(Ebook) Topics in Measure Theory and Real Analysis by

Alexander B. Kharazishvili (auth.) ISBN 9789078677208,

9789491216367, 9078677201, 9491216368 Pdf Download

https://ebooknice.com/product/topics-in-measure-theory-and-real-
analysis-2299726

★★★★★
4.9 out of 5.0 (78 reviews )

DOWNLOAD PDF

ebooknice.com
 (Ebook) Topics in Measure Theory and Real Analysis by
Alexander B. Kharazishvili (auth.) ISBN 9789078677208,
9789491216367, 9078677201, 9491216368 Pdf Download

EBOOK

Available Formats

■ PDF eBook Study Guide Ebook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

 We believe these products will be a great fit for you. Click
the link to download now, or visit ebooknice.com
to discover even more!

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason; Viles,

James ISBN 9781459699816, 9781743365571, 9781925268492,
1459699815, 1743365578, 1925268497

https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

(Ebook) Matematik 5000+ Kurs 2c Lärobok by Lena Alfredsson, Hans

Heikne, Sanna Bodemyr ISBN 9789127456600, 9127456609

https://ebooknice.com/product/matematik-5000-kurs-2c-larobok-23848312

(Ebook) SAT II Success MATH 1C and 2C 2002 (Peterson's SAT II

Success) by Peterson's ISBN 9780768906677, 0768906679

https://ebooknice.com/product/sat-ii-success-math-1c-and-2c-2002-peterson-
s-sat-ii-success-1722018

(Ebook) Set Theoretical Aspects of Real Analysis by Alexander B.

Kharazishvili ISBN 9781482242010, 9781482242027, 148224201X,
1482242028

https://ebooknice.com/product/set-theoretical-aspects-of-real-
analysis-4748354
(Ebook) Strange Functions in Real Analysis, Third Edition by
Alexander Kharazishvili ISBN 9781498773140, 1498773141

https://ebooknice.com/product/strange-functions-in-real-analysis-third-
edition-6838826

(Ebook) Master SAT II Math 1c and 2c 4th ed (Arco Master the SAT
Subject Test: Math Levels 1 & 2) by Arco ISBN 9780768923049,
0768923042

https://ebooknice.com/product/master-sat-ii-math-1c-and-2c-4th-ed-arco-
master-the-sat-subject-test-math-levels-1-2-2326094

(Ebook) Cambridge IGCSE and O Level History Workbook 2C - Depth

Study: the United States, 1919-41 2nd Edition by Benjamin
Harrison ISBN 9781398375147, 9781398375048, 1398375144,
1398375047
https://ebooknice.com/product/cambridge-igcse-and-o-level-history-
workbook-2c-depth-study-the-united-states-1919-41-2nd-edition-53538044

(Ebook) Strange Functions in Real Analysis by Kharazishvili A.B.

ISBN 9781584885825, 1584885823

https://ebooknice.com/product/strange-functions-in-real-analysis-4679128

(Ebook) Real Analysis : Theory of Measure and Integration by J.

Yeh ISBN 9789814578530, 9814578533, B00S4DHB7K

https://ebooknice.com/product/real-analysis-theory-of-measure-and-
integration-5290884
ATLANTIS S TUDIES IN M ATHEMATICS
VOLUME 2
S ERIES E DITOR : J. VAN M ILL
 Atlantis Studies in Mathematics

Series Editor:
J. van Mill, VU University Amsterdam, Amsterdam, the Netherlands

(ISSN: 1875-7634)

Aims and scope of the series

The series ‘Atlantis Studies in Mathematics’ (ASM) publishes monographs of high quality
in all areas of mathematics. Both research monographs and books of an expository nature
are welcome.

All books in this series are co-published with World Scientific.

For more information on this series and our other book series, please visit our website at:

www.atlantis-press.com/publications/books

A MSTERDAM – PARIS


c ATLANTIS PRESS / WORLD SCIENTIFIC
Topics in Measure Theory
and Real Analysis

Alexander B. Kharazishvili
A. Razmadze Mathematical Institute
Tbilisi
Republic of Georgia

A MSTERDAM – PARIS
Atlantis Press
29, avenue Laumière
75019 Paris, France

For information on all Atlantis Press publications, visit our website at:
www.atlantis-press.com

Copyright
This book, or any parts thereof, may not be reproduced for commercial purposes in any
form or by any means, electronic or mechanical, including photocopying, recording or any
information storage and retrieval system known or to be invented, without prior permission
from the Publisher.

Atlantis Studies in Mathematics

Volume 1: Topological Groups and Related Structures - A. Arhangel’skii, M. Tkachenko

ISBN: 978-90-78677-20-8 e-ISBN: 978-94-91216-36-7

ISSN: 1875-7634


c 2009 ATLANTIS PRESS / WORLD SCIENTIFIC
Preface

This book is concerned with questions of classical measure theory and related topics of
real analysis. At the beginning, it should be said that the choice of material included in the
present book was completely dictated by research interests and preferences of the author.
Nevertheless, we hope that this material will be of interest to a wide audience of mathe-
maticians and, primarily, to those who are working in various branches of modern math-
ematical analysis, probability theory, the theory of stochastic processes, general topology,
and functional analysis. In addition, we touch upon deep set-theoretical aspects of the top-
ics discussed in the book; consequently, set-theorists may detect nontrivial items of interest
to them and find out new applications of set-theoretical methods to various problems of
measure theory and real analysis. It should also be noted that questions treated in this book
are related to material found in the following three monographs previously published by
the author.

1) Transformation Groups and Invariant Measures, World Scientific Publ. Co., London-
Singapore, 1998.
2) Nonmeasurable Sets and Functions, North-Holland Mathematics Studies, Elsevier, Am-
sterdam, 2004.
3) Strange Functions in Real Analysis, 2nd edition, Chapman and Hall/CRC, Boca Raton,
2006.

For the convenience of our readers, we will first, briefly and schematically, describe the
scope of this book.
In Chapter 1, we consider the general problem of extending partial real-valued functions
which, undoubtedly, is one of the most important problems in all of contemporary math-
ematics and which deserves to be discussed thoroughly. Since the satisfactory solution to
this task requires a separate monograph, we certainly do not intend on entering deeply into

v
vi Topics in Measure Theory and Real Analysis

various aspects of the problem of extending partial functions, but rather we restrict our-
selves to several examples that are important for real analysis and classical measure theory
and vividly show the fundamental character of this problem. The corresponding examples
are given in Chapter 1 and illustrate different approaches and appropriate research methods.
Notice that some of the examples presented in this chapter are considered in more details
in subsequent sections of the book.

Chapter 2 is devoted to a special, but very important, case of the extension problem for
real-valued partial functions. Namely, we discuss therein several variants of the so-called
measure extension problem and we pay our attention to purely set-theoretical, algebraic and
topological aspects of this problem. In the same chapter, the classical method of extending
measures, developed by Marczewski (see [234] and [235]), is presented. Also, a useful
theorem is proved which enables us to extend any σ -finite measure μ on a base set E to
a measure μ  on the same E, such that all members of a given family of pairwise disjoint
subsets of E become μ  -measurable (see [1] and [13]). This theorem is then repeatedly
applied in further sections of the book.

