Visit [Link]

com to download the full version and

explore more ebook

The Psychology of Aging An Introduction 4th Edition

Ian Stuart-Hamilton

_____ Click the link below to download _____

[Link]
aging-an-introduction-4th-edition-ian-stuart-hamilton/

Explore and download more ebook at [Link]

Here are some recommended products that might interest you.
You can download now and explore!

Evolutionary Psychology An Introduction 1st Edition Lance

Workman

[Link]
introduction-1st-edition-lance-workman/

[Link]

The Person An Introduction to the Science of Personality

Psychology 5th Edition Dan P. Mcadams

[Link]
science-of-personality-psychology-5th-edition-dan-p-mcadams/

[Link]

Accounting An Introduction 4th Edition Peter Atrill

[Link]
edition-peter-atrill/

[Link]

Handbook of Health Psychology and Aging 1st Edition Ronald

P. Abeles

[Link]
aging-1st-edition-ronald-p-abeles/

[Link]
Introduction to Nanoscience Stuart Lindsay

[Link]
lindsay/

[Link]

Diseases Of Sheep 4th Edition Ian Aitken

[Link]
aitken/

[Link]

Digital Photography An Introduction 4th Edition Tom Ang

[Link]
introduction-4th-edition-tom-ang/

[Link]

Film art an introduction 4th Edition David Bordwell

[Link]
david-bordwell/

[Link]

Psychopharmacology An Introduction 4th Edition René

Spiegel

[Link]
introduction-4th-edition-rene-spiegel/

[Link]
The Psychology of Aging An Introduction 4th Edition Ian
Stuart-Hamilton Digital Instant Download
Author(s): Ian Stuart-Hamilton
ISBN(s): 9781846425011, 1846425018
Edition: 4
File Details: PDF, 1.49 MB
Year: 2006
Language: english
Conference Board of the Mathematical Sciences

CBMS
Regional Conference Series in Mathematics
Number 121

Toeplitz Approach
to Problems of the
Uncertainty Principle
Alexei Poltoratski

American Mathematical Society

with support from the
National Science Foundation
 [Link]

Conference Board of the Mathematical Sciences

CBM S
Regional Conference Series in Mathematics
Number 121

Toeplitz Approach
to Problems of the
Uncertainty Principle
Alexei Poltoratski

Published for the

Conference Board of the Mathematical Sciences
by the
American Mathematical Society
Providence, Rhode Island
with support from the
National Science Foundation
 NSF-CBMS Regional Conference in the Mathematical Sciences:
Uncertainty Principles in Harmonic Analysis: Gap and Type Problems,
held at Clemson University, Clemson, South Carolina, August 12–16, 2013.
Partially supported by the National Science Foundation
2010 Mathematics Subject Classification. Primary 30-XX, 33-XX, 34-XX, 42-XX.

For additional information and updates on this book, visit

[Link]/bookpages/cbms-121

Library of Congress Cataloging-in-Publication Data

Poltoratski, Alexei, 1966–
Toeplitz approach to problems of the uncertainty principle / Alexei Poltoratski.
pages cm. — (CBMS regional conference series in mathematics ; number 121)
“Partially supported by the National Science Foundation.”
Based on the NSF-CBMS Regional Conference in the Mathematical Sciences on Uncertainty
Principles in Harmonic Analysis: Gap and Type Problems, held at Clemson University, Clemson,
South Carolina, August 12–16, 2013.
Includes bibliographical references.
ISBN 978-1-4704-2017-8 (alk. paper)
1. Heisenberg uncertainty principle—Congresses. 2. Functional analysis—Congresses.
3. Toeplitz, Otto, 1881–1940. I. National Science Foundation (U.S.) II. Title.
QC174.17.H4P65 2015
530.12—dc23
2014043343

Copying and reprinting. Individual readers of this publication, and nonprofit libraries
acting for them, are permitted to make fair use of the material, such as to copy select pages for
use in teaching or research. Permission is granted to quote brief passages from this publication in
reviews, provided the customary acknowledgment of the source is given.
Republication, systematic copying, or multiple reproduction of any material in this publication
is permitted only under license from the American Mathematical Society. Permissions to reuse
portions of AMS publication content are handled by Copyright Clearance Center’s RightsLink
service. For more information, please visit: [Link]
Send requests for translation rights and licensed reprints to reprint-permission@[Link].
Excluded from these provisions is material for which the author holds copyright. In such cases,
requests for permission to reuse or reprint material should be addressed directly to the author(s).
Copyright ownership is indicated on the copyright page, or on the lower right-hand corner of the
first page of each article within proceedings volumes.

c 2015 by the American Mathematical Society. All rights reserved.
The American Mathematical Society retains all rights
except those granted to the United States Government.
Printed in the United States of America.

∞ The paper used in this book is acid-free and falls within the guidelines
established to ensure permanence and durability.
Visit the AMS home page at [Link]
10 9 8 7 6 5 4 3 2 1 20 19 18 17 16 15
 “We demand rigidly defined areas
of doubt and uncertainty!”

Douglas Adams

iii
 Contents

Thanks vii
Chapter 1. Mathematical Shapes of Uncertainty 1
1. Basic notations 3
2. Variety of mathematical forms of UP 4
3. Function theoretic background 10
Chapter 2. Gap Theorems 17
1. Classical Gap Theorems 18
2. Spectral gap as a property of the support 22
3. Toeplitz kernels and uniform approximation 23
4. A formula for the gap characteristic of a set 24
5. Examples and applications 27
6. Appendix: Proof of the gap formula 29
7. Appendix: Technical lemmas 51
8. Appendix: De Branges’ theorem 66 in Toeplitz form 61
Chapter 3. A Problem by Pólya and Levinson 67
1. Pólya sequences 67
2. A theorem on existence of a de Branges’ space in L2 68
3. Beurling–Malliavin densities 69
4. Two theorems on Toeplitz kernels 70
5. A description of Pólya sequences 71
6. Technical lemmas 72
7. Proofs of theorems 74
Chapter 4. Determinacy of Measures and Oscillations of High-pass Signals 77
1. Sign changes of a measure with a spectral gap 77
2. Entire functions and densities 79
3. A lemma on d-uniform sequences 80
4. M. Riesz-type criterion and its consequences 81
5. Extreme measures in the indeterminate case 85
6. Measures annihilating Paley-Wiener spaces 90
7. Sign changes of measures with spectral gap 93
Chapter 5. Beurling–Malliavin and Bernstein’s Problems 95
1. A problem on completeness of exponentials 95
2. Structure of proof of BM Theorem: BM theory 99
3. Bernstein’s problem 101
4. Semi-continuous weights 102
5. Characteristic sequences 103
v
vi CONTENTS

6. Equivalence between weighted uniform and Lp -approximation 104

7. A criterion for completeness of polynomials in CW 105
8. Lemmas and proofs 106
9. Examples and corollaries 114
Chapter 6. The Type Problem 119
1. General case p = 2 120
2. Known examples 121
3. Polynomial decay 122
4. Completeness of exponentials in Lp and CW 123
5. Main results 124
6. Classical results and further corollaries 126
7. Auxiliary statements 130
8. Proofs of main results 139
Chapter 7. Toeplitz Approach to UP 143
1. Spaces of entire functions and their zero sets 143
2. Model spaces and Toeplitz operators 144
3. Spectral theory 145
4. Outer, inner and Herglotz functions 146
5. Model spaces 148
6. Weyl inner functions 148
7. Modified Fourier transform 149
8. Entire functions 151
9. Square root transformation 153
10. Dimension and triviality of Toeplitz kernels 153
11. General form of Levinson’s completeness theorem 161
12. Applications to UP 164
Chapter 8. Toeplitz Version of the Beurling–Malliavin Theory 183
1. The language of Toeplitz kernels 184
2. Super-exponential case 186
3. Sub-exponential case 187
4. The structure of BM theory 189
5. One-sided Lipschitz condition for the Hilbert transform 189
6. Triviality of Toeplitz kernels 193
7. Non-triviality of Toeplitz kernels in Smirnov–Nevanlinna class 197
8. Multiplier theorem 200
9. Non-triviality of Toeplitz kernels in Hardy spaces 204
Bibliography 211
 Thanks

These notes are based on a Conference Board of the Mathematical Sciences

mini-course given in August of 2013 at Clemson, South Carolina. I am deeply
grateful to Constanze Liaw, Mishko Mitkovski and Brett Wick for inviting me to be
a CBMS speaker and for all the hard work they put into organizing the conference.
More than 40 researchers attended the conference, several of whom contributed
outstanding lectures. The lectures were given by Michael Lacey, Doron Lubinsky,
Shahaf Nitzan, Bill Ross, Carl Sundberg and Sergei Treil.
On behalf of the organizers and participants of the conference, I wish to thank
the National Science Foundation and Conference Board of the Mathematical Sci-
ences for their generous support.
Much credit is due to Nikolai Makarov, whose ideas initiated the Toeplitz ap-
proach in the area of the Uncertainty Principle. The last two chapters of the
notes are devoted to our joint work, where these methods were developed. I also
would like to thank Misha Sodin for convincing me to apply this approach to the
Gap and Type problems and for numerous invaluable discussions. My interest in
the Pólya-Levinson problem and Bernstein’s weighted uniform approximation also
stems from communications with Nikolai and Misha. I am grateful to Alexander
Eremenko for bringing the problem on oscillations of Fourier integrals to my atten-
tion. Thought-provoking collaboration with my former student Mishko Mitkovski
produced chapters 3 and 4.
Original results discussed in these notes were obtained with support of the NSF
through grants DMS-1101278, DMS-0800300 and DMS-0500852.
I am especially indebted to my wife Svetlana for helping me find the classical
texts mentioned in the notes and for constant moral support.

vii
 [Link]

CHAPTER 1

Mathematical Shapes of Uncertainty

For a non-mathematical audience, the term ‘Uncertainty Principle’ is unequiv-

ocally associated with Werner Heisenberg’s 1927 statement, which has become one
of the fundamental constituents of quantum mechanics. In his ground-breaking pa-
per [63] Heisenberg explained the new principle in physical terms. A mathematical
justification quickly followed via the work of Kennard, Pauli and Weyl [75, 148].
The so-called Heisenberg Inequality, see (1.3), contained in Weyl’s work and attrib-
uted to Pauli, illustrates the mathematical meaning of the principle in terms of one
of the fundamental notions of Harmonic Analysis, the Fourier transform.
From 1924 to 1927 Heisenberg held a position of Privatdozent in Göttingen,
where together with Max Born and Pascual Jordan he worked on the foundations of
quantum mechanics. In 1925, Göttingen was visited by Norbert Wiener who gave a
seminal lecture in Harmonic Analysis. The lecture intrigued the famous Göttingen
scientists including Hilbert, Courant and Born. In his memoir [149] Wiener writes:
“The talk which I gave the Göttingen people on my work on general harmonic
analysis was very well received. Hilbert, in particular, showed great interest in the
subject, but what I did not realize at that time was that my talk was closely keyed
to the new ideas of physics which were about to burst into bloom at Göttingen in
the form of what is now known as quantum mechanics.”
Wiener’s lecture concerned the basic question of Harmonic Analysis, the prob-
lem of decomposing a wave into a linear combination of simple harmonics. One of
the main points of the lecture, according to Wiener’s popular-science account in
[149], was the statement that a function and its Fourier transform cannot both be
localized. Aiming at a broader audience, Wiener discusses this statement using the
example of a musical note which cannot be both short in time and low in frequency.
From the point of view of the Fourier analysis, Heisenberg’s inequality expresses a
very similar property which looks like the next step in Wiener’s line of reasoning.
Although the exact contents of Wiener’s Göttingen lecture seem to be lost, it may
be viewed as one of the early declarations of the Uncertainty Principle in Harmonic
Analysis, which happened two years before Heisenberg’s paper was published. A
possible conclusion to this historical remark is that, perhaps, it would be incorrect
to think that the Uncertainty Principle came to Harmonic Analysis from Quantum
Mechanics, or vice versa. Rather, like many fundamental ideas, it seems to have
emerged as a result of interaction between two adjacent areas of science.
Mathematical aspects of the Uncertainty Principle were studied by many promi-
nent mathematicians throughout the twentieth century. To this day it remains one
of the classic, yet active and exciting areas of research. New methods that ap-
peared in this area in the last 10-12 years have produced solutions to long standing
problems and sparked new interest towards the Uncertanty Principle from analysts
worldwide. Samples of this recent activity will be discussed in the present text.

