 281
N
identity operator, 20
Nevanlinna–Fock class, 211
identity theorem, 3
normalized reproducing kernel, 35
infinite order, 6
infinite rank, 5
infinite type, 6
initial condition, 102 O
integer group, 9 odd function, 14
integer lattice, 9 optimal rate of growth, 36
integral operator, 43 order, 6
integral pairing, 53 orthogonal projection, 34
integral representation, 35 orthonormal basis, 33
intermediate value theorem, 166
interpolating sequence for Fαp , 143
inverse Bargmann transform, 223 P
inverse Fourier transform, 20 parallelogram, 9
isometry, 76 parametrized Bargmann transform, 222
iterates of the Berezin transform, 110 parametrized Berezin transforms, 105
pathological properties, 199
period, 13
J periodicity, 13
Jensen’s formula, 4 permutation, 255
John–Nirenberg correspondence, 20 permutation invariance, 255
permutation invariant, 257
perturbation, 154
K Planck’s constant, 19, 87
Korenblum’s maximum principle, 87 pseudodifferential operator, 19
L Q
Lagrange-type interpolation formula, 164 quantum physics, 19
Laplacian, 104 quasi-periodic, 15
lattice, 9, 142, 277, 312 quasi-periodicity, 13, 201, 202
344 Index
R trace, 98
rank, 5 trace class, 96
rank of an operator, 258 trace class operator, 96
rank-one operator, 98 trace formula, 213
rank-two operator, 98 translation, 75
rate of growth, 36 translation invariance, 10, 75, 145
relatively closed set, 151 translation operator, 76
reproducing formula, 36
reproducing kernel, 34
Riemann ζ -function, 14
U
Riesz representation, 34
uniform density, 139
uniformly close, 160
S uniqueness sequence, 165
sampling sequence for Fαp , 144 uniqueness set, 165
sampling set, 145 unit mass, 22
Schatten class, 96, 97, 275 unitary operator, 26, 76
Schatten class Hankel operator, 301 unitary representation, 25, 76
Schatten class operator, 96 upper density, 139
Schatten class Toeplitz operator, 96
Schatten classes, 24
Schrödinger representation, 26
Schur’s test, 43 V
Schwarz lemma, 84 Vandermonde determinant, 258
semi-group property, 101 vanishing average, 130
separated sequence, 143 vanishing Carleson measure, 118
separation constant, 143 vanishing Fock–Carleson measure, 118
set of uniqueness, 165 vanishing mean oscillation, 130
small Hankel operator, 267 vanishing oscillation, 130
square lattice, 11, 16 vertices, 9
stability, 151
stable interpolation, 144
standar factorization, 5
Stein–Weiss interpolation theorem, 59 W
Stirling’s formula, 169 weak convergence, 151
Stone–Weierstrass theorem, 258 weak limit, 165
strong convergence, 151 Weierstrass σ -function, 13, 201
strong limit, 165 Weierstrass factorization, 4
strong operator topology, 265 Weierstrass factorization theorem, 202
sub-lattice, 10 Weierstrass functions, 15
symbol, 19 Weierstrass product, 16, 197
symbol calculus, 19 weighted translation operator, 76
symbol function, 19 Weyl pseudodifferential operator, 20
symmetric function, 255 Wick correspondence, 20
symmetric polynomial, 255 Wyle operator, 76
symmetrization, 255
T Z
telescoping decomposition, 306 zero sequence, 193
Toeplitz operator, 213 zero set, 193

Analysis On Fock Spaces, Kehe Zhu (2012)

Uploaded by

Copyright:

Available Formats

Analysis On Fock Spaces, Kehe Zhu (2012)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Analysis On Fock Spaces, Kehe Zhu (2012)

Uploaded by

Copyright:

Available Formats

Graduate Texts in Mathematics 263

Graduate Texts in Mathematics

Colin Adams, Williams College

For further volumes:

Analysis on Fock Spaces

Mathematics Subject Classification (2010): 30H20

© Springer Science+Business Media New York 2012

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Albany, NY, USA Kehe Zhu

4.3 Stability Under Weak Convergence . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 151

In this chapter, we collect several preliminary results about entire functions,

K. Zhu, Analysis on Fock Spaces, Graduate Texts in Mathematics 263, 1

1.1 Entire Functions

has a limit point in C, then f ≡ 0 on C.

for all a ∈ C and all r ∈ [0, ∞).

Theorem 1.4. A bounded entire function is necessarily constant. More generally,

where {z1 , · · · , zN } are zeros of f in 0 < |z| < r.

for all r > 0. Furthermore, the infinite product

f (z) = zm P(z)eg(z) . (1.8)

M(r) = M f (r) = sup{| f (z)| : |z| = r}.

We say that f is of order ρ if

It is clear that 0 ≤ ρ ≤ ∞. When ρ < ∞, f is said to be of finite order; otherwise, f

| f (z)| < exp(|z|a ), |z| > r.

will be denoted by ρ1 = ρ1 ( f ). The smallest positive integer s satisfying the

1.2 Lattices in the Complex Plane

is then called the lattice generated by ω , ω1 , and ω2 .

for every f ∈ L1 (C, dA).

Then each Λ j is a sublattice of Λ ; the distance between any two points in Λ j is

∑ ∑ e−σ |z−w| χγ (z,w) (u) ≤ C

which implies that the sum

∑ ∑ e−σ |z−w| χγ (z,w) (u)

is actually independent of u. For convenience, we will assume that u = 0.

n=0 m=0 k=−∞

n=0 m=0 k=−∞

This proves the lemma.

1.3 Weierstrass σ -Functions

Λ = Λ (0, ω1 , ω2 ) = {ωmn }, ωmn = mω1 + nω2 .

To every such lattice, we associate a function P(z) = PΛ (z) as follows:

P(z + ω1 ) = P(z), P(z + ω2 ) = P(z), (1.10)

Uδ = {z ∈ C : d(z, Λ ) > δ , |z| < 1/δ }.

It is clear that for z ∈ Uδ we have

when |ωmn | is large. Since

the series in (1.9) converges uniformly and absolutely to an analytic function in Uδ .

To see that P is doubly periodic with periods ω1 and ω2 , we differentiate

ζ (z + ωk ) = ζ (z) + ηk , z ∈ C −Λ, (1.12)

where ηk = 2ζ (ωk /2).

as |ωmn | → ∞. It follows that the series in (1.11) converges to an analytic function

If we integrate the equation P(z + ωk ) = P(z) on the connected region C − Λ , we

Proof. If we pull the center c = (ω1 + ω2 )/2 of the parallelogram spanned by ω1

σ (z + ωk ) = −eηk (z+(ωk /2)) σ (z), (1.15)

where k = 1, 2 and ηk are the constants from the previous proposition.

If we integrate the equation

for all z ∈ C and ωmn ∈ Λα .

where m = −m. It follows that

ωmn = mω1 + nω2.

It follows from (1.15) and induction that

σ (z + mω1 ) = (−1)m σ (z)emη1 z+ 2 m

σ (z + nω2 ) = (−1)n σ (z)enη2 z+ 2 n

(z, s) ⊕ (w,t) = (z + w, s + t − Im (z, w)),