In Chapters 3 and 4 we primarily deal with those measures on E which are invariant or
quasi-invariant with respect to a certain group of transformations of E. It is widely known
that invariant and quasi-invariant measures play a central role in the theory of topological
groups, functional analysis, and the theory of dynamical systems. We discuss some gen-
eral properties of invariant and quasi-invariant measures that are helpful in various fields
of mathematics. First of all, we mean the existence and uniqueness properties of such
measures. The problem of the existence and uniqueness of an invariant measure naturally
arises for a locally compact topological group endowed with the group of all its left (right)
translations. In this way, we come to the well-known Haar measure. The theory of Haar
measure is thoroughly covered in many text-books and monographs (see, for instance, [80],
[83], [182], [202]), so we leave aside the main aspects of this theory. But we present the
classical Bogoliubov-Krylov theorem on the existence of a dynamical system for a one-
parameter group of homeomorphisms of a compact metric space E. More precisely, we
formulate and prove a significant generalization of the Bogoliubov-Krylov statement: the
so-called fixed-point theorem of Markov and Kakutani ([93], [168]) for a solvable group of
affine continuous transformations of a nonempty compact convex set in a Hausdorff topo-
logical vector space. In the same chapters, we distinguish the following two situations: the
case when a given topological space E is locally compact and the case when E is not locally
compact. The latter case involves the class of all infinite-dimensional Hausdorff topolog-
Preface vii

ical vector spaces for which the problem of the existence of a nonzero σ -finite invariant
(respectively, quasi-invariant) Borel measure needs a specific formulation. Some results in
this direction are presented with necessary comments.
Chapter 5 is concerned with measurability properties of real-valued functions defined on
an abstract space E, when a certain class M of measures on E is determined. We introduce
three notions for a given function f acting from E into the real line R. Namely, f may be

(a) absolutely nonmeasurable with respect to M,

(b) relatively measurable with respect to M,
(c) absolutely (or universally) measurable with respect to M.

We examine these notions and show their close connections with some classical construc-
tions in measure theory. It should be pointed out that the standard concept of measurability
of f with respect to a fixed measure μ on E is a particular case of the notions (b) and (c).
In this case, the role of M is played by the one-element class {μ }.
In Chapter 6 we discuss, again from the measure-theoretical point of view, some properties
of the so-called step-functions. Since step-functions are rather simple representatives of
the class of all functions (namely, the range of a step-function is at most countable), it is
reasonable to consider them in connection with the measure extension problem. It turns out
that the behavior of such functions is essentially different in the case of ordinary measures
and in the case of invariant (quasi-invariant) measures.
In Chapter 7, we introduce and investigate the class of almost measurable real-valued func-
tions on R. This class properly contains the class of all Lebesgue measurable functions
on R and has certain interesting features. A characterization of almost measurable func-
tions is given and it is shown that any almost measurable function becomes measurable
with respect to a suitable extension of the standard Lebesgue measure λ on R.
Chapter 8 focuses on several important facts from general topology. In particular, Kura-
towski’s theorem (see, for instance, [58], [101], [149]) on closed projections is presented
with some of its applications among which we especially examine the existence of a co-
meager set of continuous nowhere differentiable functions in the classical Banach space
C[0, 1]. Also, we prove a deep theorem on the existence of Borel selectors for certain
partitions of a Polish topological space, which is essentially used in the sequel.
In Chapter 9 the concept of the weak transitivity of an invariant measure is considered
and its influence on the existence of nonmeasurable sets is underlined. Here it is vividly
shown that some old ideas of Minkowski [173] which were successfully applied by him in
viii Topics in Measure Theory and Real Analysis

convex geometry and geometric number theory, are also helpful in constructions of various
paradoxical (e.g., nonmeasurable) sets. Actually, Minkowski had at hand all the needed
tools to prove the existence of those subsets of the Euclidean space Rn (n  1), which are
nonmeasurable with respect to the classical Lebesgue measure λn on Rn .

Chapter 10 covers bad subgroups of an uncountable solvable group (G, ·). The term ”bad”,
of course, means the nonmeasurability of a subgroup with respect to a given nonzero σ -
finite invariant (quasi-invariant) measure μ on G. We establish the existence of such sub-
groups of G and, moreover, show that some of them can be applied to obtain invariant
(quasi-invariant) extensions of μ . So, despite their bad structural properties, certain non-
measurable subgroups of G have a positive side from the view-point of the general measure
extension problem.

The next two chapters (i.e., Chapters 11-12) are devoted to the structure of algebraic sums
of small (in a certain sense) subsets of a given uncountable commutative group (G, +).
Recall that the first deep result in this direction was obtained by Sierpiński in his classical
work [219] where he stated that there are two Lebesgue measure zero subsets of the real line
R, whose algebraic (i.e., Minkowski’s) sum is not Lebesgue measurable. Let us stress that
in [219] the technique of Hamel bases was heavily exploited and in the sequel such an ap-
proach became a powerful research tool for further investigations. We develop Sierpiński’s
above-mentioned result and generalize it in two directions. Namely, we consider the purely
algebraic aspect of the problem and its topological aspect as well. The difference between
these two aspects is primarily caused by two distinct concepts of “smallness” of subsets
of R.

In Chapters 13 and 14 we turn our attention to Sierpiński-Zygmund functions [225] and

study them from the point of view of the measure extension problem. It is well known
that the restriction of a Sierpiński-Zygmund function to any subset of R of cardinality
continuum is discontinuous. This circumstance directly implies that a Sierpiński-Zygmund
function is nonmeasurable in the Lebesgue sense and, moreover, is nonmeasurable with
respect to the completion of an arbitrary nonzero σ -finite diffused Borel measure on R. In
addition, no Sierpiński-Zygmund function has the Baire property.

We give two new constructions of Sierpiński-Zygmund functions.

(1) The construction of a Sierpiński-Zygmund function which is absolutely nonmeasurable

with respect to the class of all nonzero σ -finite diffused measures on R. (Notice that
this result needs some extra set-theoretical axioms.)
Preface ix

(2) The construction of a Sierpiński-Zygmund function which is relatively measurable with

respect to the class of all translation-invariant extensions of the Lebesgue measure λ
on R. (This result does not need any additional set-theoretical hypotheses.)

Chapters 15-17 are similar to each other in the sense that the main topics discussed therein
are connected with different constructions of nonseparable extensions of σ -finite measures.
Among the results presented in these chapters, let us especially mention:

(i) the construction (assuming the Continuum Hypothesis) of a nonseparable extension

of the Lebesgue measure without producing new null-sets;
(ii) the construction (also under some additional set-theoretical axioms) of nonseparable
invariant extensions of σ -finite invariant measures by using their nontrivial ergodic
components;
(iii) the construction (assuming again the Continuum Hypothesis) of a nonseparable non-
atomic left invariant σ -finite measure on any uncountable solvable group.