1
2 1. MATHEMATICAL SHAPES OF UNCERTAINTY

Following the ideas of Wiener, the original mathematical version of the Uncer-
tainty Principle can be formulated as follows: A function (measure, distribution)
and its Fourier transform cannot be simultaneously small. The informal term ‘small’
in this statement can be given a variety of mathematical meanings, each leading
to a deep and important problem of analysis, see for instance [60]. Examples of
such problems will be discussed in section 2. As it often happens in mathematics,
these problems can then be viewed in more general and abstract settings, sometimes
developing into a whole branch of analysis.
One of such branches is the study of ‘basis’ properties of more and more gen-
eral families of functions. Via its very definition, the Fourier transform is directly
related to the family of complex exponentials and questions on bases, frames and
completeness of families of such functions (some of them will be discussed below).
But the exponentials are not the only natural family for such studies and similar
questions can be asked about every new set of ‘harmonics.’ Following this direction
one may arrive at other classical problems, such as the moment problem, prob-
lems of completeness of various families of polynomials and special functions, and
proceed further to wavelet expansions or other bases and frames, see for instance
[24, 39].
Another example of broader meaning of the Uncertainty Principle is the clas-
sical theory of zero sets of entire functions. If a system of exponentials E = {eiλn t }
is incomplete in L2 ([−1, 1]) then this space must contain a non-trivial function f
orthogonal to all the exponentials in the family. Then the Fourier transform fˆ of
this function is an entire function that vanishes at {λn }. To establish completeness
of E in L2 ([−1, 1]), one needs to prove a version of the Uncertainty Principle: f and
fˆ cannot be simultaneously small, in the sense that f cannot have a small support,
contained in the interval [−1, 1], while fˆ is being small by vanishing on the large
set {λn }. Problems of this sort were studied in the 1930’s and 40’s by Paley and
Wiener, Levinson and many other prominent mathematicians, see for instance [97].
The smallness of the support of f immediately translates into the exponential type
of the entire function fˆ, i.e. into a restriction on its growth in the complex plane.
After that one can use use standard analytic tools, such as Jensen’s inequalty, to
obtain restrictions on density and balance of the zero set {λn }. More advanced
methods give further, more subtle results, while the problem gradually expands be-
yond the Paley-Wiener classes, i.e. beyond Fourier transforms of functions from L2
on an interval. A more general question is how the growth estimates translate into
restrictions on the zero sets for broader classes of entire functions. Even though
the Fourier transform disappears from this general formulation, it is clearly still a
version of the Uncertainty Principle. Such problems were at the center of complex
analysis for a large part of the 20th century, e.g. [94].
Further generalizations of the Uncertainty Principle (UP) include Fourier anal-
ysis on groups (see [49, 50]), problems in signal processing and time-frequency anal-
ysis stemming from a seminal paper by Gabor [51] (see also [11]), and many other
important problems of analysis. At present, a meaningful survey of all branches of
UP would require several volumes. In these notes we certainly do not aim at giving
such a survey. Fortunately, many good texts on various topics of UP are available.
Let us mention the book by Havin and Jöricke [60] and the paper by Folland and
Sitaram [50] (which, to a certain extent, complement each other) as excellent initial
 1. BASIC NOTATIONS 3

sources of results and references. Further references will be given in section 2 and
throughout these notes.
Our goal is to look only at a small sample of classical problems of Fourier anal-
ysis that stem from Wiener’s original ideas of 1920’s, together with their modern
extensions and generalizations. Among them are Beurling’s Gap Problem on the
Fourier transform, classical completeness problems, such as the Beurling–Malliavin
problem and the Type Problem on completeness of trigonometric polynomials in
L2 , a problem on oscillations of Fourier integrals with a spectral gap, etc. The
discussion will include a brief outline of the Toeplitz approach to the problems of
UP developed recently in our joint work with Nikolai Makarov [103, 104]. One of
the extensions of UP that we will consider involves the so-called Weyl-Titchmarsh
transform, a generalization of the Fourier transform that appears in spectral prob-
lems for differential operators. Via the Weyl-Titchmarsh transform, the traditional
Payley-Wiener spaces are replaced with de Branges spaces of entire functions or
with model subspaces of the Hardy space in the upper half-plane. The traditional
completeness problems of systems of complex exponentials become problems on
completeness of special functions or spectral problems for Schrödinger equations
and Krein’s canonical systems. The Toeplitz approach allows one to use methods
of complex function theory and harmonic analysis to solve some of such problems.
One of my goals when writing these notes was to make it possible to read each
chapter from 2 to 8 independently. This led to numerous repetitions in definitions
and arguments, for which I apologize in advance. Additionally, after giving it some
thought, I have decided against including a list of open problems. The main reason
for this omission was that most of the problems I had in mind were still raw and
uneven in quality. At the same time, I hope very much that an interested reader
can find plenty of open problems on her/his own while looking through the text,
especially in the last two chapters.

1. Basic notations
If f is a function from L1 (R) we denote by fˆ its Fourier transform

(1.1) ˆ
f (z) = f (t)e−2πizt dt.
R

The function fˆ(z) is well defined for z ∈ R and, under various additional
conditions on f , may be defined in a larger subset of the complex plane C.
Let M be a set of all finite Borel complex measures on the real line. Similarly,
for μ ∈ M we define 
μ̂(z) = e−2πizt dμ(t).
R
Via Parseval’s theorem, the Fourier transform may be extended to be a unitary
operator from L2 (R) into itself and can be defined for even broader classes of
distributions. It can also be extended to functions (measures, distributions) in Rn
using the same formula

ˆ
f (s) = f (t)e−2πi<s,t> dmn (t),
Rn

where mn is the Lebesgue measure in Rn , s, t ∈ Rn and < s, t > is the standard

scalar product.
4 1. MATHEMATICAL SHAPES OF UNCERTAINTY

If one extends all functions from L2 (−a, a) as 0 to the rest of the line, one
can apply the Fourier transform to all such functions and obtain the Paley-Wiener
space of entire functions
PWa = {fˆ|f ∈ L2 (−a, a)}.
By the Paley-Wiener theorem, an alternative definition of PWa is that it is the
space of all entire functions F such that F has exponential type at most 2πa, i.e
satisfies
|F (z)| < Ce2πa|z| , z ∈ C,
for some C > 0, and F ∈ L2 (R).
In addition to the Lebesgue measure on R that is infinite, the standard finite
measure on the real line is the Poisson measure Π,
dx
dΠ(x) = .
1 + x2
We call functions from L1Π = L1 (R, dx/(1 + x2 )) Poisson-summable and measures
satisfying 
d|μ|(x)
<∞
1 + x2
Poisson-finite.
In our estimates we write a(n)  b(n) if a(n) < Cb(n) for some positive constant
C, not depending on n, and large enough |n|. Similarly, we write a(n)  b(n) if
ca(n) < b(n) < Ca(n) for some C ≥ c > 0. Some formulas will have other
parameters in place of n or no parameters at all. For instance,  may be put
between two improper integrals to indicate that they either both converge or both
diverge.
If I is an interval on R and C > 0 we denote by |I| the length of I and by CI
the interval with the same center as I of length C|I|. If A is a subset of Rn we
write |A| for the Lebesgue measure of A.

2. Variety of mathematical forms of UP

In this section we discuss several mathematical statements stemming from the
general formulation of UP given in the introduction. The purpose of this section
is to give the reader an idea of the variety of mathematical reformulations of UP.
As was mentioned before, a full survey of such results would require a much larger
text. The selection presented here is somewhat random and only represents the
results closely related to the topics discussed in the main body of the text. To see a
more complete picture the reader may look at a number of excellent existing books
and papers, such as [10, 11, 39, 50, 82, 58, 60, 83].

2.1. Heisenberg’s inequality. If f ∈ L2 (R) is a function of unit norm, then

|f (x)|2 dx is a probability measure on R. In physical settings, f may appear as a
wave function, in which case |f |2 is the probability density of the position of the
corresponding wave/particle,
 in the sense that the probability of finding the particle
in a Borel set E is E |f (x)|2 dx.
The Fourier transform fˆ of the wave function f is the representation of the
wave in the momentum space. Since by Parseval’s theorem fˆ has the unit L2 -norm
as well, |fˆ(x)|2 dx is another probability measure.
 2. VARIETY OF MATHEMATICAL FORMS OF UP 5

This interpretation remains meaningful even in certain non-L2 cases. For in-
stance, when f (x) = eiλx is a single-moded wave (harmonic) with the wave number
λ ∈ R, then |f (x)|2 dx is the infinite Lebesgue measure but the momentum of the
wave can be calculated as p = λ. The Fourier transform fˆ, understood in the
sense of distributions, is a point-mass placed at p, which reflects the fact that in
this particular situation the moment of the particle can be calculated precisely.
In the L2 -case, f is an infinite combination of harmonics and its portrait in the
momentum space cannot be localized. The classical Heisenberg’s inequality [148]
expresses this fact in terms of the variance of the measures |f (x)|2 dx and |fˆ(x)|2 dx.
If μ is a probability measure on R we define its variance V (μ) as

inf (x − a)2 dμ(x)
a∈R

with an agreement that for absolutely continuous measures μ = ρ(x)dx we write

V (ρ) instead of V (μ).
Heisenberg’s inequality says that if f ∈ L2 (R) then
||f ||42
(1.2) V (|f |2 )V (|fˆ|2 ) ≥ .
16π 2
A simpler (but easily equivalent) version of the same inequality estimates the
product of second moments of the measures:
 
||f ||42
(1.3) x2 |f (x)|2 x2 |fˆ(x)|2 dx ≥ .
R R 16π 2
The equation occurs when f is a Gaussian. The proof is straight-forward via inte-
gration by parts, Parseval’s theorem and the Cauchy-Schwarz inequality.
A stronger inequality was conjectured by Hirschman [67] and Everett [47] in
1957 and proved by Beckner [9] in 1975. If μ, dμ(x) = ρ(x)dx, is a probability
measure, its Shannon entropy H(μ) (H(ρ)) is defined as

H(ρ) = − ρ(x) log ρ(x)dx.
R
If f ∈ L2 (R), ||f ||2 = 1 then the Hirschman uncertainty of f , defined as H(|f |2 ) +
H(|fˆ|2 ), is non-negative. Moreover, it has a lower bound
e
H(|f |2 ) + H(|fˆ|2 ) ≥ log
2
with the equality, once again, achieved when f is a Gaussian.
Many further variations and generalizations of Heisenberg’s inequality can be
found in mathematical literature, see [50] for a detailed discussion and further
references.
An integral inequality is not the only way to express UP in the settings of
Harmonic Analysis. As was mentioned in the introduction, the following more
abstract statement is universally considered to be its proper formulation: f and
fˆ cannot be simultaneously small.
The ‘smallness’ in this statement can be understood in a variety of ways, each
producing its own problem. If this statement is applied to an L2 -function, the norm
must of course be fixed beforehand, as was done in our discussion above.
Heisenberg’s inequality interprets smallness in the sense of the second moment,
while Hirschman’s inequality measures smallness with entropy. Many other natural
meanings of ‘smallness’ of f and fˆ have been considered by mathematicians, some
6 1. MATHEMATICAL SHAPES OF UNCERTAINTY

of them launching whole areas of Harmonic and Complex Analysis. For instance,
it is natural to consider the support of a measure and pronounce it small if it is
bounded, or has finite volume (length) or if it is thin (porous) in some sense. The
most basic statement of that type, the one that Wiener relays to non-mathematical
audience using musical analogies in his memoir [149], is that the supports of f and
fˆ cannot both be compact. Further statements in the same direction are given in
the next subsection.
Another natural way to understand the smallness of f is in terms of its decay
near infinity. In many situations the Fourier transform fˆ is an analytic function that
can be considered to be small if it has a large zero set. Of course, the smallness of
f and fˆ does not have to be understood in the same way and many deep problems
appear when the criteria for smallness applied to these two objects is different.
As was mentioned above, the Fourier transform does not have to be a part of
the statement and similar questions can be formulated about other transforms or,
more generally, a variety of pairs of ‘dual’ properties of f . Despite rather extensive
literature, many of such questions remain open.
In the rest of this section we give further examples of particular problems of UP,
with smallness in the general statement understood in terms of support, decay or
the size of the zero set. We then discuss an immediate connection of such questions
with completeness problems.

2.2. Support/Support. As was mentioned above, perhaps the simplest

statement of that type is the observation that the supports of f and fˆ on the
real line cannot both be bounded. A more advanced statement is a corollary of a
theorem by F. and M. Riesz that says that if μ ∈ M has semi-bounded support
then the support of μ̂ contains the whole line R.
The next significant step in the same direction begins with results of Amrein-
Berthier [5] and Benedicks [12] that imply that the Lebesgue measures of the
supports of f and fˆ cannot both be finite. These results were later improved by
Nazarov [114] who showed that for any two sets T, S ∈ Rn

||f ||L2 (Rn ) < C exp (C|S||T |)(||f ||L2 (S c ) + ||fˆ||L2 (T c ) ),

where T c , S c denote the complements of T, S correspondingly. If one of the sets is
convex, the product |S||T | in the above inequality may be raised into the power
1/n. It is unknown if the same is true in general.
As we can see, even in this case of UP there are subcases corresponding to
different meanings of smallness of the supports of f and fˆ. Another set of problems
appears if one understands such smallness in terms of porosity. One of the classical
results in this group is the following theorem by Beurling from 1961 [14].
A collection of disjoint intervals {In } on R (or the union of these intervals,
when it is more convenient) is called long if
 |In |2
=∞
n
1 + dist2 (In , 0)

or short if the sum is finite. Here |In | denotes the length of In . Such collections of
intervals appear in many results in the area of UP including those discussed further
in this text.
 2. VARIETY OF MATHEMATICAL FORMS OF UP 7

Theorem 1. [Beurling’s Gap Theorem] Let μ ≡ 0 be a finite complex measure

on R. If the complement of the support of μ is long then the support of μ̂ does not
have any gaps, i.e. there is no interval on R where μ̂ is identically 0.
We will further discuss this result in chapter 2. Beurling’s Gap Theorem is one
of the original results on the Gap Problem, the problem of finding the maximal size
of the gap in the support of μ̂ in terms of the metric characteristics of the support
of μ. The Gap Problem is one of the main topics of these notes.
2.3. Decay/Decay. This version of UP says that f and fˆ cannot both decay
fast near infinity. A classical result of this type was proved by Hardy in 1933 [59].
Assume that
|f (x)| < C(1 + |x|n )e−aπx , and |fˆ(x)| < C(1 + |x|n )e−bπx
2 2

for some a, b, C > 0, n ∈ N. If ab > 1 then f ≡ 0. If ab = 1 then f = pe−aπx for

2

some polynomial p of degree at most n.

A year later Morgan [110] replaced the right-hand sides of the inequalities
p q
with eax and ebx respectively and proved that if p1 + 1q < 1 then f ≡ 0. He
also described the constants a, b that give the same conclusion when p1 + 1q = 1.
More general functions in the right-hand sides, with integral inequalities in place
of pointwise ones were later considered by Dzhrbashyan [44].
In the same section of UP one may mention another classical result by Beurling
that says that if f ∈ L1 (R) and

|f (x)||fˆ(y)|e|x||y| dxdy < ∞
R×R

then f ≡ 0. An extension of this result was obtained by Hedenmalm in 2012 [62].