In Chapter 18, we consider universally measurable additive functionals. The universal

measurability is treated here in a generalized sense, namely, a real-valued functional f on
a Hilbert space E is universally measurable if and only if for any σ -finite Borel measure
μ given on E, there exists an extension μ  of μ such that f becomes measurable with
respect to μ  . It is established that there are universally measurable additive functionals
which are everywhere discontinuous. This result may be regarded as a counter-version to
the well-known statement (see, e.g., [97], [153], [154]), according to which any universally
measurable (in the usual sense) additive functional on E is necessarily continuous.
Chapter 19 is devoted to certain strange subsets of the Euclidean plane R2 . We discuss
various properties of these paradoxical sets from the measure-theoretical view-point. In
particular, a subset Z of R2 is constructed which is almost invariant under the group of all
translations of R2 , is λ2 -thick (where λ2 stands for the two-dimensional Lebesgue measure
on R2 ) and, in addition, has the property that for each straight line l in R2 , the intersection
l ∩ Z is of cardinality strictly less than the cardinality of the continuum. By using this set
Z, we define translation-invariant extensions of λ2 for which no analogue of the classical
Fubini theorem can be valid.
The final chapter is connected with certain restrictions of functions acting from R into R.
The first examples of those restrictions of measurable functions, which are defined on suf-
ficiently large subsets of R and have various nice properties, were given in widely known
x Topics in Measure Theory and Real Analysis

statements of real analysis. For instance, in accordance with the classical Luzin theorem
(see, e.g., [16], [26], [65], [80], [161], [183], and [192]), every real-valued Lebesgue mea-
surable function restricted to a certain set of strictly positive λ -measure becomes continu-
ous (and an analogous purely topological result holds true in terms of the Baire property
and category). We touch upon some other results in this direction. In particular, it is proved
that every Lebesgue measurable function g : R → R is monotone on a nonempty perfect
subset of R. At the same time, such a perfect set does not need to be of strictly posi-
tive λ -measure. The last circumstance is shown by considering Jarnik’s [88] continuous
nowhere approximately differentiable function whose existence is a rather deep theorem of
real analysis (cf. [33], [34], and [167]).

In general, we tried to present the material in a self-contained form completely accessible

to graduate and post-graduate students. Moreover, for the reader’s convenience, six Ap-
pendices are attached to the main text of this book, which can be read independently of the
material covered in the chapters.

In Appendix 1 some auxiliary set-theoretical facts and constructions are considered, which
are essential in various sections of the book. Namely, elements of infinite combinatorics
(e.g., infinite trees and König’s lemma), several delicate set-theoretical statements, and the
existence of an uncountable universal measure zero subset of R are discussed.

Appendix 2 is devoted to various theorems on the existence of measurable selectors. Re-

sults of this type are important and attractive and have found applications in numerous
branches of modern mathematics. We begin with the Choquet theorem on capacities (see
[24], [52], [187]) and show its close connection with statements about measurable selectors.
Also, we present the following two fundamental results in this topic: the theorem of Kura-
towski and Ryll-Nardzewski [151] and, as one of its consequences, the Luzin-Jankov-von
Neumann theorem (see, e.g., [99]).

In Appendix 3 deep properties of σ -finite Borel measures on metrizable topological spaces

are examined. It is proved that if the topological weight of a metric space E is not measur-
able in the Ulam sense, then any σ -finite Borel measure on E admits a separable support
(cf. [192]). This important fact is essentially used in Chapter 12.

Appendix 4 contains a detailed proof of the existence of a continuous function f : R → R

which is nowhere approximately differentiable. As mentioned above, the first example of
such a function was constructed by Jarnik in his remarkable work [88]. A function of this
type is needed for considerations in Chapter 20.
Preface xi

In Appendix 5 general properties of commutative groups (primarily, infinite commutative

groups) are examined and one useful theorem, due to Kulikov, on the algebraic structure
of such groups is proved. This theorem is utilized in several parts of the main text of the
book; see especially Chapters 10 and 11.
Appendix 6 presents elements of classical descriptive set theory. Namely, we touch upon
certain properties of Borel and analytic (Suslin) subsets of uncountable Polish spaces and
apply those properties to the question of measurability of sets or functions. (In this con-
text, Appendix 2 is also helpful.) The so-called separation principle, first introduced and
extensively studied by Luzin and Sierpiński, receives special attention. Of course, our pre-
sentation of this material is concise and superficial. The standard monographs or text-books
devoted to classical descriptive set theory are [99], [148], [150], [160], and [162]; see also
Martin’s article in [10].
Finally, we would like to note that all sections of the book, including the Appendices, are
provided with exercises which contain additional information concerning the questions un-
der discussion. Some of the exercises are quite easy but some of them involve difficult
mathematical facts and need intensive efforts for their solution. These more difficult exer-
cises are marked by the symbol ∗ and we recommend that the reader solve them in order to
better understand the subject. Finally, we also state in the text several unsolved problems
which are motivated by (or closely connected with) topics presented in this book.
The Bibliography consists of 251 titles and contains only the most relevant ones. Of course,
it is far from being complete but rather provides a basic orientation to the subject in order
to stimulate further interest of our readers in various questions of measure theory and real
analysis.

A.B. Kharazishvili
Contents

Preface v
1. The problem of extending partial functions . . . . . . . . . . . . . . . . . . 1
2. Some aspects of the measure extension problem . . . . . . . . . . . . . . . 19
3. Invariant measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4. Quasi-invariant measures . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5. Measurability properties of real-valued functions . . . . . . . . . . . . . . . 79
6. Some properties of step-functions connected with extensions of measures . . 97
7. Almost measurable real-valued functions . . . . . . . . . . . . . . . . . . . 111
8. Several facts from general topology . . . . . . . . . . . . . . . . . . . . . . 125
9. Weakly metrically transitive measures and nonmeasurable sets . . . . . . . 145
10. Nonmeasurable subgroups of uncountable solvable groups . . . . . . . . . 159
11. Algebraic sums of measure zero sets . . . . . . . . . . . . . . . . . . . . . 177
12. The absolute nonmeasurability of Minkowski’s sum of certain universal
measure zero sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
13. Absolutely nonmeasurable additive Sierpiński-Zygmund functions . . . . . 215
14. Relatively measurable Sierpiński-Zygmund functions . . . . . . . . . . . . 227
15. A nonseparable extension of the Lebesgue measure without new null-sets . 241
16. Metrical transitivity and nonseparable extensions of invariant measures . . 257
17. Nonseparable left invariant measures on uncountable solvable groups . . . 269
18. Universally measurable additive functionals . . . . . . . . . . . . . . . . . 281
19. Some subsets of the Euclidean plane . . . . . . . . . . . . . . . . . . . . . 297
20. Restrictions of real-valued functions . . . . . . . . . . . . . . . . . . . . . 313
Appendix 1. Some set-theoretical facts and constructions . . . . . . . . . . . . 339
Appendix 2. The Choquet theorem and measurable selectors . . . . . . . . . . 359

xiii
xiv Topics in Measure Theory and Real Analysis

Appendix 3. Borel measures on metric spaces . . . . . . . . . . . . . . . . . . 383

Appendix 4. Continuous nowhere approximately differentiable functions . . . 397
Appendix 5. Some facts from the theory of commutative groups . . . . . . . . 411
Appendix 6. Elements of descriptive set theory . . . . . . . . . . . . . . . . . 423

Bibliography 447

Subject Index 457

Chapter 1

The problem of extending partial functions

There are several general concepts and ideas in contemporary mathematics which play a
fundamental role in almost all of its branches. Among ideas of this kind, the concept of
extending a given partial function is of undoubted interest and of paramount importance for
various domains of mathematics. For instance, every working mathematician knows that
the problem of extending partial functions is considered and intensively studied in universal
algebra, general and algebraic topology, mathematical and functional analysis, as well as
other fields. Of course, this problem has specific features in any of the above-mentioned
disciplines and it frequently needs special approaches or appropriate research tools which
are suitable only for a given situation and are applicable to concrete mathematical objects,
for example, groups, topological spaces, ordered sets, differentiable manifolds, and other
structures.
However, this problem can also be examined from the abstract view-point and method-
ological conclusions of a general character can be made. Below, we touch upon different
aspects of the problem and illustrate them by relevant examples. Some of those examples
will be envisaged more thoroughly in subsequent sections of this book. The main goal
of our preliminary consideration is to demonstrate how the problem of extending partial
functions accumulates ideas from different areas of modern mathematics.
The best known example of this type is the famous Tietze-Urysohn theorem which states
that every real-valued continuous function defined on a closed subset of a normal topo-
logical space (E, T ) can be extended to a real-valued continuous function defined on the
whole space E (see, for instance, [58], [101], and [148]).
Another example of this kind is the classical Hahn-Banach theorem which states that a
continuous linear functional defined on a vector subspace of a given normed vector space
(E, || · ||) can be extended to a continuous linear functional having the same norm and
defined on the whole E (see any text-book of functional analysis, for instance, [56], [57],

A.B. Kharazishvili, Topics in Measure Theory and Real Analysis, Atlantis Studies in Mathematics 2, 1
DOI 10.1007/978-94-91216-36-7_1, © 2009 Atlantis Press/World Scientific
2 Topics in Measure Theory and Real Analysis

or [209]).
The third example of this sort, although far from the main topics of topology and analysis,
is the problem of extending a given partially recursive function to a recursive function.
As known, the latter should be defined on the set N (= ω ) of all natural numbers. The
existence of a partially recursive function that does not admit an extension to a recursive
function is crucial for basic statements of mathematical logic and the theory of algorithms.
It suffices to mention Gödel’s incompleteness theorem of the formal arithmetic (see, for
instance, [10] and [215]).
Obviously, many other interesting and important examples can be pointed out in this con-
text.
The present book contains selected topics of measure theory which is a necessary part of
modern mathematical and functional analysis. As is well known, ordinary measures are
real-valued functions defined on certain classes of subsets of a given base set E and having
the countable additivity property. Of course, in contemporary mathematics the so-called
vector-valued measures and operator-valued measures are also extremely important and are
used in many questions of analysis and the theory of stochastic processes, but we do not
touch them in our further considerations. Here we would like to stress especially that topics
presented in this book are primarily concentrated around the measure extension problem
which plays a significant role in numerous questions of real analysis, probability theory,
and set-theoretical topology. Actually, the measure extension problem will be central for
us in most sections of the book.
Consequently, it is reasonable to begin our preliminary discussion by considering several
facts from mathematical analysis, which are closely connected to extensions of partial
real-valued functions. Some of the facts listed below are fairly standard and accessible
to average-level students. But among the presented facts the reader will also encounter
those which are more important and deeper and which find applications in various domains
of modern mathematics.

Let R denote the real line and let X be an arbitrary subset of R.

A function f : X → R is called a partial function acting from R into R.
For this f , we may write f : R → R saying that f is a partial function whose domain is
contained in R. As usual, we denote

dom( f ) = X.

If X = R, then we obviously have an ordinary function f : R → R.

The problem of extending partial functions 3

The symbol ran( f ) denotes the range of a partial function f , i.e.,

ran( f ) = { f (x) : x ∈ dom( f )}.

If Y is any subset of R, then the symbol f |Y stands for the restriction of a partial function
f to Y .
As a rule, people working in classical mathematical analysis are often interested in the
following general question.
Does there exist an extension f ∗ : R → R of a partial function f : R → R such that f ∗ is
defined on the whole R and has certain ”nice” properties?
In particular, we may require that f ∗ should be differentiable or continuous or semicontin-
uous or monotone or convex or Borel measurable or Lebesgue measurable or should have
the Baire property. (Notice that the Baire property can be regarded as a topological version
of measurability; extensive material about this property is contained in remarkable books
[148], [176], and [192].)
An analogous question arises in a more general situation, e.g., for partial functions f acting
from subsets of an abstract set E into R, where E is assumed to be endowed with some
additional structure. In such a case an extension

f∗ : E → R

with dom( f ∗ ) = E must preserve a given structure on E or should be compatible, in an

appropriate sense, with this structure. It is clear that questions of the above-mentioned kind
quite frequently arise in mathematical analysis, general topology, and abstract algebra.
Therefore, this topic is of interest for large groups of the working mathematicians.
Below, we have made a small list of results in this direction, have commented on each of
them or have given a necessary explanation, and have referred the reader to other related
works in which extensions of partial functions are considered more thoroughly (see, [58],
[70], [83], [101], and [148]).
For the convenience of potential readers, the material below is presented in the form of
examples of statements about extensions of real-valued partial functions. By the way, we
think that in various lecture courses for students it is useful to provide them with additional
information concerning extensions of partial functions. Such an approach essentially helps
them to see more vividly deep connections and interactions between distinct fields of con-
temporary mathematics. In addition, the students should know that the general problem of
extending partial functions is important for all mathematics because this problem almost
permanently occurs in different mathematical branches and finds numerous applications.
4 Topics in Measure Theory and Real Analysis

We start with a fundamental result of purely set-theoretical flavor.

Example 1. Let {Xi : i ∈ I} be an arbitrary family of nonempty sets. A partial selector of

(for) {Xi : i ∈ I} is any family {xi : i ∈ J} where J is a subset of I and xi ∈ Xi for all indices
i ∈ J. If J = I, then {xi : i ∈ J} is said to be a selector of (for) the given family {Xi : i ∈ I}.
The natural question arises whether every partial selector of {Xi : i ∈ I} can be extended to
a selector of the same family. As is well known, this question is solved positively if and
only if the Axiom of Choice (denoted by AC) is assumed.
At the same time, even in a very special case