2.4. Decay/Support or Support/Decay. If f decays fast near infinity then
fˆ cannot have small support. One of the classical theorems in this direction belongs
to Levinson [97]. This result will be further discussed in chapter 2.
Theorem 2. [Levinson’s Gap Theorem] Let μ be a finite measure on R. Denote
M (x) = |μ|((x, ∞)).
Suppose that log M is not Poisson-summable on R+ . If μ̂ vanishes on an interval
then μ ≡ 0.
This statement was later improved by Beurling [14] who replaced the interval
with any set of positive Lebesgue measure. A theorem by Volberg [146, 147, 23]
showed that under some regularity conditions on M , the conclusion can be further
strengthened to say that log |μ̂| must be Poisson summable. Further results in this
direction are due to Borichev [20, 23].
Using the fact that the Fourier transform of fˆ is f (−x), one can now approach
the same problem from the opposite direction and look for statements which say
that if f has small support then fˆ cannot decay too fast. Another consequence
of the F. and M. Riesz theorem tells us that if supp f is semibounded then log |fˆ|
is Poisson-summable. The theorem by Volberg mentioned in the last subsection is
a vast generalization of this statement. Once again, smallness of the support can
be understood in many different ways. For instance, a special place in this part of
UP is occupied by problems about measures whose ‘essential’ support has measure
8 1. MATHEMATICAL SHAPES OF UNCERTAINTY

zero, i.e. measures mutually singular with Lebesgue measure (singular measures).
Let us start with the following classical result.
A sequence an ≥ 0 is calm if can ≤ am ≤ Can for all m ∈ [n, 2n] and some
fixed c, C > 0.
Theorem 3 (Ivashov-Musatov, 1952). Let μ be a finite singular measure on
[0, 1]. If μ̂(n) ∈ l2 then μ ≡ 0. However, if an ≥ 0 is any calm sequence that is not
in l2 then there exists a singular measure μ with |μ̂(n)| < an .
The calmness condition can be replaced with other regularity requirements. In
this regard let us mention the result by Körner [90, 1987] where an ∈ l2 is even
and decreasing for positive n.
The fact that a singular measure may have a Fourier transform that tends to
zero at infinity was known much earlier. It belongs to an interesting branch of
Fourier analysis that we would like to discuss in the remainder of this subsection.
We will call a finite complex measure on R an R-measure if μ̂(t) → 0 as t → ±∞.
Similarly, for measures on the unit circle T, an R-measure is the measure whose
Fourier coefficients μ̂(n) = z n dμ(z) tend to zero as |n| → ∞. In this name ‘R’
stands for A. Rajchman who first undertook a systematic study of such measures
in 1920’s.
The basic facts about R-measures are as follows. Every finite absolutely con-
tinuous measure, i.e. a measure of the form f (x)dx, f ∈ L1 (R), is an R-measure.
This was proved by Riemann in 1854 for Riemann-integrable f and extended by
Lebesgue in 1903 to Lebesgue integrable densities.
The property of being and R-measure is ‘spectral,’ i.e. if μ is an R-measure
then for all f ∈ L1 (|μ|) , f μ is an R-measure. Hence, any measure splits into R-
and non-R- parts: μ = μR + μnR where μR is an R-measure and μnR is completely
non-R, in the sense that any measure absolutely continuous with respect to μnR
is not an R-measure, unless it is trivial. Such a decomposition appears in physical
settings: if μ is a spectral measure of a Hamiltonian then μR and μnR correspond
to bound and transient states of the quantum system, see for instance [124].
Another important property is that μ̂(t) → 0 as |t| → ∞ if and only if μ̂(t) → 0
as t → ∞ (−∞) and similarly for the circle case.
The first UP-style problem in this area, motivated by the Riemann theorem
above, is whether measures suppported on zero sets (singular measures) can be R-
measures. The answer is positive and the first example of a singular R-measure was
given by Menshov in 1916 [111]. On the other hand, we have the following theorem
by Wiener, 1924 [151]. We denote by μp the pure-point part of the measure μ.
Theorem 4. Let μ be a finite measure on R. Then
1 
n
lim |μ̂(k)|2 = ||μp ||.
n→∞ 2n + 1
−n
In particular, the measure is continuous if and only if the limit is 0.
Hence, measures with point masses cannot be R-measures. Wiener’s theorem
immediately raises another historical problem: do there exist continuous measures
that are not R-measures? To answer this question in the positive F. Riesz developed
his products in 1918 [131]. He seemingly overlooked a simple fact that the standard
3 -Cantor measure satisfies μ̂(3n) = μ̂(n) = 0 and therefore provides an example of
1

such a measure.
 2. VARIETY OF MATHEMATICAL FORMS OF UP 9

Connections with adjacent fields, such as problems on convergence of trigono-

metric polynomials, made description of all R-measures an important classical prob-
lem. A solution suggested by Shreĭder in 1950 [141] involved the so-called Weyl
sets. The statement that he published without a proof, following the style of the
Soviet magazine “Doklady,” said that R-measures are exactly those measures that
do not see Weyl sets. The proof was given by Lyons in 1983 [101].
Lyons’ theorem answers a long standing question on the possibility of a de-
scription of R-measures in terms of a collection of sets, as stated above. Such
a set-theoretic answer was needed in particular due to the connection with con-
vergence problems for trigonometric polynomials. It is still unclear if an analytic
condition, in terms of local symmetry and mass distribution of the measure, can be
found.
For more on R-measures see [81, 83, 102].
2.5. Support/Sampling sets. In this version of UP we claim that if f has
small support then fˆ cannot be small in L2 norm on a fixed subset of Rn , if that
subset is large enough.
More precisely, a set S ⊂ Rn is determining, if for any bounded Σ ⊂ Rn there
exists C = C(Σ) > 0 such that

(1.4) supp f ⊂ Σ ⇒ ||fˆ||22 < C |fˆ|2 dmn .
S
A natural problem is to describe all determining sets. The following theorem
gives such a description.
A set S ⊂ Rn is relatively dense if there exists r, δ > 0 such that
B(x, r) ∩ S > δ
for all x.
Theorem 5 (Paneah 1961, Logvinenko-Sereda, 1974). S is determining if and
only if it is relatively dense.
Paneah [121] proved this result in one dimension and one of the implications
in all dimensions, Logvinenko and Sereda finished the theorem [98].
If the support of f : R → R is contained in [−1, 1] then its Fourier transform
is an entire function of Paley-Wiener class and ||fˆ||L2 can be calculated as the l2 -
norm of fˆ(n) by Parseval’s theorem. The Paneah-Logvinenko-Sereda theorem may
be viewed as an extension of that property. A natural question about sequences
that can replace Z in Parseval’s theorem, with an equivalence of norms or a one
sided inequality, becomes a question on frames, sampling sequences and Riesz bases
in Paley-wiener spaces. Naturally, similar questions can be considered in other
spaces of analytic functions. For examples of deep results in this direction and
further references see a paper by Hruschev, Nikolski and Pavlov on Riesz bases of
reproducing kernels and exponential functions [68] or a paper by Ortega-Cerda and
Seip on Fourier frames [118]. The problem on Pólya sequences discussed in chapter
3 can be viewed as an L∞ -version of the same problem.
2.6. Support/Zero sets. Uniqueness and completeness problems. If
the support of a non-zero function f is bounded then fˆ is an entire function whose
zero set cannot be too large. The problem of obtaining quantative estimates on
the density and balance of such zero sets is treated in many classical texts, see for
10 1. MATHEMATICAL SHAPES OF UNCERTAINTY

instance [97]. Basic results in this direction use the growth estimates of Paley-
Wiener functions as functions of exponential type. Originating from this approach
is a whole branch of complex analysis dealing with distributions of zero sets of
entire functions in different classes of growth near infinity, see the well-known book
by Levin [94].
The study of zero sets is closely related to the study of uniqueness sets via a
trivial observation that if F is not uniquely determined in its class by its values on
Λ = {λn } ⊂ C, then Λ is in the zero set of F − G for some G from the same class.
Another close relative is the set of completeness problems.
We say that a system of vectors in a Banach space is complete if finite linear
combinations of these vectors are dense in the space. Given a space and a system of
vectors, an important problem that comes from many applications is to determine
if the system is complete in the space. The space usually appears as a space of
functions and the system is formed by some special or elementary functions often
called harmonics.
A general problem of completeness of harmonics in a Banach space gave Har-
monic Analysis its name. The area of UP contains many completeness problems,
some of which will be discussed in detail in this text.
If a system is complete, then the existence of a function orthogonal to all of
the harmonics violates the uncertainty principle in some form. To give an example
of this relation let us consider one of the most celebrated of such problems, the
Beurling–Malliavin problem.
Let Λ = {λn } ⊂ C be a sequence of complex numbers and let EΛ = {e−2πiλn }
be the system of complex exponentials with frequencies from Λ. One of the funda-
mental questions of analysis asks when is EΛ complete in L2 (−a, a). By the duality
of L2 , one may equivalently ask when does there exist a function in f ∈ L2 (−a, a)
orthogonal to all exponentials from EΛ . If such a function f exists, then it’s Fourier
transform is an entire function of the Paley-Wiener class PWa vanishing on Λ.
Conversely, the Paley-Wiener theorem says that any PWa -function vanishing on
Λ is a Fourier transform of such an f . Hence, the Beurling–Malliavin problem of
completeness of EΛ in L2 (−a, a) is equivalent to a problem of UP on zero sets of
entire functions with bounded spectrum.
The Beurling–Malliavin problem was solved by Beurling and Malliavin in a
series of paper in the early 1960’s. In these notes we discuss this problem together
with its recent extensions and further completeness problems, see chapters 5, 7 and
8.

3. Function theoretic background

In this section we provide the minimal function theoretic background necessary
for the rest of the text. The definitions and statements are given in a very condensed
form. A researcher who wants to work in the presented area of UP will certainly
need a deeper knowledge of these basics. In that regard let us mention two books
on H p -theory, [52, 84], a book on model spaces KΘ , [116], a survey paper on Clark
theory, [126], and a book and a chapter on de Branges’ spaces [26], [43, Chapter
6]. A text that will include all these topics at once is currently in preparation [105].
A standard subject of complex function theory is the study of linear spaces of
analytic functions in a fixed complex domain. In view of the Riemann conformal
mapping theorem, all simply connected domains split into two (unequal) classes
 3. FUNCTION THEORETIC BACKGROUND 11

of conformal equivalence, the whole plane C and the rest of such domains. The
standard choice for the representative of the latter class is the unit disk D or the
upper half-plane C+ . Correspondingly, most of complex function theory splits into
two main branches: the theory of entire functions and the theory of functions in D
or C+ .
Function theory in D and C+ can be developed in parallel, due to their con-
formal equivalence. It is convenient, however, to keep some minor differences that
prove useful in applications. For instance, as the reader will see below, the Hardy
spaces in C+ are not mere conformal images of the Hardy spaces in D. At the same
time, the difference between the two versions of the same space is one dimensional
(i.e. H 2 in the disk contains constants, while H 2 in the half-plane does not) and all
the basic results are true for both spaces with minor adjustments. The same holds
for the model spaces, Toeplitz kernels, etc.
In the main part of the text, we will mostly need the spaces defined in the half-
plane. Hence, below we will only provide the C+ -versions of some of the definitions.

3.1. Hardy spaces. For 0 < p ≤ ∞ we denote by H p (D) and H p (C+ ) the
Hardy spaces of analytic functions in the unit disk D and the upper half-plane C+
correspondingly. These spaces are defined as follows.
By Hol(Ω) we denote the set of all functions holomorphic in a complex domain
Ω. For p < ∞,

H p (D) = {f ∈ Hol(D) | sup0<r<1 |f (rξ)|p dm(ξ) < ∞},

where m is the normalized Lebesgue measure on the unit circle T, m(T) = 1, and

H p (C+ ) = {f ∈ Hol(C+ ) | sup0<y |f (x + iy)|p dx < ∞}.

For p ≥ 1 the supremum from the definition raised into the power 1/p presents
the norm in H p . For 0 < p < 1, the supremum gives a metric in the corresponding
space. For p = ∞ both spaces are defined as spaces of bounded holomorphic
functions in the corresponding domains with sup-norms.
By the theorem of Fatou, all H p functions possess non-tangential boundary
limits on the boundary of the domain. Via these limits, each H p -function can be
uniquely identified with an Lp -function on the boundary. Moreover, by a version
of the maximum principle, if f ∈ H p (C+ ) (H p (D)) for p ≥ 1 and f ∗ are the
non-tangential boundary values of f on R (T) then ||f ||H p (C+ ) = ||f ∗ ||Lp (R,dx)
(||f ||H p (D) = ||f ∗ ||Lp (T,dm) ). Similar equations hold for metrics for p < 1.
This way every Hardy space H p can be identified with a closed subspace of Lp
on the boundary. In the case p = 2, H 2 becomes a Hilbert space with the inner
product inherited from the corresponding L2 :

f, g ∈ H 2 (C+ ), < f, g >H 2 = f ∗ (x)ḡ ∗ (x)dx,

f, g ∈ H 2 (D), < f, g >H 2 = f ∗ (ξ)ḡ ∗ (ξ)dm(ξ).

A standard approach is to think of f as of a function defined inside the domain

and almost everywhere on the boundary and write f instead of f ∗ in the above
formulas and in similar situations.
12 1. MATHEMATICAL SHAPES OF UNCERTAINTY

3.2. Schwarz integrals, Outer functions, Nevanlinna classes. If h ∈ L1Π

(where Π is the Poisson measure on the real line dΠ(x) = dx/(1 + x2 )) then its
Schwarz integral is
  
1 1 t
Sh(z) = − h(t)dt.
πi t−z 1 + t2
If h is a real-valued function then real and the imaginary parts of Sh are the Poisson
and the conjugate Poisson integrals of h:
Sh = Ph + iQh,

1 y
Ph(x + iy) = h(t)dt.
π (t − x)2 + y 2
  
1 x−t t
Qh(z) = + h(t)dt.
π (x − t)2 + y 2 1 + t2
Outer functions in the upper half-plane are holomorphic functions of the form
H = eSh , h ∈ L1Π .
A similar definition can be given for the unit disk.
The Nevanlinna class N (C+ ) is the class of all functions of the form f /g where
f, g ∈ H ∞ (C+ ). A slightly smaller class is the Smirnov class N + (C+ ) consisting of
all functions of the form f /g where f, g ∈ H ∞ (C+ ) and g is outer. All H p -functions
are contained in N + . The definitions for N (D) and N + (D) are analogous.
Note, that since all non-trivial H ∞ -functions have non-tangential boundary
values almost everywhere with respect to the Lebesgue measure on the boundary,
which are not equal to zero except possibly on zero sets, so do all the functions from
N and N + . Moreover, H p (C+ ) = Lp (R)∩N + , in the sense that a function f ∈ N +
belongs to H p if and only if its boundary values fall into Lp on the boundary.