card(I) = 2c , (∀i ∈ I)(card(Xi ) = 2),

where c denotes the cardinality of the continuum, the question of the existence of a selector
of {Xi : i ∈ I} is highly nontrivial. For instance, as shown by Sierpiński, in this case the
positive answer to the question necessarily implies the existence of a subset of R which is
not measurable in the Lebesgue sense (see [223] and references therein; cf. also [90]).
The formal Zermelo-Fraenkel set theory (denoted usually by ZF) is the standard system
of set-theoretical statements (axioms) without the Axiom of Choice (see, e.g., [10], [91],
[145], [150], and [215]). The symbol ZFC stands for the theory ZF & AC. There are
many set-theoretical assertions equivalent (within ZF) to AC, for instance, the Zorn lemma,
Zermelo’s theorem on the existence of a well-ordering of any set, the equality a2 = a for
all infinite cardinal numbers a, and so on. (In this connection, see especially [90], [150],
and [223].)
It is remarkable that some nontrivial and interesting equivalents of AC can be formulated
in concrete mathematical disciplines. For example, in general topology we have the fun-
damental Tychonoff theorem stating that the product of any family of quasicompact spaces
is a quasicompact space, too. As demonstrated by Kelley [100], this theorem is equivalent
to AC within ZF theory. In linear algebra we have a very important theorem stating that
every vector space (over an arbitrary field) possesses at least one basis. As shown by Blass
[14], this theorem is also equivalent to AC within ZF theory.
Additionally, the importance of AC in classical mathematical analysis is well known (see,
e.g., the old extensive work by Sierpiński [217]). It suffices to remind that even a proof of
the equivalence of the two standard definitions, due to Cauchy and Heine respectively, of
the continuity at a point x ∈ dom( f ) of a partial function f : R → R needs some weak form
of AC.

The following simple result can be included in a beginner lecture course of real analysis.
The problem of extending partial functions 5

Example 2. Let a partial function f : R → R be given. Then it admits a continuous exten-

sion f ∗ : R → R defined on R if and only if for each open interval ]a, b[ ⊂ R, the restriction
of f to the set dom( f ) ∩ ]a, b[ is uniformly continuous. More generally, let {]ai , bi [ : i ∈ I}
be a family of open intervals in R such that

R = ∪{]ai , bi [ : i ∈ I}.

We can assert that a partial function f : R → R admits a continuous extension f ∗ with

dom( f ∗ ) = R if and only if the restrictions of this f to all sets

dom( f ) ∩ ]ai , bi [ (i ∈ I)

are uniformly continuous.

The proof of this fact is quite easy. It suffices to take into account that any continuous
real-valued function defined on a closed bounded subinterval of R is uniformly continuous.
Also, it is not hard to give an example of a partial function f : R → R for which there exists
a countable family {[ai , bi ] : i ∈ I} of segments such that

R = ∪{[ai , bi ] : i ∈ I}

and all restrictions f |(dom( f ) ∩ [ai , bi ]) are uniformly continuous, but f does not admit a
continuous extension f ∗ with dom( f ∗ ) = R.

The next example is far from trivial, however.

Example 3. Let X and Y be two metric spaces and let Y be complete. Suppose also that
f : X → Y is a continuous partial mapping. Then there exists a continuous partial mapping
f ∗ : X → Y extending f and defined on some Gδ -subset of X; consequently, f ∗ can be
extended to a Borel mapping acting from the whole X into Y .
This result is due to Lavrentiev and has numerous applications in descriptive set theory and
general topology (see [58], [148], [157], and Exercise 13 for Chapter 8 of this book).

Example 4. Recall that a partial function f : R → R is upper (respectively, lower) semi-

continuous if for each t ∈ R, the set

{x ∈ dom( f ) : f (x) < t}

(respectively, the set {x ∈ dom( f ) : f (x) > t}) is open in dom( f ). For f with dom( f ) = R,
this definition is equivalent to the following: f is upper (respectively, lower) semicontinu-
ous if and only if

limsupy→x f (y) = f (x)

6 Topics in Measure Theory and Real Analysis

(respectively, liminfy→x f (y) = f (x)) for all x ∈ R (see, e.g., [58], [101], [148], and [183]).
It is interesting to notice that every bounded upper (lower) semicontinuous partial function
admits a bounded upper (lower) semicontinuous extension defined on R. Let us formulate
a more precise result in this direction.
First, recall that a partial function g : R → R is locally bounded from above (from below)
if for each point x ∈ R, there exists a neighborhood U(x) such that g|U(x) is bounded from
above (from below).
A partial function g : R → R is locally bounded if it is locally bounded from above and
from below simultaneously.
We also recall that an upper semicontinuous function can take its values from the set R ∪
{−∞} and a lower semicontinuous function can take its values from the set R ∪ {+∞}.
(These assumptions are convenient in numerous topics of mathematical analysis.)
Now, let f : R → R ∪ {−∞, +∞} be any partial function. The following two assertions are
equivalent:
(a) f admits an upper (lower) semicontinuous extension f ∗ whose domain coincides with
the closure of dom( f );
(b) f is upper (lower) semicontinuous and locally bounded from above (from below).
The equivalence of (a) and (b) implies the validity of the next two statements.
(i) Let f : R → [a, b] be a partial upper semicontinuous function. Then there exists an upper
semicontinuous function f ∗ : R → [a, b] extending f .
(ii) Let f : R → [a, b] be a partial lower semicontinuous function. Then there exists a lower
semicontinuous function f ∗ : R → [a, b] extending f .
It should be mentioned that the same results hold true in a more general situation, namely,
for partial semicontinuous bounded functions acting from a normal topological space E
into the real line R. Of course, in this generalized case, the Tietze-Urysohn theorem should
be applied to E in order to obtain the required result.

The next example deals with monotone extensions of partial functions acting from R into
R.

Example 5. Let f : R → R be a partial function increasing on its domain. It is easy to

show that f can always be extended to an increasing function f ∗ defined on some maximal
(with respect to inclusion) subinterval of R. Let us denote the above-mentioned maximal
subinterval by T and let
a = inf(T ), b = sup(T ).
If a = −∞ and b = +∞, then f ∗ is the required increasing extension of f defined on the
The problem of extending partial functions 7

whole R. If a = −∞, then in view of the maximality of T , we must have

inf{ f (t) : t ∈ T } = −∞

and, therefore, f cannot be extended to an increasing function acting from R into R. Anal-
ogously, if b = +∞, then in view of the maximality of T , we must have

sup{ f (t) : t ∈ T } = +∞

and therefore f cannot be extended to an increasing function acting from R into R. We see
that in both of these cases our partial function f is not locally bounded.
A similar result holds true for any decreasing partial function f : R → R. We thus obtain
a necessary and sufficient condition for extending a given partial function to a monotone
function acting from R into R. Namely, the following two assertions are equivalent:
(a) f is extendable to a monotone function f ∗ with dom( f ∗ ) = R;
(b) f is monotone and locally bounded.
Of course, there is no problem connected with extending monotone partial functions if we
admit infinite values of functions under consideration (cf. Example 4). Indeed, in such a
case any monotone partial function

f : R → R ∪ {−∞, +∞}

can be extended to a monotone function f ∗ with dom( f ∗ ) = R.

Example 6. Let f : R → R be a partial function. Suppose that f is Borel on its domain, i.e.,
for every Borel set B ⊂ R, the pre-image f −1 (B) is a Borel subset of dom( f ) where dom( f )
is assumed to be endowed with the induced topology. It can be proved that f always admits
a Borel extension f ∗ : R → R with dom( f ∗ ) = R. However, the proof of this fact is far
from being easy. It needs a certain classification of all Borel partial functions acting from
R into R. This classification is due to Baire (see [3], [4], and [5]). According to it, any
Borel partial function f has its own Baire order α = α ( f ), where α is some countable
ordinal number. For instance, the equality α ( f ) = 0 simply means that f is continuous on
its domain. Taking into account the above-mentioned classification and Lavrentiev’s result
mentioned earlier (see Example 3), the existence of f ∗ can be established by using the
method of transfinite induction on α (for more details, see [148] or Exercise 16 for Chapter
8).

Example 7. Let f : R → R be a partial function. The following two assertions are equiva-
lent:
(a) f admits an extension f ∗ defined on R and measurable in the Lebesgue sense;
8 Topics in Measure Theory and Real Analysis

(b) there exists a Lebesgue measure zero set A ⊂ R such that the restriction of f to the set
dom( f ) \ A is a Borel function on dom( f ) \ A.
This fact can be proved by using the well-known Luzin criterion for the Lebesgue measur-
ability of real-valued functions (see, e.g., [161], [183], and [192]).
A parallel fact holds true for partial functions having the Baire property. Similarly to the
measurability in the Lebesgue sense, for a partial function f : R → R, the following two
assertions are equivalent:
(c) f admits an extension f ∗ defined on R and having the Baire property;
(d) there exists a first category set B ⊂ R such that the restriction of f to the set dom( f ) \ B
is a Borel function on dom( f ) \ B.

In connection with the latter fact, let us remark that if X is a second category subset of
R, then there always exists a function f : X → R which cannot be extended to a function
f ∗ : R → R having the Baire property. This deep result is due to Novikov (see [188]). It
is essentially based on the Axiom of Choice and some special facts from descriptive set
theory (e.g., the separation principle for analytic sets). A detailed discussion of this result
is also given in Chapter 14 of [122].

Example 8. If g : R → R is a Lebesgue measurable function (respectively, function hav-

ing the Baire property), then there exists a nonempty perfect set P ⊂ R such that g|P is
monotone on P (see Chapter 20 of this book).
Sierpiński and Zygmund proved in their joint work [225] that there exists a function

h:R→R

satisfying the following condition: for any set X ⊂ R of cardinality continuum, the restric-
tion h|X is not continuous.
In particular, this condition readily implies that for any set X ⊂ R of cardinality continuum,
the restriction h|X is not monotone on X. Indeed, it suffices to use the fact that the set of
all discontinuity points of any monotone partial function acting from R into R is at most
countable.
Assuming the Continuum Hypothesis or, more generally, Martin’s Axiom (see [10], [40],
[67], [91], and [145]), we have the following two statements.
(a) If X ⊂ R is of second category, then h|X cannot be extended to a function on R having
the Baire property (cf. Novikov’s result mentioned above).
(b) If X ⊂ R is of strictly positive outer Lebesgue measure, then h|X cannot be extended to
a function on R measurable in the Lebesgue sense.
The problem of extending partial functions 9

Notice that the validity of (a) and (b) does not need the full power of Martin’s Axiom.
Actually, it suffices to assume that any subset of R of cardinality strictly less than c is of
first category (respectively, of Lebesgue measure zero) in R.
Statement (a) directly implies that no Sierpiński-Zygmund function has the Baire property.
Statement (b) directly implies that no Sierpiński-Zygmund function is measurable in the
Lebesgue sense. Moreover, we can assert that a Sierpiński-Zygmund function is nonmea-
surable with respect to the completion of any nonzero σ -finite diffused (i.e., vanishing at
all singletons) Borel measure on R (see Exercise 2 for Chapter 13).
In addition to the above, no Sierpiński-Zygmund function is countably continuous. (A
partial function f : R → R is called countably continuous if dom( f ) admits a representation
in the form dom( f ) = ∪{Xn : n < ω }, where all restrictions f |Xn (n < ω ) are continuous.)
Sierpiński-Zygmund functions have other interesting and important properties. Many
works were devoted to functions of Sierpiński-Zygmund type (see [7], [42], [123], [124],
[148], [185], and [199]). In the sequel, we will be dealing with Sierpiński-Zygmund
functions possessing some additional properties which are of interest from the measure-
theoretical point of view and are closely connected with the measure extension problem
(see Chapters 13 and 14).

Example 9. Consider the set R as a vector space over the field Q of all rational numbers.
Let f : Q → Q denote the identity mapping. Since Q is a vector subspace of R (actually,
Q can be treated as a ”line” in the infinite-dimensional vector space R) and f is a partial
linear functional, it admits a linear extension

f∗ : R → Q

with dom( f ∗ ) = R. Of course, the construction of f ∗ is not effective because it is based on

the Axiom of Choice or on the Zorn lemma (cf. Exercise 5 for this chapter). In fact, the
obtained extension f ∗ is a solution of the Cauchy functional equation, that is we have

f ∗ (x + y) = f ∗ (x) + f ∗ (y)

for all x ∈ R and y ∈ R. At the same time, taking into account the relation

ran( f ∗ ) = Q,

we see that f ∗ is discontinuous at all points of R or, equivalently, f ∗ is a nontrivial solution

of the Cauchy functional equation (cf. [18], [82], [108], [143], and [176]). It is well
known that all nontrivial solutions of the Cauchy functional equation are nonmeasurable
10 Topics in Measure Theory and Real Analysis

with respect to the Lebesgue measure on R and do not possess the Baire property (see, for
instance, [143]). In our case, it can be observed that the set

X = {x ∈ R : f ∗ (x) = 0}

is not Lebesgue measurable and does not have the Baire property. Notice also that among
nontrivial solutions of the Cauchy functional equation, we can encounter some Sierpiński-
Zygmund functions (see Chapters 13 and 14). In addition, it should be pointed out that
certain nontrivial solutions of the Cauchy functional equation are successfully applied in
some deep geometrical questions concerning equidecomposability of polyhedra lying in
a finite-dimensional Euclidean space (see [18] where Hilbert’s third problem and related
topics are discussed in detail).

Example 10. In fact, the preceding example is purely algebraic. Another example of a
similar type is the following. Let (G, +) be a commutative group and let (H, +) be a
divisible commutative group, i.e., any equation of the form

nx = h (n ∈ N \ {0}, h ∈ H)

has a solution in H. Suppose that a partial homomorphism φ : G → H is given, which

means that φ is a homomorphism from some subgroup of G into H. Then we can assert
that there always exists a homomorphism φ ∗ : G → H extending φ (see [70], [83], [152],
or Exercise 2 in Appendix 5).
The above-mentioned result is extremely useful in all theory of commutative groups. For
instance, by applying it, we may readily prove that every commutative group can be isomor-
phically embedded in some divisible commutative group. Since all divisible commutative
groups admit a visual algebraic characterization (cf. [70] or [152]), the importance of this
result becomes quite evident.
An analogous statement concerning extensions of partial continuous homomorphisms can
be formulated in the case of a compact commutative group (G, +) and the one-dimensional
unit torus (S1 , ·) which is a canonical representative of the class of all divisible commutative
compact groups (see [83], [177], and [202]).