3.3. Inner/outer factorizations. Inner functions are functions from H ∞

whose boundary values are equal to 1 by absolute value a.e. on the boundary.
Every inner function I can be represented as a product
I = B Λ Jμ ,
where BΛ is the Blaschke product corresponding to the sequence Λ = {λn } ⊂ C+
of zeros of I and Jμ is a singular inner function corresponding to a positive finite
singular measure μ on the boundary of the domain. In the case of C+ , μ is a
measure on R̂, μ = ν + cδ∞ where ν is Poisson-finite on R, i.e.

dν(x)
< ∞,
1 + x2
and c ≥ 0 is the mass at infinity. The singular function Jμ is defined as
Jμ = e−Sμ = e−Sν+icz ,
where Sμ is the Schwarz integral of μ. The Blaschke Product BΛ for Λ = {λn } is
defined as
 z − λn
BΛ = cn ,
z − λ¯n
 3. FUNCTION THEORETIC BACKGROUND 13

i−λn
where cn are unimodular constants chosen so that cn i− λ¯n
> 0. If Λ is an infinite
sequence then the necessary and sufficient condition for the normal convergence of
the partial products of BΛ is that Λ satisfies the Blaschke condition
 λn
< ∞.
1 + |λn |2
Every function F in N + has a unique factorization F = IH, where H is the
outer function with modulus equal to |F | a.e. on the boundary and I is an inner
function.
A mean type of a function f (z) in N + (C+ ) is defined as
τ = lim sup log |F (iy)|/y.
y→∞

It is easy to see that every function in N + (C+ ) has a non-positive mean type which
is exactly the exponent a of the singular inner function S a (z) = eiaz in the inner-
outer factorization of f (z) taken with a negative sign. Here and throughout the
text S denotes the singular inner function in C+ , S(z) = eiz .
Similar statements and formulas are true for the case of the unit disk, see for
instance [52, 84].
A special role in our notes will be played by meromorphic inner functions
(MIF) which are inner functions in the upper half-plane that can be extended
meromorphically to the whole plane. The above formulas imply that an inner
function is a MIF if and only if its Blaschke factor corresponds to a discrete sequence
Λ ⊂ C (a sequence without finite accumulation points) and the measure in the
singular factor is a point mass at infinity, i.e. Jμ = S a = eiaz for some non-negative
a.

3.4. Model spaces and Clark theory. For each inner function Θ(z) one
may consider a model subspace
KΘ = H 2  ΘH 2
of the Hardy space H 2 (C+ ) Here ‘’ stands for the orthogonal difference, i.e. KΘ
is the orthogonal complement of the space ΘH 2 = {Θf |f ∈ H 2 } in H 2 . These
subspaces play an important role in complex and harmonic analysis, as well as in
operator theory, see [116].
Each inner function Θ(z) defines a positive harmonic function
1 + Θ(z)

1 − Θ(z)
and, by the Herglotz representation, a positive measure σ such that

1 + Θ(z) 1 ydσ(t)
(1.5)  = py + , z = x + iy,
1 − Θ(z) π (x − t)2 + y 2
for some p ≥ 0. The number p can be viewed as a point mass at infinity. The
measure σ is a singular Poisson-finite measure, supported on the set where non-
tangential limits of Θ are equal to 1. The measure σ + pδ∞ on R̂ is called the Clark
measure for Θ(z).
Following standard notations, we will sometimes denote the Clark measure
defined in (1.5) by σ1 . If α ∈ C, |α| = 1 then σα is the measure defined by (1.5)
with Θ replaced by ᾱΘ.
14 1. MATHEMATICAL SHAPES OF UNCERTAINTY

Conversely, for every positive Poisson-finite singular measure σ and a number

p ≥ 0, there exists an inner function Θ(z) satisfying (1.5).
Every function f ∈ KΘ has non-tangential boundary values σ-a.e. and can be
recovered from these values via the formula
 
p 1 − Θ(z) f (t)
(1.6) f (z) = (1 − Θ(z)) f (t)(1 − Θ(t))dt + dσ(t)
2πi 2πi t−z
see [122]. If the Clark measure does not have a point mass at infinity, the formula
is simplified to
1
(1.7) f (z) = (1 − Θ(z))Kf σ
2πi
where Kf σ stands for the Cauchy integral

f (t)
(1.8) Kf σ(z) = dσ(t).
t−z
This gives an isometry of L2 (σ) onto KΘ . The Clark measure σ1 has a point mass
at infinity if and only if 1 − Θ(t) ∈ L2 (R).
Similar formulas can be written for any σα corresponding to Θ. For any α, |α| =
1 and any f ∈ KΘ , f has non-tangential boundary values σα -a.e. on R. Those
boundary values can be used in (1.6) or (1.7) to recover f .
In the case of meromorphic Θ(z) (MIF), every function f ∈ KΘ also has a
meromorphic extension in C, and it is given by the formula (1.6). The corresponding
Clark measure is discrete with masses at the points of the set {Θ = 1} given by
2π
σ({x}) = .
|Θ (x)|
Each meromorphic inner function Θ(z) can be written as Θ(t) = eiφ(t) on
R, where φ(t) is a real analytic and strictly increasing function. The function
φ(t) = arg Θ(t) is a continuous branch of the argument of Θ(z).
For any inner function Θ in the upper half-plane we denote by specΘ the closure
of the set {Θ = 1}, the set of points on the line where the non-tangential limit of Θ
is equal to 1, plus the infinite point if the corresponding Clark measure has a point
mass at infinity, i.e. if p in (1.5) is positive. If specΘ ⊂ R, then p in (1.5) is 0.
We call a sequence of real points discrete if it has no finite accumulation points.
Note that {Θ = 1} is discrete if and only if Θ is meromorphic.
If Λ ⊂ R (R̂) is a given discrete sequence, one can easily construct a meromor-
phic inner function Θ satisfying {Θ = 1} = Λ by considering a positive Poisson-
finite measure concentrated on Λ and then choosing Θ to satisfy (1.5). One can
prescribe the derivatives of Θ at Λ with a proper choice of pointmasses.
The same construction shows that an arbitrary continuous growing function
γ on R can be approximated, up to a bounded function, by the argument of a
meromorphic inner function. If Λ = {γ = 2πn} then Θ constructed as above with
{Θ = 1} = Λ satisfies |γ − arg Θ| < 2π on R. Furthermore, if Γ = {γ = (2n + 1)π}
one can easily construct Θ so that {Θ = 1} = Λ and {Θ = −1} = Γ and achieve
an even better approximation |γ − arg Θ| < π. The last construction is discussed
in sections 12.5 and 9.1 of chapter 7 in connection with the two spectra problem.
We will return to Clark theory in chapter 7. For more information and further
references see [126].
 3. FUNCTION THEORETIC BACKGROUND 15

3.5. Toeplitz kernels. Recall that the Toeplitz operator TU with a symbol
U ∈ L∞ (R) is the map
TU : H 2 → H 2 , F → P+ (U F ),
where P+ is the Riesz projection, i.e. the orthogonal projection from L2 (R) onto
the Hardy space H 2 = H 2 (C+ ). As was mentioned before, via non-tangential
boundary values H 2 (C+ ) can be identified with a closed subspace of L2 (R) which
makes such a projection correctly defined.
We will use the following notation for kernels of Toeplitz operators (or Toeplitz
kernels in H 2 ):
N [U ] = kerTU .
An important observation is that N [Θ̄] = KΘ if Θ is an inner function. Along with
H 2 -kernels, one may consider Toeplitz kernels N p [U ] in other Hardy classes H p ,
the kernel N 1,∞ [U ] in the ‘weak’ space H 1,∞ = H p ∩ L1,∞ , 0 < p < 1, or the kernel
in the Smirnov class N + (C+ ), defined as
N + [U ] = {f ∈ N + ∩ L1 (R) : Ū f¯ ∈ N + }
loc

for N and similarly for other spaces.

+
+
If Θ is a meromorphic inner function, KΘ = N + [Θ̄] can also be considered.
For more on such kernels see chapters 7 and 8.
3.6. Entire functions and de Branges’ spaces. Recall that an entire func-
tion F (z) is said to be of exponential type at most a > 0 if |F (z)| = O(ea|z| ) as
z → ∞. The infimum of such a is the exponential type of F .
A classical theorem of Krein gives a connection between the Smirnov-
Nevanlinna class N + (C+ ) and the Cartwright class Ca consisting of all entire func-
tions F (z) of exponential type ≤ 2πa that satisfy
log |F (t)| ∈ L1Π .
An entire function F (z) belongs to the Cartwright class Ca if and only if
F (z) F # (z)
−2πa
∈ N + (C+ ), and ∈ N + (C+ ),
S (z) S −2πa (z)
where F # (z) = F (z̄).
As an immediate consequence one obtains a connection between the Hardy
space H 2 (C+ ) and the Paley-Wiener space PWa . Namely, an entire function F (z)
belongs to the Paley-Wiener class PWa if and only if
F (z) F # (z)
∈ H 2 (C+ ), ∈ H 2 (C+ ).
S −2πa (z) S −2πa (z)
The definition of the de Branges spaces of entire functions may be viewed as
a generalization of the last definition of the Payley-Wiener spaces with S −a (z) re-
placed by a more general entire function. Consider an entire function E(z) satisfying
the inequality
|E(z)| > |E(z̄)|, z ∈ C+ .
Such functions are usually called de Branges functions. The de Branges space BE
associated with E(z) is defined to be the space of entire functions F (z) satisfying
F (z) F # (z)
∈ H 2 (C+ ), ∈ H 2 (C+ ).
E(z) E(z)
16 1. MATHEMATICAL SHAPES OF UNCERTAINTY

It is a Hilbert space equipped with the norm F E = F/EL2 (R) . If E(z) is of

exponential type then all the functions in the de Branges space BE are of exponential
type not greater then the type of E(z) (see, for example, the last part in the proof
of lemma 3.5 in [42]). A de Branges space is called short (or regular) if together
with every function F (z) it contains (F (z) − F (a))/(z − a) for any a ∈ C.
One of the most important features of de Branges spaces is that they admit
a second, axiomatic, definition. Let H be a Hilbert space of entire functions that
satisfies the following axioms:
• (A1) If F ∈ H, F (λ) = 0, then F (z)(z−
z−λ
λ̄)
∈ H with the same norm;
• (A2) For any λ ∈ R, point evaluation at λ is a bounded linear functional
on H;
• (A3) If F ∈ H then F # ∈ H with the same norm.
Then H = BE for a suitable de Branges function E. This is theorem 23 in [26].
Usually, for a given Hilbert space of entire functions it is not difficult to verify
the above axioms and conclude that the space is a de Branges space. It is however
a challenging problem in many situations to find a generating function E. This
problem can be viewed as a deep and abstract generalization of the inverse spectral
problem for second order differential operators.
Every de Branges function E(z) gives rise to a meromorphic inner function
Θ(z) = E # (z)/E(z) and a model space KΘ that this inner function generates.
There exists a well known isometric isomorphism between BE and KΘ given by
F → F/E.
Conversely, every meromorphic Θ can be represented as Θ(z) = E # (z)/E(z)
for some de Branges function E. As was mentioned above, in this case all Clark
measures σα of Θ are discrete and their point masses can be computed by σα (λ) =
2π/ |Θ (λ)| for λ ∈ {Θ = α}.
We will call the measures |E|2 σα , where σα is a Clark measure for Θ(z) =
#
E (z)/E(z), spectral measures of the corresponding de Branges space. It is well
known that for any spectral measure ν of a de Branges space BE the natural em-
bedding gives an isometric isomorphism between BE and L2 (ν). This isomorphism
generalizes the Parseval theorem.
As we mentioned before, on the real line each inner Θ(z) coming from a
de Branges function can be written as Θ(t) = eiθ(t) , t ∈ R, where θ(t) is real
analytic strictly increasing function, a continuous branch of the argument of Θ(z)
on R. The phase function of the corresponding de Branges space is defined by
φ(t) = θ(t)/2 and is equal to − arg E.
Function theory discussed in this section will be revisited in slightly more detail
in chapter 7.
 [Link]

CHAPTER 2

Gap Theorems

In this chapter we discuss the problem of finding the maximal possible size of
the gap in the Fourier spectrum of a measure. The Fourier spectrum of a finite
measure is the closed support of its Fourier transform on R. Conditions imposed on
the measure may take several different forms, as was discussed in the first chapter.
Classical theorems by Krein-McKean, Levinson and de Branges show that a measure
that decays fast at infinity and satisfies some regularity conditions may not have
any spectral gaps. Our first goal in this chapter is to review these classical results
and give them simple proofs following one universal approach, see section 1.
The rest of the chapter is devoted to the Gap Problem stemming from Beurling’s
theorem discussed in subsection 2.2 in chapter 1. Following the general scope of
UP, the problem asks to show that if the support of the measure is porous then its
Fourier spectrum cannot have large gaps.
Beurling’s theorem, as well as other classical gap theorems mentioned above,
give sufficient conditions for the absence of spectral gaps. Our methods allow us to
take on the next logical stage of the problem and find the maximal possible size of
the spectral gap for a measure with a given support. More precisely, for a closed set
X on the real line we give a formula for its gap characteristic, the supremum of the
size of the spectral gap taken over all non-zero finite complex measures supported
on X, see section 4.
The Gap Problem has many connections and applications in adjacent fields
of UP. In particular, the formula for the gap characteristic of a set proves to be
useful in several classical problems discussed in other chapters of these notes: a
problem on completeness of complex exponentials in L2 -spaces (the Type Problem,
chapter 6), a problem on sampling sets for entire functions of zero exponential type
(the Polya-Levinson problem, chapter 3) and a problem on oscillations of high-pass
signals (the problem by Grinevich, chapter 4).
We utilize close connections between most problems from this area of harmonic
analysis and the problem of injectivity of Toeplitz operators. Such connections are
further discussed in chapters 7 and 8. In the case of the Gap Problem, this connec-
tion is expressed by theorem 8 below. The Toeplitz approach for similar problems
was first suggested by Nikolski [116], see also [68]. Our main proof utilizes several
important ideas of the Beurling–Malliavin theory, including its famous multiplier
theorem.
One of the advantages of the Toeplitz approach is that it reveals hidden connec-
tions between various problems of analysis and mathematical physics. The relations
between the Gap Problem and the Beurling–Malliavin theory on completeness of
exponentials in L2 on an interval have been known to experts, at a rather intuitive
level, for several decades. Now we can see this connection formulated in precise
mathematical terms. Namely, the Beurling–Malliavin problem is equivalent to the

17
18 2. GAP THEOREMS

problem of triviality of the kernel of a Toeplitz operator with the symbol

φ = exp(−iax)Θ,
for a suitable meromorphic inner function Θ, while the Gap Problem reduces to
the triviality of the kernel of the Toeplitz operator with the symbol
φ̄ = exp(iax)Θ̄,
see chapter 8 and theorem 8 below.
This chapter is based on [127].