Example 11. We may expect that if a partial function f : R → R is defined on a small

(in an appropriate sense) subset of R, then f admits extensions with ”nice” properties as
well as extensions with ”bad” properties. Indeed, if dom( f ) is of Lebesgue measure zero
(respectively, is of first category), then f trivially can be extended to a Lebesgue measurable
function (respectively, to a function possessing the Baire property). Thus, in both of these
cases, we come to extensions with “nice” properties.
The problem of extending partial functions 11

On the other hand, it is possible to present an example of an extension of a partial func-

tion which is defined on a small subset of R, with an extremely “bad” property from the
measure-theoretical view-point. For this purpose, let us first introduce two auxiliary no-
tions which play a significant role in the sequel.
We recall that a set X ⊂ R is universal measure zero if there exists no nonzero σ -finite
diffused Borel measure on X.
Evidently, if X ⊂ R is countable, then X is universal measure zero. The question of the
existence of uncountable universal measure zero subsets of R was posed many years ago
and turned out to be closely connected with some rather delicate set-theoretical construc-
tions due to Hausdorff, Mahlo, Luzin and other authors. For instance, every Luzin subset
of R is uncountable and universal measure zero (various interesting properties of Luzin
sets are considered in [143], [148], [159], [176], and [192]; see also Chapters 5, 12 and
Appendix 1).
We shall say that a function g : R → R is if there exists no nonzero σ -finite diffused measure
μ on R such that g is measurable with respect to μ . (It is assumed in this definition that the
domain of μ may be an arbitrary σ -algebra of subsets of R, containing all singletons.)
We thus see that absolutely nonmeasurable functions (if they exist) are of more pathological
nature than well-known examples of Lebesgue nonmeasurable real-valued functions.
It can be proved that the following two assertions are equivalent:
(a) a function g : R → R is absolutely nonmeasurable;
(b) the set ran(g) is universal measure zero and for each t ∈ R, the set g−1 (t) is at most
countable.
For the proof of the equivalence (a) ⇔ (b), see Chapter 5.
Starting with this result, it is not hard to show that an absolutely nonmeasurable function
acting from R into R exists if and only if there exists a universal measure zero subset of R
of cardinality continuum. Also, we can infer the validity of the next statement.
Under the Continuum Hypothesis (or, more generally, under Martin’s Axiom), for a partial
function f : R → R, the following two assertions are equivalent:
(c) f admits an absolutely nonmeasurable extension f ∗ with dom( f ∗ ) = R;
(d) the set ran( f ) is universal measure zero and the sets f −1 (t) are countable for all t ∈ R.
From the equivalence of (c) and (d) we also obtain, under the same additional set-theoretical
assumptions, that any partial function f : R → R defined on a countable subset of R admits
an extension to an absolutely nonmeasurable function defined on the whole R.
Let us remark that according to Blumberg’s fundamental theorem [15], any function acting
12 Topics in Measure Theory and Real Analysis

from R into R can be regarded as an extension of some continuous partial function whose
domain is a countable everywhere dense subset of R (for the proof, see [15] or Exercises
21 and 22 of Chapter 8).
The existence of a Sierpiński-Zygmund function h : R → R shows that under the Contin-
uum Hypothesis, there is no uncountable set X ⊂ R such that h|X is continuous and, con-
sequently, h cannot be considered as an extension of a continuous partial function defined
on an uncountable subset of R.
On the other hand, according to a recent result of Roslanowski and Shelah [208], it is
consistent with ZFC that any function f : R → R may be regarded as an extension of a
continuous function defined on a Lebesgue nonmeasurable subset of R which, obviously,
is necessarily uncountable.
Let us also notice that under Martin’s Axiom, there exist additive absolutely nonmeasurable
functions which simultaneously are Sierpiński-Zygmund functions. One construction of
such functions will be given later in this book (see Chapter 13). It is based on the fact that
there exists a generalized Luzin subset of R which simultaneously is a vector space over
the field Q of all rational numbers.

The next example is concerned with extensions of real-valued partial functions of two vari-
ables.

Example 12. Let λ (= λ1 ) denote the Lebesgue measure on the real line R (= R1 ). Con-
sider a function of two real variables
Φ : R × [0, 1] → R.
Recall that Φ satisfies the Carathéodory conditions if the following two relations hold:
(i) for each x ∈ R, the function Φ(x, ·) : [0, 1] → R is continuous;
(ii) for each y ∈ [0, 1], the function Φ(·, y) : R → R is λ -measurable.
Functions of this type play a prominent role in mathematical analysis, the theory of ordinary
differential equations, optimization theory, and probability theory.
It is well known that if Φ satisfies the Carathéodory conditions, then Φ is measurable
with respect to the product σ -algebra dom(λ ) ⊗ B([0, 1]), where B([0, 1]) denotes the
σ -algebra of all Borel subsets of [0, 1].
Now, take a partial function of the form
F : R × [0, 1] → R,
i.e., suppose that dom(F) ⊂ R × [0, 1]. In addition, suppose that F is measurable with
respect to the product σ -algebra dom(λ ) ⊗ B([0, 1]). Then the following two assertions
are equivalent:
The problem of extending partial functions 13

(a) F admits an extension F ∗ : R × [0, 1] → R satisfying the Carathéodory conditions and

such that dom(F ∗ ) = R × [0, 1];
(b) for each x ∈ R, the partial function F(x, ·) is uniformly continuous on its domain.
This result is rather deep. (The reader may compare it with the simplest Example 2.)
Indeed, to establish the equivalence of (a) and (b), we have to apply the Choquet theorem on
capacities and the theorem on measurable selectors due to Kuratowski and Ryll-Nardzewski
(see Appendix 2).
It should be mentioned that the same result remains true if we replace the Lebesgue measure
space (R, dom(λ ), λ ) by an arbitrary σ -finite complete measure space (Ω, dom(μ ), μ ),
simultaneously replace the segment [0, 1] by a compact metric space Y and consider partial
functions of the form

F : Ω × Y → R.

For more details, see Exercise 13 of Appendix 2. Notice also that the standard Wiener
process

W : R[0,1] × [0, 1] → R,

which is regarded as a reasonable mathematical model of the Brownian motion, yields a

good example of a function of two variables, satisfying the Carathéodory conditions. Here
R[0,1] stands, as usual, for the space of all real-valued functions defined on the segment
[0, 1], and this space is assumed to be equipped with the completion of the Wiener proba-
bility measure μw . More precisely, we are dealing with a certain set Ω ⊂ R[0,1] such that:
(c) μw (Ω) = 1;
(d) for any t ∈ [0, 1], the partial function W (·,t) is measurable with respect to μw ;
(e) for any ω ∈ Ω, the trajectory W (ω , ·) is continuous on [0, 1].
Extensive information about stochastic processes and corresponding probability measures
may be found in [16], [26], [52], [187], [205], and [226].

Now, we would like to consider a more difficult special case of the problem of extending
partial functions. In fact, our last example will be concerned with extensions of real-valued
partial set-functions.
We recall that a set-function is any function whose domain is some family of sets. Equiva-
lently, we may say that a set-function f is a function whose domain is a subset of the power
set P(E) of some base set E. Thus, f can be treated as a partial function acting from the
set E  = P(E). Of course, partial functions of the form

f : P(E) → R
14 Topics in Measure Theory and Real Analysis

are of prime interest in this book because they include measures and measure type func-
tionals on E, for instance, capacities in the Choquet sense (see Appendix 2).

Example 13. The Lebesgue measure λ is a real-valued function defined on some class of
subsets of R. It is well known that the class dom(λ ) is properly contained in P(R), i.e.,
there are λ -nonmeasurable sets in R. However, the proof of this fact needs uncountable
forms of the Axiom of Choice. Thus, λ may be regarded as a partial function acting from
the power set of R into R ∪ {+∞}, i.e., λ is a partial function of the form

λ : P(R) → R ∪ {+∞}.

Several constructions of λ -nonmeasurable subsets of R are presented in the literature which

give us Vitali sets, Bernstein sets, nontrivial ultrafilters in the set N of all natural num-
bers, and other pathological sets of real numbers. Compare also Example 9, where the
λ -nonmeasurable set X associated with a certain nontrivial solution of the Cauchy func-
tional equation was pointed out.
It is natural to ask whether there exists a measure μ on R extending λ and defined on the
family of all subsets of R. This problem was originally posed by Banach and then was
studied by many famous mathematicians. Under some additional set-theoretical assump-
tions (e.g., the Continuum Hypothesis or Martin’s Axiom), the answer is negative (see, for
instance, [67], [69], [78], [79], [91], [192], [222], [224], and [238]). But it seems that this
question is undecidable within the standard system of axioms of contemporary set theory,
e.g., within the Zermelo-Fraenkel formal system ZFC.
It can be shown (see [1], [13], and Chapter 2) that for any countable disjoint family {Xi :
i ∈ I} of subsets of R, there exists a measure λ  on R extending λ and such that

{Xi : i ∈ I} ⊂ dom(λ  ).

We thus see that the countable additivity property can be preserved by λ  under the assump-
tion that the ”points” Xi (i ∈ I) onto which we extend λ are pairwise disjoint. Actually, it
was established in [1] and [13] that the same result remains valid for any σ -finite measure
μ given on an abstract set E and for any disjoint family {Y j : j ∈ J} of subsets of E. In
particular, we easily obtain from this result that for every finite family {Z1 , Z2 , ..., Zn } of
subsets of E, there exists a measure μ  extending μ and satisfying the relation

{Z1 , Z2 , ..., Zn } ⊂ dom(μ  ).

However, if we have an arbitrary infinite sequence {Z1 , Z2 , ..., Zn , ...} of subsets of E, then
we cannot assert, in general, that μ is extendable to a measure μ  for which all these subsets
The problem of extending partial functions 15

become μ  -measurable. In other words, sometimes there are countably many ”points” in
the power set of E, which all together do not admit further extensions of a given nonzero
σ -finite measure μ . For instance, we always have such ”bad” points Z1 , Z2 , ..., Zn , ... in the
power set of E where E is an arbitrary uncountable universal measure zero subset of R.
It should be noticed that the existence of an uncountable universal measure zero set E ⊂ R
is well known and can be proved within ZFC theory. In Appendix 1 of the present book
we give a construction of E starting with the classical Sierpiński partition of the product
set ω1 × ω1 , where ω1 denotes, as usual, the least uncountable cardinal number.
In this context, the works [197] and [250] should also be mentioned, in which analogous
and stronger results are obtained stating the existence of uncountable universally small
subsets of R.

Finishing this chapter, we hope that the examples just considered and concerning extensions
of partial real-valued functions are sufficiently illustrative to show the reader the importance
of the problem of extending partial functions. We will continue our discussion of this
problem in the following sections. As already said, the measure extension problem touched
upon in Example 13 will be of special interest in our further considerations.

EXERCISES

1. Give a proof of the statement presented in Example 1. Namely, verify that the following
two assertions are equivalent within ZF theory:
(a) the Axiom of Choice;
(b) if {Xi : i ∈ I} is an arbitrary family of nonempty sets, then any partial selector of {Xi :
i ∈ I} can be extended to a selector of this family.
Moreover, by using AC demonstrate that every infinite set contains a countably infinite
subset. (Note that this fact is deducible with the aid of some weak forms of AC but is not
provable within ZF theory.)

2. Let ]a, b[ be an open subinterval of R and let f : ]a, b[ → R be a convex function on

]a, b[. Show that the following two statements are equivalent:
(1) there exists a convex extension f ∗ : R → R of f defined on the whole R;
(2) f admits a continuous extension defined on [a, b] and this extension has the right-hand
side derivative at a and the left-hand side derivative at b.
Notice that the above-mentioned derivatives are assumed to be finite.

3. Give a proof of the equivalence (a) ⇔ (b), where (a) and (b) are assertions from Example
4.
16 Topics in Measure Theory and Real Analysis

4. Give a proof of the equivalence (a) ⇔ (b), where (a) and (b) are assertions from Example
5.

5∗ . Show that the set X = {x ∈ R : f ∗ (x) = 0} described in Example 9 is not Lebesgue

measurable and does not possess the Baire property.

6∗ . Let (E, d) be a metric space and let f : E → R be a partial function satisfying the
Lipschitz condition with a Lipschitz constant L, that is

| f (x) − f (y)|  Ld(x, y) (x ∈ dom( f ), y ∈ dom( f )).

Prove that there exists an extension f ∗ : E → R of f also satisfying the Lipschitz condition
with the same constant L and defined on the whole space E.

7. Let f : R → R be a function and let x ∈ R. Recall that t ∈ R is a Dini right derived

number of f at x if there exists a sequence {hn : n ∈ N} of strictly positive real numbers
such that

limn→+∞ hn = 0, t = limn→+∞ ( f (x + hn ) − f (x))/hn .

In this case, the notation t ∈ D+f (x) is frequently used.

Let f : R → R be a continuous function. Suppose that for any x ∈ R, this f admits a
nonnegative Dini right derived number at x. Show that f is increasing on R.
Starting with this result, prove that the following two assertions are equivalent for a contin-
uous function f acting from R into R:
(a) f satisfies the Lipschitz condition;
(b) there is a constant L  0 such that for any point x ∈ R, the function f has at least one
Dini right derived number whose absolute value does not exceed this L.

8. Let Bn denote the unit ball in the Euclidean space Rn and let Sn−1 denote the boundary
of Bn . Show that the following two assertions are equivalent:
(a) the identity embedding of Sn−1 into Bn does not admit a continuous extension defined
on the whole Bn ;
(b) Bn has the fixed-point property which means that for every continuous mapping f :
Bn → Bn , there exists a point x ∈ Bn such that f (x) = x.

As is well known, assertion (b) is valid and was first proved by Brower (for the proof,
see [56], [58], and [148]). Consequently, (a) is valid, too. Brower’s theorem (b) is very
deep, has numerous applications, and was generalized in many directions (see, e.g., [56]
and [57]). One of important generalizations of this theorem is due to Kakutani and states
the existence of fixed-points for certain set-valued mappings.
Other documents randomly have
different content
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!

ebooknice.com