1. Classical Gap Theorems

The goal of this section is to discuss gap theorems by Krein, Levinson-McKean,
Beurling and de Branges. We formulate theorem 6, that can be viewed as a hybrid
of Beurling’s and Levinson’s theorems, and give it a short elementary proof. We
then show how to deduce the classical results from theorem 6. In some cases, instead
of deducing classical theorems from each other we prefer to give each a direct closed
proof through theorem 6.
Recall that a sequence of disjoint intervals {In } on the real line is called long
(in the sense of Beurling and Malliavin) if
 |In |2
(2.1) =∞
n
1 + dist2 (0, In )
where |In | stands for the length of In . If the sum is finite we call {In } short. If I
is an interval on R and C > 0 we denote by CI the interval with the same center
as I of length C|I|. Sometimes we call ∪In long if {In } is long, see the statement
of our next theorem.
This definition appears in many results including several statements discussed
in these notes. For readers unfamiliar with this notion of ‘longness,’ let us point
out some of its simple properties. If the intervals In are uniformly bounded  1 in
length then the sequence is short because the sum in the definition is  n2 . At
kn kn +1
the same time, if In is a subsequence of diadic intervals, < 2 , 2 >, then the
sum is infinite, no matter how rare the subsequence is. More subtle examples lie in
between these two extreme cases. The condition of shortness of {In } is equivalent
to Poisson-summability of the sum of ‘tent-functions’

Tn , Tn (x) = dist(x, R \ In ).
If the sequence is long then not only is the sum of tent functions Poisson-
unsummable, but it is not a harmonic conjugate of a Poisson-summable function
(this is a version of one of the BM theorems, e.g. lemma 26 in chapter 6). Via
this property the longness condition enters into proofs, including its use in the
Beurling–Malliavin theory.
As was mentioned in chapter 1, Beurling’s version of a Gap Theorem says that
if the support of μ is porous, then the support of μ̂ cannot have any gaps. Note
that a support of a measure on R is a closed set and its complement is a union of
disjoint open intervals (gaps).
Theorem 1. [Beurling’s Gap Theorem] Let μ ≡ 0 be a finite complex measure
on R. If the complement of the support of μ is long then the support of μ̂ does not
have any gaps, i.e. there is no interval on R where μ̂ is identically 0.
 1. CLASSICAL GAP THEOREMS 19

Levinson’s version of the theorem says that if instead of having porous support
μ decays fast at infinity, one can arrive at the same conclusion:

Theorem 2. [Levinson’s Gap Theorem] Let μ be a finite measure on R. Denote

M (x) = |μ|((x, ∞)).

Suppose that log M is not Poisson-summable on R+ . If μ̂ vanishes on an interval

then μ ≡ 0.

We will deduce the last two theorems from the following more technical ‘hybrid’
statement:

Theorem 6. Let μ be a finite measure on R whose Fourier transform vanishes

on an interval. Suppose that there exists a sequence of disjoint intervals {In } such
that
 |In | 1
(2.2) min |In |, log = ∞.
1 + dist2 (In , 0) |μ|(In )

Then μ ≡ 0.

In comparison with Beurling’s theorem, our statement says that the measure
does not have to vanish on a long sequence of intervals, it just has to be small on
it. In comparison with Levinson’s version, we say that the measure does not have
to decay fast on the whole line, just on a large enough subset of the line. These
remarks are made precise in the proofs of theorems 1 and 2 below.
The proof of theorem 6 borrows an idea from the proof of Beurling’s Gap
Theorem by Benedicks in [12].

Proof of theorem 6. Without loss of generality |In | > 1 for all n, because
the sum in (2.2) taken over all intervals of length less than 1 is finite. Suppose that
μ̂ vanishes on [−a, a]. Then its Cauchy integral Kμ (defined as in (1.8)) is divisible
by e2πiaz in C+ , in the sense that

Kμ = e2πiaz Kν,

where ν is a finite measure, ν = e−2πiaz μ, see for instance lemma 11 in chapter 3

in addition to the results from [122].
Denote by Jn the interval on R + i:
1
Jn = z| z = 1, z ∈ In .
2
Denote by μn the restriction of μ on In and put ηn = μ − μn . Notice that
Kηn (z) is holomorphic in (C \ R) ∪ In . Hence − log |Kηn (z)| is superharmonic in
{|z − ξ|  |In |/4} for any ξ ∈ Jn . Since

− log |Kμ(z)| = − log |Kν(z)| − log |e2πiaz |  a|In |

in the half-plane
{ z > |In |/8},
20 2. GAP THEOREMS

we obtain
 2π
1 |In | iφ
− log |Kηn (ξ)|  − log Kηn ξ + e dφ
2π 0 4
 2π
1 |In | iφ |In | iφ
=− log Kμ ξ + e − Kμn ξ + e dφ
2π 0 4 4
|μ|(In )
 min a|In |, − log
|In |
for any ξ ∈ Jn . On the other hand,
|Kμ(ξ)| = |Kηn (ξ) + Kμn (ξ)|  |Kηn (ξ)| + |μ|(In )|
and
|μ|(In )
− log |Kμ(ξ)|  min(|In |, − log , − log |μn |(In ))  min(|In |, − log |μ|(In ))
|In |
(recall that |In | > 1|).
Now (2.2) implies that log |Kμ| is not Poisson-summable on the line
{ z = 1}.
Therefore it is identically zero in C+ . Similarly, it is zero in C− . If Kμ is zero in
both half-planes, μ ≡ 0. 
Now Beurling’s theorem follows immediately:
Proof of theorem 1. Assume that the complement of supp μ is long. Then
the complement can be taken as {In } in (2.2). 
Next, we deduce Levinson’s theorem:
Proof of theorem 2. Suppose that log M is not Poisson-summable on R+ .
Without loss of generality, M (0) = 1. Let 0 = a0 < a1 < a2 < ... be the points
such that M (an ) = 2−n and denote by In = (an , an+1 ] the corresponding partition
of R+ . If
 n|In |
<∞
1 + dist2 (In , 0)
then log M is Poisson-summable and we have a contradiction.
If the last sum is infinite, but the sum in 2.2 is finite, i.e. the partition In
is short, then any long super-partition of In will satisfy (2.2). If the last sum is
infinite and In is long, then (2.2) is satisfied. 
As we discussed in section 2.4 of chapter 1, Levinson’s result above was later
improved by Beurling [14] who showed that an interval can be replaced with a set
of positive Lebesgue measure. It seems logical to consider such an improvement
after every gap result. Note that Beurling’s theorem itself does not admit such
a strengthening, see [85]. A further step would be to improve the condition of
vanishing on a set of positive measure to summability of the logarithm of the
density, which is possible in some situations, see for instance [146]. In this text we
will not follow this path focusing just on the Gap Problem.
A stronger gap theorem, that implies the classical statements discussed above,
belongs to de Branges (theorem 63 [26]):
 1. CLASSICAL GAP THEOREMS 21

Theorem 7 (de Branges’ Gap Theorem). Let K(x) be a continuous function

on R such that K(x)  1, log K is uniformly continuous and Poisson-unsummable.
Then there is no nonzero finite measure μ on R such that
 ∞
(2.3) Kd|μ| < ∞
−∞
and μ̂ vanishes on an interval.
Proof. Without loss of generality K  2 and K is Poisson-unsummable on
R+ . Choose points a0 , a1 , ... on R+ in the following way. Put a0 = 0. After an−1
is chosen, choose an to be the smallest point greater than an−1 such that
log K(an−1 )
log K(an ) ∈ , 2 log K(an−1 ) .
2
Note that such an always exists because K is unbounded on any ray [x, ∞). Denote
by L the step function, minorating log K defined as
L(x) = Ln = min log K
In

on each In = (an−1 , an ]. Notice that by the choice of {In }, log L  log K. In

particular, log L is Poisson-unsummable. By (2.3), μ(In )  1/Ln . Also, because of
uniform continuity of log K, log Ln  |In |. Hence the sum in (2.2) is minorated by
 |In | log Ln 
dx
2  log L(x) = ∞.
1 + dist (In , 0) 1 + x2

For a finite positive measure μ on R define

(2.4) Gpμ = sup{a > 0| ∃ f ∈ L (μ), ∀ λ ∈ [0, a],
p
f (x)e2πiλx dμ(x) = 0},

or 0 if there are no such a. By duality, for 1 < p  ∞, Gpμ can be defined as the
infimum of a such that the system of exponential functions
Ea = {e2πiλt }λ∈[0,a]
is complete in Lq (μ), p1 + 1q = 1. For p = 2, the problem of finding G2μ constitutes
the well-known Type Problem discussed in chapter 6. Cases p = 2 were considered
in several papers, see for instance articles by Koosis [87] or Levin [96] for the case
p = ∞. As will be discussed in the next section, the case p = 1 is an equivalent
reformulation of the Gap Problem.
Theorem 6 has the following partial inverse.
Proposition 1. Let μ = w(x)dx be an absolutely continuous finite measure
with w > 0 and log |w| absolutely continuous. Suppose that the sequence of intervals
In satisfying (2.2) does not exist. Then G∞μ = ∞.

Proof. Similarly to the last proof, it is not difficult to show that log |w| is
Poisson-summable. After that for any C > 0 consider the measure uμ with
u = eiCx F/w,
where F is the outer function in the upper half-plane satisfying |F | = w. 
We will return to the discussion of the classical gap theorems of this section in
chapter 4 to formulate stronger versions of these results.
22 2. GAP THEOREMS

2. Spectral gap as a property of the support

The classical Gap Theorems of the last section give sufficient conditions for the
absence of gaps in the Fourier spectrum of a measure. In this section we continue
in the direction of Beurling’s result and study further relations between gaps in
the supports of μ and μ̂, thus following the ‘support/support’ version of UP as
discussed in section 2.2 of chapter 1.
As usual, the ultimate challenge in the problems of UP is to obtain a quantative
estimate for the uncertainty. In our settings, we would like to find a formula for
the maximal possible size of the gap in the Fourier spectrum of a non-zero measure
in terms of the metric properties of its support.
Recall that M denotes the set of all finite Borel complex measures on the real
line. If X is a closed subset of the real line we denote by GX its gap characteristic
defined as

(2.5) GX = sup{a | ∃ μ ∈ M, μ ≡ 0, supp μ ⊂ X, such that μ̂ = 0 on [0, a]}.

Our goal for the remainder of this chapter will be to find a formula for the gap
characteristic of X. First, let us discuss our choice of the set-up in more details.
If μ ∈ M it seems natural, for instance, to look at the quantity Gμ = G1μ as
defined in the last section and try to find it instead of GX . However, one soon
realizes that the latter problem reduces to the former:
Proposition 2.
(2.6) Gμ = Gsupp μ .

Proof. Obviously, Gsupp μ ≥ Gμ . To prove the opposite inequality, notice

that by de Branges theorem 66 (see theorem 10 discussed in section 8), there exists
a finite discrete measure

ν= αn δxn , {xn } ⊂ supp μ,

such that μ̂ has a gap of the size greater than Gsupp μ − ε (see the remark before
lemma 9). Around each xn choose a small neighborhood Vn = (an , bn ) so that for
any sequence of points
Y = {yn }, yn ∈ Vn

there exists a non-trivial measure ηY = βn δyn such that η̂ has a gap of the size
greater than Gsupp η −ε. The existence of such a collection of neighborhoods follows
from the results of [12] (for some sequences) and [21] as well as from theorem 9
below.
Now one can choose a family of finite measures ητ , τ ∈ [0, 1] with the following
properties:

• for each τ ,

ητ = βnτ δynτ
such that ynτ ∈ Vn and η̂τ has a gap of the size greater than Gsupp η − ε
centered at 0;
 3. TOEPLITZ KERNELS AND UNIFORM APPROXIMATION 23

• the measure  1
γ= ητ dτ
0
is non-trivial and absolutely continuous with respect to μ.
It remains to notice that then the support of γ̂ has a gap of the size at least
Gsupp η − ε. 

3. Toeplitz kernels and uniform approximation

Recall that for a function φ ∈ L∞ (R) we denote by N [φ] the kernel of the
Toeplitz operator Tφ in H 2 (C+ ). The notation S a is used for the singular inner
function S a (z) = eiaz in C+ . For a meromorphic inner function Θ in C+ , specΘ
denotes the set {Θ = 1} ⊂ R̂.
One of the standard classes of Toeplitz operators consists of those with symbols
¯ where I and J are inner functions. A subclass with J = S a plays
of the form IJ,
a key role in the Gap Problem.
Definition 1. If X ⊂ R is a closed set, denote
TX = sup{ a | N [Θ̄S 2πa ] = 0 for some meromorphic inner Θ, specΘ ⊂ X}.
The following theorem will be proved in section 5 chapter 3. It shows the
connection between the Gap Problem and the problem of triviality of Toeplitz
kernels which will be used throughout these notes.

Theorem 8.
GX = TX .

The Gap Problem is closely related to problems of weighted uniform approx-

imation of continuous functions by trigonometric polynomials. This topic will be
discussed in chapters 5 and 6.
To give a simple example of such a connection we consider the following version
of the problem.
Let again X be a closed subset of the line. Denote by C0 (X) the space of
all continuous functions on X tending to 0 at infinity, with the usual sup-norm.
It is not possible to discuss approximation by trigonometric polynomials in this
particular space directly, since finite linear combinations of exponential functions
do not belong to C0 (X). The standard solution is to consider ‘generalized’ linear
combinations of exponentials, i.e. the Payley-Wiener space
PWa = {fˆ | f ∈ L2 ([−a, a])}.
Let us define
AX = inf{ a > 0 | PWa is dense in C0 (X)}
or ∞ if the set is empty. The following statement is a product of the standard
duality argument.

Proposition 3.
GX = 2AX .
24 2. GAP THEOREMS

Together with theorem 9 this statement gives a formula for AX .

4. A formula for the gap characteristic of a set

Until recently we could calculate the gap characteristic GX only for a small
collection of sets X. For instance, to calculate GZ , notice that the function z(z −
1) csc πz is a Cauchy integral Kμ of a finite measure μ supported on Z \ {0, 1}. The
Cauchy integral Kμ(iy) decays as e−|y| along the imaginary axis. It is well known
that such a decay indicates a spectral gap, see lemma 11 in section 6, chapter 3.
Hence, GZ ≥ 1. The opposite inequality we leave to the reader as an exercise. The
conclusion is that GZ = 1.
Furthermore, the gap characteristic easily adjusts to the affine transformations
of the set, which allows us to calculate the gap characteristic of any infinite arith-
metic progression P = {an + C}n∈Z , a > 0, C ∈ R:
GP = 1/a.
Our next trivial observation is that the gap characteristic is monotone, i.e.
GX ≥ GY if Y ⊂ X. Hence for an arbitrary closed X,
GX ≥ sup{a| X contains {an + C}n∈Z }.
As it turns out, our simple observations are the first step towards a formula
for GX . The last inequality can be made an equation if instead of arithmetic
progressions we consider a wider class of sequences, whose gap characteristic we
are able to evaluate. The proper generalization of an arithmetic progression in this
context is a sequence that we call d-uniform. Before we give a definition of such
sequences we need some preparation.

Let Λ = {λ1 , ..., λn } be a finite set of points on R. Consider the quantity


(2.7) E(Λ) = log |λk − λl |.
λk ,λj ∈Λ, λk =λj

Physical interpretation:
According to the 2-dimensional Coulomb’s law, E(Λ) is the potential energy of
a system of ‘flat’ electrons placed at the points of Λ. The 2D Coulomb-gas formalism
corresponds to the planar potential theory with logarithmic potential and assumes
the potential energy at infinity to be equal to −∞, e.g. [37, 113, 137].
Physically, the 2D Coulomb’s law can be derived from the standard 3D law via
a method of ‘reduction.’ According to this method, one replaces each electron in
the plane with a uniformly charged string orthogonal to the plane. After that one
applies the 3D law and a renormalization procedure.
Thus, if Λ and Γ are two finite sequences with equal number of points, the
difference E(Λ) − E(Γ) presents work needed to move the the electrons from Γ to
Λ.

The following example will help us interpret the main definition.

 4. A FORMULA FOR THE GAP CHARACTERISTIC OF A SET 25

Key example:
Let I ⊂ R be an interval, C > 0 and let Λ be a set of k points uniformly spread
across I:
Λ = I ∩ CZ = {(n + 1)C, (n + 2)C, ..., (n + k)C}.
Then
  
(2.8) E(Λ) = log C k−1 (m − 1)!(k − m)! = k2 log |I| + O(|I|2 )
1≤m≤k

as follows from Stirling’s formula. Here the notation O(|I|2 ) corresponds to the
direction |I| → ∞ (with C remaining fixed).

Note that the energy of k points on I will never exceed the main term k2 log |I|
in the last equation as the energy is the sum of less than k2 terms each no greater
than log |I|. Thus, even though the uniform distribution of points on the interval
does not maximize the energy E(Λ), it comes within O(|I|2 ) from the maximum,
which is negligible for our purposes, see the main definition and its discussion below.
It is interesting to observe that the true maximum for the energy of k electrons on
I is achieved when they are placed at the endpoints of I and the zeros of the Jacobi
(1, 1)-polynomial of degree k − 2, see for example [74].

Let
... < a−2 < a−1 < a0 = 0 < a1 < a2 < ...
be a two-sided sequence of real points. We say that the intervals In = (an , an+1 ]
form a short partition of R if |In | → ∞ as n → ±∞ and the sequence {In } is short,
i.e. the sum in (2.1) is finite.

As usual, a sequence of points Λ = {λn } ⊂ C is called discrete if it has no finite

accumulation points.

Main definition:

Let Λ = {λn } be a discrete sequence of distinct real points and let d be a

positive number. We say that Λ is a d-uniform sequence if there exists a short
partition {In } such that
(2.9) Δn = #(Λ ∩ In ) = d|In | + o(|In |) as n → ±∞ (density condition)
and
 Δ2 log |In | − En
(2.10) n
<∞ (energy condition)
n
1 + dist2 (0, In )
where

En = E(Λ ∩ In ) = log |λk − λl |.
λk ,λl ∈In , λk =λl

This definition will play a key role not only in our gap theorem below but in
the main results of chapters 4 and 3. Roughly, it can be summarized as follows: a
sequence is d-uniform if it is close to the arithmetic progression d1 Z in the precise
sense given by the two conditions.
26 2. GAP THEOREMS

The density condition says that on each interval of the short partition In our
sequence has about the same number of points, d|In |, as the said arithmetic pro-
gression. The example before the definition shows that the numerator in the energy
condition is equal (up to lower order terms) to the difference between the energy of
the arithmetic progression and the energy of Λ on In . Thus the energy condition is
a requirement that the placement of the points of Λ is close to uniform, in the sense
that the total work needed to move the points of Λ into the points of an arithmetic
progression on each interval is finite (with respect to the Poisson weight).
Notice that the series in the energy condition is positive since every term in the
sum defining En is at most log |In | and there are less than Δ2n terms. Convergence
of positive series is usually easier to analyze.
One of the features of the Gap Problem is that the density of a sequence by
itself turns out to be insufficient to find the correct formula. Out of all the sequences
with correct densities we consider only those satisfying the energy condition that
prohibits clustering of points.
The following remarks will be used in the proofs.
Remark 1. The simple inequality log |In | − log |λk − λj | > 0, which holds for
any λk , λj ∈ In , also implies that if a sequence Λ satisfies the energy condition
(2.10), then any subsequence of Λ also satisfies (2.10) on {In }. A deletion of points
from Λ will eliminate some of such positive differences from the numerator in (2.10)
which can only make the sum smaller.
Remark 2. We say that a partition {In } is monotone if |In | ≤ |In+1 | for n ≥ 0
and |In+1 | ≤ |In | for n < 0. Corollary 3 in section 7 shows that in the above defi-
nition the words “short partition” can be replaced with “short monotone partition.”
Since monotone partitions are easier to work with, this modified definition will be
used in the proof of theorem 9.
Remark 3. The requirement that the partition In = (an , an+1 ] satisfied |In | →
∞ is not essential and can be omitted if one slightly changes the definitions of Δn
and En in (2.10). One could, for instance, use
Δn = #(Λ ∩ (an − 1, an+1 + 1])
and 
En = log |λk − λl |.
λk ,λl ∈(an −1,an+1 +1], λk =λl

In [127] the density condition (2.9) was given in a slightly weaker form:
(2.11) Δn = #(Λ ∩ In ) ≥ d|In |
for all n. In view of remark 1, one can easily see that either one of the definitions
can be used in the statement of theorem 9 below.
Those readers familiar with various definitions of densities used in the area of
UP may notice similarities with our density condition. In fact, existence of a short
partition {In } on which Λ satisfies (2.11) is equivalent to the property that the
lower (interior) density of the sequence in the sense of Beurling and Malliavin, is
at least d. Such a density can be defined in several different ways.
• If Λ is a real sequence define d1 (Λ) to be the supremum of all a such that
there exists a short monotone partition In satisfying (2.11).
 5. EXAMPLES AND APPLICATIONS 27

• Denote by d2 (Λ) the supremum of all a such that there exists a short (not
necessarily monotone) partition In satisfying (2.11).

• Define d3 (Λ) to be the supremum of all a such that there exists a subse-
quence of Λ whose counting function n(x) satisfies

|n − ax|
dx < ∞.
1 + x2
This definition was used in [26].

• Finally, define d4 (Λ) to be the infimum of all a such that there exists a
long sequence of disjoint intervals In satisfying
#(Λ ∩ In ) < a|In |.
This definition will be used in chapter 3.

One can easily show that all these definitions are equivalent, i.e.
d1 (Λ) = d2 (Λ) = d3 (Λ) = d4 (Λ).
As was mentioned above, such a density d was introduced in [17], where it was
called interior density. A closely related notion of exterior density appears in the
Beurling–Malliavin theorem on completeness of exponential functions in L2 on an
interval discussed in chapter 5. We give a definition of exterior density in section
7. For further discussion of BM densities see section 3 in chapter 3.

Now we are ready to return to the Gap Problem and give the following formula
for the gap characteristic GX of a closed subset of R.
Theorem 9.
GX = sup{d | X contains a d-uniform sequence},
if the set on the right is non-empty and GX = 0 otherwise.
It will be proved in section 6.

5. Examples and applications

In this section we discuss examples related to theorem 9, including some of its

relations with existing results.

Example 1. As discussed above,  if the points of the sequence are spread uni-
formly over the interval, then En = λi ,λj ∈In log |λi −λj | is roughly (up to O(|In |2 ),
which is small for short sequences of In ) equal to Δ2n log |In |. This happens for in-
stance when the sequence Λ is separated, i.e. satisfies |λn − λn+1 | > δ > 0 for all
n. Thus for separated sequences Λ the energy condition disappears and
GΛ = di (Λ)
where di , i = 1, 2, 3, 4 is any of the equivalent densities defined in the last section,
i.e. the interior density of Λ. This is one of the results of [129] discussed in chapter
3 of these notes.
28 2. GAP THEOREMS

For example, as follows from proposition 2, if the support of a measure μ con-

tains a separated sequence of interior density D, then for any ε > 0 there exists a
non-zero function f ∈ L1 (|μ|) such that f
μ = 0 on [0, D − ε].
Example 2. Let Λ be a real sequence such that the density condition (2.9)
holds for some d > 0 and some partition {In } that satisfies a stronger log-shortness
condition:
 |In |2 log |In |
< ∞.
n
1 + dist2 (0, In )
Then we will automatically have that

 Δ2n log |In | − λi ,λj ∈In log+ |λi − λj |
< ∞.
n
1 + dist2 (0, In )
Accordingly, condition (2.10) will be significantly simplified and one will only need
to check that
 log− |λi − λj |
<∞
1 + λ2j
λi ,λj ∈Λ, λi =λj

to conclude that GΛ = d. Note that since log− is supported on [0, 1], only points
that are close to each other contribute to the last sum.
Consider, for instance, log-short partition
I0 = (−1, 1], In = (nα , (n + 1)α ], I−n = (−(n + 1)α , −nα ], n = 1, 2, ...
for some α > 1. Let an increasing discrete sequence Λ = {λn } be such that
(2.12) α|n|α−1 ≤ #(Λ ∩ In ) ≤ α|n|α−1 + 1
for all n and
λk+1 − λk  e−|k|/ log |k|
2
(2.13)
for all k, |k| > 1. Then by the previous discussion GΛ = 1.
Similarly to the last example, if μ is a finite measure whose support contains
Λ, then for any ε > 0 there exists f ∈ L1 (|μ|) such that f
μ = 0 on [0, 1 − ε].
On the other hand, if (2.12) holds, but instead of (2.13) we have that on each
In ,
λk+1 − λk  e−|k|/ log |k|
for any λk , λk+1 ∈ In , |k| > 1, then the interior density of Λ is still 1, but the
energy condition is not satisfied by any subsequence of Λ of positive interior density
on any short partition. Thus GΛ = 0.

Theorem 9 easily implies Beurling’s Gap Theorem 1. Notice that if the support
of μ has long complement, then for any short partition of R infinitely many intervals
of the partition will be contained in the gaps of X = supp μ. Hence X does not
contain a sequence Λ that satisfies the density condition (2.9) on a short partition
with d > 0. Therefore GX = 0.

Another application of theorem 9 in the ‘positive’ direction produces examples

given by the result of Benedicks in [12]. Benedicks’ theorem provided some of
the very few examples of sets with positive gap characteristic that existed in the
literature before [127]. This result is discussed in chapter 6.
 6. APPENDIX: PROOF OF THE GAP FORMULA 29

6. Appendix: Proof of the gap formula

The rest of this chapter is occupied by the proof of theorem 9 and can be
skipped by those readers not willing to go into hard technical details. The first step
of the proof follows the approach of de Branges which relies on the Krein-Milman
theorem on the existence of extremal points of a star-weakly compact convex set.
That step is not very technical but quite important for several results in these notes.
For that purpose we formulate a suitable version of de Branges’ theorem 66 from
[26], see theorem 10 in section 7, which may be of independent interest. Further
versions of that result are included in other chapters of these notes. Switching our
attention to extreme points of the set of measures with a fixed spectral gap allows
us to discretize the problem and ultimately reduce it to measures concentrated on
d-uniform sequences. That reduction is made via the Toeplitz approach with the
use of the real Dirichlet space in the upper half-plane and the Beurling–Malliavin
multiplier theorem.

Before starting the proof, let us introduce the following notations. If f is a

function on R and I ⊂ R we denote by f |I the function that is equal to f on I and
to 0 on R \ I.
Recall that by Π we denote the Poisson measure dx/(1 + x2 ) on the real line.
In particular, LpΠ = Lp (R, dx/(1 + x2 )).
We will denote by D(R) the standard Dirichlet space on R (in C+ ). Recall that
the Hilbert space D = D(R) consists of functions h ∈ L1Π such that the harmonic
extension u = u(z) of h to C+ has a finite gradient norm,

2 def
hD ≡ u∇ =
2
|∇u|2 dA < ∞,
C+

where dA is the area measure. If h ∈ D(R) is a smooth function, then we also have

h2D = h̄ 
h dx,
R

where 
h denotes the derivative of a harmonic conjugate function 
h.
The proof relies on several technical lemmas supplied in section 7.

Proof of theorem 9, part I:

First suppose that X contains a d-uniform sequence Λ = {λn } with d > 2π 1

.
We will show that GX ≥ 2π . (Obviously, it is enough to prove this for any positive
1
1
constant in place of 2π ; our choice is due to purely technical reasons.)

Choose ε > 0. Let

In = (an , an+1 ]
be the short monotone partition from the definition of d-uniform sequences corre-
sponding to Λ, see remark 2. We will assume that
1 1
|In | < #(Λ ∩ In ) ≤ |In | + 1
2π 2π
for all n, which can be achieved by deletion of some of the points from Λ, see
remark 1, and uniting finitely many In into one, if necessary. We will also assume
that |In | >> 1/ε >> 1 for all n.
30 2. GAP THEOREMS

By lemma 1 and corollary 2 we can suppose that the lengths of the intervals
(λn , λn+1 ) are bounded from above. It will be convenient for us to assume that
the endpoints of In belong to Λ, i.e. that an = λkn and an+1 = λkn+1 for some
λkn , λkn+1 ∈ Λ. We will also include the endpoints of the intervals into the energy
condition by defining En as

(2.14) En = log |λk − λl |
λkn ≤λk ,λl ≤λkn+1 , λk =λl

and assuming that (2.10) is satisfied with these En . Such an assumption can be
made because if the sum in (2.10) becomes infinite with En defined by (2.14), one
can, for instance, delete the first point λkn +1 from Λ on all In for large n. After the
addition of λkn and deletion of λkn +1 in the sum defining En , each term in (2.10)
will become smaller and the sum will remain finite. At the same time, since
|In |  #(Λ ∩ In ) → ∞,
the subsequence will still satisfy the density condition.
Our goal is to show that GΛ ≥ 1 by producing a measure on Λ with spectral
gap of the size arbitrarily close to 1. Due to connections discussed in section 3,
existence of such a measure will follow from non-triviality of a certain Toeplitz
kernel.

Since the lengths of (λn , λn+1 ) are bounded from above, we can apply lemma
5. Denote by Θ the corresponding meromorphic inner function with specΘ = Λ.
Let u = arg(ΘS̄) = arg Θ−x. First, we choose a larger partition Jn = (bn , bn+1 )
and a small ‘correction’ function v so that u − v becomes an atom on each Jn :

Claim 1. There exists a subsequence {bn } of the sequence {an } and smooth
functions v1 , v2 such that:
1) |v1 | < ε/2 and u − v1 = 0 at all an ;
2) Jn = (bn , bn+1 ) is a short monotone partition;

3) |v
 2 | < ε/2 and u − v = u − (v1 + v2 ) = 0 at all bn ;
4) Jn
(u − v)dx = 0 for all n;
5) ũ − ṽ ∈ L1Π .

Proof of claim. First, choose a smooth function v1 satisfying 1. Such a function

exists because
ε
|2πΔn − |In || ≤ 2π << |In |.
2
Notice that because the sequence In is short and
ε
(u − v1 ) > −1 − ,
2
condition 1 implies
(2.15) u − v1 ∈ L1Π .
Choose b0 = a0 = 0. Choose b1 = an1 > b0 to be the smallest element of {ak }
satisfying  an1
ε
(u − v1 )dx < (an1 − b0 )2 .
b0 8
 6. APPENDIX: PROOF OF THE GAP FORMULA 31

Notice that because of (2.15) such an an1 will always exist. After that proceed
choosing b2 , b3 , ... in the following way: If bi is chosen, choose bi+1 = ani+1 to be
the smallest element of {ak } satisfying ani+1 > bi ,
 ani+1
ε
(2.16) (u − v1 )dx < (ani+1 − bi )2
bi 8
and
ani+1 − bi ≥ bi − bi−1 .
Choose bk , k < 0 in the same way.
We claim that the resulting sequence Jk = (bk−1 , bk ) forms a short monotone
partition.
Let k be positive. By our construction, Ink is the last (rightmost) among the
intervals In contained in Jk . Notice that because of monotonicity Ink is the largest
interval among the intervals In contained in Jk . We will show that for each k
 
10
(2.17) |Jk | < + 1 |Ink |
ε
where [·] stands for the integer part of a real number.
This can be proved by induction. The basic step: By our construction b1 = an1
and  an1 −1
ε
(u − v1 )dx ≥ (an1 −1 − b0 )2 .
b0 8
Since (u − v1 ) > −1 − ε and u − v1 = 0 at all an , |u − v1 | ≤ (1 + ε)|In1 −1 | on
(b0 , an1 −1 ). Hence
 an1 −1
ε
(1 + ε)|In1 −1 |(an1 −1 − b0 ) ≥ (u − v1 )dx ≥ (an1 −1 − b0 )2
b0 8
and
1+ε
(an1 −1 − b0 ) ≤ 8 |In1 −1 |.
ε
It follows that
10
(2.18) |J1 | = (an1 −1 − b0 ) + |In1 | ≤ 9ε−1 |In1 −1 | + |In1 −1 | ≤|In1 −1 |
ε
(if ε is small enough) . For the inductional step, assume that (2.17) holds for
k = l − 1. For Jl = (bl−1 , bl ), bl = anl there are two possibilities:
 an −1
l ε
(u − v1 )dx ≥ (anl −1 − bl−1 )2
bl−1 8
or
anl −1 − bl−1 < bl−1 − bl−2 .
In the first case we prove (2.18) in the same way as in the basic step. In the
second case we notice that, by monotonicity of In , the number of intervals In
inside (bl−1 , anl −1 ) is at most (anl −1 − bl−1 )/|Inl−1 | which is strictly less than
|Jl−1 |/|Inl−1 | ≤ [10/ε] + 1. Accordingly, the number of intervals in (bl−1 , anl −1 )
is at most [10/ε]. Therefore the number of intervals in Jl = (bl−1 , bl ) is at most
[10/ε] + 1. Now, since Inl is the largest interval in Jl , we again get (2.17), which
implies shortness of Jn . The monotonicity follows from our construction.
Exploring the Variety of Random
Documents with Different Content
Thomas Coram
Adam Anderson
Gen. Booth
John Howard
Tuke (family)
Baron de Montyon
Granville Sharp
Johann Beckmann
Sir Thomas Bernard
Robert Owen
François Charles Marie Fourier
George Birkbeck
Elizabeth Fry
Sir M. H. Montefiore
Sir Thomas F. Buxton
Theobald Mathew
Lucretia Mott
Joseph Sturge
Sir Rowland Hill
B. N. M. Appert
Gerrit Smith
Framjee Nasarwanjee Patel
Victor P. Considerant
E. Vansittart Neale
Baroness Burdett-Coutts
Grace Horsley Darling
J. B. A. Godin
John B. Gough
George Jacob Holyoake
Madhowdas Vurjcevandas
Clara Barton
Louis Adolphe Bertillon
Henri Cernuschi
Mary Ashton Livermore
Sir Francis Galton
Geo. Thorndike Angell
Sir D. M. Petit
George Smith of Coalville
M. E. L. Walras
Emily Faithfull
Lyman Judson Gage
Octavia and Miranda Hill
A. Carnegie
Baron Rowton
J. D. Rockefeller
Benjamin Waugh
Frances E. Willard
F. A. Bebel
Charles Booth
Gabriel Tarde
Laurence Gronlund
Samuel Gompers
Sidney Webb
Jane Addams
Helen Gould
 CHAPTER XLIX
HEALTH AND DISEASE

You may have happened to glance at one of the text-books written for
the use of medical students and of doctors, and found that you could
hardly understand a word of it. And yet you have found, when you
consulted a specialist, and he wanted to explain to you just what was
wrong with some part of your body, that he could make it all quite
clear to you. The six hundred articles on health and disease in the
Britannica are written by specialists, most of them, indeed, by
professors in the leading medical schools; and these contributors to
the Britannica are also the authors of many of the best text-books
that practising physicians and surgeons habitually use. But in the
Britannica the specialists were writing for the general public; and for
that reason they have taken care not to be too technical either in
their point of view or in the language they use.
Right and Wrong In this present chapter of the
Way to Read Reader’s Guide, the subject of health
and disease is treated just as the Guide
treats any other department of knowledge. You may want to learn
something about it because it is one of the most wonderful branches
of science, just as you would take up the course of reading on
astronomy. Or you may feel that you ought to know more than you
do about your own body, about the way you should live in order to
preserve your health, and about the causes of the diseases to which
you are exposed. Some people will tell you that it is unwise to read
about the subject at all. That is absurd. There are no doubt
exceptional people, with unsound nerves, who will imagine they
must take every patent medicine they see advertised, and who long to
try every newly discovered serum that the newspapers tell them
about.
The Danger of Again, you may be told that if you try
“Doctoring” to learn something about health and
Yourself disease, you will be tempted to think
you know as much as the doctor; and
so neglect to go to him when you need his advice. But this objection,
again, applies only to people who lack good sense. For example, if
you read the article on Dentistry, by Dr. E. C. Kirk, dean of the
Dental Faculty of the University of Pennsylvania, it will help you to
understand whatever your dentist may be doing for you. But it will
certainly not give you the idea that you could fill your own teeth.
When you find your watch has stopped, you wind it. Then, if it
does not start, you take it to the watchmaker. If, instead of doing
that, you tried to tinker with it yourself, you would soon be in
trouble. On the other hand, it would be ridiculous to go to the
watchmaker without first finding out whether the watch merely
wanted winding, and a man ought to know enough about his watch
to connect the fact that it has stopped with the probability that he
has forgotten to wind it. The daily winding is his work, not the
watchmaker’s. The chemical and mechanical work that is going on
within you is as complicated as anything in a watch or anything that
you could see in a laboratory or factory. It is your business (and your
most important business, for if you neglect it, you will not be able to
do anything properly, for yourself or for anybody else) to keep this
machinery running, and to do that is not so simple as to wind a
watch. Your body needs food and warmth. It very probably gets too
much of both. Furthermore, the food is often unwholesome, and the
warmed air is often bad air. But unless you are a millionaire invalid,
you do not have a private doctor with you at all hours to watch the
food put on your plate and to ventilate your room.
The Kind of The average watch is better treated
Knowledge You than the average human body, and
Need when the average body goes wrong,
through the average man’s
thoughtlessness, he proceeds, without in the least knowing what is
wrong, to take violent medicines, or to experiment with some fad
about diet or underclothing or gymnastics, and to make matters very
much worse. The knowledge he can gain from the Britannica will
tend to keep him from being careless, and also from trying to doctor
himself when he needs professional care. Whether you undertake a
complete course of medical reading or not, it is certainly worth your
while to read the first group of articles mentioned in this chapter—
those which have to do with the healthy routine of life.
Eating and You will find the best introduction to
Drinking the subject of diet in general in a
section (Vol. 26, p. 799) of the article
Therapeutics, by Sir Lauder Brunton. He is one of the most famous
consulting physicians in the world, and he gives you advice which
your own doctor will certainly confirm when he tells you that the way
to avoid indigestion is to masticate your food well and sip half a pint
of hot water four times a day—when you go to bed, when you get up,
and again about an hour before luncheon and dinner, instead of
drinking anything with any meal except your breakfast. If you try
that treatment for a week, you will be glad that you looked at this
chapter of the Guide. Nutrition (Vol. 19, p. 920), by Prof. Noel
Paton and Dr. Cathcart, describes the process of nourishment and
shows how important it is to chew the food thoroughly, not only in
order to break it up, but also in order to combine with it a sufficient
supply of the chemical juices which come from the glands in the
mouth. Dietetics (Vol. 8, p. 214) shows what use your body makes of
each kind of food that you eat. This article, by the late Dr. Atwater of
the United States Department of Agriculture, who conducted the
famous government investigation of diet, and R. D. Milner, also of
the Department, contains tables showing the amount of nourishment
required by persons who are doing light or heavy muscular work, as
well as by those who lead a sedentary life. It will interest you to see
(p. 218) how the food of an American business man compares with
that of an American working in a lumber camp. The article Dietary
(Vol. 8, p. 212), describing the food given to prisoners, soldiers and
sailors in various parts of the world, contains some striking
information as to the possibilities of the simple life. In Sweden
prisoners get only two meals a day, and those consisting chiefly of
porridge or gruel; and the “punishment diet” in English prisons is
one pound of bread a day, and nothing else but water. The article
Water Supply (Vol. 28, p. 387), by G. F. Deacon, deals with the
storage and distribution of water, and shows how it should be filtered
for drinking. Sewerage (Vol. 24, p. 735) describes the sanitary
systems which prevent the pollution of streams and wells. Mineral
Waters (Vol. 18, p. 517) describes the great variety of springs from
which the table-waters in general use are obtained. Their medicinal
values are also indicated, and in the table which classifies thirty of
the most important American springs it is curious to see that nearly
all of them lie in the Appalachian Mountain chain.
Hurtful Foods Vegetarianism (Vol. 27, p. 967), by
Dr. Josiah Oldfield, describes the
various systems of diet which reject flesh, the most extreme of which
exclude everything but nuts, fruit and cereals, all to be eaten raw.
Cookery (Vol. 7, p. 74) shows how the digestibility of food is
influenced by methods of cooking, and unhesitatingly condemns the
general practice of baking meat. Adulteration (Vol. 1, p. 218), by
Dr. Otto Hehner, describes the dangers to health which arise from
the use of preservatives as well as substitutes. For the use of boracic
acid, which has been proved to be slightly unwholesome, but not
really dangerous, there is at any rate the excuse that it keeps food
from spoiling, but the article has nothing but blame for the
“coppering” of vegetables. “Many years ago some artful, if stupid,
cook found that green vegetables like peas or spinach, when cooked
in a copper pan, by preference a dirty one, showed a far more
brilliant colour than the same vegetable cooked in earthenware or
iron. The manufacturer who puts up substances like peas in pots or
tins for sale produces the same effect which the cook obtained by the
wilful addition of a substance known to be injurious to health,
namely, sulphate of copper.” Food Preservation (Vol. 10, p. 612)
also shows the risks of using carelessly canned goods. Temperance
(Vol. 26, p. 578), by Dr. Arthur Shadwell, tells the story of the
reforms that have been effected since the 18th century days when
London bars used to put up signs inviting customers to get “drunk
for one penny” or “dead drunk for twopence;” and Liquor Laws (Vol.
16, p. 759) describes temperance legislation in all parts of the world,
with a most interesting section on prohibition in the United States.
Drunkenness (Vol. 8, p. 601) deals specifically with the effects of
excess on the health.
Alimentary Canal (Vol. 1, p. 663), by Dr. Chalmers Mitchell,
describes all the organs of the body that deal with food. Digestive
Organs (Vol. 8, p. 262), by Dr. Andrew Gillespie, shows how
indigestion arises, and Dyspepsia (Vol. 8, p. 786) describes the
symptoms caused by habitual indigestion. Metabolic Diseases (Vol.
18, p. 195), by Dr. Noel Paton, covers all the maladies arising from
defective nutrition. Corpulence (Vol. 7, p. 192) tells about the
reduction of superfluous fat, while Fasting (Vol. 10, p. 193) and
Hunger and Thirst (Vol. 13, p. 931) discuss the intentional or
accidental cutting down of the usual food supply. Famine (Vol. 10, p.
166) gives a most interesting account of the disasters with which crop
failures still threaten Asiatic countries. The feeding of young children
is, of course, a distinct subject, and is treated in great detail in the
article Infancy (Vol. 14, p. 513), by Dr. Harriet Hennessy.
Sleep and the Sleep (Vol. 25, p. 238), by Prof.
Want of it McKendrick, is an elaborate study of
the curious changes in the action of the
brain and other organs which take place during slumber. Insomnia
(Vol. 14, p. 644) is a practical article on the causes and treatment of
sleeplessness. Between absolutely lying awake and obtaining a really
good night’s rest there are many intermediate stages, and the article
Dream (Vol. 8, p. 558) contains a great deal of curious information
about disturbed sleep. Somnambulism (Vol. 25, p. 393) shows that
when dreams are vivid enough to produce sleepwalking there must
be nervous trouble calling for immediate treatment. Narcotics (Vol.
19, p. 239) describes the dangers of the drugs to produce sleep; and
in Hypnotism (Vol. 14, p. 201) and Suggestion (Vol. 26, p. 48) there
is a full account of the treatment frequently used for sleeplessness
and other nervous disorders.
The Right Kind The effect of climates upon health is
of Air the subject of a special section (Vol. 6,
p. 526) of the article Climate.
Ventilation (Vol. 27, p. 1008) shows how to secure fresh air in the
house without draughts. Dust (Vol. 8, p. 713), by Dr. Aitken, the
inventor of the ingenious machine for counting the particles of dust
floating in the atmosphere, gives a very full account of the impurities
in the air. Heating (Vol. 13, p. 160) contains descriptions and
diagrams of the best methods of warming houses, and there is at the
end of the article an account of the system of steam heating
employed at Lockport, N. Y., where buildings anywhere within three
miles of the central plant are heated at a very moderate cost.
General Hygiene Baths (Vol. 3, p. 514), and
Hydropathy (Vol. 14, p. 165), and
Balneotherapeutics (Vol. 3, p. 284) describe all the bathing
treatments in which water, steam and hot air are employed. Electric
baths are described in Electrotherapeutics (Vol. 9, p. 249), and
Aerotherapeutics deals with compressed air baths. Massage (Vol.
17, p. 863), by Dr. Arthur Shadwell, describes all the systems of
rubbing. Gymnastics (Vol. 12, p. 752) gives an account of the
Swedish and other systems of hygienic exercise; and out-door
exercises of every kind are described in the articles mentioned in the
chapter of Readings in Connection with Recreations and Vacations.
Two other articles which relate to general hygiene are Disinfectants
(Vol. 8, p. 312) and Antiseptics (Vol. 2, p. 146). The proper care of
the hair is indicated in the article Baldness (Vol. 3, p. 243), where
prescriptions for lotions are given.
Various Diseases The articles already named cover
very fully the application of medical
science to the ordinary routine of life, and will help you to regulate
wisely your habits in regard to eating, sleeping and to the general
care of your body. It may be the case that you wish, for your own
sake, or for the sake of some member of your family, to carry your
reading further in respect to some one disease or some one part of
the body. In the list of articles at the end of this chapter you will find
more than two hundred, each of which deals with one disease, such
as rheumatism, catarrh, malaria or neuralgia. In the case of a very
simple trouble you will find directions for treatment, as for example
in the article Corn, where you are advised to use a solution of
salicylic acid in collodion, or, for a soft corn, to paint it with spirits of
camphor. Where the trouble is anything more serious, you should of
course consult a doctor, but you will understand what he tells you all
the better, and worry less, if you have read an article, which describes
the usual course of the disease.
Parts of the Body Again, you may have a special reason
for wishing to learn all you can about
some one part of the body: the eye, the ear, or the heart. There are
fifty articles, in the list below, each dealing with some one organ or
part of the body. The illustrations in these articles will help you to
understand the exact position of any trouble which you have read
about in the article on a disease affecting that particular part.
Another set of articles divides the body into groups of organs, one
dealing with the Nervous System, another with the Muscular
System, another with the Respiratory System, and so on. Then you
have the five general articles: Anatomy, Physiology, Pathology,
Therapeutics and Surgery, which outline all medical science. The
article Medicine gives a complete history of medical science, and its
section on Modern Progress reviews all that has been accomplished
within recent years.
More Advanced Beginning with the six articles just
Study mentioned, and then taking the more
detailed articles in the groups into
which their subjects divide them, it is quite possible to follow in the
Britannica a complete course of reading on medicine and surgery,
and you may desire to do that, just as someone else likes to read
about geology or astronomy. But do not forget that no amount of
reading can give you more than a theoretical knowledge. When your
doctor discovers what is the nature of your illness (which is much the
most difficult part of his work), and when he gives you the treatment
you need, his eye is comparing what it sees in your case, and his
hand is comparing what it touches in your case, with the thousands
of observations that he has made in the wards and in the operating
theatre of the hospital. Without going through the course that he has
gone through in the dissecting room, and studying the living body as
he has studied it, you can never know what he knows. But you will be
a more understanding patient, and a better nurse, if occasion brings
nursing for you to do, if you have learned something of medical
science from the Britannica.

ALPHABETICAL LIST OF ARTICLES IN THE

ENCYCLOPAEDIA BRITANNICA RELATING TO MEDICAL
SCIENCE

Abattoir
Abdomen
Abortion
Abscess
Abscission
Acne
Aconite
Acromegaly
Actinomycosis (Streptotrichosis)
Acupressure
Acupuncture
Adam’s Apple
Addison’s Disease
Adenoids
Adulteration
Aerotherapeutics
Ague
Ala
Albuminuria
Alienist
Alimentary Canal
Amaurosis
Ambulance
Anaemia
Anaesthesia and Anaesthetics
Anatomy
Aneurysm, or Aneurism
Angina Pectoris
Animal Heat
Ankle
Ankylosis, or Anchylosis
Ankylostomiasis
Anodyne
Antiseptics
Aphasia
Aphemia
Apnoea
Aponeurosis
Apophysis
Apoplexy
Apothecary
Appendicitis
Apyrexia
Araroba Powder
Arm
Arnica
Arteries
Arthritis
Articulation
Arytenoid
Asafetida
Ascites
Asphyxia
Asthma
Athetosis
Atrophy
Auscultation
Autopsy
Bacteriology
Baldness
Balneotherapeutics
Balsam
Baths
Bedsore
Belladonna
Beriberi
Bhang
Bibirine
Bilharziosis
Blackwater Fever
Bladder
Bladder and Prostate Diseases
Blindness
Blister
Blood
Blood-letting
Boil
Bone
Bow-leg
Brain
Breast
Bright’s Disease
Bronchiectasis
Bronchitis
Bronchotomy
Bunion
Burns and Scalds
Caesarean Section
Caisson Disease
Cajaput Oil
Calabar Bean
Cancer, or Carcinoma
Cantharides
Capsicum
Carbuncle
Cartilage
Castor Oil
Catalepsy
Catarrh
Caul
Chicken-pox
Chilblains
Chirurgeon
Cholera
Chamomile, or Camomile Flowers
Climacteric
Cleft Palate and Hare-Lip
Clinic
Club-foot
Cod-Liver Oil
Coelom and Serous Membranes
Colic
Colon
Coma
Connective Tissues
Constipation
Convulsions
Corn
Corpulence or Obesity
Cramp
Cremation
Cretinism
Croton Oil
Croup
Cubebs
Cupping
Delirium
Dengue
Dentistry
Diabetes
Diaphoretics
Diaphragm
Diarrhoea
Dietary
Dietetics
Digestive Organs
Digitalis
Dilatation
Dill
Diphtheria
Dipsomania
Disinfectants
Diuretics
Dropsy
Drowning and Life Saving
Drug
Drunkenness
Ductless Glands
Dysentery
Dyspepsia
Ear
Eczema
Elaterium
Elbow
Electrotherapeutics
Elephantiasis
Emetics
Emphysema
Enteritis
Epilepsy
Epistaxis
Epithelial, Endothelial and Glandular Tissues
Equilibrium
Ergot, or Spurred Rye
Erysipelas
Eucalyptus
Euphorbium
Excretion
Eye
Face
Fauces
Favus
Fever
Fibrin
Filariasis
Finger
Fistula
Food
Foot
Frostbite
Fumigation
Galangal
Galbanum
Gall
Gamboge
Gangrene
Gastric Ulcer
Gastritis
Gelsemium
Ginseng
Goitre
Gout
Guaco, Huaco, or Guao
Guaiacum
Guarana
Guinea-Worm
Gynaecology
Haematocele
Haemophilia
Haemorrhage
Haemorrhoids
Hammer-toe
Hand
Hashish
Hay Fever, or Summer Catarrh
Head
Heart
Heel
Hernia
Herpes
Hip
Homoeopathy
Hospital
Humane Society, Royal
Hunger and Thirst
Hydrocele
Hydrocephalus
Hydropathy
Hydrophobia, or Rabies
Hygiene
Hypertrophy
Hypnotism
Hypochondriasis
Hysteria
Icthyosis or Xeroderma
Imbecile
Infancy
Influenza
Insanity
Insomnia
Intestinal Obstruction
Intestine
Intoxication
Ipecacuanha
Jaborandi
Jalap
Jaundice
Jaw
Joints
Kámalá
Kala-Azar
Kidney Diseases
Kino
Knee
Kousso
Laryngitis
Laudanum
Lead Poisoning
Leg
Leontiasis Ossea
Leprosy
Lethargy
Lichen
Ligament
Lip
Liver
Lobe
Locomotor Ataxia
Lumbago
Lung
Lupus
Lymph
Lymphatic System
Malaria
Malta Fever
Mammary Gland
Massage
Matrix
Measles
Medicine
Medical Education
Medical Jurisprudence
Ménière’s Disease
Meningitis
Metabolic Diseases
Midwife
Mineral Waters
Morphine
Mortification
Mouth and Salivary Glands
Mumps
Muscle and Nerve
Muscular System
Myelitis
Myxoedema
Naevus
Narcotics
Navel
Necrosis
Nepenthes
Nerve
Nervous System
Nettlerash, or Urticaria
Neuralgia
Neurasthenia
Neuritis
Neuropathology
Nose
Nosology
Nostalgia
Nursing
Nutrition
Nux Vomica
Obstetrics
Oesophagus
Officinal
Olfactory System
Ophthalmology
Opium
Ovariotomy
Pain
Palate
Pancreas
Paralysis or Palsy
Paranoia
Parasitic Diseases
Pathology
Pediculosis
Pellagra
Pelvis
Pemphigus
Pepsin
Peritonitis
Perspiration
Phagocytosis
Pharmacology
Pharmacopoeia
Pharmacy
Pharyngitis
Pharynx
Phlebitis
Phrenology
Phthisis
Physiology
Picrotoxin
Pinto
Pityriasis Versicolor
Placenta
Plague
Pleurisy or Pleuritis
Pneumonia
Podophyllin
Poison
Polypus
Poultice
Prognosis
Pruritus
Psoriasis
Psorospermiasis
Ptomaine Poisoning
Puberty
Public Health, Law of
Puerperal Fever
Purpura
Pulse
Quarantine
Quassia
Quinine
Quinsy
Raynaud’s Disease
Relapsing Fever
Reproductive System
Respiratory System
Rhamnus Purshiana
Rhatany, or Krameria Root
Rheumatism
Rheumatoid Arthritis
Rhubarb
Rickets
Ringworm
St. Vitus’ Dance, or Chorea
Salep
Salicin, Salicinum
Sanatorium
Sandalwood
Sandarach
Santonin
Sarsaparilla
Scabies, or Itch
Scalp
Scarlet Fever
Sciatica
Scrofula, or Struma
Scurvy
Sea-Sickness
Seborrhoea
Senega
Senna
Sepsis
Sewerage
Shock or Collapse
Shoulder
Sinew
Skeleton
Skin and Exoskeleton
Skin Diseases
Skull
Slaughter-house
Sleep
Sleeping-sickness
Smallpox
Sneezing
Somnambulism
Spinal Cord
Spleen
Sprue
Squill
Stammering or Stuttering
Starvation
Stethoscope
Stomach
Stramonium
Strychnine
Sumbul, or Sumbal
Sunstroke
Supra-renal Extract
Surgery
Surgical Instruments
Sweating-sickness
Sweetbread
Sympathetic System
Syncope
Taraxacum
Teeth
Temperance
Tetanus
Therapeutics
Thorax
Throat
Thyroid
Tongue
Tonsillitis
Toxicology
Tracheotomy
Trachoma
Trichinosis
Tuberculosis
Tumour
Typhoid Fever
Typhus Fever
Ulcer
Upas
Urinary System
Vaccination
Valerian
Varicose Veins
Vascular System
Vegetarianism
Veins
Venereal Diseases
Viburnum
Vivisection
Voice
Wart
Water-supply
Whitlow
Whooping-Cough
Windpipe
Wound
Wrist
Wry-neck
X-Ray Treatment
Yaws
Yellow Fever
Zymotic Diseases
 CHAPTER L
GEOGRAPHY AND EXPLORATION

A Library of The Britannica devotes nearly one

Geography fourth of all its space to geographical
subjects. You may miss the full
significance of this statement; therefore let us put it differently. The
matter in the Britannica on geography is equivalent to more than 100
ordinary volumes each containing 100,000 words, which, put on
shelves about 5 feet long, would fill a section in your library 5 shelves
high. But by the use of new India paper, this same material on
geography, combined with three times as much on other subjects of
importance, occupies in the Britannica less than 3 feet of shelf space.
The unity of plan and treatment and the high authority of the
Britannica in these articles are far beyond comparison with that you
could get in the most wisely and carefully selected hundred volumes
on Geography that would give an equivalent number of words.
A Science as well Geographical information is so
as a Body of useful that the student is likely to
Facts overlook the scientific importance of
geography in itself. The articles in the
Encyclopaedia Britannica described in this chapter, besides giving
the fullest information on countries, cities, towns, rivers, mountains,
etc., trace the development of the science from its beginning; and the
gradual increase of geographical knowledge, as told in the
Britannica, is a story of fine out-of-door adventure, of just the kind of
spirited action that has supplied the theme of the most popular
works of fiction.
This chapter will suggest an outline course of reading in
geography, systematically grouping the more important articles in
the Britannica.
 The starting point for this course of study is the article Geography
(Vol. 11, p. 619), equivalent in length to 70 pages of this Guide,
written by Hugh R. Mill, author of Hints on the Choice of
Geographical Books, etc. The story that it tells us is a most
interesting one.
What Early The early Greeks thought of the
Writers Taught earth as a flat disk, circular or elliptical
about the Earth in outline; and even in Homeric times
this supposition had “acquired a
special definiteness by the introduction of the idea of the ocean river
bounding the whole.” Hecataeus recognized two continents on the
circular disk. Herodotus, traveler and historian both (see the article
Herodotus Vol. 13, p. 381, by George Rawlinson and Edward M.
Walker), who knew only the lands around the roughly elliptical
Mediterranean Sea, was certain that the earth was not a circle
because it was longer from east to west than from north to south, and
he distinguished three continents, adding Africa to Europe and Asia.
“The effect of Herodotus’s hypothesis that the Nile must flow from
west to east before turning north in order to balance the Danube
running from west to east before turning south lingered in the maps
of Africa down to the time of Mungo Park.” Aristotle (see also the
article Aristotle, Vol. 2, p. 501, by Thomas Case, president of
Corpus Christi College, Oxford, and author of Physical Realism,
etc.,) was the real founder of scientific geography. “He demonstrated
the sphericity of the earth by three arguments, two of which are
important ... only a sphere could always throw a circular shadow on
the moon during an eclipse; and that the shifting of the horizon and
the appearance of new constellations ... as one travelled from north
to south, could only be explained on the hypothesis that the earth
was a sphere.... He formed a comprehensive theory of the variations
of climate with latitude and season ... speculated on the differences
in the character of races of mankind living in different climates, and
correlated the political forms of communities with their situation on
a seashore, or in the neighborhood of natural strongholds.” The
article Ptolemy (Vol. 22, p. 618), equivalent to 27 pages of this
Guide, by the late Sir Edward Herbert Bunbury, the historian of
ancient geography, and Dr. C. R. Beazley, author of The Dawn of
Modern Geography, etc., should be studied in conjunction with the
summary, in the article Geography, of Ptolemy’s achievements. “He
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

[Link]